Cover page and artwork designed by Matt Czapanskiy using nine overlaid flatfielded
images of Saturn in the FQCH4N-D (methane) filter on WFPC2 (Karkoschka & Koekemoer,
p. 315). This program is an example of our new category of Calibration Outsourcing
proposals.
The 2002 HST Calibration Workshop
Hubble After the Installation of the ACS
and the NICMOS Cooling System
Proceedings of a Workshop held at the
Space Telescope Science Institute
Baltimore, Maryland
October 17 and 18, 2002
Edited by Santiago Arribas, Anton Koekemoer, and Brad Whitmore
Published and distributed by the Space Telescope Science Institute
3700 San Martin Drive, Baltimore, MD 21218, USA
iii
Contents
Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Organizing Committee . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Participant List . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
vii
ix
xi
Part 1. ACS
Status of the Advanced Camera for Surveys . . . . . . . . . . . . . . . . . . . . . . .
3
M. Clampin, M. Sirianni, J. P. Blakeslee, and R. L. Gilliland
Astrometry with the Advanced Camera: PSFs and Distortion in the WFC and HRC
13
J. Anderson
ACS Flat Fields and Low-order “L-flat” Corrections from Observations of 47 Tucanae 23
J. Mack, R. C. Bohlin, R. L. Gilliland, R. van der Marel, G. de Marchi, and J. P. Blakeslee
On-orbit Sensitivity of ACS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
31
M. Sirianni, G. de Marchi, R. L. Gilliland, R. C. Bohlin, C. Pavlovsky, and J. Mack
The Wavelength Calibration of the WFC Grism . . . . . . . . . . . . . . . . . . . .
38
A. Pasquali, N. Pirzkal, and J. R. Walsh
Growth of Hot Pixels and Degradation of CTE for ACS . . . . . . . . . . . . . . . .
47
A. Riess
ACS Calibration Software . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
53
W. B. Sparks
The Effect of Velocity Aberration on ACS Image Processing . . . . . . . . . . . . . .
58
C. Cox and R. L. Gilliland
Extreme Red Sensitivity of ACS/WFC . . . . . . . . . . . . . . . . . . . . . . . . . .
61
R. L. Gilliland and A. Riess
Calibration of Geometric Distortion in the ACS Detectors . . . . . . . . . . . . . . .
65
G. R. Meurer, D. Lindler, J. P. Blakeslee, C. Cox, A. R. Martel, H. D. Tran, R. J. Bouwens,
H. C. Ford, M. Clampin, G. F. Hartig, M. Sirianni, and G. de Marchi
Drizzling Dithered ACS Images—A Demonstration . . . . . . . . . . . . . . . . . . .
70
M. Mutchler, A. M. Koekemoer, and W. Hack
Flat-fielding of ACS WFC Grism Data . . . . . . . . . . . . . . . . . . . . . . . . . .
74
N. Pirzkal, A. Pasquali, and J. R. Walsh
Statistical Analysis of ACS Data without Covariance in Errors . . . . . . . . . . . .
78
K. U. Ratnatunga
Bias Subtraction and Correction of ACS/WFC Frames . . . . . . . . . . . . . . . . .
82
M. Sirianni, A. R. Martel, M. J. Jee, D. Van Orsow, and W. B. Sparks
On-Orbit Performance of the ACS Solar Blind Channel . . . . . . . . . . . . . . . .
86
H. D. Tran, G. R. Meurer, H. C. Ford, A. R. Martel, M. Sirianni, R. C. Bohlin, M. Clampin,
C. Cox, G. de Marchi, G. F. Hartig, R. A. Kimble, and V. Argabright
Modelling the Fringing of the ACS CCD Detectors . . . . . . . . . . . . . . . . . . .
J. R. Walsh, N. Pirzkal, and A. Pasquali
90
iv
Contents
Part 2. STIS
STIS Calibration Status . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
97
C. R. Proffitt, P. Goudfrooij, T. M. Brown, J. E. Davies, R. I. Diaz-Miller, L. Dressel,
J. Kim Quijano, J. Maı́z-Apellániz, B. Mobasher, M. Potter, K. C. Sahu, D. J. Stys,
J. Valenti, N. R. Walborn, R. C. Bohlin, P. Barrett, I. Busko, and P. Hodge
Correcting STIS CCD Photometry for CTE Loss . . . . . . . . . . . . . . . . . . . .
105
P. Goudfrooij and R. A. Kimble
STIS Flux Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
115
R. C. Bohlin
STIS Echelle Blaze Shift Correction . . . . . . . . . . . . . . . . . . . . . . . . . . .
127
C. W. Bowers and D. Lindler
Coronagraphic Imaging with HST and STIS . . . . . . . . . . . . . . . . . . . . . .
137
C. A. Grady, C. R. Proffitt, E. M. Malumuth, B. E. Woodgate, T. R. Gull, C. W. Bowers,
S. R. Heap, R. A. Kimble, D. Lindler, and P. Plait
The STIS CCD Spectroscopic Line Spread Functions . . . . . . . . . . . . . . . . . .
148
T. R. Gull, D. Lindler, D. Tennant, C. W. Bowers, C. A. Grady, R. S. Hill, and E. M. Malumuth
FOS Post-Operational Archive and STIS Calibration Enhancement . . . . . . . . . .
162
M. R. Rosa, A. Alexov, P. Bristow, and F. Kerber
Accuracy and Precision of Measuring Emission Line Velocities with the Space Telescope Imaging Spectrograph . . . . . . . . . . . . . . . . . . . . . . . . . . . .
171
T. R. Ayres
Modelling Charge Transfer on the STIS CCD . . . . . . . . . . . . . . . . . . . . . .
176
P. Bristow, A. Alexov, F. Kerber, and M. R. Rosa
STIS Status after the Switch to Side 2 . . . . . . . . . . . . . . . . . . . . . . . . . .
180
T. M. Brown and J. E. Davies
Optimal Extraction with Sub-sampled Line-Spread Functions . . . . . . . . . . . . .
184
N. R. Collins, T. R. Gull, C. W. Bowers, and D. Lindler
Recent Improvements to STIS Pipeline Calibration . . . . . . . . . . . . . . . . . . .
189
R. I. Diaz-Miller, J. Kim Quijano, J. Valenti, C. R. Proffitt, K. C. Sahu, R. C. Bohlin,
T. M. Brown, and D. Lindler
Autofilet.pro: An Improved Method for Automated Removal of Herring-bone Pattern
Noise from CCD Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
193
R. A. Jansen, N. R. Collins, and R. A. Windhorst
Removing Fringes from STIS Slitless Spectra and WFC3 CCD Images . . . . . . . .
197
E. M. Malumuth, R. S. Hill, T. R. Gull, B. E. Woodgate, C. W. Bowers, R. A. Kimble,
D. Lindler, R. J. Hill, E. S. Cheng, D. A. Cottingham, Y. Wen, and S. D. Johnson
Absolute Flux Calibration of STIS Imaging Modes . . . . . . . . . . . . . . . . . . .
201
C. R. Proffitt, J. E. Davies, T. M. Brown, and B. Mobasher
Sensitivity Monitor Report for the STIS First-Order Modes . . . . . . . . . . . . . .
205
D. J. Stys, N. R. Walborn, I. Busko, P. Goudfrooij, C. R. Proffitt, and K. C. Sahu
2-D Algorithm for Removing STIS Echelle Scattered Light
. . . . . . . . . . . . . .
J. Valenti, I. Busko, J. Kim Quijano, D. Lindler, and C. W. Bowers
209
Contents
v
Part 3. NICMOS
NICMOS Status . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
215
A. B. Schultz, D. Calzetti, S. Arribas, T. Böker, M. Dickinson, S. Malhotra, L. Mazzuca,
B. Mobasher, K. Noll, E. W. Roye, M. Sosey, T. Wiklind, and C. Xu
NICMOS Detector Performance in the NCS Era . . . . . . . . . . . . . . . . . . . .
222
T. Böker, L. E. Bergeron, L. Mazzuca, M. Sosey, and C. Xu
NICMOS Photometric Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . .
232
M. Dickinson, M. Sosey, M. Rieke, R. C. Bohlin, and D. Calzetti
NICMOS Grism Calibrations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
241
R. I. Thompson and W. Freudling
Coronagraphy with NICMOS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
249
G. Schneider
Polarimetry with NICMOS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
259
D. C. Hines
NICMOS Cycle 10 and Cycle 11 Calibration Plans . . . . . . . . . . . . . . . . . . .
263
S. Arribas, S. Malhotra, D. Calzetti, L. E. Bergeron, T. Böker, M. Dickinson, L. Mazzuca,
B. Mobasher, K. Noll, E. W. Roye, A. B. Schultz, M. Sosey, T. Wiklind, and C. Xu
NCS NICMOS Focus and Coma Analysis . . . . . . . . . . . . . . . . . . . . . . . .
267
E. W. Roye and A. B. Schultz
Combining NICMOS Parallel Observations . . . . . . . . . . . . . . . . . . . . . . .
271
A. B. Schultz and H. Bushouse
NICMOS User Tools and Calibration Software Updates . . . . . . . . . . . . . . . .
275
M. Sosey
Part 4. WFPC2
WFPC2 Status and Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
281
B. C. Whitmore
WFPC2 Calibration and Close-Out . . . . . . . . . . . . . . . . . . . . . . . . . . . .
291
A. M. Koekemoer
WFPC2 CTE Characterization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
301
A. E. Dolphin
An Improved Distortion Solution for WFPC2 . . . . . . . . . . . . . . . . . . . . . .
311
I. R. King and J. Anderson
WFPC2 Flatfields with Reduced Noise and an Anomaly of Filter FQCH4N-D . . . .
315
E. Karkoschka and A. M. Koekemoer
Using MultiDrizzle to combine Dithered WFPC2 Images . . . . . . . . . . . . . . . .
325
G. Brammer, A. M. Koekemoer, and B. Kiziltan
WFPC2 Pointing Uncertainties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
329
G. Brammer, B. C. Whitmore, and A. M. Koekemoer
The Accuracy of WFPC2 Photometric Zeropoints . . . . . . . . . . . . . . . . . . .
I. Heyer, M. Richardson, B. C. Whitmore, and L. M. Lubin
333
vi
Contents
MultiDrizzle: An Integrated Pyraf Script for Registering, Cleaning and Combining
Images . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
337
A. M. Koekemoer, A. S. Fruchter, R. N. Hook, and W. Hack
WFPC2 Re-Commissioning After Servicing Mission 3B . . . . . . . . . . . . . . . .
341
A. M. Koekemoer, S. Gonzaga, I. Heyer, L. M. Lubin, V. Kozhurina-Platais, and B. C. Whitmore
Photometry of Saturated Stars in CCD Images . . . . . . . . . . . . . . . . . . . . .
346
J. Maı́z-Apellániz
Updated Contamination Rates for WFPC2 UV Filters . . . . . . . . . . . . . . . . .
350
M. McMaster and B. C. Whitmore
Toward a Multi-Wavelength Geometric Distortion Solution for WFPC2 . . . . . . .
354
V. Kozhurina-Platais, S. Casertano, and A. M. Koekemoer
Charge Transfer Efficiency for Very Faint Objects and a Reexamination of the Longvs.-Short Problem for the WFPC2 . . . . . . . . . . . . . . . . . . . . . . . .
359
B. C. Whitmore and I. Heyer
Part 5. Other Instruments
Optical Interferometry with HST /FGS at V > 15 . . . . . . . . . . . . . . . . . . .
367
E. Nelan and R. Makidon
The Optical Field Angle Distortion Calibration of HST Fine Guidance Sensors 1R
and 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
373
B. McArthur, G. F. Benedict, W. H. Jefferys, and E. Nelan
Wide Field Camera 3: Design, Status, and Calibration Plans . . . . . . . . . . . . .
383
J. W. MacKenty
Calibration Status of the Cosmic Origins Spectrograph Detectors . . . . . . . . . . .
390
S. V. Penton, S. Béland, and E. Wilkinson
Coronagraphic Imaging: Keck II AO and HST ACS Compared . . . . . . . . . . . .
394
P. Kalas, D. Le Mignant, F. Marchis, and J. R. Graham
Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Subject Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
399
401
vii
Preface
More than a dozen years have passed since the launch of the Hubble Space Telescope
(HST ). The telescope, like a fine wine, continues to improve with age. The installation of
the Advanced Camera for Surveys (ACS), with its factor of ten improvement in discovery efficiency, and the NICMOS Cooling System (NCS), which resucitated HST’s IR capabilities,
means that the telescope is currently more capable than it has ever been.
However, with new instruments come new challenges. Charge-transfer efficiency, pointspread functions, pedestal effects, instrumental throughputs, scattered light, line-spread
functions, cosmic-rays... whether we like it or not, the astronomical knowledge that will
appear in tomorrow’s textbooks hinges on understanding our sometimes imperfect sensory
apparatus. In addition, pushing the forefronts of science often means pushing the instruments to their limits, where all kinds of calibration “gotchas” may be hiding.
The fourth HST Calibration Workshop was held October 17–18, 2002 at the Space
Telescope Science Institute to help address these new challenges. The workshop featured
reports from the commissioning of ACS and the re-commissioning of NICMOS. New calibrations and advances in the understanding of STIS, WFPC2, FOS, and the FGS were also
presented, as well as previews of calibration plans for COS and WFC3 which are scheduled
to be launched in approximately two years. The workshop was designed to foster the sharing
of information and techniques between observers, instrument developers, and instrument
support teams. Roughly 120 astronomers attended the workshop which included approximately 30 invited talks, 40 posters, and time for demos and splinter groups on various
topics.
An electronic copy of these proceedings is available at the 2002 Calibration Workshop
Web site.1 The Abstract Booklet for the workshop can also be found at this site. We note
that in a few cases, a talk or poster was presented at the workshop that is not represented in
these proceedings. We also remind our readers that as our understanding of the instruments
continue to improve, and as the instruments themselves evolve, some of the information in
these proceedings will be superseeded. For the latest calibration information the reader
should check the instrument Web sites.2
The workshop represents a great deal of work by a number of dedicated and talented
individuals. Foremost amongst these were Dixie Shipley, the primary contact and logisitical
coordinator for the meeting, and Matt Lallo, the technical support coordinator. We would
also like to thank Robin Auer, Stefano Casertano, Matt Czpanskiy, Helmut Jenkner, Steve
Hulbert, Calvin Tullos, Ed Weibe, and Mike Wiggs for a wide range of items ranging
from WWW support to web-casting to general consulting. In addition, we thank Harry
Payne, Susan Rose (lead), and Sharon Toolan for an excellent job supporting the production
of this book, as well as all the people that gave talks, presented posters, and wrote up
the contributions that made these proceedings possible. We also wish to thank NASA
Headquarters, the HST Project at the Goddard Space Flight Center, the Johnson Space
Center, and the astronauts, for supporting the servicing mission activities. Finally, we
would like to dedicate these proceedings to the Instrument Definition Teams that built the
incredible instruments onboard the Hubble Space Telescope.
The Editors
Santiago Arribas, Anton Koekemoer, and Brad Whitmore
January 2003
1
2
http://www.stsci.edu/stsci/hst/HST_overview/documents/calworkshop/workshop2002
http://www.stsci.edu/hst/HST_overview/instruments
ix
The Organizing Committee
Santiago Arribas (co-chair)
Rosa Diaz-Miller
Harry Ferguson
Ron Gilliland
Anton Koekemoer
Ed Nelan
Charles Proffitt
Dixie Shipley
Brad Whitmore (co-chair)
xi
Participant List
Anastasia Alexov
Sahar Allam
Jay Anderson
Solomon Kwao Annan
Santiago Arribas
Thomas Ayres
Stephane Beland
Fritz Benedict
David Bennett
Louis Bergeron
Chris Blades
Torsten Boeker
Ralph Bohlin
Charles Bowers
Art Bradley
Gabriel Brammer
Paul Bristow
Thomas Brown
Marc Buie
John Caldwell
Daniela Calzetti
Bill Carithers
Mark Clampin
Nicholas Collins
Colin Cox
Duilia de Mello
Susana Deustua
Rosa Diaz-Miller
Mark Dickinson
Andrew Dolphin
Daniel Durand
Dennis Ebbets
Annette Ferguson
Anthony Ferro
Wolfram Freudling
Ron Gilliland
David Golimowski
Shireen Gonzaga
Paul Goudfrooij
Carol Grady
Theodore Gull
Jonas Haase
George Hartig
Inge Heyer
Robert Hill
Dean Hines
Jay Holberg
Richard Hook
Rolf Jansen
Myungkook Jee
ESO, ST-ECF
Fermilab
U. Califronia, Berkeley
U. Ghana Legon
STScI
U. Colorado (CASA)
U. Colorado
McDonald Observatory, U. Texas
U. Notre Dame, Physics Dept.
STScI, JWST
STScI, INS
STScI, ACS
STScI, ACS
GSFC/NASA
Spacecraft System Eng. Services
STScI, WFPC2
ESO, ST-ECF
STScI, SPG
Lowell Observatory
STScI, SD
STScI, NICMOS
LBNL, Berkeley
STScI, ACS
SSAI/GSFC
STScI, ACS
Onsala Space Observ/JHU
American Astronomical Society
STScI, SPG
STScI, NICMOS
NOAO
National Research Council
Ball Aerospace
Kapteyn Inst.
U. Arizona
ESO
STScI, ACS
Johns Hopkins University
STScI, WFC3
STScI, SPG
NOAO, GSFC, Eureka Scientific
NASA’s GSFC LASP
ESO, ST-ECF
STScI, ACS
STScI, WFPC2
SSAI/GSFC
Steward Observatory, U. Arizona
Lunar & Planetary Lab, U. Arizona
ESO, ST-ECF @ STScI
Arizona State University
Johns Hopkins University
aalexov@eso.org
sallam@fnal.gov
jay@cusp.berkeley.edu
lalablay@yahoo.com
arribas@stsci.edu
ayres@origins.colorado.edu
sbeland@colorado.edu
fritz@clyde.as.utexas.edu
bennett@nd.edu
bergeron@stsci.edu
blades@stsci.edu
boeker@stsci.edu
bohlin@stsci.edu
bowers@band2.gsfc.nasa.gov
abradley@hst.nasa.gov
brammer@stsci.edu
bristowp@eso.org
tbrown@stsci.edu
buie@lowell.edu
caldwell@stsci.edu
calzetti@stsci.edu
WCCarithers@lbl.gov
clampin@stsci.edu
collins@stis.gsfc.nasa.gov
cox@stsci.edu
demello@pha.jhu.edu
deustua@aas.org
rmiller@stsci.edu
med@stsci.edu
dolphin@noao.edu
Daniel.Durand@nrc.ca
debbets@ball.com
ferguson@astro.rug.nl
tferro@as.arizona.edu
wfreudli@eso.org
gillil@stsci.edu
dag@pha.jhu.edu
shireen@stsci.edu
goudfroo@stsci.edu
cgrady@echelle.gsfc.nasa.gov
gull@sea.gsfc.nasa.gov
jhaase@eso.org
hartig@stsci.edu
heyer@stsci.edu
hill@tophat.gsfc.nasa.gov
dhines@as.arizona.edu
holberg@vega.lpl.arizona.edu
hook@stsci.edu
Rolf.Jansen@asu.edu
mkjee@pha.jhu.edu
xii
Participant List
Michael Jones
Paul Kalas
Erich Karkoschka
Stephen Kent
Ivan King
Anton Koekemoer
John Krist
Wayne Landsman
Bryan Laubscher
Lori Lubin
Jennifer Mack
John MacKenty
Jesús Maı́z-Apellániz
Russell Makidon
Eliot Malumuth
Peter McCullough
Matt McMaster
Gerhardt Meurer
Alberto Micol
Bahram Mobasher
Nick Mostek
Stuart Mufson
Max Mutchler
Ed Nelan
Jan Noordam
Susan Parker
Anna Pasquali
Steven Penton
Francesco Pierfederici
Nor Pirzkal
Imants Platais
Vera Platais
Marc Postman
Charles Proffitt
Kavan Ratnatunga
Swara Ravindranath
Jason Rhodes
Michael Richmond
Adam Riess
Michael Rosa
Jessica Rosenberg
Erin Roye
Abi Saha
Glenn Schneider
Al Schultz
Hsien Shang
Murray Silverstone
Marco Sirianni
Petr Skoda
Allyn Smith
George Sonneborn
Megan Sosey
Bill Sparks
GSFC
U. California, Berkeley
Lunar & Planetary Lab, U. Arizona
Fermilab
U. Washington
STScI, WFPC2
STScI
SSAI/GSFC
Los Alamos National Laboratory
STScI, WFPC2
STScI, ACS
STScI, WFC3
STScI, SPG
STScI, JWST
SSAI/GSFC
STScI, WFPC3
STScI, COS
Johns Hopkins University
ESO, ST-ECF
STScI, SPG
Indiana University
Indiana University
STScI, ACS
STScI, JWST
ASTRON
Inst. for Astronomy, Hilo, Hawaii
ESO, ST-ECF
U. Colorado
ESO, ST-ECF
ESO, ST-ECF
Johns Hopkins University
STScI, WFPC2
STScI, ODM
STScI, SPG
Carnegie Mellon University
Carnegie Institution of Washington
GSFC/NASA
Rochester Inst. of Technology
STScI, ACS
ESO
U. Colorado
STScI, NICMOS
NOAO
U. Arizona
Computer Sciences Corporation
ASIAA
U. Arizona
Johns Hopkins University
Astronomical Inst. Czech Republic
U. Wyoming
GSFC/NASA
STScI, NICMOS
STScI, ACS
michael.r.jones@gsfc.nasa.gov
kalas@astron.berkeley.edu
erich@lpl.arizona.edu
skent@fnal.gov
king@astro.washington.edu
koekemoe@stsci.edu
krist@stsci.edu
landsman@mpb.gsfc.nasa.gov
blaubscher@lanl.gov
lml@stsci.edu
mack@stsci.edu
mackenty@stsci.edu
jmaiz@stsci.edu
makidon@stsci.edu
eliot@barada.gsfc.nasa.gov
pmcc@stsci.edu
mcmaster@stsci.edu
meurer@pha.jhu.edu
Alberto.Micol@eso.org
mobasher@stsci.edu
nmostek@indiana.edu
mufson@indiana.edu
mutchler@stsci.edu
nelan@stsci.edu
noordam@astron.nl
parker@ifa.hawaii.edu
apasqual@eso.org
spenton@casa.colorado.edu
fpierfed@eso.org
npirzkal@eso.org
imants@astro.yale.edu
verap@stsci.edu
postman@stsci.edu
proffitt@stsci.edu
kavan@cmu.edu
swara@ociw.edu
jrhodes@howdy.gsfc.nasa.gov
mwrsps@rit.edu
ariess@stsci.edu
mrosa@eso.org
jrosenbe@origins.colorado.edu
roye@stsci.edu
saha@noao.edu
gschneider@stsci.edu
schultz@stsci.edu
shang@asiaa.sinica.edu.tw
murray@as.arizona.edu
sirianni@pha.jhu.edu
skoda@adara.asu.cas.cz
jasmith@uwyo.edu
george.sonneborn@gsfc.nasa.gov
sosey@stsci.edu
sparks@stsci.edu
Participant List
Karl Stapelfeldt
Elizabeth Stobie
David Stys
Rodger Thompson
Hien Tran
David Trilling
Douglas Tucker
Jeff Valenti
Roeland van der Marel
Jeremy Walsh
Brad Whitmore
Tommy Wiklind
Jennifer Wiseman
Bruce Woodgate
Haojing Yan
David Zurek
Jet Propulsion Laboratory
U. Arizona
STScI, SPG
U. Arizona
Johns Hopkins University
U. Pennsylvania
Fermilab
STScI, SPG
STScI, ACS
ESO
STScI, WFPC2
STScI, NICMOS
Johns Hopkins University
GSFC/NASA
Arizona State University
American Museum Nat. History
xiii
krs@exoplanet.jpl.nasa.gov
bstobie@as.arizona.edu
stys@stsci.edu
rthompson@as.arizona.edu
tran@pha.jhu.edu
trilling@hep.upenn.edu
dtucker@fnal.gov
valenti@stsci.edu
marel@stsci.edu
jwalsh@eso.org
whitmore@stsci.edu
wiklind@stsci.edu
jwiseman@pha.jhu.edu
woodgate@stars.gsfc.nasa.gov
yhj@asu.edu
dzurek@amnh.org
Part 1. ACS
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
Status of the Advanced Camera for Surveys
M. Clampin and G. Hartig
Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218
H. C. Ford, M. Sirianni, G. Meurer, A. Martel and J. P. Blakeslee
Department of Physics and Astronomy, The Johns Hopkins University, 3400 North
Charles Street, Baltimore, MD 21218
G. D. Illingworth
UCO/Lick Observatory, University of California, Santa Cruz, CA 95064
J. Krist, R. Gilliland and R. Bohlin
Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218
Abstract. The Advanced Camera for Surveys (ACS), installed in the Hubble Space
Telescope in March 2002, will significantly extend HST ’s deep, survey imaging capabilities. ACS has met, or exceeded all of its key performance specifications. In
this paper we present an introductory review of the in-flight performance of the
instrument.
1.
Introduction
The Advanced Camera for Surveys (ACS) is a third generation instrument for the Hubble
Space Telescope (HST ). It was installed in HST during the fourth servicing mission (SM3B)
in March 2002. ACS replaced a first generation axial bay instrument, the Faint Object
Camera (FOC). ACS has three channels, shown schematically in Figure 1, the Wide Field
Camera (WFC), the High Resolution Camera (HRC) and the Solar Blind Camera (SBC).
WFC is a high-throughput, wide field imager (202×202 ) designed for deep imaging surveys
in the near-IR. WFC provides a factor of 10 gain in discovery efficiency at 800 nm, compared
to the Wide Field Planetary Camera-2 (WFPC2). In this context, discovery efficiency is
defined as the product of field of view (FOV) and instrumental throughput. WFC is an
f/25 camera which employs three reflective optics. The first mirror in the optical chain is a
spherical mirror IM1, which images the HST pupil onto the mirror IM2. The mirror IM2
is an anamorphic asphere figured for the inverse conic error on the HST primary mirror, in
order to correct spherical aberration on the HST primary, and field dependent astigmatism
at the center of the ACS field of view. IM2 images onto mirror IM3, a Schmidt-like plate,
which corrects astigmatism over the field of view, and images the beam through two filter
wheels onto the WFC focal plane. The focal plane detector array is a mosaic of two Scientific
Imaging Technologies (SITe) 2048 × 4096 CCDs (Sirianni et al. 2000, Clampin et al. 1998).
The primary WFC design goal is to maximize the instrument throughput in the near-IR,
and has been achieved by minimizing the number of optical elements in the design, and
coating the mirrors with Denton protected-silver. The combined reflectivity of three silver
coated mirrors at 800 nm is 98%, compared to 61% for three MgF2 over coated aluminum
mirrors. In the near-UV (>370 nm) the reflectivity of the silver coating falls rapidly. The
3
4
Clampin et al.
Figure 1. Schematic showing the optical designs for the WFC (left) and the
HRC/SBC (right).
plate scale of the WFC is 0.05 pixel−1 , which delivers near-critical sampling at the near-IR
wavelengths for which the camera is optimized.
The HRC is a near-UV to near-IR imager, which provides critically sampled images in
the visible, over a 29 × 26 field of view. HRC is also equipped with a true coronagraphic
mode for high contrast imaging of the circumstellar environments of bright stars. HRC is
a f/70 camera which shares two of its three mirrors with the SBC. The third mirror M3
is a fold mirror which is inserted into the beam to direct it through the two filter wheels
onto the HRC focal plane array. The focal plane detector array is a SITe 1024 × 1024 CCD
(Sirianni et al. 2000). The HRC shares the two filter wheels with the WFC and is capable of
operating simultaneously with WFC. The HRC and SBC mirrors M1 and M2 are aluminum
coated with MgF2 overcoating, and optimized for maximum reflectivity at 121.6 nm. The
HRC mirror M3 is optimized for 200 nm and an incidence angle of 45◦ . The HRC focal
plane detector is a SITe 1024 × 1024 CCD detector, based on the Space Telescope Imaging
Spectrograph (STIS) CCD (Kimble et al. 1998). The HRC plate scale is 0.027 pixel−1 ,
which yields fully sampled images in the visible. The SBC is selected when M3 is moved
out of the light beam.
In order to maximize far-UV throughput, the SBC optical design is a two mirror optical
system, with its own independent filter wheel. The SBC is a far-UV imager optimized
for high throughput at 121.6 nm, with a field of view of 31 × 35 , and a plate scale of
0.032 pixel−1 . Its focal plane detector is a photon-counting CsI photocathode MAMA
previously designated as the STIS flight spare detector.
2.
Detectors
The ACS CCD detector systems are performing nominally. The detector read noise figures
for WFC and HRC are summarized in Table 1. Both detector’s are unchanged within
their respective uncertainties, demonstrating the high degree of noise isolation achieved
during ground testing, and the excellent on-orbit shielding from noise sources in the HST.
Consequently, WFC broadband science observations will be typically, sky limited, while
HRC science programs are read-noise limited, due to the smaller pixel size. The WFC
Status of the Advanced Camera for Surveys
Table 1.
WFC1
WFC1
WFC2
WFC2
5
Comparison of Pre-launch and Post-launch CCD Readout Noise
Amp.
Gain
A
B
C
D
1
1
1
1
Read Noise
e− RMS
pre post
4.8
4.9
4.7
4.8
5.2
5.2
4.7
4.8
HRC
HRC
HRC
HRC
Amp.
Gain
A
B
C
D
2
2
2
2
Read Noise
e− RMS
pre post
4.6
4.6
4.4
4.7
4.7
4.7
5.0
4.9
Figure 2. The growth of WFC hot pixels since launch from Riess (2002), illustrating the effect of monthly anneals on the long term evolution of hot pixels in
the WFC. (Courtesy A. Riess 2002)
CCDs are read out simultaneously through all four amplifiers, while the HRC is read out
though amplifier C.
The measured dark currents, excluding hot pixels (>0.04 e− pixel−1 s−1 ) are 7.5 e− pix−1 −1
e s (−77◦ C) and 9.1 e− pixel−1 s−1 (−80◦ C), for the WFC and HRC respectively. These
temperatures are achieved without the aft-shroud cooling system, which will be installed
during the next (SM4) servicing mission. Hot pixels are a result of high energy proton
displacement damage. The primary technique for moderating the hot pixel growth rate is
annealing of the CCDs at the instrument’s ambient “power off” temperature. Typically,
the ACS detectors reach ∼20◦ C when CCD cooling is switched off. The SBC’s MAMA
detector has also exceeded expectations for dark current, since pre-launch predictions of its
operating temperature proved pessimistic. The SBC’s measured dark current is 1.2 × 10−5
photons s−1 across the detector.
Hot pixel evolution has been evaluated over several ACS annealing cycles by Riess
(2002). Hot pixels in the WFC appear at a rate of ∼1230 pixels day−1 . In Figure 2, we
show the evolution of hot pixels and the effect of the monthly annealing. WFC hot pixels
are annealed at a rate of ∼60%, in contrast to the factor of ∼80% for the HRC detector. In
subsequent WFC anneals, existing hot pixels are annealed at very low rates such that after
7 to 8 anneal cycles the cumulative fraction of annealed pixels reaches a plateau at ∼70%
(Riess 2002). Consequently, ∼1.5% of the WFC mosaic will be covered by hot pixels after
6
Clampin et al.
24 months. This corresponds to the fraction of the WFC mosaic covered by cosmic rays
in a 1000 second WFC exposure. The WFPC2 and STIS CCDs are similar to the HRC in
annealing at a rate of ∼80%. Currently, the WFPC2 focal plane array has ∼2.5% coverage
by hot pixels, where WFPC2 hot pixels are defined as >0.02 e− s−1 . While the WFC hot
pixel evolution rate is a concern it can be handled during science operations by obtaining
daily hot pixel calibration images, and dithering observations so that optimal combination
of science frames can be used to eliminate hot pixels.
Initial post-launch measurements indicated that charge transfer efficiency (CTE) for the
WFC and HRC detectors was consistent with pre-launch calibration data. CTE is expected
to degrade with time since the radiation environment experienced in HST ’s orbit has caused
long-term degradation of CTE in previous HST instruments including STIS (Kimble et al.
2000), and WFPC2 (Whitmore et al. 1999). Preliminary calibration using the extended edge
pixel response (EPER) method on internal WFC flat field images, indicates a degradation
of parallel CTE from 0.999999 to 0.999991 (amplifier D) during the first six months of
operation. EPER measurements help to track CTE degradation, but are not a good measure
for assessing the scientific impact of CTE degradation observations. Observational factors
such as the target size, density of sources in the field and sky background levels influence
the impact of CTE on a given target.
3.
Image Quality
The image quality for each of the three cameras is summarized in Table 2 from measurements
by Hartig et al. (2002). In Figure 3 we show images of a star taken through the filters
covering the spectral range of the WFC and HRC. The WFC F850LP image shows one
artifact, a faint ehancement of the horizontal diffraction spike at wavelengths longer than
∼800 nm. In order to rectify the long wavelength halo observed in these CCDs (Sirianni
et al. 1998), a reflective coating was applied to the frontside of the CCD prior to thinning.
This coating appears to give rise to a diffraction artifact, which is seen in the F850LP image
as an enhancement in the brightness in one of the four diffraction spikes.
Figure 5 shows the normalized, azimuthally averaged ACS and WFPC2 profiles for
stars observed through a F555W filter. Although the ACS WFC and the WFPC2-PC
cameras have very nearly the same physical and spatial pixel sizes, the WFPC2-PC half
width at half maximum (HWHM) is ∼20% narrower than the ACS WFC HWHM. This is
likely due to slightly more charge diffusion in the backside-illuminated ACS pixels than in
the frontside-illuminated WFPC2 pixels.
In the case of the HRC images in Figure 3, it can be seen that the NUV image through
the F220W filter exhibits a small “spur.” The feature is independent of field position
and is due to a moderate amount of several low order aberrations in the optical system.
The aberration also impacts the SBC images so that at 122 nm they fall slightly below
their specified encircled energy (EE), most likely as a combined result of this uncorrectable
aberration, the large halo induced by the MAMA detector, and the mid-frequency figure
error of the OTA optics. The SBC radius for EE = 0.3, derived from F125LP images of hot
stars, is 0.06.
Figure 4 shows the encircled energy plots for the HRC and WFC point spread functions
(PSFs) derived from F555W images. With allowance for telescope jitter, the WFC meets
the image specification (Hartig et al. 2002). Even with image jitter included, the HRC
exceeds the image specification. The encircled energy values for ACS and WFPC2 at
550 nm are also compared in Table 2. The WFPC2 values were taken from the WFPC2
Instrument Handbook. The ACS WFC and the WFPC2 PC have nearly the same angular
pixel sizes (0.050 and 0.046 respectively) and the same 50% and 80% EE values. Because
the half width at half maximum (HWHM) is a better measure of image resolution than
EE, Table 3 includes the HWHMs for the ACS and WFPC2 cameras derived from F555W
Status of the Advanced Camera for Surveys
Table 2.
555 nm
Comparison of ACS and WFPC2 HWHMs and Encircled Energies at
HWHM (Radius)
50% EE (Radius)
80% EE (Radius)
ACS WFC
( ) pixels
0.044 0.88
0.06
1.4
0.13
2.6
WFPC2 WFC
( )
pixels
0.096
0.96
0.11
1.1
0.24
2.4
ACS HRC
( ) pixels
0.025 1.02
0.05
2.0
0.11
4.4
WFPC2 PC
( ) pixels
0.034 0.75
0.07
1.4
0.13
2.6
Figure 3. HRC images taken through the filters, F220W, F435W and F850LP
are shown in the top three images (3.3 × 3.3 ). The bottom three images show
WFC images taken through F435W, F606W and F850LP (5.9 × 5.9).
Figure 4.
The fraction of the total light (“encircled energy” EE) enclosed within
an aperture versus radius for WFC and HRC images of stars taken through an
F555W filter.
7
8
Clampin et al.
Figure 5.
The normalized line profiles for ACS and WFPC2 PSFs derived from
F555W images of stars. The profiles are “Moffat” profiles derived with the iraf
task “imexamine.”
Figure 6.
The net efficiency of the WFC plus the HST OTA versus wavelength.
The predicted values were derived by combining the preflight component calibrations. The observed efficiency was derived from observations of standard stars
used for on-orbit calibration of previous HST instruments.
Status of the Advanced Camera for Surveys
9
Figure 7.
The net efficiency of the HRC plus the HST OTA versus wavelength.
The observed sensitivity is 5 to 10% higher than predicted at wavelengths between
550 nm and 800 nm, and ∼25% lower than predicted at 250 nm. In spite of the
lower than expected sensitivity in the near ultraviolet, the HRC meets or exceeds
its design specifications at all wavelengths.
Figure 8.
SBC plus HST OTA net efficiency versus wavelength. The predicted
values were derived by combining the preflight component calibrations. The observed efficiency was derived from observations of standard stars used for on-orbit
calibration of previous HST instruments.
images of stars. The Rayleigh criterion (1.22l/D) for an unobstructed 2.4 aperture is 0.058
at 555 nm, and the full width at half maximum (FWHM) of the Airy function is 0.05.
Table 2 shows that the observed ACS HRC FWHM is 0.05 (2 pixels). The HRC meets the
design goal of being critically sampled at wavelengths λ > 500 nm.
4.
Sensitivity
The on-orbit performance of the WFC exceeds the preflight predictions by a substantial margin. The pre-flight sensitivity of ACS was determined from component level measurements
of the CCD quantum efficiencies, mirror reflectivities and filter throughputs. Systematic
errors and measurement uncertainties are the most likely explanation for this unexpected
gain in sensitivity. In Figure 6 we show the measured on-orbit net throughput of the WFC
based on measurements of spectrophotometric standard stars through broad band filters.
The WFC is near 50% overall efficiency at 650 nm. The HRC exhibited the same gain
10
Clampin et al.
Figure 9.
Radial surface brightness profiles of a star observed through the filter
F435W. The top line is the predicted HRC profile (direct, no coronagraph). The
middle line is with the coronagraph (1.8 occulting spot), and the bottom line is
the absolute value of the residual coronagraphic profile after the star is subtracted
by an image of itself taken during a separate visit. Note that the Lyot stop in the
coronagraph reduces the throughput by 53%.
in performance at visible wavelengths as the WFC, but showed a small decrease in NUV
sensitivity compared to pre-launch predictions. The HRC measurements are shown in Figure 7. The SBC’s peak net throughput is shown in Figure 8, with a comparision to STIS.
The superior throughput of ACS results from the fact that it has only two reflections in the
optical path to the SBC, compared to four in STIS.
5.
ACS Coronagraph
The ACS HRC coronagraph comprises two opaque circular stops that are positioned in the
HST aberrated focal plane and a Lyot stop that is simultaneously positioned immediately
in front of the HRC M2 mirror, and thus close to the pupil image. The smallest stop has
a diameter of 1.8 and is positioned in the center of the HRC field. The largest stop has a
diameter of 3.0 and is 5.25 from one edge of the HRC. The 1.8 and 3.0 spots block about
88% and 95% of the aberrated PSF, respectively. The Lyot stop reduces the throughput of
non-occulted sources by ∼48%. In addition to the two circular stops, there is a 0.8 wide
opaque finger that extends 5.5” over the HRC window at an angle of ∼74◦ to the edge.
Because the finger is not in the focal plane, there is a small amount of vignetting around
its edges. During assembly of the ACS the tip of the finger was aligned with the center of
the 3.0 spot, with the goal of blocking light diffracted into the geometrical shadow from
bright stars centered on the spot. However, after launch “gravity release” caused the finger
and shadow of the large mask to misregister by ∼ 1 .
Figure 9 shows the azimuthally averaged radial surface brightness profiles for a simulated direct F435W image of a star, the observed F435W profile when the star is centered
on the small spot (1.8), and the observed radial profile when two sequential coronagraphic
images are subtracted. The simulated image includes diffraction from the HST pupil and
from the residual polishing errors on the HST primary and secondary mirrors. The figure
shows that the coronagraph reduces the background by ∼6. If a matching PSF is subtracted
(e.g., by rolling the telescope and taking another image), the background is reduced by a
factor of ∼1000.
Status of the Advanced Camera for Surveys
11
Figure 10. The left panel shows a 5-orbit direct image (2 orbits F775W + 3 orbits
F850LP) image of an emission line galaxy in the HDFN. The F775W filter (Sloan i)
includes the redshifted [OIII] λ5007 emission line and the F850LP filter (Sloan z)
includes redshifted Hα emission. The right panel shows a 3-orbit grism image of
the galaxy. The arrows mark the bright emission line regions in the image; the
vertical lines mark the emission lines [OIII] λ5007 and Hα 6563 in the spectrum.
6.
ACS Grism
The first order WFC dispersion, which depends on the position in the focal plane, varies from
3.63 nm pixel−1 in one corner to 4.55 nm pixel−1 in the other corner, with an average value
of 3.95 nm pixel−1 . The spectral resolution is the product of the monochromatic FWHM
and the dispersion. We assume that a monochromatic image has the same dimensions in
the spatial and spectral directions. Table 3 gives the spatial and spectral resolution derived
from an average of the five measurements at positions near the center and ends of the
spectra. The resolution R = λ/δλ varies from ∼65 at the blue end of the spectrum to
∼78 at the red end. The resolutions at 800 and 1000 nm are very close to the values for
a diffraction-limited image sampled with 0.05 pixels. The spectral resolution achieved on
extended sources will be proportional to the square root of the quadratic sum of the image
size in the dispersion direction and the FWHM for a point source.
Table 3.
ACS Grism Spectral and Spatial Resolution for Stellar Sources
Avg. Wavelength
(nm)
593.8 ± 8.2
801.6 ± 6.9
977.6 ± 13.3
Avg. Cross Dispersion FWHM
(pixels)
2.30 ± 0.2
2.33 ± 0.3
3.16 ± 0.7
Avg. Resolution
(nm)
9.07
9.21
12.49
Resolution
(λ/δλ)
65
87
78
Figure 10 shows a 5-orbit direct image (2 orbits F775W + 3 orbits F850LP) and a 3orbit grism image of a galaxy in the HDFN13. The grism image shows that the prominent
knots at each end of the galaxy are star forming regions with two strong emission lines
([OIII] λ5007 and Hα l6563). Figure 11 shows the spectra extracted at the positions of the
two knots and the nucleus. The nucleus also shows strong emission at Hα. The observed
wavelengths of the emission lines agree with the published redshift of the galaxy, z = 0.319.
The grism’s high sensitivity and low resolution make it particularly suitable for observations
of stellar sources with broad spectral features, such as supernovae and brown dwarfs, and
compact star forming regions that have strong emission lines.
Acknowledgments. ACS was developed under NASA contract NAS 5-32865, and this
research is supported by NASA grant NAG5-7697. We are grateful for an equipment grant
from the Sun Microsystems, Inc.
12
Clampin et al.
Figure 11. Spectra extracted from the grism image shown in Figure 10. The
two strong emission lines in the knots at ∼660 nm and 866 nm are [OIII] λ5007
and Hα 6563 at a redshift z = 0.3229.
References
Clampin, M., et al. 1998, SPIE 3356, 332
Hartig, G., et al. 2002, SPIE 4854, in press
Kimble, R. A., Goudfrooij, P., & Gilliland, R. L. 2000, SPIE 4013, 532
Kimble, R. A., et al. 1998, ApJ 492, L83
Riess, A. 2002, Instrument Science Report ACS 02-06 (Baltimore: STScI)
Sirianni, M., et al. 2000, SPIE 4008, 669
Sirianni, M., et al. 1998, SPIE 3355, 608
Whitmore, B., Heyer, I., & Casertano, S. 1999, PASP 111, 1559
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
Astrometry with the Advanced Camera: PSFs and Distortion in
the WFC and HRC
Jay Anderson
Department of Physics and Astronomy MS-108, Rice University, Houston TX
77005
Abstract. Before ACS can be useful for astrometry, we must first determine how
best to model the PSF and how to correct for the camera’s considerable distortion.
I analyze WFC and HRC images taken of the core of 47 Tuc and find that the
same effective-PSF-based approach that works for WFPC2 produces excellent results
with ACS. Positions of reasonably bright stars can be measured with a random
error of better than 0.01 pixel in each coordinate. Distortion is another matter.
It is well known that the Advanced Camera for Surveys suffers from significant
linear and higher-order distortion. I find that the ACS also suffers from some finescale distortion that appears to be different for each filter. This fine-scale distortion
perturbs a polynomial solution by about 0.05 pixel and is coherent on spatial scales
of about 200 pixels. I find that the distortion in each chip can be modeled with a
4th-order polynomial and a separate look-up table for each filter. With such a model,
the distortion residuals are typically ∼0.01 pixel.
1.
Introduction
We have found in our research with WFPC2 (see references by Anderson & King) that the
two keys to high-precision astrometry with HST are (1) careful treatment of the undersampled point-spread function (PSF) and (2) accurate modeling of the geometric distortion.
Once these two issues are addressed, it is possible to attain differential astrometry to a
fraction of a milli-arcsecond in a well-dithered set of images.
With its large field of view and better sampling, the Advanced Camera for Surveys has
the potential to measure positions at least a factor of two better than WFPC2. However,
before we can realize this potential we must first learn how to model the PSF and remove
the distortion in this new camera.
With regards to the PSF, I have examined some of the early ACS images and find that
the same modeling techniques that worked for WFPC2 work quite well for ACS. In fact, in
many ways, the ACS PSF is better behaved than that for WFPC2.
It was well known that distortion in the Advanced Camera for Surveys would be much
greater than that in WFPC2 (50 pixels as compared with 5 pixels). It was not known
how accurately this distortion could be modeled and removed. I have examined several
of the data sets available at the center of 47 Tuc and characterized the various sources of
distortion present in the two cameras. I find that in addition to the expected large geometric
distortion, there appears to be some unexpected component of distortion that is introduced
by the filter itself.
2.
Available Data Sets
There are two main data sets available to explore the PSF and distortion properties in
the WFC and the HRC: GO-9028 (PI Meurer) and GO-9443 (PI King). These two sets
13
14
Anderson
Figure 1. Contour plots of the super-sampled F475W PSFs. These PSFs represent the inner 5 × 5 pixels of a star and are sampled every 0.25 pixel in x and y.
The heavy contours are separated by 0.5 dex.
were both taken at the center of globular cluster 47 Tuc. The main set (9028) consists
of 20 exposures through F475W with a range of offsets all at the same orientation. GO9443 contains a tightly dithered set of pointings through the same filter at an orthogonal
orientation. This program also has several single exposures taken through other filters to
allow us to investigate how the solution may vary with filter.
Other data sets cover the same field and an outer field of the same cluster, and will
provide valuable checks on the solution.
3.
3.1.
The PSF
Lessons from WFPC2
Our experience with WFPC2 (Anderson & King 2000) has shown us that the main sources
of PSF-related astrometric error are due to undersampling and spatial variability. The
undersampled nature of HST images does not mean that we cannot measure accurate
positions for stars, but it does mean that the accuracy of our positions will depend critically
on our model of the PSF.
We deal with the undersampling by taking several observations of the same star field
with a variety of sub-pixel offsets. This dithered set allows us to reconstruct a properly
sampled version of the PSF. Our PSF model is typically super-sampled by a factor of four
with respect to the image pixels. We find that without such a carefully constructed PSF,
position measurements contain systematic errors of up to 0.05 pixel. But with an accurate
PSF model, it is possible to measure positions with a random accuracy of 0.02 pixel and no
significant systematic errors.
The WFPC2 PSF changes shape with location in the field, so we derive a different PSF
for nine fiducial points in each chip. We interpolate among these fiducial PSFs to construct
a PSF that is appropriate for the location of each star we measure.
3.2.
ACS Results
I constructed a PSF for the Advanced Camera for Surveys using the same techniques as
above for WFPC2. To start, I found a single PSF for each chip. These PSFs are shown in
Figure 1. (Note that the images used were the flt or crj images, not the drz images; the
latter images are not well suited for astrometry as they have been re-sampled and much of
the positional information has been blurred out.)
Astrometry with ACS
15
Figure 2. The single-coordinate astrometric precision as a function of instrumental magnitude, −2.5·log10 (DN). Stars brighter than −13.5 in the WFC and
−14.5 in the HRC are saturated and have greatly increased errors.
Astrometric Accuracy. I find that the 4×-supersampled PSF model can measure reasonably bright, isolated stars with a random accuracy of better than 0.01 pixel in each
coordinate and with no significant systematic errors (see Figure 2). This accuracy with
respect to the pixel grid holds true for the WFC and the HRC, so that the HRC angular
precision is a factor of two better than in the WFC.
Thus far, I have used the same 5×5-pixel PSF model as we used for WFPC2. Since even
in the HRC most of the flux is contained within this aperture, photometry and astrometry
should not require a larger-format PSF. However, if one wants to do PSF subtraction, a
larger-format PSF might be useful.
Spatial Variation of the PSF. The WFPC2 PSF changes shape appreciably from the center
of the chip to the corners. We found that if we didn’t adequately model the PSF variation,
we saw systematic errors in astrometry that correlated with the pixel phase of the star
being measured. I therefore examined the astrometric residuals from different regions of
the ACS chips in search of such tell-tale signs that the PSF was changing. I found only a
very slight (0.002 pixel) hint of this effect in the very corner of the WFC chips. This is a
factor of five less variation than was seen across the (smaller) WFPC2 chips. For almost
all ACS projects, it should therefore be safe to use a spatially constant PSF—a welcome
simplification.
The random accuracy shown above is a factor of two better than we have been able to
achieve for WFPC2. This improvement is a result of several factors: (1) we only need to
solve for a single PSF over a much larger area, so we have many more PSF samplings from
which we can construct a model; (2) the deeper CCD wells make each observed sampling
of the PSF more accurate; and (3) the deeper wells also give the bright stars much higher
signal-to-noise so that we can fit for a more accurate position.
Photometric Accuracy. While my focus has been more on astrometry than photometry, I
have examined the photometric residuals as well and find that the best stars can be measured
differentially to about 0.005 magnitude in each exposure. However there do appear to be
some small systematic errors (∼0.003 mag) related to intra-pixel sensitivity variations. In
the PSF constructed above, I constrained the pixel-response function to be flat, but more
16
Anderson
realistic constraints should be possible once we have better observations of the intra-pixel
sensitivity function for the various filters.
Stability of the PSF over Time. Finally, I looked at how the PSF changes over the longterm. I measured the GO-9443 images (taken in July 2002) with the PSF from the GO-9028
images (taken in April 2002) and found that the systematic position errors incurred were on
the order of 0.005 pixel. This error is a factor of five smaller than we found with a similar
WFPC2 comparison (Anderson & King 1999).
This apparent stability of the PSF should be confirmed with additional observations,
but if the PSF is typically this stable, then it would be worthwhile to construct a library
of PSFs, one for each filter/chip combination. The single-pointing observations of GO9443 will allow us to find a preliminary PSF for several filters. However to get the best
possible PSFs, and to constrain the pixel-response function for each filter, we will need more
observations. A tightly dithered set of observations through each filter of a reasonably dense
star field should allow us to constrain the pixel-response function and develop an exquisitely
accurate PSF for each filter. This would require about an orbit per filter.
PSF Summary. The raw astrometric quality of these early ACS images is extremely encouraging. If this precision can be matched with an equally good distortion solution, then
the Advanced Camera for Surveys will open up many new avenues for astrometry.
4.
4.1.
Distortion
Different Needs
Different applications make different demands on a distortion solution. The pipeline requires
a solution that is accurate to about 0.2 pixel, so that the drizzle procedure can create a
rectified image that can be easily compared with undistorted images. The solution does
not need to be more accurate than this because the resampling inherent in the drizzling
process introduces errors of about this size. For this reason, the spec demanded of the
initial distortion solution was ∼0.2 pixel.
Applications which produce a mosaic from a set of offset pointings (perhaps to cover the
gap between the chips or to panel-image a large field) are particularly sensitive to distortion
errors at the edges of the chips. Such errors can produce additional blurring in the interfaceregion, resulting in a variation of the resolution with location in the mosaic. Sometimes
polynomial solutions can have their largest errors at the edges, so for these applications it
may be worthwhile to use a more elaborate solution.
To do differential astrometry, we need to measure relative positions of nearby stars to
better than 0.01 pixel. For this, we do not need a solution that is globally accurate at this
precision (namely, knowledge of precisely where a pixel in one corner of an image is relative
to the pixel in an opposite corner); but we do have to trust that the solution is locally flat
to very high precision.
Fortunately all applications will benefit from the best possible distortion solution on
both local and global scales. Thus, our focus has been to examine all sources of distortion, large-scale and small-scale, so that we would arrive at the best possible model for all
applications.
4.2.
How to Solve for Distortion
The easy way to solve for distortion is to observe an astrometrically-calibrated star field
with the detector. The detector distortion then shows up directly as position residuals.
Unfortunately, there do not exist any astrometric-standard fields with the precision and
density that would be useful to help us calibrate HST. Hopefully by the end of the ACS
calibration procedure the core of 47 Tuc will be able to serve such a purpose.
Astrometry with ACS
17
The hard way to solve for distortion is to do a self-calibration. Here, we observe the
same star field through the same detector at various offsets and orientations. We then
relate the images to each other, solving simultaneously for the transformations between the
images and the distortion. (Usually the telescope pointing is not precise enough to allow
us to take the offsets between the images as known.)
4.3.
Lessons from WFPC2
We have recently completed a new self-calibration of the distortion in WFPC2 (see Anderson
& King 2003). While WFPC2 and ACS are different instruments, many of the issues
involved in self-calibration will be applicable here as well. Some issues we uncovered in our
work with WFPC2 are:
1. In order to solve for distortion, we obviously need observations of the same field at
different offsets. If these offsets are too large, however, it can be very difficult to
tease apart the solution for the offsets between the pointings from the solution for the
distortion itself. It is easiest to use only image pairs with at least 50% overlap.
2. We have found that there are several aspects to a distortion solution: periodic irregularities in the detector, a polynomial solution, variations with focus and filter,
additional effects near the chip-edges, the inter-chip solution, etc. It can be extremely
hard to solve for all aspects of the solution at once, thus it is useful to examine data
sets which can isolate one or two of these effects, and then construct a piecemeal
solution that includes all of the effects.
3. The only way to solve for the linear terms is to observe the same field at different
orientations. However, if the two observations do not have enough overlap, then
the solved-for linear terms are very sensitive to errors in the higher-order terms. In
general, the low-order terms are harder to solve for than the higher-order ones, since
a global solution can only be as accurate as the local solution.
4. Finally, changes in focus can introduce additional distortions, so that the solution
may change over time, both long-term and short-term.
4.4.
Starting with the WFC
With all these cautions in mind, I began to examine the ACS distortion. I started with the
GO-9028 data set, which consists of 20 exposures taken through F475W, with a variety of
offsets but all with essentially the same orientation. Such a set of parallel displacements
will not allow us to solve for the linear terms, but it is an excellent opportunity to isolate
the higher-order terms.
I used the PSFs from the previous section to measure every star in every image. I then
cross-identified all the stars in all the images, creating a master list of all the stars in the
field and recording the observed location for each star in each image. (This initial task of
cross-identification is not easy with images that suffer from as much distortion as ACS!)
The Polynomial Solution. The first task is to find the best 4th-order solution. This requires many iterations between determining the inter-image offsets and solving for the distortion itself. In this initial polynomial solution, I set the linear terms to be zero, since
we know that parallel displacements cannot constrain them. I found a 4th-order solution
very similar to what Meurer et al. (2003) have found: The WFC non-linear terms have an
amplitude of about 50 pixels (from center to edge). The HRC non-linear distortion amounts
to about 3 pixels in amplitude.
Like Meurer et al., I also found that there are some systematic residuals from this simple
polynomial solution. These systematic residuals are typically 0.05 pixel in amplitude and
18
Anderson
are coherent on a spatial scale of about 200 pixels. It is clear that we must somehow treat
this fine-scale distortion if we desire to make use of the 0.01-pixel precision that is possible
with the ACS PSF.
The Supplementary Look-up Table. It was surprising to find such fine-scale structure in
the solution. The WFPC2 solution had no such fine-scale variation and was almost perfectly
fit with a 3rd-order polynomial. It was not practical to model this with a polynomial of even
higher order, so I decided to model the additional distortion by means of a supplemental
table, sampled every 64 pixels in the WFC (65×33 entries for each chip) and every 16 pixels
in the HRC (65 × 65 entries). To evaluate the distortion at any point in a chip, I interpolate
this table and add the table result to the polynomial solution to get the total distortion
correction. Solving for the table is also an iterative procedure, and after each iteration I
constrain the table to be smooth by convolving it with a 5 × 5 quadratic smoothing kernel.
The table for the bottom WFC chip is shown graphically in Figure 3.
The residuals from this polynomial-plus-table solution are typically less than 0.01 pixel,
both for the HRC and for the WFC. There remains some quasi-periodic high-frequency
variation with a scale of ∼120 pixels and with an amplitude of about 0.005 pixel. This
is likely due to limitations in the look-up table formulation. The smoothing I perform
means that the table cannot correct for variations on a scale smaller than about 150 pixels.
Improving the distortion solution further may require observations with an array of dither
offsets that are better spaced to sample the distortion at this spatial frequency.
Inter-chip Solution. The gap between the WFC chips is qualitatively different from the
gap between the WFPC2 chips. Each of the WFPC2 chips is independently imaged by
different optics, whereas the ACS/WFC chips are all in a common image field and the
gap is simply a physical offset between the chips. This offset should not change with time
and breathing as much as the WFPC2 offsets do, so it should be safer to use a meta-chip
solution for the WFC. The meta-chip solution is included as part of my standard distortion
correction. The errors in the inter-chip solution appear to be no larger those in the global
single-chip solutions.
Distortion from Detector Defects. Because of a manufacturing defect, every 34th row in
WFPC2 is slightly (3%) narrower than average. This was noticed early-on in the flat
fields, but it was not initially known if it was a geometric effect or a pixel-sensitivity effect
(Holtzman et al. 1995). We demonstrate in Anderson & King (1999) that it is indeed
a geometric effect—it produces a 0.03-pixel skip every ∼34 pixels. We provide a simple
correction for this.
The WFC flat fields are seen to exhibit a similar striping every ∼68 columns, so I
looked for an astrometric signature here as well. I plotted the localized relative intensity of
the flat field as a function of column number (see Figure 4) and noticed not only a bright
column, but a regular pattern with an rms amplitude of 0.1% and maximum amplitude of
0.8%. The expected astrometric signature of this is 0.008 pixel, four times smaller than with
WFPC2. Nonetheless, I examined the position residuals as a function of this 68.270-column
phase and, sure enough, the observed trend matched up almost exactly with the predictions
based on the flat fields. It should be relatively straightforward to come up with a correction
for this; but since it is a small effect, I will put off deriving a correction until the other
(larger) sources of distortion are properly treated.
4.5.
Comparing with the GO-9443 data set
In a self-calibrated solution, we need to have at least one pointing that is rotated with
respect to the others in order to solve for the linear terms. That was the idea behind taking
GO-9443 with an orthogonal orientation to the many GO-9028 images.
Astrometry with ACS
Figure 3. Graphical presentation of the WFC[1] (bottom chip) table of corrections. I plot each of the 33 rows along with its baseline. The y coordinate
corresponding to the center of each row is labeled up the middle. The separation between rows is 0.05 pixel. The table corrections are generally smaller than
0.05 pixel, but can be larger than 0.1 pixel.
19
20
Anderson
Figure 4. For each pixel in the WFC[1] flat, I took the ratio of the pixel value to
the median of the 30 pixel values on either side. I then took a median of this ratio
for all 2048 pixels in each column. This is plotted for each of the 4096 columns in
a phase diagram.
Introducing the Orthogonal Pointing. In my initial attempt to solve for the distortion,
I took the orthogonal pointing of GO-9443 and compared it against the twenty GO-9028
pointings, thinking that would allow me to solve for the linear and the higher-order distortion at the same time.
Using the residuals from these 1-versus-20 image comparisons, I found a solution which
did a very good job reducing the residuals. But when I used this solution to compare the 20
GO-9028 pointings with each other, I found significant systematic residuals. This indicated
that there was some change in the solution between the two observations. So, I decided
to focus first on the non-linear terms (as above) using exclusively the parallel GO-9028
observations, and only later compared with the orthogonal pointing.
Comparing with the Orthogonal Pointing. I correct both the GO-9028 and orthogonal GO9443 observations with the above polynomial-plus-table solution and examine the residuals
between the two sets of positions. The most obvious difference comes from the whopping
linear terms, which of course have not yet been solved for. These linear terms correspond
to a shear of over 500 pixels from bottom to top in the WFC field.
Once the linear terms are solved for and removed, however, I find that the positions in
the two frames still do not match up perfectly, indicating that there is some distortion that
has not been removed. In fact the non-linear distortion difference between the two images
appears to be almost entirely quadratic. Figure 5 shows the position residuals between the
GO-9443 and GO-9028 observations.
Note that this smooth behavior would be impossible to see without an exquisite highorder solution. If we had left out the fine-tuning tabular portion of the solution, the
quadratic behavior would be completely washed out by the interplay of the fine-scale distortions in the two images being compared. Also, if we had not bundled the two chips together
but had transformed them separately, then the quadratic behavior would have been much
harder to see—much of it would be absorbed in the fit for the linear terms in the four
half-chip overlaps.
The smoothness of this relationship means that the high-order and fine-scale solutions
are essentially constant over time. The residuals are even continuous across the chip gaps
(Y ∼ 2048), which means that the inter-chip solution is extremely stable as well. Only the
Astrometry with ACS
21
Figure 5. The position residuals between the central GO-9028 pointing and
the orthogonal GO-9443 pointing for horizontal (left) and vertical (right) strips
through the center of the image. The high-frequency striping is likely related to
the smoothing length of the table solution.
low-order terms appear to change with time. (We should note that we cannot rule out a
variation of the linear terms of the solution along with the quadratic. Since we had to use
these observations to constrain the linear terms, we cannot detect a change in the linear
terms.)
This observed quadratic variation is probably due to changes in focus. It should not
constitute a major limitation to WFC astrometry. However, it will force us either to use
more local transformations or to transform with 2nd-order rather than linear transformations.
Variation of the Solution with Filter. The GO-9443 data set had a dithered set of images
taken through F475W (the filter that was chosen for the initial distortion mapping), but
also had several single exposures at the same pointing through different filters: F435W,
F555W, F606W, and F814W. By comparing the positions in the F475W images with the
positions measured for the other filters, we can see how much the distortion solution varies
with filter.
Whereas in WFPC2, the image-scale had a strong correlation with filter wavelength,
the ACS filters all seem to have the same global solution. The different filters do have
different fine-scale solutions, however, which will require a separate look-up table for each
filter. This is the case for both the WFC and the HRC. Without treatment of the filterdependence, the distortion solution will only be accurate to 0.1 pixel or so.
(While it is true in general that the different filters share the same backbone polynomial
solution and do not have markedly different scales, it is worth noting that F814W images
22
Anderson
do seem to have a slightly different scale than the other filters, amounting to almost a half
pixel from top to bottom of the WFC. This is accounted for in my F814W look-up table
supplement, and could be important for those programs that need to create co-registered
mosaics for a variety of filters.)
An Independent Check. The program of GO-9018 (PI De Marchi) images an intermediate
field in 47 Tuc with the WFC. The program was taken to study the accuracy of the lowfrequency flat fields and consists of several pointings with large offsets through a variety of
filters.
While the star density in this outer field is not high enough to constrain the distortion
solution, there are enough stars in the images to allow us to test the solution. I compared the
distortion-corrected star positions in images taken at different offsets and through different
filters and found that the systematic position residuals are generally less than 0.01 pixel,
which demonstrates that the filter-specific distortion corrections are indeed good to about
this accuracy. Evidently, the breathing-state of the telescope during GO-9018 was similar
to that during the GO-9028 observations, since no variation of the solution is seen here.
GO-9019 (PI Bohlin) has re-observed the center of 47 Tuc with the HRC in a program
analogous to GO-9018. This data set will provide a valuable check on the HRC solution.
5.
Recommendations
There are many aspects to the ACS distortion solution: a polynomial backbone, a fine-scale
solution, a dependence on filter, and a low-order variation with breathing. Not all of these
aspects will be of concern to all observers. Many applications require a solution which is
accurate to only 0.1 or 0.2 pixel, so that basic image rectification and reconstruction can be
done. The polynomial solution produced by Meurer et al. should be entirely adequate for
these purposes. Other applications, such as high-precision differential astrometry or careful
mosaicking, may require a more elaborate solution, such as the one I have constructed here.
6.
End-products of this Analysis
This analysis will produce several products that could be useful to the community. Thus
far, I have created a FORTRAN subroutine that computes the WFC distortion correction
for the five filters in GO-9443: F435W, F475W, F555W, F606W, and F814W. This routine
can be downloaded from my anon-ftp site (ftp://cusp.berkeley.edu/pub/jay/ACS).
Additional products of this analysis will be: the HRC corrections, PSFs for a variety
of ACS filters, and a list of ∼150,000 stars at the center of 47 Tuc with coordinates in a
distortion-free system, so that in the future distortion corrections can be done the easy way.
Acknowledgments. This research was supported by STScI grant GO-9443.
References
Anderson, J. & King, I. R. 1999, PASP, 111, 1095
Anderson, J. & King, I. R. 2000, PASP, 112, 1360
Anderson, J. & King, I. R. 2003, PASP, (in press)
Holtzman, J. A., Burrows, C. J., Casertano, S., Hester, J. J., Trauger, J. T., Watson, A. M.,
& Worthey, G. 1995, PASP, 107, 1065
Meurer, G., Blakeslee, J., Lindler, D., & Cox, C. 2003, this volume, 65
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
ACS Flat Fields and Low-order “L-flat” Corrections from
Observations of 47 Tucanae
J. Mack, R. C. Bohlin, R. L. Gilliland, R. van der Marel, G. de Marchi
Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218
J. P. Blakeslee
Johns Hopkins University, Dept. of Physics & Astronomy, Baltimore, MD 21218
Abstract. The uniformity of the WFC and HRC detector response has been assessed using multiple dithered pointings of 47 Tucanae. By placing the same stars
over different portions of the detector and measuring relative changes in brightness,
low frequency spatial variations in the response of each detector have been measured.
The original WFC and HRC laboratory flat fields produce photometric errors of 6
to 18 percent from corner-to-corner. The required low-order correction (L-flat) has
been applied to the lab flats, and new flat fields have been delivered for use in the
calibration pipeline. Initial results indicate that the photometric response for a given
star is now the same to ∼1% for any position in the field of view. A comparison of
the improved flat fields with high signal observations of the bright earth at 3300 Å
and with preliminary skyflats at 7700 Å also shows agreement to within ∼1%.
1.
Introduction
In 2001, flat field images for the ACS detectors were produced on the ground using the
RAS/HOMS (Refractive Aberrated Simulator/Hubble Opto-Mechanical Simulator) with a
continuum light source. These flats include both low frequency (L-flat) and high frequency
pixel-to-pixel (P-flat) structure. The RAS/HOMS provides an external, OTA-like illumination above its refractive cutoff wavelength of ∼3500 Å. Because the RAS/HOMS optics
are opaque below 3500 Å on-orbit observations of the bright earth are used to create the
UV flats for the HRC. The intrinsic pixel-to-pixel rms of the detector is ∼0.5%. Thus, to
avoid any significant loss of signal-to-noise when applying the flat fields to the science data,
the Poisson counting statistics of the external flats are ∼0.3%, i.e., at least 100,000 electrons/pixel. For a detailed discussion of creating the ground flat reference files, see Bohlin
et al. 2001.
To assess the accuracy of the ground flats, multiple dithered pointings of the globular
cluster 47 Tucanae were made. By placing the same star over different portions of the
detector and measuring its relative changes in brightness, errors in the laboratory flats have
been discovered. This paper is devoted to a discussion of the L-flat corrections derived for
the WFC and HRC detectors. For more a more detailed description of the L-flat program
for the WFC, see Mack et al. 2002.
23
24
Mack, et al.
Sky flats may eventually replace the corrected ground flats and will be built up from
numerous GO images of sparse fields. To achieve the required 100,000 electrons/pixel at
F606W, for example, ∼3 weeks of constant observing would be required. Preliminary lowsignal sky flats for F775W are compared below with the derived stellar L-flat.
Internal lamp flats are also part of the ACS flat fielding program but are relevant for
monitoring purposes only, since these flats have a different illumination than the external
flats. Specifically, the internal lamps blur the shadows of the dust motes and cannot properly
correct for new motes or for existing motes which may have shifted.
2.
Observations
For the WFC and HRC detectors, the accuracy of the ground flats is assessed using observations of the globular cluster 47 Tucanae (NGC 104). The WFC observations are offset
6 arcmin from the cluster center to minimize the effects of crowding which can complicate
the sky subtraction when performing aperture photometry. The L-flat observing program is
summarized in Table 1 which includes the target RA and Dec, the dither step size, and the
filters used for imaging. Each WFC image contains ∼6000 stars of sufficient signal-to-noise
at each of the 9 dither positions, while each HRC image contains ∼3000 stars.
Table 1.
L-Flat Observations of 47 Tucanae
Detector
WFC
Program
9018
Dither Step
22 (11% FOV)
RA
00:22:37.20
Dec
−72:04:14.0
SDSS Filters
F775W
F850LP
BVI Filters
F435W
F555W
F606W
F814W
HRC
9019
6 (23% FOV)
00:24:06.52
−72:05:00.6
F475W
F625W
F775W
F850LP
F435W
F555W
F606W
F814W
For each detector/filter combination, the dither pattern consists of 9 pointings along
a diagonal cross, where the step size in X and Y is 22 for the WFC and 6 for the HRC.
Since the field size is ∼ 200 for WFC and ∼ 25 for the HRC, each step is a large fraction
of the detector field of view, as indicated in Table 1.
The 9-point dither pattern is illustrated in Figure 1 for each detector, where the size
of each dither in pixels is shown with respect to the size of the WFC and HRC detectors.
The central dither position for each detector is shown in bold. Stars which were observed
in at least three images are overplotted on the diagram.
3.
Matrix Solution
A matrix-solution program was developed by R. van der Marel for deriving the low-frequency
flat field corrections from the dithered, stellar point-source photometry. Separate solutions
for each of the filters listed in Table 1 were derived using this algorithm. The details of this
code are described in a separate HST Instrument Science Report, currently in progress.
To summarize, the observed magnitude of a star at a given dither position is assumed
to be the sum of the true magnitude of the star plus a correction term that depends on the
position on the detector. The correction term represents the L-flat. The L-flat, when given
in magnitudes, can be expressed as the product of fourth-order polynomials in the detector
ACS Flat Fields
Figure 1.
Nine-point dither pattern, in geometrically-corrected pixel coordinates, for the WFC and HRC L-flat observations. The central dither for each
detector is shown in bold. Each WFC step (dotted and dashed lines) is ∼440 pixels (11% FOV) in x and in y. Each HRC dither is ∼240 pixels (23% FOV). Stars
which were observed at least three times are plotted.
25
26
Mack, et al.
x and y coordinates. When a set of multiple dithered observations is available for a given
star field, the determination of both the L-flat and the true instrumental magnitude of each
star can be written as an overdetermined matrix equation. This equation has a unique
minimum χ2 -solution that can be efficiently obtained through singular-value-decomposition
techniques.
The matrix solution requires as input the magnitude of each star and its uncertainty.
The appropriate photometric aperture is chosen which includes ∼80% of the encircled energy. This radius corresponds to 5 pix for the WFC and 7 pix for the HRC. Uncertainties
from the sky background and from neighboring stars increase at larger radii. Sigma clipping
is employed to reject stars having large photometric residuals with respect to the variation
in the L-flat. These rejected outliers are mostly due to cosmic rays, image saturation, or
stars falling at the edge of the detector.
In Figure 2, the L-flat solutions derived for F555W are shown for the WFC and HRC.
The residual of the stellar magnitude with respect to the predicted magnitude is displayed,
where stars in the upper left of the WFC (HRC) are too faint (bright) after pipeline calibration with the original ground flats, and stars in the lower right are too bright (faint). There
is a continuous gradient in the L-flat along the diagonal of the detector, which corresponds
to the axis of the maximum geometric distortion. This gradient is of order 10% from corner
to corner and varies with wavelength (see Table 2.) The diagonal in the opposite direction
shows no systematic gradient, and stars falling along this diagonal require no correction.
Because the detectors are rotated approximately 180 degrees with respect to one another
on the sky, the direction of the L-flat gradient for the WFC and HRC detectors is reversed.
Assuming a simple linear dependence on wavelength, the L-flat correction for the remaining wide, medium, and narrow-band WFC filters is derived. The pivot wavelength of
each filter is used for the interpolation, where the resulting L-flat correction is equal to the
weighted average of the L-flat correction for the two filters nearest in wavelength.
A comparison of ACS F550M with WFPC2 F547M photometry indicates that the
“interpolated” L-flat has errors that are no larger than ∼2–3%. Other flats which were derived via interpolation are the HRC and WFC narrow-band filters: F502N, F550M, F658N,
F660N, F892N and the WFC broad-band filters: F475W, F625W. Further study is required
to achieve uncertainties of 1% for these flats.
Table 2. Required Corrections to the Ground Flats, Expressed As a Percent
Gradient from the Upper-left to the Lower-right Corner
Filter
F435W
F475W
F555W
F606W
F625W
F775W
F814W
F850LP
4.
WFC HRC
16%
7%
–
7%
10%
6%
14%
7%
–
7%
13%
8%
15%
9%
18% 12%
Comparison with In-Flight Sky Flats
To verify the L-flats derived from point source photometry, sky flats were created by
J. Blakeslee using WFC exposures of the Hubble Deep Field- North which contains a large
ACS Flat Fields
Too Faint
27
Too Bright
WFC
HRC
Too Bright
Too Faint
Figure 2. Low-order flats derived for F555W from the matrix-solution code. The
WFC (HRC) variation from corner-to-corner is 10% (6%). Black indicates that
the ground flats produce photometry which is too faint with respect to the true
stellar magnitude; white indicates that photometry is too bright.
amount of “blank” sky. For F775W, the combined exposure is 4500 sec. Because this exposure is not high enough signal-to-noise to reproduce the pixel-to-pixel structure of the
detector, SExtractor was used to fit a smooth, low-order bicubic spline to the sky background, after first masking all detected sources. The resulting image was then normalized
by its mean countrate to produce the sky flat field. The ratio of the F775W sky flat to the
ground flat, corrected using the photometrically derived L-flat, shows residuals of less than
1% which are relatively flat across the detector. Thus, the L-flat derived using point-source
photometry is nearly identical to the L-flat derived from an extended source.
For the ultraviolet HRC filters, the pixel-to-pixel flats are derived from observations
of the bright earth. To verify the L-flat technique at visible wavelengths, stellar L-flat
observations using the F330W filter were obtained. Then, the ground flats are corrected by
the derived ultraviolet L-flat and divided by the earth flat to create a residual image. Again,
residuals are less than 1% over the detector. The corner-to-corner gradient in the L-flat
correction is ∼5% for F330W, consistent with the trend in L-flat gradient with wavelength
seen in Table 2.
As more precise, high signal-to-noise sky flats are built up over the next year, detailed
comparisons with L-flat solutions from the 47 Tuc data will be possible, with the anticipated
redelivery of improved flat fields using the sky flats themselves.
5.
L-flat Verification Using 47 Tuc Color-Magnitude Diagrams
To verify the photometry derived after the L-flat corrections are applied, features in the color
magnitude diagram for 47 Tucanae are examined. More than 3500 stars were matched in
the F435W, F555W and F814W filters for the WFC, from one magnitude above the cluster
turnoff to four magnitudes down the main sequence. Zero points were added to each bandpass to approximately match the position of the 47 Tuc color-magnitude diagram (CMD)
from numerous published sources using standard Johnson-Cousins B, V , and I photometry.
In Figure 3, the V vs. B −V and V vs. V −I color magnitude diagrams are shown. The
relevant comparison is not the consistency of the zero points (which are arbitrary for this
28
Mack, et al.
test), but rather the consistency of CMD features for stars observed over different portions
of the detector. In particular, the L-flat corrections are largest in the upper-left and lowerright corners of the detector, so it is natural to consider whether CMDs for these regions
are consistent.
The nearly horizontal CMD region, which is associated with the transition to subgiants,
can be used to test for any positional dependence. Twenty stars were selected from each
corner of the 47 Tuc image for the region near V = 17.15 in the CMD. The difference in
V magnitude measured between the upper-left and lower-right corners is zero to within
0.6% for regions which are separated by more than 2000 pixels along the diagonal. A test
along the opposite diagonal gives a similarly small change.
The nearly vertical CMD region near V = 17.7, which is associated with stars nearing
the main sequence turnoff, can be used to check for positional dependencies in B and I
relative to V . These tests confirm that the photometry is consistent to within several
tenths of a percent over the chip.
Photometry using the new LP-flats for 47 Tuc in the B, V , and I bands yield excellent
CMDs. The ACS CMDs are noticeably tighter than CMDs published from much more
extensive WFPC2 observations within the same field (Zoccalli et al. 2001). These results
confirm that any field dependence in the point source photometry, caused by inaccurate flat
fields, geometric distortion effects, and aperture corrections, is now corrected to within 1%.
6.
Summary
• Low-frequency flat field corrections have been created for most ACS modes, where
the required correction to ground flats is ∼6–18%. The reference files now in the
calibration pipeline are accurate to ∼1% over the FOV for most modes. These files
are revised LP-flats, derived by dividing the original ground P-flats by the derived
L-flat corrections.
• For filters which were not directly observed, an interpolated L-flat was derived from
the weighted average of the L-flats using the two filters nearest in wavelength. Testing
indicates that the interpolated flats are accurate at the 2–3% level. Further study is
required to correct these filters to the 1% level.
• L-flat corrections for the SBC are underway. Observations of the UV-bright star
cluster NGC 6681 were taken in program 9024 for filters F125LP and F150LP. The
corrected SBC reference files should be available in the calibration pipeline in early
2003.
• Determination of the L-flats for other modes, including the ramp filters, the polarizers,
the coronagraph, and the dispersers is still required.
• In the ultraviolet, Earth flats will be used to create high signal-to-noise P-flats for the
HRC. Earth flats for the bluest HRC and WFC filters will be used to verify the stellar
L-flats derived for these filters.
• Sky flats will be built-up over time to compare with the corrected ground flats and may
eventually replace them. To achieve the required signal to reproduce the pixel-to-pixel
structure of the flats, a total of several weeks of integrated time is required.
• Observers are encouraged to check the ACS web site for the latest flat field reference
files available for recalibration.
http://www.stsci.edu/hst/acs/analysis/reference files/flatimage list.html
ACS Flat Fields
Figure 3.
Color magnitude diagrams for WFC observations of 47 Tucanae derived from B, V , and I photometry after applying the L-flat correction. Each
WFC exposure is 60 seconds. A precise transformation to the standard B, V ,
I system has not been attempted, and small color terms may be unaccounted for
in the ACS plots. The WFPC2 CMDs are shown for the same field (Zoccali et al.
2001).
29
30
Mack, et al.
Acknowledgments. We would like to thank the ACS Calibration and Photometry
Working Group for valuable brainstorming sessions related to this work. We also express
thanks to Don Lindler for creating the matched-coordinate lists, Gehrhardt Meuer for useful
comments and suggestions, Colin Cox for input on velocity aberration, and Tom Brown for
sharing his insights on creating L-flats for STIS.
References
Bohlin, R. C., Hartig, G., & Martel, A. 2001, Instrument Science Report ACS 01-11 (Baltimore: STScI)
Mack, J., Bohlin, R. C., Gilliland, R. L., van der Marel, R., Blakeslee, J. P., & de Marchi,
G. 2002, Instrument Science Report ACS 02-08 (Baltimore: STScI)
Zoccali, M., Renzini, A., Ortolani, S., Bragaglia, A., Bohlin, R., Carretta, E., Ferraro, F. R.,
Gilmozzi, R., Holberg, J. B., Marconi, G., Rich, R. M., & Wesemael, F. 2001, ApJ,
553, 733
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
On-orbit Sensitivity of ACS
M. Sirianni
Astronomy Department, Johns Hopkins University, Baltimore, MD 21218
G. De Marchi,1 R. Gilliland, R. Bohlin, C. Pavlovsky, J. Mack
Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218
and The ACS photometric Calibration Working Group
Abstract. Ground measurements of all the components of the Advanced Camera
for Surveys (ACS) allow one to predict the sensitivity of each instrument. Soon after
the installation of ACS we tested the on-orbit sensitivity. We observed spectrophotometric standard stars with the three channels of ACS to calculate the observed-topredicted count rates ratios. We performed a first order correction of the pre-flight
quantum efficiency curve of the detectors to reflect the on-orbit sensitivity measurements. The new curves have been implemented in SYNPHOT which is used by the
Exposure Time Calculator. We report the analysis performed for the first order
corrections of the sensitivity of the three cameras and the progress in developing an
improved sensitivity correction.
1.
Introduction
It is important to determine the observed throughput of all three cameras of ACS to determine accurate photometric zero points, to determine the feasibility of and exposure times
for science programs and finally to calculate transformations to and from other instruments
photometric systems.
The STSDAS package SYNPHOT can calculate the predicted count rates from sensitivity curves for the telescope, ACS mirrors and windows, filters and detectors. Each camera
of ACS consists of several components like mirrors, windows, filters and the detectors. Each
system was carefully characterized and tested as part of the ground calibration of the instrument. The results in terms of reflectivity, transmittance or quantum efficiency have been
used within SYNPHOT to estimate the exposure time for the first ACS observations during
the Servicing Mission Orbital Verification (SMOV). However in some cases the pre-flight
measurements could have been done with a fairly sparse wavelength resolution, therefore
extrapolation, interpolation and sampling errors can be important. In addition, calibration
instruments could have had systematic offsets. A reality check is therefore required for each
instrument. With the data acquired in the first part of the SMOV, pre-launch estimates
of count rates have been compared with observations to derive modification to the input
sensitivity curves. These first observations have been used to derive rough corrections to
the sensitivity curves. These correction have been implemented into SYNPHOT at the end
of August 2002. Further observations obtained during the summer will permit a fine tuning
of the corrections and a better estimate for exposure time prediction.
1
European Space Agency
31
32
2.
Sirianni, et al.
Observations
The observed throughput has been determined using observations of flux standards through
a variety of filters (Proposal ID 9029 and 9654, P.I. De Marchi). In particular the spectrophotometric standards GD 71 and GRW +70 5824 have been observed with the WFC
and the HRC in April 2002 and July 2002, while for SBC observation of HS 2027+0651 and
NGC 6681−STAR 1 have been executed in the same period of time. All the filters have
been used for the observations of GD 71, all the broad band filters and the major narrow
band filters for the observations of GRW+70. The full set of filters was used for the SBC
observation in both epochs.
The star was placed at the center of the aperture and two images have been taken
through each filter. The exposure times have been selected to reach, on average a signal to
noise ratio of ∼350 in the central pixel for broad band filters.
The data acquired in April and July have been used to estimate the first order corrections which have been released at the end of August for all three cameras as new SYNPHOT
tables. Subsequent observations of GRW +70 and NGC 6681 in August and September 2002
have been used to improve such corrections and they will be available within the end of
2002.
3.
Data Reduction and Analysis
All the images have been processed using the STScI standard ACS pipeline CALACS and
the geometric distortion has been corrected running PyDrizzle. Even if a first assessment
of the sensitivity of the camera was done just after the data collection, the final reduction
was repeated after the L flats were made available (see Mack et al., this volume). Aperture
photometry has been performed in the reduced data, the total counts in a 2.5 radius
aperture have been corrected by background contamination using a sky level measured in
an annulus between 3.5 and 5 . Predicted count rates have been estimated using the prelaunch response curves and the spectra of the standard stars. The spectra give photon rates
as a function of wavelength, which are multiplied by the response curves and are integrated
over wavelength.
3.1.
WFC
For the first order correction the data of April and July 2002 have been used. The application of the L flats greatly reduced the observed discrepancies between WFC1 and WFC2
response which should be the same after the flat field normalization (Bohlin et al. 2002).
The observed count rates relative to predicted rates are shown in Figure 1, where each
bar represents the average result for the two spectrophotometric standards. The figure
shows the presence of systematic errors as a function of wavelength. Ground measurement
underestimated the performance of the camera from a minimum of 2% in the red to a
maximum of 22% in the blue. The results of WFC1 and WFC2 agree within a couple
of percent. Since there is a fairly smooth variation with wavelength we believe that the
discrepancy is mostly due to an incorrect measurement of the CCD quantum efficiency or
mirrors and windows throughput more than to errors in the filter transmission curves.
In order to calculate a correction factor to apply to the sensitivity curve we assign
to each ratio in Figure 1 the pivot wavelength of each filter. With such transformation
(Figure 2) we can now calculate a correction curve to apply to the predicted response of
the camera. We averaged the results obtained with the two standard stars in the two CCDs
and fitted a spline function though the points. We needed to contain the fit at the two
edges of the spectral range. In the blue side the trend of the two points at λ < 5000 Å
suggested a constant value of 1.23 for λ < 4000 Å. In the red side we extrapolate the trend
On-orbit Sensitivity of ACS
33
Figure 1.
Ratio of observed-to-predicted count rates using our original estimates
of response curves. Each bar represents the average results for the two standard
stars observed in June and July.
of the three measurements at λ > 7500 Å and set the ratio at 11000 Å to 0.88. Figure 2
shows the derived correction curve for the response.
We have attempted to make the derived response curve smooth; however, in some
wavelength regions, there could be a few percent error in the derived curve. We then used
the derived curve to correct the pre-flight Quantum Efficiency (QE) curve of the detector.
The new CCD QE curve has been implemented in SYNPHOT since late August. The
current version of the exposure time calculator (ETC) will use this new curve. The overall
accuracy is now better than 5% in all filters.
Subsequent observations of the star GRW+70 in August and September 2002 gave us
the opportunity to test the time stability of the sensitivity and to improve the statistics
on our measurements. Figure 4 shows the ratio of observed to predicted count rates after
application of the correction curve for the response. The residual errors are in general less
than 1 or 2%. The two most deviant filters are F606W which shows a residual of almost
3% and the filter F850LP marked as a box in Figure 4. The reason why the filter F850LP
shows a residual of almost 6% is that the count rate predictions for the first order correction
made in August 2002 where calculated using spectra of the standard stars that did not
extend to 11000 Å, the sensitivity limit of the camera, but they were instead truncated at
approximatively 10,500 Å. For the new corrections, that will be implemented in SYNPHOT
by the end of 2002, a theoretical model of the spectra have been used to cover the missing
spectral range.
3.2.
HRC
As for the WFC, only the data collected in April and July were used to calculate the first
order correction for the sensitivity curve. Figure 5 shows the observed count rates relative
to the predicted rates using the pre-flight response curves. The panel in the left shows the
response in each filter, while the right panel shows the same results as a function of the
pivot wavelength of the filters. In general the on-orbit sensitivity is higher than expected;
between 5 and 14% in the range λλ 4500–8500 Å. There is however a well defined dip
34
Sirianni, et al.
Figure 2.
Ratio of observed-to-predicted count rates using our original estimates
of the WFC response curves. Circles show the average ratio for the two standard
stars and for the two CCDs. The dashed line shows the derived correction curve
for the response.
Figure 3.
Quantum efficiency curve of the WFC CCDs. Dotted line shows the
pre-flight measurement. Solid line shows the on-orbit curve after the sensitivity
correction.
On-orbit Sensitivity of ACS
Figure 4.
QE curve
35
Ratio of observed to predicted count rates using the new WFC detector
Figure 5.
Left: ratio of observed-to-predicted count rates using the pre-flight
estimates of the response curves for HRC. Right: Circle show the average ratio for
the two standard stars observed in June and July. The solid line shows the derived
correction curve for the response.
between 2500 and 3500 Å where the sensitivity is ∼15% lower than expected. The solid
line in the right panel of the figure shows the derived correction curve for the response. The
fitted curve nicely reproduce the overall variations in sensitivity, but residual of 5–8% are
still possible in some filters.
As in the previous case we decided to modify the CCD response to reflect the observed
sensitivity. The panel in the left in Figure 6 shows the pre-flight detector QE curve and the
on-orbit curve after the sensitivity correction. This curve was implemented in SYNPHOT
at the end of August 2002.
We repeated the observation of GRW+70 in August and September 2002 with the same
instrument configuration. The analysis of the new data shows that the in-flight sensitivity
is not affected by variation with time. We used the new observations also to re-test the
sensitivity curve with the goal to reduce the residuals and produce a better correction. The
right panel of Figure 6 shows the residuals using the average observed counts of GRW+70
in the three repeated observations. The residuals are usually less than 3%. The biggest discrepancy is the filter F850LP for the problem with the spectra used in the initial calibration
as explained in the previous paragraph.
36
Sirianni, et al.
Figure 6.
Left: pre-flight and corrected quantum efficiency curve for the HRC
CCD. Right: Ratio of observed to predicted count rates using the new detector
QE curve.
Figure 7. Left: Preliminary comparison between observed and predicted count
rates using the pre-flight response curve for the SBC. Solid line shows the derived
correction. Right: Pre-flight and corrected quantum efficiency for the SBC MAMA.
New curves will take into account the modification to the spectra of the standard stars
at λ > 10500 Å and the derived correction will be calculated iteratively to reduce the
residuals. As for WFC the new sensitivity curve is expected to be available by the end of
year 2002.
3.3.
SBC
Observations of HS 2027+0651 and the cluster NGC 6681 have been used to check the onorbit sensitivity of the SBC. At the moment of writing there are no L FLAT available for
the SBC. The corrections reported in this paper and implemented in SYNPHOT at the end
of August are only approximated and might be different from the final version by several
percent. Figure 7 shows the observed-to-predicted count rates ratio and the correction for
the MAMA quantum efficiency curve.
4.
Conclusion
We observed spectrophotometric standard stars with the three cameras of ACS to measure
the on-orbit sensitivity. The response of the WFC is higher than expected from ground
measurement; from a few percent in the red up to ∼ 20% in the blue. The HRC sensibility
On-orbit Sensitivity of ACS
37
is higher than expected in the visual and red but it shows an unpredicted dip in the blue.
Finally, preliminary analysis on SBC data shows that the on-orbit sensitivity could be higher
in the far UV and slightly lower in the near UV with respect to pre-flight estimations.
Corrections of the sensitivity have been applied to the detector quantum efficiency
curves. The first corrections, implemented in SYNPHOT at the end of August 2002, reduce
the observed-to-predicted discrepancy to less than 5%. Follow up observations of the same
standard star in August and September are being used to improve the sensitivity corrections.
Repeated observation the spectrophotometric standards gave us also the opportunity
to calculate the on-orbit encircled energy profile on most of the ACS filters. Since all the
stars are isolated we were able to perform aperture photometry with aperture radii from 0
to 4 arcsec. The sky level was determined in the external annulus between 5 and 6 arcsec.
The result will be used to update the prediction of the ETC and to provide users an estimate
of the aperture correction.
Once the final sensitivity correction are available, synthetic zero points will be computed for all filters and transformations to and from other instruments photometric systems
will be also calculated (Sirianni et al. 2003).
References
Bohlin, R. C., Hartig, G., & Sparks, Wm. 2002, Instrument Science Report ACS 02-03
(Baltimore: STScI)
Mack, J., et al. 2003, this volume, 23
Sirianni, M., et al. 2003, in preparation
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
The Wavelength Calibration of the WFC Grism
A. Pasquali, N. Pirzkal and J. R. Walsh
ESO/ST-ECF, Karl-Schwarzschild-Strasse 2, D-85748 Garching bei München,
Germany
Abstract. We present the wavelength solution derived for the G800L grism with
the Wide Field Channel from the spectra of two Galactic Wolf-Rayet stars, WR 45
and WR 96. The data were acquired in-orbit during the SMOV tests and the early INTERIM calibration program. We have obtained an average dispersion of 39.2 Å/pix
in the first order, 20.5 Å/pix in the second and −42.5 Å/pix in the negative first
order. We show that the wavelength solution is strongly field-dependent, with an
amplitude of the variation of about 11% from the center of the WFC aperture to the
corners. The direction of the field-dependence is the diagonal from the image left
top corner (amplifier A) to the bottom right corner (amplifier D). These trends are
observed for all grism orders. We also describe the calibration files derived from the
SMOV and INTERIM data which are used by the ST-ECF slitless extraction code
aXe.
1.
Introduction
The Advanced Camera for Surveys (ACS) has been designed to perform low-resolution,
slitless spectroscopy over a wide range of wavelengths, from the Lyα line at λ = 1216 Å to
∼ 1 µm. One optical grism, one blue prism and two near-UV prisms cover this range and
are coupled with the Wide Field (WFC) and High Resolution (HRC) Channels, the HRC
and the Solar Blind Channel (SBC) respectively.
The WFC and the HRC make use of the same grism, G800L which works between
∼ 5500 Å and ∼ 1 µm. Its nominal dispersion is ∼ 40 Å/pix and ∼ 29 Å/pix in first order
for the WFC and HRC respectively.
The HRC also features a prism, PR200L, which covers the spectral range between
∼ 2000 Å and 5000 Å, with a non linear dispersion varying from 2.6 Å/pix at λ = 1600 Å
to 91 Å/pix at λ = 3500 Å and 515 Å/pix at λ = 5000 Å.
The SBC is equipped with two prisms, PR110L and PR130L which range from
∼ 1150 Å and 1250 Å to 2000 Å with a resolving power of ∼ 80 and ∼ 100, respectively, at λ = 1600 Å. In particular, PR130L does not include the geocoronal Lα line for
low background measurements.
Pasquali et al. (2001b) showed that the high angular resolution of the ACS may easily
decrease the effective resolution of the grism, since, when no slit is used, the grism nominal
resolution is convolved with the object size along the dispersion axis. The extension of any
grism spectrum along the cross-dispersed direction is set by the size of the object which
acts as an extraction aperture. This is also an additional source of degradation when the
whole spectrum is summed along the cross-dispersion axis.
The amplitude of these effects was investigated by simulating with SLIM 1.0 (Pirzkal
et al. 2001b) the spectrum of the Galactic planetary nebula NGC 7009, and by increasing the
linear size of the object as well as its orientation in the sky. The simulated grism spectra
indicated that line blending becomes severe when objects are observed with a diameter
38
The Wavelength Calibration of the WFC Grism
39
larger than 2 pixels (0.1) and with a major axis at PA > 45o with respect to the dispersion
axis (cf. Pasquali et al. 2001b).
These limits pose strong constraints on the selection of targets for the in-orbit wavelength calibration of the ACS spectral elements. Indeed, such calibrators have to be sorted
by:
1. high brightness, to allow for short exposure times and time-series observations across
the field of view;
2. a large number of emission lines in their spectra;
3. the absence of an extended nebula, which would otherwise degrade the spectral resolution;
4. negligible spectro-photometric variability, to be able to identify emission lines at any
observation epoch;
5. minimum field crowding, to avoid contamination by spectra of nearby stars;
6. visibility, to allow repeated HST visits.
The above set of requirements rules out planetary nebulae (PNe) as possible wavelength
calibrators, at least in the case of the optical G800L grism. Indeed, PNe are resolved by
HST up to the Large Magellanic Clouds and hence do not meet requirement #3, while PNe
in M31 are compact enough but faint and therefore can not fulfil requirements #1, 6 and 5
as they also lie in crowded fields (Pasquali et al. 2001a).
Wolf-Rayet stars (WRs) of spectral type WC have been shown to satisfy all the requirements, at the expense of introducing a further constraint which concerns the velocity
of their stellar wind. Indeed, the wind velocity in WRs can be as slow as 700 km s−1 and
as fast as 3300 km s−1 (cf. van der Hucht 2001). A typical wind speed of 2000 km s−1
produces a line broadening of about 1.3 and 1.9 pixels in the grism first order with the
WFC and the HRC, respectively. Therefore, to limit the loss of resolution due to objects
with broad emission lines, WR stars should be selected with Vwind ≤ 2000 km s−1 (Pasquali
et al. 2001a).
2.
The Observational Strategy
We eventually selected two Galactic WR stars from the VIIth Catalogue by van der Hucht
(2001) which meet the listed criteria. Their basic properties, coordinates, V magnitude and
wind velocity are in Table 1.
Table 1.
The WR Stars Selected for the Wavelength Calibration of the ACS Grism
Star
WR 45
WR 96
Spectral
type
WC6
WC9
RA (2000)
DEC (2000)
V mag
11:38:05.2
17:36:24.2
−62:16:01
−32:54:29
14.80
14.14
Wind speed
(km s−1 )
2100
1100
Both stars had been observed from the ground with the ESO/NTT EMMI spectrograph
with the purpose to acquire high-resolution spectra which would be later used as templates
for the comparison with the ACS grism observations. The EMMI spectra of WR 45 and
WR 96 are plotted in Figure 1, where the dispersion is 1.26 Å/pix.
40
Pasquali, et al.
Figure 1. The spectra of WR 45 and WR 96 acquired with the ESO/NTT EMMI
spectrograph with a dispersion of 1.26 Å/pix.
2.1.
Observations During the SMOV Tests
WR 45 was observed as part of the Servicing Mission Orbital Verification (SMOV) tests
(ID 9029, PI Pasquali), at the end of April to early May 2002. Spectra were taken at
five different pointings across the field of view (f.o.v) of the WFC: W1 close to the center
of chip 2, W3 and W5 close to amplifiers C and D of chip 2 and W7 and W9 close to
amplifiers A and B in chip 1. These pointings are shown in Figure 2.
At each position, a pair of direct and grism images were acquired, and repeated two to
four times, either in the same visit or in a subsequent one to verify the stability of the filter
wheel positioning. The direct image, which provides the zero point of the grism dispersion
correction, was taken in the F625W and F775W filters, in order to check the target position
stability with wavelength. The adopted exposure times were 1 s for the direct imaging and
20 s for the grism.
2.2.
Observations During the INTERIM Program
WR 96 was observed during the INTERIM calibrations (ID 9568, PI Pasquali) in June 2002.
The observational strategy was similar to WR 45, but the number of individual pointings
was increased to 10 by adding to the SMOV positions the W2, W4, W8, W10 pointings
and the centre of chip 1 (cf. Figure 2).
Monodimensional spectra of WR 45 and WR 96 were extracted from the raw, non
drizzled images using the ST-ECF slitless spectra extraction code, aXe (Pirzkal et al. 2001a,
http://www.stecf.org/software/aXe/index.html).
3.
The Grism Characteristics
The extraction of slitless spectra relies on a number of parameters:
1. the shift in the X and Y coordinates between the position of the target in the direct
image and the position of the zeroth order in the grism image;
2. the tilt of the spectra;
The Wavelength Calibration of the WFC Grism
41
Figure 2.
The pointings across the f.o.v of the WFC used during the SMOV and
INTERIM observations.
3. the separation in pixels along the dispersion axis of the nth grism order from the
zeroth;
4. the length in pixels along the dispersion axis of each grism order.
While the shift allows a spectrum to be identified in the grism image given the coordinates of the target in the direct image, the tilt enables it to be traced. The separation
and the length of the grism orders are used to set the extraction aperture for each order
spectrum.
Because of the severe geometric distortions in the WFC, these quantities are expected
to be field-dependent.
3.1.
The X- and Y -shifts
The X- and Y -shifts could be measured for all the pointings, except W7 and W8 whose
zeroth orders fall outside the physical boundaries of the grism image. The remaining are
listed in Table 2 in units of pixels; they are the difference between the target position
in the direct image and the zeroth order coordinate in the grism image. The values are
averages among multiple measurements available for each pointing. The standard deviation
is typically 0.1 pixels.
A decrease can be recognized for the Y -shift along the diagonal from amplifier A
(W7 position) to D (W3 position).
3.2.
The Tilt
The tilt of the spectra was derived by fitting the (X, Y ) coordinates of the emission line
peaks along the dispersion axis, measured from the negative third to the positive third order
(the negative orders at smaller x pixels than the zeroth order x coordinate, the positive ones
at larger x pixels). A first order polynomial was used to determine the slope of the spectra
with respect the X axis. Repeated measurements were averaged and the standard deviation
of the spectra tilt was determined to be typically of 0o.02. The average tilt is shown in Table 2
as a function of position in the WFC.
On average, the tilt of the grism spectra in the WFC is −1o.98, and it is field-dependent
as it increases along the W7–W3 diagonal by 1o.1.
42
Pasquali, et al.
Table 2.
The X- and Y -shifts and the Spectra Tilt across the WFC Aperture
Position
W1
W2
W3
W4
W5
W7
W8
W9
W10
Chip1 center
3.3.
X-shift
(in pixels)
113.02
120.02
102.02
106.32
115.24
Y -shift
(in pixels)
−3.14
−3.56
−1.85
−2.88
−2.77
112.83
108.15
117.32
−4.57
−3.39
−4.18
Tilt
(in degrees)
−1.91
−1.95
−1.43
−1.85
−1.53
−2.52
−2.11
−2.32
−1.95
−2.25
The Separation and Length of the Grism Orders
The distance in pixels of the grism orders from the zeroth order and their approximate
length were measured by counting the pixels along the X axis whose counts are 3σ above
the background level. The mean length and FWHM of the zeroth order are 23 and 4.4 pixels,
respectively. The average (across the f.o.v of the WFC) order separations and lengths are
listed in Table 3 in pixels.
Table 3.
The Separation from the Zeroth Order and the Length in Pixels of the
Grism Orders (from the SMOV data)
Parameter
Separation
Length
4.
1st ord.
93
156
2nd ord.
251
125
−1st ord.
−122
102
−2nd ord.
−247
111
The Method of Wavelength Calibration
The grism spectra were extracted in both units of pixels and wavelength adopting for
the latter the wavelength solutions derived from the ground calibrations of the ACS. This
allowed us to derive the mean FWHM in Å of the lines in each grism order. The NTT/EMMI
template spectra of WR 45 and WR 96 were then convolved by these mean FWHMs, their
lines reidentified and the line wavelengths re-measured. The position in pixels of the same
lines was measured in the ACS grism spectra with respect to the target X position in the
direct image and tables of pixels vs. wavelengths were built. Each table was then fitted with
the routine POLYFIT in IRAF and a wavelength solution was determined for each grism
order.
This procedure was applied to each grism spectrum in all positions across the f.o.v of
the WFC.
The Wavelength Calibration of the WFC Grism
43
Figure 3.
The grism first order spectra acquired for WR 45 at position W1 (left)
and WR 96 at the centre of chip 1 (right).
5.
The Wavelength Solutions for the WFC and G800L Grism
In this proceeding we present the wavelength solutions (and their field-dependence) computed for the grism first, second and negative first orders which are the brightest. A field
map of the dispersion correction for the grism third, negative second and negative third
orders can be found in Pasquali et al. (2003).
5.1.
The Grism First Order
An example of the grism first order as obtained in position W1 and at the centre of chip 1
is shown in Figure 3 for WR 45 (left) and WR 96 (right).
The dispersion correction for the grism first order is reproduced by a second order
polynomial of the form: λ = λ0 + ∆λ0 X + ∆λ1 X 2 , where X is the distance along the
dispersion axis from the target X position in the direct image.
The wavelength solutions obtained for the ten pointings are reported in Table 4. The
tabulated values are averages of multiple measurements of the dispersion parameters derived
for each pointing. The typical RMS associated with the fits is 3 Å/pix, while the typical error
on λ0 is 7 Å. The uncertainty in the first-order term of the dispersion (∆λ0 ) is 0.2 Å/pix.
Table 4.
The Wavelength Solutions Obtained for the Grism First Order as a
Function of Position across the f.o.v of the WFC
Position
W1
W2
W3
W4
W5
W7
W8
W9
W10
chip 1 center
λ0
(Å)
4815.25
4777.62
4811.80
4760.06
4803.95
4800.86
4772.51
4795.39
4777.90
4787.27
∆λ0
(Å/pix)
39.79
37.28
44.03
41.94
39.13
35.09
36.23
39.64
40.83
37.82
∆λ1
(Å/pix2 )
0.0099
0.0098
0.0096
0.0108
0.0097
0.0068
0.0098
0.0095
0.0130
0.0112
The first order dispersion (∆λ0) is clearly field-dependent: it worsens along the diagonal from the W7 (the pointing with the highest dispersion) to the W3 position (lowest
44
Pasquali, et al.
Figure 4.
The grism second order spectra acquired for WR 45 at position W1
(left) and WR 96 at the centre of chip 1 (right).
dispersion) by 22% of the value at position W1. Alternatively, it can be said that the amplitude of the field-dependence between the centre of chip 2 and the W3 and W7 corners is
11% of the value in the W1 position.
5.2.
The Grism Second Order
The grism second order spectra obtained in the W1 position and at the centre of chip 1 are
plotted in Figure 4 for both WR 45 and WR 96.
The second order overlaps with the first order at λ 5400 Å and does not extend
beyond 9000 Å. For these reasons, the wavelength solution was determined with a first-order
polynomial fit, i.e., λ = λ0 + ∆λ0 X, where X is again the distance along the dispersion axis
from the target X position in the direct image.
The results are listed in Table 5. As for Table 4, these values are averages among multiple measurements available at each pointing. Typical standard deviations are 0.1 Å/pix
and 7 Å on the dispersion and zero point, respectively. The RMS values of the fits are
about 3 Å.
Table 5.
The Wavelength Solutions Obtained for the Grism Second Order as a
Function of Position across the f.o.v of the WFC.
Position
W1
W2
W3
W4
W5
W7
W8
W9
W10
chip 1 center
λ0
(Å)
2432.38
2400.40
2445.01
2411.01
2405.74
2411.48
2391.95
2418.70
2409.33
2406.73
∆λ0
(Å/pix)
20.75
19.63
22.85
21.89
20.54
18.33
19.13
20.72
21.56
19.98
The field dependence noticed earlier for the grism first order is also present in the
dispersion of the second. The amplitude of the dispersion variation from center to the
W7 and W3 corners is about 11% of the dispersion in the W1 position. Once again, the
dispersion decreases along the diagonal from W7 to W3.
The Wavelength Calibration of the WFC Grism
45
Figure 5. The grism negative first order spectra acquired for WR 45 at position
W1 (left) and WR 96 at the centre of chip 1 (right).
5.3.
The Grism Negative First Order
The negative first order spectra of WR 45 and WR 96 are shown in Figure 5.
Since the resolution is here lower than for the positive first order and the noise higher,
the wavelength solution of the negative first order was fitted with a first-order polynomial,
λ = λ0 + ∆λ0 X where X is the distance in pixels along the dispersion axis from the target
X position in the direct image. The measurements of dispersion and zero point obtained
from multiple spectra acquired at the same pointing were averaged and are presented in
Table 6.
Table 6.
The Wavelength Solutions Obtained for the Grism Negative First Order
as a Function of Position across the f.o.v of the WFC.
Position
W1
W2
W3
W4
W5
W7
W8
W9
W10
chip 1 center
λ0
(Å)
−4820.51
−5026.08
−4882.90
−4808.92
−4995.32
∆λ0
(Å/pix)
−41.71
−40.05
−46.48
−44.37
−41.54
−4858.67
−4862.51
−4784.49
−41.96
−43.81
−40.14
Since W7 and W8 positions are closer to the edge of the field than W5 (cf. Figure 2),
the negative first order falls physically outside the frame. Nevertheless, a variation in the
dispersion of about 11% of the value in W1 is still detected between the W1 and W3
positions. The standard deviation is 0.1 Å/pix and 27 Å on the dispersion and the zero
point, respectively. The typical RMS of the first-order polynomial fits is 9 Å.
6.
Products Delivered to Users
The average dispersion coefficients derived for the ten positions across the f.o.v of the
WFC have to be parametrized as a function of position in order to extract and calibrate
46
Pasquali, et al.
spectra anywhere within the WFC aperture. We thus derived a two dimensional fit for
each parameter of the dispersion solutions of each grism order, where each parameter is a
function of the (X, Y ) coordinates of the target in the direct image. These 2D fits were
perfomed by adopting surface fits polynomials. The same was also done for the X- and
Y -shifts, the tilt of the spectra, the orders separation and length.
The above fits are stored in calibration files used by the ST-ECF slitless spectra extraction code aXe (Pirzkal et al. 2001a) and are delivered together with the software package.
Once the wavelength solution was determined, the flat-field correction and the flux
calibration using the SMOV and INTERIM spectra of two white dwarfs, GD 153 and
G191B2B could be formalised. This is fully described in Pirzkal et al., this volume, p. 74.
It is also possible, at this stage of the calibrations, to correct the extracted spectra for CCD
fringing. The modeling of the WFC fringing is explained in detail in Walsh et al., this
volume, p. 90.
References
van der Hucht, K. A. 2001, The VIIth Catalogue of Galactic Wolf-Rayet Stars, New AR,
45, 135
Pasquali, A., Pirzkal, N., & Walsh, J. R. 2001a, Selection of Wavelength Calibration Targets
for the ACS Grism, ST-ECF Instrument Science Report ACS 2001-04
Pasquali, A., Pirzkal, N., & Walsh, J. R. 2003, The In-orbit Wavelength Calibration of the
WFC Grism, ST-ECF Instrument Science Report ACS 2003-01, in preparation
Pasquali, A., Pirzkal, N., Walsh, J. R., Hook, R. N., Freudling, W., Albrecht, R., Fosbury,
R. A. E. 2001b, The Effective Spectral Resolution of the WFC and HRC Grism,
ST-ECF Instrument Science Report ACS 2001-02
Pirzkal, N., Pasquali, A., & Demleitner, M. 2001a, ST-ECF Newsletter, 29, 5
Pirzkal, N., Pasquali, A., Walsh, J. R., Hook, R. N., Freudling, W., Albrecht, R., & Fosbury, R. A. E. 2001b, ACS Grism Simulations using SLIM 1.0, ST-ECF Instrument
Science Report ACS 2001-01
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
Growth of Hot Pixels and Degradation of CTE for ACS
A. Riess
Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218
Abstract. The anneal rate of hot pixels on the ACS WFC is ∼60%–65%, significantly lower than the characteristic anneal rate of 80–85% seen for other CCDs flown
on HST (i.e., WFPC2, STIS, and ACS HRC). The ACS WFC is annealed in the
same way as the other HST CCDs and there is no firm understanding at this time of
the source of the difference. After ∼7–8 successive anneals, the cumulative fraction
of annealed pixels reaches an apparent plateau at ∼70%. The fitted, successive annealing function is used to project forward in time the expected fractional coverage
of the CCD by hot pixels. Approximately 2 years after launch the coverage by hot
pixels is expected to exceed that by cosmic rays in a ∼1000 sec exposure. At the
nominal end of the HST mission (2010) the coverage by hot pixels would be ∼6%,
i.e., one out of every 16 pixels. Because hot pixels are readily flagged and corrected or
discarded they do not pose a serious threat to science observations, but their growing
presence require careful dithering and consideration. For CTE, internal tests such
as the cosmic ray tail measurements show the degradation of CTE on ACS which is
most pronounced for WFC. Simple scaling from WFPC2 provides some quantitative
estimates for photometry.
1.
Introduction
The anneal rate of new hot pixels (dark current > 0.04 e/s) on the ACS WFC has been
a disappointingly-low ∼60% in the first 8 monthly anneal cycles of the instrument (Riess,
Mutchler, Van Orsow 2002, Instrument Science Report 02-06). This rate is significantly
lower than the observed and characteristic value for other HST CCDs; 80–85% for WFPC2,
STIS, and ACS HRC. The anneal rate is significantly less than 60% for pixels which are
much hotter than 0.04 e/s. The likely consequence of poor annealing is a greater fractional
coverage of the camera by pixels with elevated dark current than the experience of other
HST CCDs.
The reason for the low anneal rate for the ACS WFC is not clear at this time. The
WFC is annealed at approximately the same temperature (∼+20 C), for no shorter a time
interval (∼24 hours) and with the same frequency (monthly) as the other instruments.
The answer may lie in the details of the manufacturing process of the CCD or in the way
in which the CCD is read-out. Expert CCD consultants have been unable to provide a
reliable explanation for the poor annealing (private communication, Blouke and Janesick).
One possibility may involve a difference in the way the chip is operated during integration
and read-out. During integration the chip is used in MPP mode, but when it is read-out,
it is switched to non-MPP mode. This results in “turning-on” the Si-SiO interface states
that are normally passivated when integrating in MPP mode. (Mark Clampin, private
communication). Unfortunately the chip electronics dictate this switching which cannot be
disabled, even for a test.
Here we seek to quantify, empirically model and project forward in time the likely
future population of hot pixels on the ACS WFC. Using our predictive model we then test
the utility of increasing the frequency of anneals to bi-monthly.
47
48
Riess
Figure 1. The growth of WFC hot pixels (dark current <0.04 e per s) with
time (log scale). The “saw-tooth” pattern reflects the continual production and
monthly annealing of hot pixels. Anneals heal ∼60% of new hot pixels.
2.
Observations
The characteristic “saw-tooth” pattern of the growth and annealing of hot-pixels seen for
ACS WFC during the first two anneal cycles (and all other CCDs flown on HST ) has
continued through the ensuing 6 anneal cycles. New hot pixels with dark current > 0.04 e/s
(a number which is ∼18 times and ∼13 standard deviations above the mean dark current)
steadily develop at a rate of ∼1200 pixels per day (see Figure 1). Once a month, the CCD
is annealed and between 60% and 65% of the hot pixels created during the preceding month
return to normal dark current production. Without any annealing, ∼10% of the CCD would
be covered by hot pixels by early 2006. For the 40% of “persistent” hot pixels which do not
anneal at their first opportunity, the likelihood that they anneal in any future anneal drops
precipitously, a result which is consistent with WFPC2 and the ACS HRC.
Current projections for the hot pixel coverage of WFC are highly dependent on both
the anneal rate of new hot pixels as well as the anneal rate of the persistent hot pixels.
New hot pixels which fail to “heal” become increasingly unlikely to heal in each successive
anneal. After approximately 7 or 8 anneals (the total number of anneal cycles we have
observed to date) the cumulative anneal rate reaches an apparent plateau for pixels which
became hot during the first or second anneal cycle (i.e., in April or May or 2002). The
“persistent-pixel-annealing function,” as seen in Figure 3, can be approximated by a simple
series of monthly anneal rates of 62%, 6%, 5%, 4%, 3%, 2%, 1%, 0%. After 7 or 8 anneal
cycles a cumulative fraction of ∼70% of hot pixels will have been annealed leaving ∼30%
to remain hot (presumably for the long term).
Using this function we can project forward in time the expected hot pixel population.
The result are Figures 1 and 2. Using as a reference the fractional WFC coverage by cosmic
rays in a 1000 sec exposure (∼1.5%), we expect to reach this coverage by hot pixels less
Growth of Hot Pixels and Degradation of CTE for ACS
Figure 2.
49
As Figure 1 on a linear scale.
than 2 years after launch or by the beginning of 2004. We would project a ∼3% coverage
by ∼4 years after launch or by early 2006. We stress that accurate projections are difficult
and grow more uncertain with the time interval of the projection.
For comparison, 2.5% of WFPC2 is currently covered by hot pixels (dark current >
0.02 e/s). However, a comparison between WFPC2 and ACS WFC coverage by hot pixels
must account for the difference in the dark current limit used to define a hot pixel. For
WFPC2 the limit is 0.02 e/s or half of the ACS WFC value. A simple correction derived
from ACS WFC is that there are 1.4 times as many hot pixels at the WFPC2 threshold
as for the ACS threshold. Using this conversion we conclude that currently ACS WFC has
1/3 the fractional coverage by hot pixels as currently exhibited by WFPC2.
2.1.
Would Bi-Monthly Annealing Help?
Would bi-monthly annealing help reduce the long term growth of hot pixels? The evidence
in hand suggests it would not. If the persistent-pixel-annealing function really does reach
a plateau (as indicated by the data; see Figure 3) then simply increasing the number of
anneals (or their frequency) would have negligible impact in the long term.
Another way in which bi-monthly annealing would differ from the current monthly
annealing (and perhaps yield better results) is by reducing the mean time interval between
the formation and attempted annealing of a hot pixel (from ∼14 days to ∼7 days). However,
comparisons of the anneal rate for hot pixels formed at the beginning and at the end of an
anneal cycle indicate that the anneal rate does not depend on this time interval. For both
the ∼30 day old and the ∼1 day old hot pixels, the anneal rate was the same ∼60% and the
cumulative anneal rate after 8 cycles was ∼70%. Therefore, simply decreasing the interval
between annealings would not appear to have an impact.
Given our generally poor understanding of the anneal process we cannot rule out the
possibility that bi-monthly annealing could impact the anneal rate in some more subtle way.
50
Riess
Figure 3.
The cumulative fraction of hot pixels which are healed after successive
anneals. Hot pixels formed during the first anneal cycle (April 2002) are shown as
open symbols while the asterisks show those formed during the second anneal cycle
(May 2002). After the first anneal which is ∼63% successful, succeeding anneals
are less successful and after ∼7–8 anneals the cumulative anneal rate reaches an
apparent plateau.
We can only say at this time the data indicates it will not help and that bi-monthly annealing
is probably not a solution to the rapid hot pixel growth on WFC. We do recommend
further study of this problem which might culminate in additional experiments. From past
experience with other HST instruments, it has been shown that the length of time of the
anneal is not an important parameter. However, other possibilities might include (but are
not limited to) warming the CCD by pointing at the bright Earth during the anneal (or
other ways of increasing the temperature of the CCD during the anneal), reading out the
CCD during the anneal to increase the energy available to the lattice to break the bonds
of the damaged site of the crystal, annealing during warm attitudes, or running the CCD
colder to reduce the dark current in the already hot pixels.
2.2.
Science Impact
Because the location of hot pixels is known from dark frames, they are readily flagged and
discarded. In principle they can be corrected without discarding, but because the noise in
hot pixels is greater than Poisson, corrections are of only limited value. The best strategy
for mitigation is dithering. For a well-dithered image, a given fractional coverage by hot
pixels of the CCD represents an equal fractional reduction in the effective exposure time.
Over the next few cycles this will result in an effective reduction of exposure time of 2%–3%
which will have little to no science impact. For searches for rare and faint transients (e.g.,
high-redshift supernovae), an additional exposure (e.g., 5 instead of 4 in an orbit) may be
required in future cycles to insure each pixel is clean from contamination. Alternatively,
contemporaneous dark frames can be used to reject transients found in the position of hot
pixels. It is possible that a reduction in the operating temperature of the ACS WFC due to
the aft-shroud cooling system could further mitigate the hot pixel problem by reducing the
dark current of hot pixels. Upcoming tests include raising the temperature of the camera.
Growth of Hot Pixels and Degradation of CTE for ACS
Figure 4.
Degradation of ACS WFC CTE from cosmic ray tails.
Figure 5.
Degradation of ACS HRC CTE from cosmic ray tails.
51
52
Riess
Stellar loss=
worst ave
e- in CTE Tails
50 WFPC2
40
el
rall
30
20
e- in CTE Tails
40% 6%
30% 5%
TE
Serial C
10
-10
50
20% 3%
10% 2%
1995 1996 1997 1998 1999
TE
ACS WFC
40
30
ed
ect
lC
alle
Par
j
Pro
20
10
Projected Serial CTE
0
2003 2004 2005 2006 2007
-10
0
2.3.
E
CT
Pa
0
Figure 6.
50% 8%
500 1000 1500 2000
Days Since Launch
Projected ACS WFC CTE degradation versus WFPC2.
CTE
It’s too early and the data is not yet available to determine the impact of imperfect CTE
on photometry. However, we have used internal diagnostics to determine the relative degradation of CTE on WFC and HRC. As seen in Figures 4, 5, 6 of this paper, only the WFC
parallel is getting markedly worse. Even this level is still not bad. Using a simple scaling
from WFPC2 we would expect typical sources in the middle of the chip with average background to have only ∼1% to 2% losses to CTE (but as much as 5% to 10% in the worst
cases such as very faint sources on little background at the edge of the chip). However, until
an external measurement is available (in early 2003) its too soon to provide a calibration
of CTE for ACS.
Acknowledgments. We wish to thank Mark Clampin, Doug van Orsow, Max Mutchler, Roeland van der Marel, Marco Sirianni and members of the ACS and ID Teams for
helpful discussions.
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
ACS Calibration Software
W. B. Sparks
Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218
Abstract. The current status of ACS calibration software is described, with emphasis on the correction of geometric distortion. Plans for the future are outlined,
as are a variety of support utilities which we anticipate providing.
1.
Overview
The Advanced Camera for Surveys (ACS) calibration pipeline has been processing science
observations for about 6 months. From the outset the pipeline processing environment has
supported On-The-Fly-Reprocessing, or OTFR. Indeed, for the majority of people, OTFR
will be the most relevant mode of operation. Clearly, not everything is done by the pipeline,
and often parameters need adjusting for specific data sets, so stand-alone use of the ACS
software tools is also encouraged. A significant number of utilities are provided within the
“pyraf” environment which also gives access to the normal suite of iraf/stsdas software.
The initial pipeline processing of ACS data is accomplished with calacs, and the subsequent correction for geometric distortion and associated image combination is accomplished
with PyDrizzle. The latter includes mapping the two 4096 × 2048 chips onto a single image.
Some of the calibration observations are obtained essentially contemporaneously with
the data, in particular the daily darks which track the growth of hot pixels. These calibration
observations require accumulation of data through a (partial) anneal cycle, processing and
delivery to support the pipeline, hence there is an unavoidable delay before the best reference
files are available for use. In addition there are occasional updates to calibration as new
information becomes available, such as photometric measurements. Information on the
appropriate reference files may be found on the ACS web page at STScI. Hence, in general
both new observers and archival investigators should expect to use OTFR when they retrieve
their data.
Below, the operations performed in the ACS calibration pipeline are described as well
as some description of the ACS stand-alone software tool usage. We recognize that there is
a need for improved cosmic ray rejection software, and we are actively working on this as
well as on providing a number of additional support utilities.
2.
ACS Pipeline Processing
At the time of writing, the current version of calacs is 4.1c. This software performs basic
standard image reductions. For the CCDs, it debiases, subtracts the dark frame, and flatfields the data. Shutter shading and pre-flash can be corrected for. Known bad pixels
and saturated data are flagged and if a CR-SPLIT has been specified, calacs will perform
cosmic ray rejection. For the MAMA detectors a global linearity correction is applied.
Finally, photometric keywords and image statistics are computed.
PyDrizzle 3.3d currently runs on the output products of calacs. PyDrizzle geometrically corrects the data and provides image combination if appropriate. The final drz image
is a single fits file containing three extensions. The first, SCI, is the (single) science image in
53
54
Sparks
units of “count rate,” and the other two images are weight and context. The weight image
provides the weight of the pixel as used by drizzle (see drizzle documentation elsewhere,
Fruchter & Hook 2002) and the context image is a bitmap giving information on which
images contributed to the output pixel.
The convention adopted for flat-fielding is that of “sky-flat”; hence the geometrical
area of a pixel imprints itself on the flat-field, in addition to the photometric “sensitivity”
of the pixel. Drizzle accounts for this effect and final output drz images are intended to be
fully photometrically corrected. By contrast, if the flat-fielded images are analysed prior
to geometric correction with drizzle, then allowance must be made explicitly for the pixel
area. A utility to provide a pixel-area map is in test.
Adopting the OTFR philosophy above, calibrated ACS data are not archived: only the
raw observations are archived.
Detailed descriptions of calacs may be found in the ACS Instrument Handbook (Pavlovsky et al. 2002), through the STScI ACS web site, and in a series of Instrument Science
Reports (ISRs) for ACS, also available through the STScI ACS web pages. These include
ACS Instrument Science Reports 99-03 (Hack 1999a), 99-04 (Mutchler et al. 1999a), 99-06
(Van Orsow et al. 1999), 99-08 (Hack 1999b), 99-09 (Mutchler et al. 1999b), 00-03 (Sparks
et al. 2000), 01-10 (Sparks et al. 2001), and a new Instrument Science Report currently
in review (Sparks et al. ISR 02-TBD). Detailed description of PyDrizzle may be found
through the STScI web site, in particular http://www.stsci.edu/hst/acs/analysis/drizzle,
with additional documentation in the ACS Instrument Handbook (Pavlovsky et al. 2002),
the ACS Data Handbook (Mack et al. 2002), and Hack (2002), Hack et al. (2003).
3.
Dithering, Drizzling and PyDrizzle
There are two related issues that can be dealt with simultaneously using drizzle. The first
is that ACS has significant internal geometrical distortion. Knowledge of the distortion
is captured in the reference file “IDCTAB” which uses fourth-order polynomials. This
distortion is implicit in the flat fields, since the flat-fielding convention we adopt is of skyflats. Drizzle corrects for this distortion in a photometrically valid fashion.
The second issue is that the observing system for ACS (and other instruments) allows
for “dithered” observations. That is, a sequence of exposures can be specified with the
telescope pointing shifted slightly between observations. These small offsets serve to help
eliminate bad pixels and hot pixels, span the gap between the chips, and improve pixel
sampling somewhat. The same guide stars are used for a dither sequence, offering the
necessary high accuracy in relative pointing for image combination. The data from a dither
sequence are formally “associated” in the pipeline, and drizzle combines all the images in
the association.
Larger regions may be covered using “mosaic” sequences, however these would not
use the same guide stars and the resulting data do not form a formal “association” in the
pipeline.
The drizzle software (Fruchter & Hook 2002) available in the stsdas.dither package combines multiple, offset images and corrects for geometric distortions. In order to implement
drizzle for ACS (and for other instruments too) a python wrapper for drizzle, “PyDrizzle,”
was written by Warren Hack. This interfaces to the core drizzle routine, and the two are
maintained separately (drizzle by Hook and Fruchter). PyDrizzle provides a versatile and
powerful but convenient interface to drizzle. It generally uses as its input an “association
table” which simply provides a set of images for processing.
The inital implementation of PyDrizzle in the pipeline aims to be robust and conservative, by means of default parameter settings. It provides no sub-sampling nor iteration
for cosmic ray rejection. It does however provide geometrical correction and photometric
correction as well as putting into place the infrastructure required for our upgrade path.
ACS Calibration Software
55
For the case of CR-SPLIT observations, cosmic ray rejection is achieved in the usual way
with acsrej . The initial implementation of PyDrizzle is described in detail in the Instrument
Science Report ACS 01-04 (Sparks, Hack & Hook 2001).
In stand-alone mode, access is provided to all parameters, so, for example, it is possible
to correct, combine, rotate, position and subsample data with a one-line command. Also,
in stand-alone mode it is possible to generate an association using data that were not
originally associated at the proposal writing stage. This is a very powerful enhancement
of analysis capability. The utility “BuildAsn,” available in pyraf and written as part of
the PyDrizzle support effort, constructs an association table from any appropriate data set.
Such a data set may be, for example, all archival observations of a particular galaxy taken
at different times, pointings and orientations, but using the same filter and substantially
overlapping. Positional misregistrations in image World Coordinate Systems are the norm
under such circumstances, arising primarily from uncertainties in the guide star catalog. It
is possible, however, to provide corrections in the association table, PyDrizzle will recognize
these offsets and adjust the positional information appropriately.
4.
Upgrade Path: MultiDrizzle
The only cosmic ray rejection currently provided is for the case where CR-SPLIT observations are acquired. However, the utilities available in the stsdas dither package offer a
much more powerful set of options, see Koekemoer et al. (2002a). A typical combination
of such utilities is available in the script “MultiDrizzle”, developed by Anton Koekemoer,
which uses capabilities that may be found in the stsdas dither package, but which interfaces
to PyDrizzle rather than drizzle directly (see Koekemoer 2002b). Specifically MultiDrizzle
performs the following steps:
• sky subtract
• drizzle individual images separately
• tweak registration information (1)
• median combine the individually drizzled images
• “Blot” the result back to the original data sets
• identify cosmic rays on individual images driz cr
• tweak registration information (2)
• drizzle the data again, combining into a single image
“Blot” is the reverse operation to drizzle and is available as part of PyDrizzle already. In
order to process data in this way, it is necessary to run calacs with the flag EXPSCORR set
to PERFORM. This then causes all individual images to be processed with calacs through
to the flat-fielding stage. Currently, if the data are CR-SPLIT (also for some other modes)
only the cosmic-ray-combined observations are processed. With EXPSCORR set, the inital
basic processing does not need to be repeated. This is a significant advantage for external
users who would otherwise need to download all relevant reference files as well as their data.
The ability of the back-end system to handle the increase in data volume is currently being
evaluated. If it is feasible, we will set this flag.
MultiDrizzle is available privately from Anton Koekemoer on a shared-risk best-effort
basis and more information about it can be obtained from the MultiDrizzle website:
http://www.stsci.edu/˜ koekemoe/multidrizzle.
56
Sparks
STScI personnel are also actively working on a trimmed down, robust version of the
script to provide much of the MultiDrizzle functionality. We anticipate offering a script
that does not attempt image registration initially, although as noted above, users may
provide such information themselves and insert it into the association table that drives the
processing.
5.
Support Utilities
A variety of support utilities have been requested for assisting with ACS analysis efforts. The
following list provides examples of requests that are in hand, and which will be implemented
as resources permit:
• velocity aberration correction to geometry
• coordinate conversion utilities, detector xy to sky, vice versa, raw xy to corrected xy,
vice versa.
• pixel area map
• exposure time image
Input on additional tools is welcome. Our current planning may be found in a new ACS
Instrument Science Report by Sparks, Hack, Hook and Koekemoer (2002), which includes
description of new tools to help with other forms of data such as ramp filter observations.
Note, however, that the software will be developed on a priority basis including maintenance
of basic existing capabilities, so this should be regarded as a “wish-list.” Finally, ST-ECF
has undertaken to provide software support for analysis of ACS grism data, using the “aXe”
package.
6.
Conclusions
Currently, ACS data are processed through the calibration pipeline, calacs, to provide basic image reductions, and then with PyDrizzle to provide geometric distortion correction
and image combination. In stand-alone mode, PyDrizzle offers a wide variety of additional
functionality, as compared to the default parameter settings adopted in the pipeline environment. A particularly powerful enhancement is the capacity to develop new “associations”
and process a group of overlapping images into a single, combined image. Ways are being explored to improve the cosmic ray rejection strategy for dithered data. As resources
permit, additional new utilities are being brought online to enhance the scientific utility of
ACS observations.
Acknowledgments. Warren Hack is the principal author of the ACS calibration and
analysis software tools. Richard Hook is the primary developer for drizzle. Their efforts are
fundamental and much appreciated.
References
Fruchter, A. S. & Hook, R. N. 2002, PASP, 114, 144
Hack, W. J. 1999a, “CALACS Operation and Implementation,” Instrument Science Report
ACS 99-03 (Baltimore: STScI)
Hack, W. J. 1999b, “CALACS reference files,” Instrument Science Report ACS 99-08 (Baltimore: STScI)
ACS Calibration Software
57
Hack, W. J. 2002, in ASP Conf. Ser., Vol. 281, Astronomical Data Analysis Software and
Systems XI, ed. D. A. Bohlender, D. Durand, & T. H. Handley (San Francisco:
ASP), 197
Hack, W. J., Busko, I., & Jedrzejewski, R. 2003, “New STScI Data Analysis Applications,”
ADASS XII Proceedings
Koekemoer, A. M., et al. 2002a, HST Dither Handbook , Version 2.0 (Baltimore: STScI)
Koekemoer, A. M., Fruchter, A. S., Hook, R., & Hack, W., 2002b, MultiDrizzle: An Integrated Pyraf Script for Registering, Cleaning and Combining Dithered Images, this
volume, 337
Mack, J., et al. 2002, in HST ACS Data Handbook, version 1.0, ed. B. Mobasher (Baltimore:
STScI)
Mutchler, M., Jedrzejewski, R., & Cox, C. 1999a, “ACS calibration pipeline testing: basic
image reduction,” Instrument Science Report ACS 99-04 (Baltimore: STScI)
Mutchler, M., Hack, W., Jedrzejewski, R., & Van Orsow, D. 1999b, “ACS calibration
pipeline testing: cosmic ray rejection,” Instrument Science Report ACS 99-09 (Baltimore: STScI)
Pavlovsky, C., et al. 2002 ACS Instrument Handook, version 3.0, (Baltimore: STScI)
Sparks, W. B., Jedrzejeski, R., Clampin, M., & Bohlin, R. C. 2000, “Software tools for ACS:
Geometrical Issues and Overall Software Tool Development,” Instrument Science
Report ACS 00-03 (Baltimore: STScI)
Sparks, W. B., Hack, W. J., & Hook, R. N. 2001, “Initial Implementation Strategy for
Drizzle with ACS,” Instrument Science Report ACS 01-04 (Baltimore:STScI)
Sparks, W. B., Hack, W. J., Hook, R. N., & Koekemoer, A. 2002, “ACS Software Tool
Development,” Instrument Science Report ACS 02-TBD (Baltimore: STScI)
Van Orsow, D., Mutchler, M., Hack, W., & Jedrzejewski, R., 1999, “ACS calibration
pipeline testing: error propagation,” Instrument Science Report ACS 99-06 (Baltimore: STScI)
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
The Effect of Velocity Aberration on ACS Image Processing
Colin Cox and Ron Gilliland
Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218
Abstract. The apparent scale change due to velocity aberration, although small,
has measurable effects on the wide field images of the ACS. Over a one orbit period
the scale can vary by as much as 5 parts in 100,000. Across a long diagonal of the
ACS field of view this amounts to about 0.3 pixels. This is sufficient to degrade the
registration needed for cosmic ray rejection. Images taken six months apart could
have scale differences as large as 12 parts in 100,000 leading to misregistrations up
to 1.4 pixels. We plan to add a velocity aberration scale correction factor to image
headers which may be used in the cosmic ray rejection algorithm and the dither
package.
1.
Introduction
The Hubble Space Telescope has an orbital speed about the Earth of about 7 km a second,
and the Earth’s orbital velocity about the Sun is approximately 30 km a second. The
net velocity causes stellar image displacements of some tens of arcseconds. In Figure 1
α represents the angle between the telescope direction of motion relative to the Sun and
the direction of a star in barycentric coordinates. α is the angle measured towards the
instantaneous apparent direction and is given by
tan α =
1 − β 2 sin α
cos α + β
Differentiating this expression gives
1 − β2
dα
=
dα
1 + β cos α
This gives the change of scale along the radial direction defined by the intersection of the
plane containing the velocity and pointing vectors with the field of view. In the tangential
direction the scale change is sin α / sin α which comes out to be exactly the same factor.
Hence the scale change is isotropic over the field of view.
When α is acute and cos α is positive, α is less than α and dα
dα is less than 1. The stars
apparently bunch together slightly, and more stars are viewed in a given pixel area. So the
plate scale increases by the reciprocal of dα
dα The magnitude of this effect is of order 1 part
in 104 and can vary by 5 parts in 105 within an orbit. Typically this can cause a difference
of order one pixel across the diagonal of the ACS Wide Field Camera, with its 4096 by 4096
pixel field of view. The figures show the pixel displacements between observations taken at
extremes of the Earth’s orbit and within a single HST orbit. A target is assumed to be
placed at the WFC reference point, which lies 200 pixels from the edge of WFC1.
58
The Effect of Velocity Aberration on ACS Image Processing
Figure 1.
59
Velocity aberration angular change.
PIXEL SHIFTS
Maximum variation due to Earth’s orbit
Maximum shift due to HST orbit
0
1024
2048
3072
4096
1024
2048
10
2048
0
0.
0.
3072
4096
40
WFC1
10
0.40
0.
0.30
0
2
0.
1024
0.0
WFC1
0.10
04
6
0.0
0.
8
0.0
1024
2048
2
0
2048
1024
WFC2
5
0.
0.
5
WFC2
1024
0.4
0.1
0
4
6
0.3
0.0
0.08
0.0
0.2
0
2048
0.
1
12
2
0.
0
0
0
Figure 2.
1024
2048
3072
4096
0
Pixel shifts caused by scale changes.
1024
2048
3072
4096
60
Cox & Gilliland
Model profile
Pixel counts
-2
-1
0
1
2
3
x
Figure 3.
2.
Pixel count errors caused by shifts.
Discussion
The measurements which brought attention to this effect occurred during a flat field monitoring program which made several observations of 47 Tucanae. Two observations 50 minutes apart indicated plate scale differences of 3.7×10−5. This was later found to be perfectly
consistent with the expected velocity aberration induced value.
The HST pointing and control system is based on guiding relative to nearby stars in the
field of view. These are displaced similarly to the target star which makes the system largely
auto-correcting. Nevertheless, small corrections are continuously made to keep the primary
target on a fixed pixel. Nothing can be done on board to correct for the scale variations
and indeed, parallel targets can easily move by several pixels during a long exposure.
The significance of this effect is relatively minor, especially when compared with the
other distortions present. However, the effect is larger than the residual error after correcting
for geometric distortion and it is easily allowed for by applying a scale factor to the images.
We intend to revisit our distortion solution and apply this velocity aberration correction to
each measurement image and obtain a new solution.
An analysis that is expected to be sensitive to small misregistrations is that of cosmic
ray rejection. In any region where the signal has a steep slope, such as in the tail of a bright
star, a displacement of one image with respect to another, even by a few tenths of a pixel,
can be seen as an amplitude difference between matching pixels and interpreted as a cosmic
ray hit. To avoid such false positives which cause us to discard good data, we have to set
a high threshold, thereby increasing the risk of false negatives, and missing genuine cosmic
ray events.
We intend to revise the cosmic ray rejection software to allow for misregistrations due to
this and other causes. A new keyword will be supplied in science image headers; namely the
factor by which the image should be corrected to a barycentric coordinate system. This will
be used in the PyDrizzle software which performs cosmic ray rejection as part of its image
combination. We might also see a slight improvement in the reconstructed stellar images
with the more accurate registration that now becomes possible throughout the image.
References
Aurière, M. 1982, A&A, 109, 301
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
Extreme Red Sensitivity of ACS/WFC
Ronald L. Gilliland and Adam G. Riess
Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218
Abstract. Establishing the instrumental sensitivity as a function of wavelength
within the domain covered by the F850LP (z) filter is critical for a number of ACS
science programs. Absolute sensitivity calibrations with hot white dwarfs showed
that substantial updates up to 20% in the blue were needed relative to pre-flight
estimates, and were suggestive of a significant gradient in the far-red (0.85–1.1 microns) QE update needed (but did not suffice to constrain this well). We report here
observations of several hot white dwarfs; as well as a M7 V, a L3.5 V and a T6.5 V
star in F625W, F814W and F850LP with ACS/WFC. The concentration of flux for
the coolest stars to the longest wavelengths probed with F850LP provide a means of
defining the quantum efficiency curve underlying this broad filter. We find, unique
to the z-band filter, that over the range O, M, L, T stars that the encircled energy
for small-to-moderate apertures can vary at the level of 10s of percent. Attempts
to separate out underlying QE variations and wavelength dependent PSF effects are
reported on here. For an L dwarf, similar in SED to a high redshift Type Ia SN,
these effects can reach half a magnitude for small apertures.
1.
Previous Sensitivity Update
Figure 1 shows the in-flight corrected throughput values for three filters with ACS/WFC
(F625W, F814W and F850LP) as well as the overall throughput (labeled as CLEAR) for all
non-filter components—with the underlying CCD Quantum Efficiency being the most limiting factor over this domain. The ratio of observed to predicted count rates for observations of
spectrophotometric standards (GD71, GRW+70D5824) using pre-launch throughput values
are shown in the lower panel for all the broadband filters. The generally smooth variation
with wavelength suggested a wavelength dependent correction was needed for the CCD QE
curve and the values shown in the upper panel incorporate such a change. QE values for
wavelengths between the pivot wavelengths of the filters were set by spline interpolation;
for lack of a better constraint updates to the QE curve used a linear extrapolation beyond
9100 Å (e.g., we assumed the pre-launch value drops by 10% by 10,400 Å). The spectrophotometric standards are hot white dwarfs, and thus have a flux distribution that tends to
weight sensitivity determinations to the short wavelength end of filters.
The remainder of this paper explores the validity of the existing throughput
curve over the F850LP—z-filter bandpass. Of particular interest is the throughput in the far red.
2.
Extreme Red Stars as Probe
Observations of very red stars with known spectrophotometry provide information on the
wavelength distribution of sensitivity within the broad F850LP filter. We chose three stars
with M, L and T spectral types for which flux-calibrated spectra are available from the
literature. These are detailed in Table 1. For the L-dwarf a STIS spectrum is used.
61
62
Gilliland and Riess
Figure 1. Left: Upper panel shows total system throughputs for CLEAR,
F625W, F814W and F850LP filters following the updates to original CCD
QE curves based on the initial observational results shown in the lower panel.
Right: Relative fluxes for the O, M, L and T stars arbitrarily normalized to 0.1 at
9000 Å, and the F850LP sensitivity. The wavelength dependent contribution to
counts changes dramatically for these stars as a function of wavelength.
Table 1.
Star
VB 8
2M0036+18
2M1237+65
Red Star Data
Type
M7
L3.5
T6.5
RA
16:55:34
00:36:16
12:37:39
DEC
V
−08:23.6 16.81
+18:21.1 21.34
+65:26.2 >28.0
Ic
12.28
16.11
21.51
z
11.73
15.21
19.62
Reference
Reid 2002
STIS(6797)
Burgasser et al. 2002
Two factors are most important in determining sensitivity to faint stellar sources:
1. Encircled energy arising from possible PSF shape differences with wavelength.
2. Establishing the correct underlying throughput as a function of wavelength.
3.
Encircled Energy Changes
We have determined encircled energies within 3 × 3, 5 × 5 and 9 × 9 pixel areas centered on
the stars. Normalization is relative to an “infinite” aperture (really a radius of 2. 8) sum,
and sky subtraction is based on the value within a surrounding annulus of 3.0–5.0 arcsec.
For these tables we include entries for ‘ETC’—the encircled energy assumed by our current
exposure time calculator, which is independent of input spectral type. The ‘Pivot lambda’
evaluates the photon flux weighted mean wavelength, i.e., the integral of QE × Fλ λ2 divided
by the integral of QE × Fλ λ, each over λ.
For F625W there is no evidence of encircled energy changes as a function of spectral
type. The drop in encircled energy at 3 × 3 pixels for the stellar sources relative to ETC
assumptions (based on modeled PSFs) is explained by the use of drizzled (geometrically
corrected) images for which the cores of PSFs experience minor smoothing. At the scale of
9×9 pixels the smoothing inherent with drizzling is unimportant and the ETC and observed
encircled energies agree well.
Extreme Red Sensitivity of ACS/WFC
Table 2.
F625W Encircled Energy
Star
ETC
O
M
L
Table 3.
63
3×3
0.671
0.591
0.590
0.582
5×5
0.834
0.791
0.792
0.782
9×9
0.880
0.882
0.886
0.877
Pivot λ
—
6249
6554
6654
5×5
0.797
0.756
0.727
0.708
0.699
9×9
0.880
0.874
0.860
0.850
0.861
Pivot λ
—
7949
8429
8597
8954
F814W Encircled Energy
Star
ETC
O
M
L
T
3×3
0.609
0.548
0.520
0.492
0.483
For F814W a progressive broadening of the PSF at the smallest aperture size is evident
and amounts to a loss of 12% in effective sensitivity from a very blue to a very red star. At
the larger 9 × 9 aperture size the differences are consistent with measurement error.
Table 4.
F850LP Encircled Energy
Star
ETC
O
M
L
T
3×3
0.588
0.481
0.430
0.425
0.355
5×5
0.757
0.679
0.632
0.614
0.526
9×9
0.880
0.829
0.800
0.783
0.692
Pivot λ
—
9020
9303
9425
9846
For the F850LP filter PSF changes on all scales are significant. In particular for the
3 × 3 aperture relevant for detection of sources near the faint limit, an L-dwarf (similar in
spectral energy distribution to a Type Ia SN with z ∼ 2) would have a throughput down
by 0.425/0.588 = 0.72 compared to the ETC PSF. Even at the larger 9 × 9 pixel aperture
the loss is 0.783/0.88 = 0.89 relative to expectations of the Exposure Time Calculator.
At the extreme-red wavelengths covered by the F850LP filter a significant
“red halo” effect has set in. Even at a radius of 1.0 arcsec the encircled energy
for the T dwarf is down by 5% relative to a blue star, perhaps more if the red
halo places flux beyond our 2.8 arcsec normalization aperture.
4.
Quantum Efficiency With λ
The second major component in setting the far-red sensitivity for the ACS/WFC is to
quantify the underlying CCD QE over the 8500 Å to 11000 Å domain. Recall (see Figure 1)
that a linear extrapolation based on observed values weighted toward shorter wavelengths
had been applied in this domain. With observations of extremely red stars we can now test
more directly for the wavelength dependence of QE.
64
Gilliland and Riess
We now take sums over large apertures, 2.8 radius in the observations, and use the
101 × 101 pixel sums from the ETC of similar size to provide expected count rates using
known spectra. As before sky is defined over a 3.0–5.0 arcsec annulus and subtracted.
The following table shows observed to expected count rate ratios in an “infinite” aperture using the current wavelength dependent sensitivities.
Table 5.
Ratio of Observed to Expected Counts—Infinite Aperture
Star
O
M
L
T
F625W
0.984
1.178
0.972
—–
F775W
0.982
1.108
0.926
1.5::
F814W
0.976
1.201
0.964
1.162
F850LP
0.942
1.125
0.859
0.971
F850LP/F814W
0.965
0.937
0.891
0.836
The ground-based M-dwarf flux is evidently ∼20% too low (a similar offset was found for
a ground-based L-dwarf spectrum—Reid et al. 2000—compared to STIS). We hypothesize
the need for linearly dropping the CCD QE over the 9500–11000 Å domain and solving for
the slope that yields near unity for all stars in the ratio F850LP/F814W (we effectively
use the F814W observations to normalize the zero point in the spectrophotometry). We
find that fixing the current value up to 9500 Å and applying a slope of −3.8 × 10−4 per Å
(i.e., sensitivity at 10000 Å is set at 81% of current value) results in F850LP/F814W ratios
of 0.995, 1.007, 0.974, 1.024 for the O, M, L, T stars respectively. The “solution” while
effective is not unique, as equally good (close to unity ratios throughout) results follow from
a range of starting λ and slopes for the linear drop of far-red QE.
5.
Summary
The far-red sensitivity needs to be adjusted for both a wavelength dependent encircled
energy (for which these observations determine well) and for a possible wavelength dependent term in the Quantum Efficiency (which these observations provide a good provisional
solution).
For a mid-L dwarf which serves as a good analogue for z ∼ 2 Type Ia Supernova
we find implied sensitivity losses of 28% due to encircled energy in a 3 × 3 pixel aperture
and 10% from an additional QE adjustment over the z-band. Combined, this equates to
a 0.47 magnitude loss in sensitivity for such an extremely red target compared to current
ETC/Synphot predictions. These results show that for the F850LP filter, count rate estimates need to take into account a PSF that is a strong function of the underlying spectral
energy distribution for the observed target. The CCD QE will be updated to reflect the
inferred drop beyond 9500 Å. The dependence of encircled energy with underlying spectrum
for F850LP will require compensation by the observer for ETC estimates and photometry.
Acknowledgments. We thank Ralph Bohlin, Guido DeMarchi, Neill Reid, Marco
Sirianni and Zlatan Tsvetanov for input at various stages.
References
Burgasser, A. J., et al. 2002, AJ, 123, 2744
Reid, I. N., et al. 2000, AJ, 119, 369
Reid, I. N. 2002, private communication
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
Calibration of Geometric Distortion in the ACS Detectors
G. R. Meurer
Department of Physics and Astronomy, The Johns Hopkins University, Baltimore,
MD 21218
D. Lindler
Sigma Space Corporation, Lanham, MD 20706
J. P. Blakeslee
Department of Physics and Astronomy, The Johns Hopkins University, Baltimore,
MD 21218
C. Cox
Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218
A. R. Martel, H. D. Tran
Department of Physics and Astronomy, The Johns Hopkins University, Baltimore,
MD 21218
R. J. Bouwens
UCO/Lick Observatory, University of California, Santa Cruz, CA 95064
H. C. Ford
Department of Physics and Astronomy, The Johns Hopkins University, Baltimore,
MD 21218
M. Clampin, G. F. Hartig
Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218
M. Sirianni
Department of Physics and Astronomy, The Johns Hopkins University, Baltimore,
MD 21218
G. de Marchi
Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218
Abstract. The off-axis location of the Advanced Camera for Surveys (ACS) is the
chief (but not sole) cause of strong geometric distortion in all detectors: the Wide
Field Camera (WFC), High Resolution Camera (HRC), and Solar Blind Camera
(SBC). Dithered observations of rich star cluster fields are used to calibrate the
distortion. We describe the observations obtained, the algorithms used to perform
the calibrations and the accuracy achieved.
65
66
Meurer, et al.
1.
Introduction
Images from the Hubble Space Telescope (HST ) Advanced Camera for Surveys (ACS) suffer
from strong geometric distortion: the square pixels of its detectors project to trapezoids of
varying area across the field of view. The tilted focal surface with respect to the chief ray is
the primary source of distortion of all three ACS detectors. In addition, the HST Optical
Telescope Assembly induces distortion as does the ACS M2 and IM2 mirrors (which are
designed to remove HST ’s spherical aberration). The SBC’s optics include a photo-cathode
and micro-channel plate which also induce distortion.
Here we describe our method of calibrating the geometric distortion using dithered
observations of star clusters. The distortion solutions we derived are given in the IDC tables
delivered in Nov 2002, and are currently implemented in the STScI CALACS pipeline. This
paper is a more up to date summary of our results than that presented at the workshop.
An expanded description of our procedure is given by Meurer (2002).
2.
Method
2.1.
Observations
The ACS SMOV geometric distortion campaign consisted of two HST observing programs:
9028 which targeted the core of 47 Tucanae (NGC 104) with the WFC and HRC, and 9027
which consisted of SBC observations of NGC 6681. Additional observations from programs
9011, 9018, 9019, 9024 and 9443 were used as additional sources of data, to check the results,
and to constrain the absolute pointing of the telescope.
The CCD exposures of 47 Tucanae were designed to detect stars on the main sequence
turn-off at mB = 17.5 in each frame. This allows for a high density of stars with relatively
short exposures. The F475W filter (Sloan g ) was used for the CCD observations so as to
minimize the number of saturated red giant branch stars in the field. For the HRC two
60s exposures were taken at each pointing, while for the WFC which has a larger time
overhead, only one such exposure was obtained per pointing. Simulated images made prior
to launch, as well as archival WFPC2 images from Gilliland et al. (2000) were used to check
that crowding would not be an issue. For calibrating the distortion in the SBC we used
exposures of NGC 6681 (300s–450s) which was chosen for the relatively high density of UV
emitters (hot horizontal branch stars). The pointing center was dithered around each star
field. For the WFC and HRC pointings, the dither pattern was designed so that the offsets
between all pairs of images adequately, and non-redundantly, samples all spatial scales from
about 5 pixels to 3/4 the detector size. For the SBC pointings, a more regular pattern of
offsets is used augmented by a series of 5 pixel offsets.
2.2.
Distortion Model
The heart of the distortion model relates pixel position (x, y) to sky position using a polynomial transformation (Hack & Cox, 2000) given by:
xc =
m
k am,n (x − xr )n (y − yr )m−n ,
m=0 n=0
yc =
m
k bm,n (x − xr )n (y − yr )m−n
(1)
m=0 n=0
Here k is the order of the fit, xr , yr is the reference pixel, taken to be the center of each
detector, or WFC chip, and xc , yc are undistorted image coordinates. The coefficients to
the fits, am,n and bm,n , are free parameters. For the WFC, an offset is applied to get the
two CCD chips on the same coordinate system:
X = xc + ∆x(chip#) ,
Y = yc + ∆y(chip#).
(2)
Calibration of Geometric Distortion in the ACS Detectors
Figure 1.
67
Non linear component to ACS distortion for WFC and SBC detectors.
∆x(chip#), ∆y(chip#) are 0,0 for WFC’s chip 1 (as indicated by the FITS CCDCHIP
keyword) and correspond to the separation between chips 1 and 2 for chip 2. The chip 2
offsets are free parameters in our fit. X , Y correspond to tangential plane positions in
arcseconds which we tie to the HST V 2, V 3 coordinate system. Next the positions are
corrected for velocity aberration: X = γX , Y = γY , where
γ=
1 + u · v/c
.
1 − (v/c)2
(3)
Here u is the unit vector towards the target and v is the velocity vector of the telescope
(heliocentric plus orbital). Neglect of the velocity aberration correction can result in misalignments on order of a pixel for WFC images taken six months apart for targets near
the ecliptic. Finally, we must transform all frames to the same coordinate grid on the sky
Xsky , Ysky :
Xsky = cos ∆θi X − sin ∆θi Y + ∆Xi ,
Ysky = sin ∆θi X + cos ∆θi Y + ∆Yi
(4)
where the free parameters ∆Xi , ∆Yi, ∆θi are the position and rotation offsets of frame i.
2.3.
Calibration Algorithm
We use the positions of stars observed multiple times in the dithered star fields to iteratively
solve for the free parameters in the distortion solution: fit coefficients am,n , bm,n ; chip 2
offsets ∆x(chip 2), ∆y(chip 2) (WFC only); frame offsets ∆Xi, ∆Yi , ∆θi; and tangential
plane position Xsky , Ysky of each star used in the fit. The stars are selected by finding local
maxima above a selected threshold. The centroid in a 7 × 7 box about the local maximum
is compared to Gaussian fits to the x, y profiles. If the two estimates of position differ by
more than 0.25 pixels, the measurement is rejected as likely being effected by a cosmic ray
hit or crowding. Further details of the fit algorithm can be found in Meurer et al. (2002).
2.4.
Low Order Terms
Originally only SMOV images taken with a single roll angle were used to define the distortion
solutions. The solution using only these data is degenerate in the zeroth (absolute pointing)
68
Meurer, et al.
Table 1.
Summary of fit results
Camera
chip
WFC
WFC
WFC
WFC
HRC
HRC
HRC
SBC
1
2
1
2
pixel
size
[arcsec]
0.05
0.05
0.05
0.05
0.025
0.025
0.025
0.03
Filter
F475W
F475W
F775W
F775W
F475W
F775W
F220W
F125LP
Pointings
25
25
10
10
20
13
12
34
N
142289
103453
31652
33834
77433
31515
14715
1561
rms(x)1 rms(y)1 Notes
[pixels] [pixels]
0.042
0.045
0.035
0.037
0.050
0.056
2
0.041
0.048
2
0.027
0.026
0.026
0.043
3
0.112
0.108
3
0.109
0.094
1
This is the rms after iteratively clipping measurements with deviations greater than 5 times the rms.
Coefficients held fixed to those found for WFC F475W.
3
Coefficients held fixed to those found for HRC F475W.
2
and linear terms (scale, skewness). So we used the largest commanded offsets with a given
guide star pair to set the linear terms. However, comparison of corrected coordinates to
astrometric positions showed that residual skewness in the solution remained. Hence, as of
November 2002, the IDC tables for WFC and SBC are based on data from multiple roll
angles. The overall plate scale is set by the largest commanded offset. For the HRC, the
linear scale is set by matching HRC and WFC coordinates, since the same field was used
in the SMOV observations. The zeroth order terms (position of the ACS apertures in the
HST V 2, V 3 frame) was determined from observations of an astrometric field.
3.
Results
The distortion in all ACS detectors is highly non-linear as illustrated in Figure 1. We find
that a quartic fit (k = 4) is adequate for characterizing the distortion to an accuracy much
better than our requirement of 0.2 pixels over the entire field of view. Table 1 summarizes
the rms of the fits to the various datasets.
The WFC and HRC fits were all to F475W data as noted above. To check the wavelength dependence of the distortion we used data obtained with F775W (WFC and HRC)
and F220W (HRC) from programs 9018 and 9019. We held the coefficients fixed and only
fit the offsets in order to check whether a single distortion solution is sufficient for each
detector. Table 2 shows that there is a marginal increase in the rms for the red data of the
WFC, little or no increase in the fit rms for the red HRC data, but a significant increase in
the rms using the UV data. An examination of the HRC F220W images reveals the most
likely cause: the stellar PSF is elongated by 0.1”. A similar elongation can also be seen
in SBC PSFs. We attribute this to aberration in the optics of either the ACS M1 or M2
mirrors or the HST OTA (Hartig et al. 2002). The aberration amounts to 0.1 waves at
1600 Å, but is negligible relative to optical wavelengths, hence it is not apparent in optical
HRC images. While it was expected that the same distortion solution would be applicable
to all filters except the polarizers, recent work (by Tom Brown, STScI, and our team) has
shown that at least one other optical filter (F814W) induces a significant plate scale change
(factor of ∼ 4 × 10−5 ). In the long term, the IDC tables will be selected by filter in the
STScI CALACS pipeline.
Calibration of Geometric Distortion in the ACS Detectors
69
Figure 2.
Binned residuals to quartic distortion fits for the WFC and HRC detectors. The large residuals in the HRC map at Xsky ≈ 5 , Ysky ≈ 10 correspond
to the Fastie Finger.
While a quartic solution is adequate for most purposes, binned residual maps (Figure 2)
show that there are significant coherent residuals in the WFC and HRC solutions. These
have amplitudes up to ∼ 0.1 pixels. The small-scale geometric distortion is the subject of
the Anderson & King contribution to this proceedings.
References
Hack, W. & Cox, C. 2000, Instrument Science Report ACS 2000-11 (Baltimore: STScI)
Hartig, G., et al. 2002, in Future EUV and UV Visible Space Astrophysics Missions and
Instrumentation, eds. J. C. Blades & O. H. Siegmund, Proc. SPIE, Vol. 4854, in
press [4854-30]
Gilliland, R. L., et al. 2000, ApJ, 545, L47
Meurer, G. R., et al. 2002, in Future EUV and UV Visible Space Astrophysics Missions
and Instrumentation, eds. J. C. Blades & O. H. Siegmund, Proc. SPIE, Vol. 4854, in
press [4854-30]
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
Drizzling Dithered ACS Images—A Demonstration
Max Mutchler, Anton Koekemoer, and Warren Hack
Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, Maryland
21218; mutchler@stsci.edu, koekemoe@stsci.edu, hack@stsci.edu
Abstract. Since the 1997 version of this poster, dithering and drizzling have evolved
from an advanced form of Hubble Space Telescope (HST ) observing and data reduction (with WFPC2), to the norm (with ACS). We demonstrate the reduction of a
typical dithered ACS dataset using the latest drizzling methods.
1.
Introduction
The drizzle task (Fruchter & Hook, 2002) is available in the IRAF/STSDAS dither
package. PyDrizzle (Hack & Jedrzejewski, 2002) is a PyRAF wrapper for drizzle, which
allowed drizzle to be incorporated into the ACS calibration pipeline. PyDrizzle combines associated HST data and corrects for geometric distortion, to produce an image
which is photometrically and astrometrically correct across the image’s entire field-of-view.
Multidrizzle (Koekemoer, 2002) encapsulates the processes of building association tables,
rejecting cosmic rays (even for singly-dithered observations), producing object catalogs,
refining shift measurements, and producing a drizzled combination of the input images.
We use a set of F814W (I-band) images of the “Tadpole” galaxy UGC 10214 (HST /ERO
program 8992, PI Holland Ford) to illustrate the use of these tools. This example can be
reproduced with the software and data available via the websites listed at the end of this
document.
2.
Pointing Patterns and Data Associations
The shifts for small-scale dither or large-scale mosaic pointing patterns can be specified
in a Phase II HST observing proposal using either POS TARG special requirements, or pattern parameter forms (Mutchler & Cox, 2001). When pattern forms are used, the entire
pointing pattern is automatically associated (except for the largest WFC mosaic patterns),
and the standard calibration pipeline is then able to process the dataset more completely.
However, this demonstration illustrates how any set of data, which may be either partially
or completely unassociated, can be associated post-facto, and reprocessed. Our sample
dataset employs a POS TARG dither to shift across the gap between the two WFC chips.
The ACS-WFC-DITHER-LINE pattern parameter form would produce the same results.
3.
Pipeline Processing: CALACS and PyDrizzle
Association tables are used to process datasets which are related, such as cosmic-ray split
(CR-SPLIT) exposures, dithered exposures, or data from different observing programs/epochs.
Our sample dataset includes CR-SPLIT associations, but we need to produce one table
which associates the entire pointing pattern. Download the following association tables
(*asn.fits), and the combined/cleaned (*crj.fits) images that PyDrizzle will use as
input:
70
Drizzling Dithered ACS Images
71
ftp://archive.stsci.edu/pub/ero/tadpole/j8cw54030_asn.fits
ftp://archive.stsci.edu/pub/ero/tadpole/j8cw54031_crj.fits
ftp://archive.stsci.edu/pub/ero/tadpole/j8cw54040_asn.fits
ftp://archive.stsci.edu/pub/ero/tadpole/j8cw54041_crj.fits
Merge the two tables, and edit to create a unique (numbered) MEMTYPE for the CR-SPLIT
pairs, and add a PROD-DTH row for the combined output, as follows:
cl> tmerge *asn.fits pipeline_asn.fits append
cl> tedit pipeline_asn.fits
# row
#
1
2
3
4
5
6
7
MEMNAME
MEMTYPE
MEMPRSNT
J8CW54P9Q
J8CW54PDQ
J8CW54031
J8CW54PPQ
J8CW54PTQ
J8CW54041
pipeline
EXP-CR1
EXP-CR1
PROD-CR1
EXP-CR2
EXP-CR2
PROD-CR2
PROD-DTH
yes
yes
yes
yes
yes
yes
no
Alternately, association tables can be built using buildasn (Hack & Jedrzejewski,
2002). This association table can be used to re-run both CALACS and PyDrizzle. We will
only run PyDrizzle here, but re-running CALACS may be necessary to recalibrate your
data, and/or to produce the flat-fielded (*flt.fits) images needed to run Multidrizzle.
Start PyRAF and define your reference file directory, where the distortion correction
table (IDCTAB) resides. Load the STSDAS dither package, run PyDrizzle using your new
association table, and display the output image:
> pyraf
--> set jref = ’/data/cdbs7/jref/’
--> stsdas
--> dither
--> pydrizzle pipeline_asn.fits bits=8578
--> display pipeline_drz.fits[sci,1] z1=0 z2=5
4.
Multidrizzle Processing
To download the latest full-featured version of Multidrizzle, contact Anton Koekemoer
(koekemoe@stsci.edu). A version will eventually be available via the IRAF/STSDAS
dither package, but as of December 2002, a scaled-down test version (with no tweakshifts
or singleCR) can be downloaded via the PyDrizzle webpage listed below. See detailed
instructions for running Multidrizzle on Anton’s webpage (also listed below).
As input, here we will use existing flat-fielded images (*flt.fits) which were created
by exposure lines with no CR-SPLIT special requirement, i.e., there is only one exposure at
each dither pointing. Download the following images to your working directory. There are
additional F814W exposures, but we will use only two here:
ftp://archive.stsci.edu/pub/ero/tadpole/j8cw54p8q_flt.fits
ftp://archive.stsci.edu/pub/ero/tadpole/j8cw54plq_flt.fits
Move to your working directory, make an input image list, and set up Multidrizzle.
Specify tweakshift1=yes to refine the shifts:
72
Mutchler, Koekemoer, and Hack
> cd /data/mymachine1/demo/
> ls *flt.fits > input.list
> source /data/wallaby1/multidrizzle/setup
> pyraf
--> pyexecute(’/data/wallaby1/multidrizzle/multidrizzle_iraf.py’,
tasknames=’multidrizzle’)
--> unlearn multidrizzle
--> multidrizzle output=’f814w’ filelist=’input.list’ tweakshift1=yes
Although we haven’t specified singleCR=yes above, the tweakshift1 step automatically runs the SExtractor version of it. This is a relatively crude rejection method, which
is used mainly to produce cleaner object catalogs for shift measurement, and the resulting
masks are not (by default) used in the final drizzle. Specifying singleCR include=yes
would include these masks in the final drizzle. Since we are using only two input images
here, this would be a way to reject cosmic rays in the gap overlap regions. While this may
produce a better result cosmetically, a more reliable result would be acheived by including
additional input frames.
Multidrizzle (using buildasn) creates the following association table, with the deltashifts and delta-rotations determined by tweakshift1 stored in additional columns:
#
Table f814w_twk1_asn.fits[1] Tue 20:02:49 05-Nov-2002
# row MEMNAME
#
1 j8cw54p8q
2 j8cw54plq
3 f814w_twk1
MEMTYPE
MEMPRSNT
XOFFSET
arcsec
EXP-DTH
EXP-DTH
PROD-DTH
yes
yes
no
0.
-0.064946
0.
YOFFSET
arcsec
ROTATION
degrees
0.
0.013180
0.
0.
0.0094
0.
Display the drizzled output: the individual frames, and the final drizzled science (sci)
and weight (wht) images:
-->
-->
-->
-->
display
display
display
display
j8cw54p8q_flt_single_sci.fits[0] 1 z1=0 z2=5
j8cw54plq_flt_single_sci.fits[0] 2 z1=0 z2=5
f814w_sci.fits[0] 3 z1=0 z2=5
f814w_wht.fits[0] 4 zr+ zs+
5.
Further Resources Available via the Web
The following web resources provide background on dithering and drizzling, and the sample
data which can be used to reproduce this demonstration, if desired:
ACS drizzling:
Andy Fruchter:
PyDrizzle:
Multidrizzle:
PyRAF:
SExtractor:
Dither Handbook:
ACS ERO data:
ACS ERO release:
www.stsci.edu/hst/acs/analysis/drizzle/
www.stsci.edu/~fruchter/dither/dither.html
stsdas.stsci.edu/pydrizzle/
www.stsci.edu/~koekemoe/multidrizzle/
pyraf.stsci.edu/
terapix.iap.fr/soft/sextractor/
www.stsci.edu/instruments/wfpc2/Wfpc2_driz/dither_handbook.html
archive.stsci.edu/hst/acsero.html
oposite.stsci.edu/pubinfo/pr/2002/11/pr-photos.html
Drizzling Dithered ACS Images
73
Figure 1. The final Multidrizzle output image of the Tadpole galaxy UGC 10214,
using only two of the available F814W exposures as input.
A “draft version” of this document was available as a handout during the workshop.
An expanded “demo version” is available via the ACS drizzling webpage (listed above). It
includes some supplemental Appendices which were excluded here due to page limitations.
This includes information on pointing patterns, using CALACS to generate the flat-fielded
images (*flt.fits) needed as input for Multidrizzle, and detailed drizzling parameters.
References
Fruchter, A. & Hook, R., 2002, Drizzle: a Method for the Linear Reconstruction of Undersampled Images, PASP, 114, 144
Hack, W. & Jedrzejewski, R., 2002, PyDrizzle User’s Manual (Baltimore: STScI)
Koekemoer, A. 2002, The Dither Handbook (Baltimore: STScI)
Koekemoer, A., 2002, this volume
Mutchler, M. & Cox, C., 2001, “ACS Dither and Mosaic Pointing Patterns,” Instrument
Science Report ACS 2001-07 (Baltimore: STScI)
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
Flat-fielding of ACS WFC Grism Data
N. Pirzkal, A. Pasquali and J. Walsh
ST-ECF, Karl-Schwarschild Strasse-2, Garching bei München, D-85748, Germany
Abstract. In direct imaging, broad band flat-fields can easily be applied to correct deviation from uniform sensitivity across the detector field. However for slitless
spectroscopy data the flat field is both field and wavelength dependent. The effect
of the wavelength dependent flat field for the slitless G800L grism mode of the ACS
Wide Field Channel (WFC) has been investigated from observations of a flux calibrator at different positions in the field. The results of various flat-fielding schemes
are presented including application of a flat field cube derived from in-orbit broad
band filter flat-fields. Excellent results are reported with deviations in the extracted
spectra 2% across the WFC field.
1.
Available WFC Flat-fields
New ACS in-orbit filter flat-fields, were recently constructed using globular cluster observations (Mack et al. 2002). These flat-fields show that there is a significant amount of large
scale structure which varies from one broad band filter to another (i.e., with wavelength).
Flat-fielding ACS G800L grism data is therefore required in order to be able to derive a
unique G800L sensitivity curve which is the same at all positions on the detector. We have
used these flat-fields and have fitted them using a 3rd order polynomial as a function of
wavelength and at each pixel. The result of this fit was the creation of a data cube enabling
aXe, the ST-ECF Slitless Extraction Software (Pirzkal et al. 2001), to compute the proper
flat-fielding coefficient at any pixel position and wavelength. Unfortunately, the in-orbit LP
flats do not offer an ideal sampling of the wavelength dependence of the flat-field because:
1) they are broad; 2) only a limited number of filters are available; and, 3) there is no filter
which reaches wavelengths beyond 8500 Å (while the G800L grism mode remains sensitive
to about 10,000 Å).
2.
Observations
Two white dwarfs have been observed using the WFC G800L grism mode: GD153 during the
ACS SMOV campaign and G191B2B during the Cycle 11 Interim Calibration. GD153 was
observed at five different positions which were read out using ACS subarrays. Unfortunately,
the target position did not always end up at the center of the subarray and this resulted in
only two useful positions where the first order spectra were completely within the subarray.
The spectra at the other positions were truncated and did not contain all the flux from
GD153 at each wavelength. G191B2B, was observed using the same technique but at
9 different positions, 5 in the top part of the detector (CHIP1) and 4 in the bottom part
(CHIP2). The G191B2B spectra were all completely within each aperture and each position
was observed twice for a total of 18 spectra across the detector. Figure 1 shows the positions
and names associated with each of the G191B2B observations.
74
Flat-fielding of ACS WFC Grism Data
Figure 1. Mosaic of the nine G800L subarray observations of the white dwarf
G191B2B. The content of this image was bias and dark subtracted and gain corrected. The first order spectrum is immediately above the labels in this figure.
The second and third orders are to the right of the first order, while the zeroth
and negative orders are to the left of the labels.
Figure 2. The fractional error in the extracted, calibrated, but not flat-fielded
spectra of G191B2B is shown. In the region where the grism is most sensitive
(6000 Å to 9200 Å) the inconsistencies in flux levels can be seen to progressively
increase from about ±1.5% to as much as ±3% with, more importantly, some
large (5%) systematic differences between different parts of the detectors. Note
that observations taken at the same positions are plotted using the same line type
and are at a closely similar flux error, demonstrating the high level of repeatability
one can expect from the G800L grism mode. The increasing errors below 5500 Å
and beyond 10000 Å correspond to wavelength ranges where the sensitivity of the
grism approaches zero. The peak around 6500 Å corresponds to the Hα absorption
feature of G191B2B. The larger errors at the position of W7 at wavelengths below
6500 Å is caused by a less well known wavelength calibration in that part of the
detector.
75
76
Pirzkal, et al.
Figure 3. The fractional error in the extracted, calibrated, flat-fielded spectra
of G191B2B (using a flat-field data cube derived directly from the latest in-orbit
flats), to be compared with Figure 1. The effect of simply applying broad band
filter flat-fields clearly introduces some rather large-scale differences across the detectors. These differences are introduced by the direct imaging flats which correct
for the field variation of the pixel effective size, which is related to the geometric
distortion of the WFC and which affects spectroscopic data differently from direct
imaging data.
3.
3.1.
aXe Extraction
No Flat-fielding
A first extraction was performed using aXe and no flat-fielding. The extracted spectra of
G191B2B were otherwise fully calibrated in physical flux units using the latest estimate
of the G800L sensitivity curve. Figure 2 shows the fractional flux difference between the
aXe extracted and calibrated spectra and a template spectrum of G191B2B. The spectra
observed at the same positions agree with the template spectrum very closely even though
they were obtained at different times. There is however an increasingly large observed
discrepancy in the measured fluxes at wavelengths beyond 7500 Å. This effect is both field
and wavelength dependent.
3.2.
Direct LP-flat Data Cube
A second extraction of the G191B2B spectra was performed using the data cube fit of the
new in-orbit flats mentioned above. The result of which is shown in Figure 3. A significant
difference is apparent between different positions on the detector. There are several reasons
why this should be expected. First, it is likely that the large scale flat (L-flat) characteristic
of the G800L grism is different than that of the broad band filter flats used to generate
the flat-field data cube. A large scale L-flat correction to these broad band filters is hence
understandable. Second, the broad band flat-fields are designed so that a direct image of
the sky will look flat, even though the effective pixel size of the WFC varies significantly
from one corner of the detector to another. Applying such a flat-field introduces a correction
which is related to the geometric distortion and which should be corrected for differently
in spectroscopic data. The effect of distortion and tilt between the grism assembly and the
detector is accounted for by a field dependence of the dispersion properties of the G800L
grism (Pasquali et al. 2003).
Flat-fielding of ACS WFC Grism Data
77
Figure 4.
The fractional error in the extracted, calibrated, flat-fielded spectra of
G191B2B (using our G800L-corrected in-orbit LP flats-field cube). At wavelengths
smaller than 8500 Å where a wavelength dependent flat-fielding was applied, the
error in measured flux is within 1% and flat-fielding has removed the systematic
differences between different positions on the detector as well as the wavelength
dependence visible in Figure 2. Using this flat-fielding scheme, a unique G800L
sensitivity curve can be computed and applied to all the extracted spectra.
3.3.
Corrected LP-flat Data Cube
The variation in observed fluxes between different positions on the detector was successfully
fitted to a quadratic surface. This relation, essentially an G800L empirical L-flat correction,
was used to generate a new, G800L-corrected flat-field data cube. Extracted G191B2B
spectra using this new cube are shown in Figure 4. Note that this flat-field cube only
allows for a wavelength variation of the flat field at wavelengths ranging from 4350 Å to
8500 Å. Beyond this range, no gain in accuracy is expected as a constant flat-field coefficient
was used. This new corrected flat-field cube produces spectra which reach flux calibration
accuracies close to the 1% level across a wide range of wavelengths, and across most of the
detector. The same G800L-corrected flat-field cube was used to extract the two un-truncated
SMOV observations of GD153, described earlier, and similar accuracy was achieved between
the extracted and flux calibrated spectra and the template spectrum of GD153.
4.
Conclusion
We have successfully constructed a G800L flat-field data cube which allows one to reach
flux calibration accuracies of about 1% at wavelengths ranging from 6000 Å to 9000 Å and
across most of the WFC field of view. This modified G800L-corrected flat-field cube can
be used directly by the extraction software aXe to extract un-drizzled, non-geometrically
corrected grism observations. It will be made available from the ST-ECF ACS spectroscopy
group at http://www.stecf.org/instrument/acs/.
References
Pasquali, A., Pirzkal, A., & Walsh, A. 2003, this volume, 38
Pirzkal, N., Pasquali, A., & Demleitner, M. 2001, ST-ECF Newsletter 29, p. 5
Mack, J., Bohlin, R., Gilliland, R., et al. 2002, Instrument Science Report ACS 2002-008
(Baltimore: STScI)
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
Statistical Analysis of ACS Data without Covariance in Errors
Kavan U. Ratnatunga
Carnegie Mellon University, Pittsburgh, PA 15213
Abstract. Careful statistical analysis is required to extract Cosmic Shear from
ACS data. Rebinning the data to remove ACS image distortion creates images with
adjacent pixels which have observational errors that are correlated. Proper statistical
analysis of the image is complicated. It is however possible in principal to avoid any
rebinning of the ACS data in the analysis of small galaxy images.
1.
Introduction
In comparison to WFPC2 the distortion across the ACS image is 5 times larger. The pixel
scale varies significantly so that full ACS images which are shifted by more than about an
arcsecond cannot be stacked without correction for plate distortion. The relative image
shift at CCD center translates to a different number of pixels close to the edge of the CCD.
Most ACS dither patterns use a larger shift to close the gap between the CCD arrays.
The on-the-fly calibration (OTFC), designed to process full ACS images, rebins the data
before stacking to identify cosmic rays. An iterative process is needed since rebinning before
cosmic ray rejection bleeds the cosmic rays. The observational errors between adjacent pixels
in resulting images have covariance complicating maximum likelihood or χ2 image fitting
which assumes independent errors in image pixels.
However, the typical faint galaxy images being surveyed with ACS in random pure
parallel fields have half-light radius typically under a half arcsecond and the region used for
image analysis is only about 6 square. Across this small region, the differential change in
pixel scale from a region even 5 away in both directions, is small. In this case, at the edge
of the 6 square region, the differential shift is about 0.2 pixels, with differential rotation
of the two regions under 0.01 degrees (see Figure 2). For most of the higher signal-to-noise
image pixels with greater weight in the analysis within the inner half arcsecond diameter
region close to the center peaked galaxy image, it is more than a factor of 10 smaller. The
changes in pixel scale and orientation within the region around the small galaxy image can
therefore be ignored.
So to better than the accuracy used in the MDS WFPC2 analysis (cf. Figure 1), small
individual galaxy images or stars from shifted ACS fields can be stacked and cosmic ray rejected without rebinning to correct for image distortion, even though the whole CCD image
cannot be similarly stacked. Only the appropriate position dependent shifts corresponding
to the shifts between images at the target reference pixel need to be computed using the
adopted ACS distortion map. The small differential change in pixel scale is negligible for
cosmic-ray rejection, which can now be redone without any image rebinning.
2.
MDS-ACS Pipeline
The MDS-WFPC2 pipeline (Ratnatunga et al. 1999) software being modified for ACS will
follow the following procedure.
1. Correct bias
78
ACS Image Covariance
79
Figure 1.
Total offset in X and Y and rotation at the corner of the 6 region
used for image analysis relative to the center of each WFPC2 CCD array for a 5
shift in pointing in both X and Y . Most shifts used for MDS WFPC2 were a lot
smaller than the extreme 7 used in this illustration.
2. Subtract dark current
3. Flag hot pixels
4. Flat field multiply by inverse sky flat and pixel area function, i.e., flux as expected
for pixel-sky background less in smaller pixels.
5. Derive a rms error image along with calibrated image incorporating all sources of
observational error.
Details of the calibrations adopted by MDS are discussed in Ratnatunga et al. (1995).
2.1.
Shifts between Images
Initially, we adopt the IDT 4th order polynomial representation of ACS distortion mapping.
This mapping is expected to line up images to about 0.1 pixels, and will be tested.
Shifts between images are determined by cross-correlation of the central region of the
images. With ACS, if a larger region is needed to get a well defined peak, a rebinned ACS
image would need to be used.
The image closest to the mean pointing is selected as the primary image to define the
coordinate frame for the pure parallel fields in the sky.
2.2.
Cosmic Ray Rejection
A minimum of 3 ACS exposures will be used in the stack. If images are all in a single
line-of-sight, we stack calibrated not-rebinned ACS images. If not, we shift rebin the ACS
images to a meta grid array of the primary image before stacking. This stacking is used
to identify pixels hit by a cosmic ray in all individual images. The regions identified in
each image as hit by a cosmic ray are used to flag pixels in the calibrated not-rebinned
80
Ratnatunga
image. These ACS images are then shifted again and rebinned to a meta grid array without
bleeding the pixels hit by cosmic rays and then stacked.
2.3.
Object Identification
The cosmic ray clean stacked image is searched to identify all faint galaxy and stellar sources.
The identified sources on this rebinned stack are used to identify a one-sigma contour around
each object in the calibrated not-rebinned primary image. Using the relative shifts between
the images and the ACS distortion map the centroid of the object can be located on each
of the calibrated not-rebinned images.
We then compute the following using the adopted ACS distortion map at the location
of the centroid of individual galaxy images.
• Pixel scale in X and Y directions
• Metapixel coordinates in arc-seconds from ACS reference pixel
• Orientation of X and Y with respect to PA V3
Using equatorial coordinates of the ACS reference pixel and PA V3 of pointing the estimated
orientation of the galaxy in observed pixel coordinates can be converted directly to the
standard frame of reference in the sky.
2.4.
Image Analysis
A fixed size pixel array is cut out of each calibrated not-rebinned image for analysis, using
the adopted ACS distortion map and the relative shifts between the images. Pre-computing
the sub-pixel shifts between these individual images, the image analysis can now be done
without stacking all of the exposures to a single image.
Assuming that the distortion is known better than the individual centers could be
independently evaluated, it is better to use all of the images in the field to define the relative
image shifts in position and orientation of the pointing, than let it be defined by only the
images of galaxy being analyzed. As the relative shifts between the frames are predefined,
it is not necessary to have extra parameters to define the centroid independently in each
image of the galaxy.
The likelihood function is integrated over all of the observed ACS images with their
individual error images in observed not-rebinned data space. The MDS analysis software
creates a sub-pixelated centered galaxy model image, then after convolution shifts the image
to the required pixel centroid. One could even convolve the galaxy model image independently based on the mean HST focus and pointing jitter during each exposure. The larger
effect of jitter on parallel observations caused by differential aberration is discussed by Ratnatunga et al. (1997). All of the images required for integration of χ2 over the pixels can
be generated by shifting and block-binning the subpixelated image. The detector distortion
caused by the non-orthogonal axes of ACS is included in this rebinning of the model image.
The same subpixelated convolved galaxy model image can be used if we ignore changes
in focus and pointing jitter between exposures, and as long as there is no significant differential rotation.
By this approach we can overcome all of the limitations and possible errors of using the
nearest integer shift in addition to the differences in telescope jitter and breathing of focus
(consequently PSF) between exposures. This approach was discussed by Ratnatunga et al.
(1994) in the Image Restoration of HST Images workshop ten years ago. It was not used
by MDS for WF/PC and WFPC2 data to avoid the increase in computation time. The
current generation of computers, which are 100 times faster than when the original MDS
software was developed, may now make this approach practical.
However, the rms cumulative error caused by both the uncertainty in image ACS
distortion maps, changes in telescope focus due to breathing and/or jitter in pointing, and
ACS Image Covariance
81
Figure 2.
Differential offset in X and Y and rotation at the corner the of 6
region used for image analysis relative to the center of each galaxy image for a 5
shift in pointing in both X and Y . The vertical scale is the same as in Figure 1.
Position dependent shifts for each image will be used for ACS stacking and analysis.
differential pixel shifts over the image needs to be estimated. If that is larger than 0.3 pixels
(15 mas), then to the known accuracy stacking cosmic ray rejected images using nearest
integer shifts as done for MDS-WFPC2 may be the safest approach, including the additional
15 mas convolution in the image analysis.
3.
Conclusion
Calibrated but not rebinned ACS images can be used directly for analysis of small galaxy
or stellar images, which then avoids using images with covariance. The same approach may
also be useful for grism images.
Acknowledgments. I am grateful to Stefano Casertano who has always been a great
help in discussing statistical issues of proper image analysis. This paper is based on observations with the NASA/ESA Hubble Space Telescope, obtained at the Space Telescope
Science Institute. The HST research is funded by STScI grant GO9488.
References
Ratnatunga, K. U., Griffiths, R. E., & Casertano, S. 1994, in Proc. The Restoration of HST
Images and Spectra II , eds. R. J. Hanish and R. L. White, p. 333
Ratnatunga, K. U., Griffiths, R. E., Neuschaefer, L. W., & Ostrander, E. J. 1995, in Proc.
HST Calibration Workshop II , eds. A. Koratkar, A. & C. Leitherer, p. 351
Ratnatunga, K. U., Ostrander, E. J., & Griffiths, R. E. 1997, in Proc. The 1997 HST
Calibration Workshop, eds. S. Casertano, et al., p. 361
Ratnatunga, K. U., Griffiths, R. E., & Ostrander, E. J. 1999, AJ, 118, 86
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
Bias Subtraction and Correction of ACS/WFC Frames
M. Sirianni, A. R. Martel, M. J. Jee
Department of Physics & Astronomy, The Johns Hopkins University, Baltimore,
MD 21218
D. Van Orsow and W. B. Sparks
Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218
Abstract. Calibrated ACS/WFC science frames processed through the CALACS
pipeline exhibit a residual offset in their absolute levels at the edge separating the
Amp A-B and Amp C-D quadrants. This effect can be attributed primarily to
uncertainties in the bias level subtraction. We present an analysis of the overscan
levels and of the amplitude and behavior of the residual offsets for a large sample of
bias frames. The scientific impact of this residual is discussed.
1.
Introduction
One of the most fundamental steps in exploiting the new capabilities of the Advanced
Camera for Surveys (ACS) consists of accurately subtracting bias frames and offsets level
from scientific images. In the routine pipeline processing (CALACS; Hack 1999), the bias
level measured in selected columns of the leading physical overscan is subtracted from the
active area, which always shows a slightly higher bias level. In principle, if the residual offset
between the imaging area and the overscan region were always of the same amplitude, a
full-frame superbias subtraction would remove any residual difference. Unfortunately, the
offset is not constant but shows random variations, resulting in a small but noticeable jump
in the middle of each chip along the amplifier edges (see Figure 1).
We present an analysis of the structure of the WFC bias frames, statistics of a sample
of bias frames acquired since ACS installation and the amplitude of their residual offset.
We look into the scientific implications of such a residual and present a possible approach
that could mitigate the problem.
2.
Description and Behavior of the Problem
Normal WFC images are read-out using the four amplifiers, two for each detector (A-B for
WFC1 and C-D for WFC2). Hence, one amplifier is used to read a single 2k × 2k quadrant
and so each quadrant needs to be treated independently in the calibration process.
Each amplifier has a specific bias level and bias structure. It is therefore normal that
the raw bias frame (and any other raw WFC frame) shows a “natural” jump in the center
of the image. Figure 2 shows an averaged horizontal profile of the WFC1 bias frame used
to calibrate the image in Figure 1 (see also Sirianni, Martel, & Hartig 2001). The plot
shows that the bias level measured in the physical overscan is lower than the bias level in
the active area. The difference between these two bias levels varies between each quadrant.
Since the bias level is measured in the leading physical overscan and subtracted from the
active area of each quadrant, the center of the resulting image shows a “residual offset”
or jump between the two adjacent quadrants, A-B or C-D (central panel in Figure 2). In
82
Bias Subtraction and Correction of ACS/WFC Frames
Figure 1.
83
A calibrated WFC1 field (F775W) showing the offset.
calibrated ACS/GTO data, the amplitude of this residual offset is typically of the order of
1% of the adjacent background level (typically < 4 DN), with extreme cases of ∼ 3%. In
principle, if the difference between the active area and the overscan regions were of the same
amplitude in all frames, the full-frame superbias subtraction performed at a later CALACS
stage would remove the residual difference between the two quadrants.
To establish the temporal dependence of the bias levels and the residual offsets, we
analyzed the stability of all the bias frames acquired as part of SMOV and monitoring
campaigns over the period Apr–Jun 2002. The results are shown in Figure 3. The top
panels show the variation of the bias overscan level with time. For all four amplifiers, the
levels are fairly stable with small variations. Each month the TECs are turned off to allow
the CCDs to warm up to ∼ 19 C and anneal hot pixels created by radiation damage. When
this occurs the bias level is higher than normal, but within a few hours it will assume the
normal value it had just before the anneal. Ignoring this monthly feature, the overscan level
for all four WFC quadrants turns out to slowly decrease with time by about 0.02–0.03 DN
per day. Finally, in addition to the monthly feature, Amp B shows an intrinsic scatter of
about 15 DN (peak to peak). Such behavior was already known from ground testing and
does not represent a concern because the bias level in the active area changes accordingly
with the same amplitude.
The two panels at the bottom of Figure 3 show the residual offset variation with time
for the four amplifiers. Each amplifier shows a positive residual offset, meaning that the
active area has a slightly higher bias level than the region used to estimate the bias offset
in the physical overscan. For all quadrants, the residual offset does not correlate with time
and varies by a few tenths of a count (Table 1). Moreover, its apparently random amplitude
does not appear to correlate with the bias level in the active area nor with the detector
temperature. On the other hand, there seems to be a correlation between the residual offset
variations in each amplifier. When the difference between the active area and the leading
physical overscan increases or decreases in one quadrant, it shows the same behavior in the
other amplifiers.
The cause of the residual offsets and their variation with time are still under investigation. The accuracy of the bias level subtraction in a single frame is limited by this random
effect.
Table 1.
Residual Offset Instability
WFC1
AMP MEAN
σ
A
1.16
0.27
B
2.13
0.23
AMP
C
D
WFC2
MEAN
3.13
0.63
σ
0.16
0.26
84
Sirianni, et al.
Figure 2.
Horizontal profile of the bias frame used in the image of Figure 1.
Figure 3.
Variation with time of the bias overscan levels and the residual offsets.
Bias Subtraction and Correction of ACS/WFC Frames
3.
85
Scientific Impact
We investigated the scientific impact of the uncertainty in the bias level subtraction. For
point sources, the local background is typically subtracted in an annular region so the
residual offset has essentially no impact on the integrated magnitudes. Extended objects
such as galaxies may be spread over two or more quadrants, so their surface brightness
profiles will suffer from the residual offset. But even at low counts near the sky levels, a
3% jump translates to a change in magnitude of only 0.03. With the addition of the object
counts superimposed on the sky, this value is correspondingly smaller. We conclude that
the scientific impact of the variation of the amplifier residual offset is negligible.
4.
Conclusion
All ACS/WFC images will suffer from small uncertainties in bias level subtraction due to
the random variation in the difference between the bias level in the leading physical overscan
and the bias level in the active area. Due to the random nature of this variation, the error
associated with reference frames, such as the superbias and the superdark, will be reduced
by the square root of the number of bias and dark images used to build the reference files.
Update on Amp B fluctuations: We recently discovered a different problem, associated with
the bias level in Amp B, that can produce a final image where the jump at the amplifier
edges is noticeably larger than the one due to the residual offset instability. Some calibrated
images show an amplifier edge jump in WFC1 of up to 0.8 DN (A-B). If the final image
is a combination of n frames (CR-SPLIT or dithered observations) the bias jump at the
center of the WFC1 frame is n × 0.8. As noted above, the bias level of the B quadrant
shows an instability peak-to-peak of ∼ 15 DN. Such an instability is also visible in the bias
level of science frames. However, the distribution of the bias level in frames with a non-zero
exposure time seems to be bimodal with a “high” and “low” status which best match the
high and low ends of the “15-DN” range. When the bias level in B is in the “high” status,
the jump between A and B is only due to the residual offset instability and can be neglected.
However, when the bias level in B is “low,” then in addition to the amplifier residual offset,
there is a contribution of ∼ 0.8 DN from the bias frame subtraction.
More studies are in progress to better characterize this new problem and find a solution
or develop a correction to apply directly into the calibration pipeline. At the moment we
suggest that ACS users fit the sky level in each quadrant separately.
References
Sirianni, M., Martel, A. R., & Hartig, G. 2001, WFC4 Overscan Analysis and Bias Subtraction, The ACS Calibration Web Pages—Results (Johns Hopkins University)
Hack, W. 1999, “CALACS Operation and Implementation,” Instrument Science Report
ACS 99-03 (Baltimore: STScI)
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
On-Orbit Performance of the ACS Solar Blind Channel
Hien D. Tran, Gerhardt Meurer, Holland C. Ford, Andre Martel, Marco Sirianni
Dept. of Physics & Astronomy, The Johns Hopkins University, Baltimore, MD
21218
Ralph Bohlin, Mark Clampin, Colin Cox, Guido De Marchi, George Hartig
Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218
Randy Kimble
Goddard Space Flight Center, Code 681, Greenbelt, MD 20771
Vic Argabright
Ball Aerospace, Boulder, CO 80301
Abstract. The ACS solar blind channel (SBC) is a photon-counting MAMA detector capable of producing two-dimensional imaging in the UV at wavelengths 1150–
1700 Å, with a field of view (FOV) of 31 ×35 . We describe the on-orbit performance
of the ACS/SBC from an analysis of data obtained from the service mission observatory verification (SMOV) programs. These data show that the detector is behaving
nominally. Images of stars with the SBC reveal an aberration in the optics similar
to that observed in the HRC at UV wavelengths.
1.
Introduction
The Solar Blind Channel (SBC) of the Advanced Camera for Survey (ACS), installed in
March 2002 on the Hubble Space Telescope (HST ) during Service Mission 3B (SM3B) is
a spare detector from the Space Telescope Imaging Spectrograph (STIS) program. It is a
Multi-Anode Microchannel Array (MAMA) photon-counting device that uses a CsI photocathode on a curved microchannel plate (MCP). To improve the quantum efficiency, the
MAMA detector has been equipped with a field electrode, or repeller wire, that repels electrons emitted away from the MCP back into the channel. Optimized for UV imaging at
wavelengths 1150 Å to 1700 Å, the SBC has an imaging area of 1024 × 1024 pixels with
a sampling of 0.030 per pixel, yielding a total field of view of 31 × 35 . Besides a set of
long-pass filters (F115LP, F125LP, F140LP, F150LP, F165LP), and a Lyα (F122M) filter
for direct imaging, it is also equipped with two prisms for low-resolution (R ∼ 100) objective prism spectroscopy. The detector is cosmetically fairly clean, the only defects being a
broken anode and three small clusters of hot pixels. As with the STIS MAMA detector, the
ACS/SBC has bright-object limits to protect it from radiation damage. For non-variable
sources, the local (per pixel) count rate cannot be over 50 counts/sec/pixel, and the global
(over the whole detector) limit is < 2 × 105 counts/s. The optical performance of the SBC
is comparable to that of the STIS FUV-MAMA. The ACS/SBC is expected to give slightly
higher quantum efficiency but lower S/N due to higher dark current than the STIS FUVMAMA. Since launch, a number of tests for performance characterization of ACS have been
carried out as part of the service mission observatory verification (SMOV) program. We
briefly report here the results of some of these tests for the SBC. Some discussion of the
86
On-Orbit Performance of the ACS Solar Blind Channel
87
ACS imaging quality is also given in Hartig et al. (2002), and the geometric distortion in
the ACS detectors is described by Meurer et al. (2002).
2.
SMOV Results
Observations of the globular star cluster NGC 6681 were used to characterize many of the
SBC properties, including flux calibration, geometric distortion, low-order flat creation,
PSF, and monitoring of contamination.
2.1.
Detector Health
An important SBC diagnostic is the so-called “fold analysis.” Individual photon events
generate charge clouds which impinge on the position-sensing anode array. The number of
anode lines that collect a charge signal is the “fold number” for the event. The distribution
of fold numbers measures the gain distribution for the MCP. A shift to lower fold numbers
would imply gain sag in the MCP, perhaps due to excessive accumulated illumination, while
a shift towards larger fold number pulses could indicate leakage of gas into or production
of gas within the sealed detector tube. Comparison of event distributions observed during
ground testing of the tube in 1997 and more recently in-flight shows that there is little
change in the distributions over a period of ∼ four years. This indicates that the detector
is in fairly good health.
2.2.
Dark Currents
The first in-flight observations for a “super dark” image were taken over a ten hour span,
near the maximum time between South Atlantic Anomaly passages. The mean dark rate in
this image is 1.049e−5 counts/sec/pixel = 0.0378 counts/hour/pixel. This fairly low rate
is comparable to those measured during the thermal vacuum ground calibration campaign
of July 2001 (Martel et al. 2001), and is largely due to a relatively cool tube temperature
TSBC = 15◦ to 27◦ C during the observations. Ground thermal modeling indicates that a
thermal balance at TSBC = 35◦ to 37◦ C may occur, in which case the dark rate will be
about three times higher.
2.3.
Image Quality
Figure 1 illustrates the PSF structure in SBC images. The right-hand panel shows an
encircled energy (EE) curve and a radial profile of a star in the first light data. The solid
curves show the profiles at nominal x-scale, whereas the scale has been stretched by a factor
of ten for the dashed curves. The EE curve shows that 28% of the light is contained within
a circular aperture 0.12 in diameter, while 51% of the light is contained within a diameter
of 0.25. The left hand panel expands a portion of the F125LP first light image. These data
have been corrected for geometric distortion. There is a small spur extending ∼ 0.13 from
the center of each PSF to the lower right. This is due to aberration in the system. A similar
aberration is seen in UV images of the high resolution channel (HRC), but the aberration
is not seen at optical wavelengths with either HRC or the wide field channel (WFC; Hartig
et al. 2002).
2.4.
Sensitivity
The SMOV data indicate that the SBC throughput is very close to pre-launch expectations.
In Figure 2 we show the average ratio of the observed over expected count rates for 10
different stars in NGC 6681, whose spectra are well-known from STIS spectroscopy, as a
function of wavelength. The on-orbit data were measured within a 0.4 radius aperture.
Data for the F122M (Lyα) filter came from the standard star HS+2027. Except for F165LP,
the sensitivity of the SBC is very close to expectations. The ∼ 16% increase in sensitivity
88
Tran, et al.
Figure 1.
ACS/SBC PSF profile (right bottom) and encircled energy curve (right
top). The left panel shows the low-level elongation to the lower right in the PSF
of each star due to aberration.
in the F122M band is probably not a result of higher efficiency of the detector, but may be
due to red leak.
2.5.
Contamination Monitor
The UV sensitivity of the ACS/SBC MAMA detector was monitored approximately once a
week for the first two months, and once a month thereafter, using observations of a field in
the globular cluster NGC 6618. This field contains several stars well-observed with STIS
for its own UV sensitivity monitoring program. The SBC observations were made through
all five longpass filters (F115LP, F125LP, F140LP, F150LP, F165LP).
The results of the UV contamination monitor show that the observed count rates are
quite stable, and behave nominally for all SBC filters. Figure 3 shows that the count rates
measured within a 0.6 radius aperture do not change significantly during the six epochs
over the first 72 days that the SBC UV fluxes were monitored. We conclude from this
behavior, and from the throughput comparison with STIS, that the SBC optics suffered no
degradation in throughput resulting from any contamination during the service mission.
Acknowledgments. ACS was developed under NASA contract NAS5-32864, and this
work was supported by a NASA grant.
References
Hartig, G. F., et al. 2002, in Future EUV and UV Visible Space Astrophysics Missions and
Instrumentation, eds. J. C. Blades & O. H. Siegmund, Proc. SPIE, Vol. 4854, in
press [4854-30]
Martel, A. R., Hartig, G., & Sirianni, M. 2001, http://acs.pha.jhu.edu/instrument/calibration/
results/by item/detector/sbc/darks jul01/
Meurer, G. R., et al. 2002, in Future EUV and UV Visible Space Astrophysics Missions
and Instrumentation, eds. J. C. Blades & O. H. Siegmund, Proc. SPIE, Vol. 4854, in
press [4854-30]
On-Orbit Performance of the ACS Solar Blind Channel
Figure 2. Average observed/predicted count rate ratios of 10 different stars in
NGC 6681 as a function of SBC bandpasses. In order from left to right, the filters
are: F122M, F115LP, F125LP, F140LP, F150LP, F165LP.
160
140
120
100
80
60
40
0
20
40
60
Time since first epoch (days)
Figure 3. Count rates versus time for six different stars in NGC 6681 through
the F115LP filter of SBC. Similar behavior is seen for other filters. No significant
changes in count rates are seen as a function of time over the first two months of
monitoring.
89
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
Modelling the Fringing of the ACS CCD Detectors
J. R. Walsh, N. Pirzkal & A. Pasquali
Space Telescope European Co-ordinating Facility, European Southern Observatory,
Karl-Schwarzschild Strasse 2, D-85748 Garching, Germany
Abstract. The fringing of CCD detectors occurs because of interference between
the incident light and the light internally reflected at the interfaces between the
thin layers of the CCD chip. Knowing the construction of the CCD, namely the
materials composing the layers, their refractive index variation with wavelength and
their thicknesses, the resulting fringe amplitude can be calculated from geometrical
optics. Malamuth et al. have applied this technique to the STIS CCD. The topmost
layer controls the high frequency fringes and the lower layers the envelope of the
fringing amplitude with wavelength. Modelling of the four-layer structure of the
ACS CCDs is described. The High Resolution Channel (SITe) CCD is a copy of the
STIS one and so has a similar structure, but the Wide Field Channel (also SiTe)
chip has proprietary construction. The observed fringe amplitude across the CCD
is used to predict the spatial variation of the thickness of the top layer, whilst the
thicknesses of the lower layers are kept fixed. By applying the model maps of the layer
thicknesses, the observed fringing in ACS can be strongly reduced from peak-to-peak
of >20% to a few percent.
1.
Fringing in CCDs
Interference in the CCD detection layer between incident light and that reflected from
interfaces of the thin layers leads to fringing. As the silicon layer becomes more transmissive
in the red, so more signal is contributed by internally reflected light. The fraction of
light internally reflected depends on the number of layers of the CCD, their thickness and
material, which enters through the complex refractive index, and the wavelength of the
incident light. Knowing in detail the bulk construction of the CCD, the resulting fringe
amplitude can be calculated from geometrical optics. The refractive index generally varies
as a function of wavelength for most materials (Palik 1985), and is a key ingredient for the
optical modelling. Malamuth et al. (2000; 2002) successfully applied this technique to the
STIS CCD. By matching the observed fringe amplitudes across the CCD, the thicknesses
of the layers can be modelled using tabulated refractive indices, although Malamuth et al.
found that the bulk Si refractive index had to be adjusted over some wavelength range to
obtain a satisfactory fit. Here we apply the fringe modelling technique to the ACS thinned,
backside-illuminated CCDs.
2.
ACS High Resolution Channel
The ACS High Resolution Channel (HRC) CCD is a SITe STIS spare, so was modelled
closely following the Malamuth et al. parameters for STIS. It is composed of 4 layers as
shown in Figure 1 (left). The lower 3 layers combine to modulate the envelope of the fringe
amplitude, whilst the top Si (detection) layer produces the high frequency fringing. The
optical modelling is performed by summation of the amplitudes of the E-vectors for the
incident and 4 returned rays at a grid of wavelengths, as for multi-layer coatings (cf. Born
90
Fringe Modelling for ACS
91
Figure 1. The layer structure of the ACS High Resolution Channel SITe CCD
is shown in the left panel and the assumed layer structure of the ACS Wide Field
Channel SITe CCD in the right panel.
& Wolf, p. 627). The data with which to compare the model consisted of 37 continuum
lamp images taken with the RASCAL simulator, illuminated by a monochromator (20 Å
bandpass), whilst the ACS was under ground testing at Goddard. ACS has no on-board
facility for narrow band illumination, so ground calibration was critical. The function to
be minimized (rms of observed-predicted fringe amplitude as a function of layer thickness)
is highly periodic and several passes were required to locate the minimum. An amplitude
factor, to allow for the fact that the geometrical optics model is not exact in terms of
transmissions and reflectivities of the layer interfaces, was also applied, but was taken as
a constant. Figure 2 shows the model compared with the observed monochromatic fringe
amplitudes at the chip center. The 10242 image of the layer thickness of the HRC Si layer
was used to compute a model fringe map for direct comparison with the observed data.
Figure 3 (left) shows the observed fringing map at 9200 Å. In the fitting procedure,
the influence of the 2nd layer (SiO2) was investigated by varying its thickness using the
initial map of the 1st (Si) layer (Figure 1). It was found that there is little to be gained by
using a pixel-by-pixel fit to the thickness of the 2nd layer. This is clear from the Figure 3
(right), which shows the observed fringing and the residuals after correcting by the model,
where the traces across the CCD were formed by averaging two rows (511–512). The strong
reduction in the observed fringing demonstrates the effectiveness of the fringe modelling as
applied to the HRC chip.
3.
Wide Field Channel
The Wide Field Channel (WFC) has two 4096 × 2048 SITe back-illuminated chips using
MPP technology, but of proprietary construction. No layer thickness or materials data
were available. The same 4 layer model as for the HRC chip was however assumed and
the thicknesses of the layers to match the observed fringing were similarly modelled (see
Figure 1, right).
46 monochromatic flats were taken during ground testing at Goddard, of width 20 and
10 Å, and the minimum rms of the observed-model fringe amplitude was better defined
than for the HRC. By minimizing the rms for each layer separately, a satisfactory fit for
the chip structure could be found (see Figure 1). Keeping the thickness of the layers 2–4
92
Walsh, et al.
Figure 2.
The model fringing (continuous line) is shown for a 2 × 2 pixel area
at the center of the HRC chip with the observed fringe amplitude, displayed by
crosses, as a function of wavelength.
Figure 3.
In the left panel is shown the observed fringing map at 9200 Å; the
peak amplitude is 0.15. In the right panel, the observed fringing at 9440 Å (upper
trace) is compared with the result of correcting the observed fringing by the model
using the pixel-to-pixel fit to the top CCD layer only (middle) and for the fit with
a pixel-by-pixel fit of the top two layers (bottom).
Fringe Modelling for ACS
93
Figure 4. In the left panel, the modelled layer thickness of the top layer of the
WFC CCD (limits 12.61 to 17.13 µm) is shown with the two chips butted together.
The right panel shows the observed fringing at 9440 Å (upper trace) compared with
the result of correcting the observed fringing by the model (lower trace).
fixed, the observed fringing of the 46 flats was modelled as a function of Layer 1 (assumed
to be Si) thickness. The result is shown in Figure 4 (left); the sampling is 1 × 1 pixels and
the ‘cosmic doughnut’ familiar from flat field images is apparent. Comparing model fringe
maps with observed shows good agreement; correcting the observed fringing by the model
can reduce the amplitude of fringing from 0.12 to ∼ 0.03 (see Figure 4 right for row 1021
of CHIP2).
So far no detectable fringing has been found in WFC 1st order spectra. This is because
the PSF modulates the fringing of the slitless spectra over the range of ∼ 2 pixels (80 Å)
reducing it to ∼ few %. For the HRC the detectable fringing for point sources is larger
since the pixels are narrower in wavelength (27 Å in 1st order).
4.
Correction for Fringing
The fits to the CCD layer thickness will be incorporated in the ACS slitless spectra reduction package, aXe (Pirzkal et al. 2001), to effectively reduce observed fringing in extracted
spectra. When wavelength is assigned to a pixel in an extracted slitless spectrum, the fringe
amplitude is computed from the model and the observed signal corrected.
This fringe modelling method is quite general and can be applied to any CCD, given
specification of its layer structure and a set of monochromatic flat fields over the wavelength
range of significant fringing. Malumuth (this volume, p. 197) has applied the method to
the WFC3 CCDs.
References
Born, M. & Wolf, E. 1975, Principles of Optics, 5th ed.
Malumuth, E. M., Hill, R. S., Gull, T. R., et al. 2000, AAS, 197, 1204
Malumuth, E. M., Hill, R. S., Gull, T. R., et al. 2002, AJ, in press
Palik, E. D. 1985, Handbook of Optical Constants of Solids
Pirzkal, N., Pasquali, A., & Demleitner, M., 2001, ST-ECF Newsletter, No. 29, p. 5.
Part 2. STIS
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
STIS Calibration Status
Charles R. Proffitt1,2, Paul Goudfrooij, Thomas M. Brown, James Davies, Rosa
Diaz-Miller, Linda Dressel, Jessica Kim Quijano, Jesús Maı́z-Apellániz, Bahram
Mobasher, Mike Potter, Kailash Sahu, David Stys, Jeff Valenti, Nolan Walborn,
Ralph Bohlin, Paul Barrett, Ivo Busko, and Phil Hodge
Space Telescope Science Institute, Baltimore, MD 21218
Abstract. Last year’s failure of the STIS Side-1 electronics temporarily suspended
use of the instrument. The Side-1 electronics are not repairable, but operations were
resumed in August of 2001 using the redundant Side-2 electronics. STIS was fully
returned to operation, with only minimal impacts on scientific performance.
MAMA detector performance continues to be very good, with sensitivity changes
of 1 to 2 percent per year. Although the detailed relation between the NUV MAMA
detector temperature and dark current has changed, typical NUV dark current levels
are similar to those in previous cycles. The FUV dark current varies irregularly, and
it is now usually significantly higher than it had been during the first two years of
STIS operations.
The effects of radiation damage on the STIS CCD detector continue to follow
previous trends, with declining charge transfer efficiency, increasing dark current,
and increasing numbers of hot pixels. We also review the use and calibration of the
E1 aperture positions which can be used to ameliorate CTE effects.
1.
Side 1 Electronics Failure
A fuse on the main STIS power bus blew on May 16, 2001, safing the STIS instrument. A
diagnostic test that was intended to repower the primary Side-1 electronics in stages by using
an alternate power bus, resulted in another blown fuse as soon as the first STIS internal
relay was closed. After review of the available telemetry and detailed engineering analyses,
the failure review board (Davis et al. 2001) identified a number of possible causes, but
concluded that the most likely cause was a shorted tantalum capacitor. There is essentially
no chance that this type of capacitor could be melted open once shorted, and on-orbit
repair appears to be impractically complex. It was concluded that no portion of the Side-1
electronics can be recovered.
Fortunately, STIS has a redundant set of electronics (Side-2), which was successfully
used to reactivate STIS in early July 2001. The MAMA detectors and most instrument
mechanisms perform much as they did on Side 1. However, because the Side-2 electronics
lack a functioning temperature sensor for the STIS CCD detector, the CCD can no longer
be operated at constant temperature. Instead, the thermo-electric cooler is operated at
constant current, and while the mean detector temperature is actually lower than the −83 C
Side-1 set point, both the CCD temperature and dark current vary significantly (Brown
2001a). The STIS CCD also suffers from a ≈1 e− /pixel increase in read noise due to
electronic pickup from the Side-2 electronics. This noise can under some circumstances be
1
2
Science Programs, Computer Sciences Corporation
Catholic University of America Institute for Astrophysics and Computational Science
97
98
Proffitt, et al.
Figure 1. The number of CCD hot pixels vs. time. The apparent drop in mid2001 is caused by the lower mean operating temperature of the CCD under Side-2
operations. Because the brightness of a given hot pixel decreases with decreasing
detector temperature, at a lower temperature fewer pixels exceed a given fixed
threshold.
ameliorated by Fourier filtering of the image (Brown 2001b). These differences are also
discussed by Brown (2003) in these proceedings.
2.
Detector Performance
Other than the differences between Side-1 and Side-2 discussed above, changes in STIS CCD
detector performance continue previous trends—apart from some scaling differences caused
by the lower mean CCD detector temperature under Side-2 operations. The number of hot
pixels (Figure 1) and the typical dark current continues to increase as radiation damage
accumulates on the detectors. As of September 2002, the mean STIS CCD dark rate is
0.026 e− /pixel/s, and, after eliminating hot pixels, the median dark rate is 0.004 e− /pixel/s.
The charge transfer efficiency (CTE) continues to degrade, and is discussed in detail by
Goudfrooij et al. (2003).
The MAMA detectors were not directly affected by the switch to the Side-2 electronics;
however, there are some long term changes in the behavior of the detectors that can affect the
calibration. Changes in the detector sensitivity over time are discussed in these proceedings
by Bohlin (2003) and by Stys (2003). Here we will review long term changes in the behavior
of the MAMA dark currents.
NUV MAMA Dark Current. The NUV MAMA dark current is dominated by phosphorescence of impurities in the MgF2 faceplate of the detector. These impurities contain
metastable states which become populated by charged particle impacts, especially during
SAA passages, which can decay several days later, emitting UV photons. A model of the
dark current was developed by Jenkins (1997, private communication) and Kimble (1997)
(see also Ferguson & Baum 1999), and predicts that over short time scales the dark current
will vary exponentially with temperature. Over longer time scales the behavior will depend
STIS Calibration Status
99
Figure 2. The NUV MAMA dark current versus detector temperature for two
different time periods.
on the detailed temperature history of the detector window as well as on any variation in
the number of charged particle impacts. The model predicts that long term increases in
detector temperature will also lead to an increase in the mean NUV dark current, although
with a slope much shallower than the short time scale variations with temperature.
Prior to SM3a in early 2000, both the mean detector temperature and the mean dark
current were increasing over time. However, since that time, the average NUV MAMA
dark current has been decreasing, even though detector temperatures have continued to
increase. During Cycle 11, the mean NUV dark current has been about 12% lower than
during Cycle 8 (Figure 2). Also note that even when detector temperatures are very low,
the dark rate no longer drops below about 0.0009 cnts/s/lo-res-pixel. The scaling formula
used in the STIS pipeline has been updated for these changes in behavior. New NUV dark
reference files are also periodically delivered to account for small changes in the distribution
of dark current across the detector.
FUV MAMA Dark Current. The FUV MAMA does not suffer from the phosphorescent
glow seen in the NUV MAMA, and as a result it has a much lower dark current than
other STIS detectors. It does, however, suffer from an intermittent glow of unknown origin
centered in the upper left quadrant of the detector (Figure 3). This glow has become
patchier and more frequent over time. The lower right corner remains free of the glow, with
a mean dark rate of 6.6 × 10−6 cnts/s/pixel.
The average FUV MAMA dark rate has been increasing over time, and, at any given
time, tends to increase with increasing detector temperature (Figure 4). However, the
strongest correlation appears to be with the length of time that the MAMA high voltage
has been turned on. Typically the MAMA high voltages are turned off prior to the block
of HST orbits that pass through the South Atlantic Anomaly (SAA), and are turned back
on after this block of orbits. The result of these policies is that FUV MAMA observations
taken on the very first orbit after the high voltage has been turned back on will usually
have very low dark currents, with little contribution from the intermittent glow. However,
as there is only one such orbit available per day, it is not practical to reserve that orbit for
100
Proffitt, et al.
Figure 3. FUV MAMA dark current. This figure is the mean of dark monitor
exposures taken between July 2001 and September 2002. Dark regions in this
figure correspond to areas of higher dark current.
FUV programs that would benefit from a very low dark current, as doing so would have
too large an impact on the flexibility needed for efficient scheduling of HST.
We instead recommend that very faint targets be placed in the darkest part of the
detector. For very faint point source spectra, putting the spectrum about 2 above the
bottom edge of the detector (POS TARG 0 -6.8), will put the spectrum in a region of much
lower dark current (see Figure 5). We plan to introduce and calibrate a new pseudo-aperture
position for this purpose during Cycle 12.
External cooling of the FUV MAMA detector would likely result in some decrease in
the typical dark current, but it probably will not completely restore the very low dark
currents that were commonly seen during the first two years of operation.
The number of hot pixels in the FUV MAMA has also been increasing with time. The
number of hot pixels tends to correlate well with the mean intensity of the glow. New dark
reference files have recently been delivered to track this increase.
3.
E1 Aperture Positions for Reducing CTE Effects
The decline in CCD charge transfer efficiency (CTE), as radiation damage has accumulated
on the detector, causes declines in the measured signal that depend both on the detected
signal level as well as on the location of the image or spectrum on the detector (see Goudfrooij 2003 elsewhere in these proceedings). Starting in Cycle 9, new “pseudo-aperture”
positions were defined for 1st order spectroscopy that allow targets to be easily placed near
the top of the CCD. This reduces the number of parallel charge transfers from ≈ 512 to
≈ 124, and proportionately reduces the electrons lost to charge traps during the readout.
STIS Calibration Status
Figure 4.
FUV MAMA dark current in glow region vs. time.
Figure 5.
The FUV MAMA mean dark current vs. detector column, in a seven
pixel high extraction box near the standard extraction position (−3 below the
detector center, dotted line) and in a box near the proposed pseudo-aperture position (6.8 further down, solid line) are compared. The data used are an average of
116 1380 second dark monitor exposures taken between July 2001 and September
2002. This illustrates the typical reduction in the dark current that will result
from putting 1st order spectra 2 above the bottom of the detector.
101
102
Proffitt, et al.
There are several advantages to using these new positions.
• Losing fewer electrons during the read-out directly increases the S/N.
• Because fractional CTE losses are larger at lower signal levels, the differential CTE
losses can distort spectra. This can, for instance, change the apparent equivalent
width of line features.
• Trapped electrons often reappear later during the read, causing “tails” to appear
below bright features. This not only distorts the spatial profiles of real features, but
these CTE tails below cosmic rays and hot pixels can be a serious source of noise for
background or dark current limited observations. Putting the spectrum near row 900
dramatically reduces this kind of noise.
There are however, a few disadvantages to using the new E1 positions.
• The E1 aperture positions are only about 6 below the top of the detector and about
6 above the upper aperture bar; there is therefore less room for extended objects.
• For point source spectra at λ > 7500 Å, the use of contemporaneous IR flats for
removing fringes (Goudfrooij & Christensen 1998) does not work as well as it does
near the center of the detector. If S/N > 50 : 1 is needed at long wavelengths, we
recommend using the regular aperture positions near row 512. For data taken at
the E1 position, consideration should also be given to the use of Malumuth’s (2003)
procedure for constructing fringe flats.
• Some calibrations are not yet as well established near the E1 position as they are
for data taken near the center of the detector. Extensive observations to improve the
calibration of the sensitivity, dispersion, line-spread-functions, point-spread-functions,
and aperture throughputs near the E1 positions have been collected during Cycles 10
and 11. We anticipate that the quality of the calibration at the E1 aperture positions
during Cycle 12 will be nearly as high as it is for spectra taken near the center of the
detector.
4.
Recent and Pending Calibration Enhancements
A number of recent improvements have been made to the STIS pipeline calibration. A
number of these are detailed in the paper by Diaz-Miller et al. (2003), and in other papers
in these proceedings. Below we will discuss the most significant of these recent changes, as
well as the improvements we plan to deliver in the near future.
Sensitivity Changes over Time. There have been clear changes in the sensitivity of the
MAMA detectors over time. These changes vary with wavelength and detector, with sensitivity declining as much as 3% per year at the long wavelength end of the G140L wavelength
range. NUV MAMA modes appear to have increased in sensitivity for the first year of STIS
operations, but have since begun to decrease in sensitivity by 1 to 2% per year (Bohlin 1999;
Stys & Walborn 2001; Stys et al. 2003). The STIS calibration pipeline now corrects 1st order
MAMA spectroscopic modes for these changes in time dependent sensitivity. MAMA imaging modes also appear to show similar time and wavelength dependent sensitivity changes.
The synphot package in the Space Telescope Data Analysis System (STSDAS) is currently
being updated to allow the proper MAMA sensitivity curves to be used for any specified
date. Evaluation of MAMA echelle modes is still in progress and we hope to include time
dependent sensitivity corrections for these modes in the near future.
Measurements of the time dependent changes in CCD sensitivity are complicated by
the degradation of CCD charge transfer efficiency (CTE). Losses due to CTE effects depend
STIS Calibration Status
103
on the signal level, the background level, and the position on the detector and need to be
handled separately from time and wavelength dependent throughput changes. Both effects
are currently being calibrated (see Goudfrooij 2003, and Bohlin 2003 in these proceedings),
and we hope, in the near future, to also include time dependent sensitivity corrections for
CCD modes in both the standard pipeline calibration and synphot.
Other recent improvements affecting flux calibration include delivery of a low order flat
for G140L observations that corrects for vignetting effects at different target positions along
the detector y axis (i.e., perpendicular to the dispersion direction). Similar flats for CCD 1st
order modes are in preparation. Improved pixel-to-pixel flats are also being generated for
both MAMA and CCD modes and will improve the calibration for very high signal-to-noise
observations.
Improvements to Echelle Calibration. Among recent enhancements to the calibration of
echelle data, the adoption in the pipeline of an improved algorithm for the subtraction of
scattered light (Valenti et al. 2002, 2003) is especially noteworthy.
Also important is the pipeline implementation of a fix for flux calibration problems
caused by shifts in the blaze function. This problem and its solution are described in detail
by Bowers & Lindler (2003) elsewhere in these proceedings. The largest part of the problem
resulted from the monthly offsetting of the location of the spectrum on the detector, and
while this offsetting was turned off for echelle spectral modes starting in August 2002, it
is still necessary to correct earlier spectra affected by this problem. Currently the pipeline
uses Bowers & Lindler’s algorithm to correct data taken at the primary echelle wavelength
settings. New dispersion relations which better account for the effects of the monthly
offsetting on the echelle wavelength solution have also been delivered to the pipeline.
Other Enhancements Under Development. A number of other improvements in both pipeline software and post-pipeline analysis tools are currently under development. Below are
some of the items we hope to complete and make available to the STIS user community in
the near future.
• CTE correction formulae for both imaging and spectroscopic observations (see Goudfrooij et al. 2003).
• Better NUV-PRISM flux and wavelength calibrations for both on-axis and off-axis
observations.
• Improved software tools for the analysis of slitless spectra, that will allow quick matching of objects observed in both dispersed and undispersed light and easier extraction
of properly calibrated 1-D spectra.
• More and better options for background smoothing or interpolation for 1st order
spectroscopic modes.
• Increased on-line selection of imaging and spectroscopic PSFs for post-pipeline analysis.
• Time dependent sensitivity corrections for all modes.
• More accurate flux calibration for all secondary wavelength settings, including echelle
blaze shift effects.
References
Bohlin, R. C. 1999, Changes in Sensitivity of the Low Dispersion Modes, Instrument Science
Report STIS 99-07, (Baltimore: STScI)
104
Proffitt, et al.
Bohlin, R. C. 2003, this volume, 115
Bowers, C. & Lindler, D. 2003, this volume, 127
Brown, T. M. 2001a, Temperature Dependence of the STIS CCD Dark Rate During Side-2
Operations, Instrument Science Report STIS 2001-003 (Baltimore: STScI)
Brown, T. M. 2001b, STIS CCD Read Noise During Side-2 Operations, Instrument Science
Report STIS 2001-005 (Baltimore: STScI)
Brown, T. M. 2003, this volume, 180
Davis, M., Campbell, D., Sticka, R., Faful, B., Leidecker, H., Kimble, R., & Goudfrooij, P.
2001, STIS Failure Review Board Final Report (Baltimore: STScI)
Diaz-Miller et al. 2003, this volume, 189
Ferguson, H. & Baum, S. 1999, Scientific Requirements for Thermal Control and Scheduling
of the STIS MAMA Detectors after SM-3, Instrument Science Report STIS 99-02
(Baltimore: STScI)
Goudfrooij, P. 2003, this volume, 105
Goodfrooij, P. & Christensen, J. A. 1998, STIS Near-IR Fringing. III. A Tutorial on the
Use of the IRAF Tasks, Instrument Science Report STIS 98-29 (Baltimore: STScI)
Kimble, R. 1997, STIS IDT Quicklook Analysis Report no. 37, Temperature/Time Modeling
of MAMA2 Phosphorescent Dark Rate
Malumuth, E. 2003, this volume, 197
Stys, D. J. & Walborn, N. R. 2001, Sensitivity Monitor Report for the STIS First-Order
Modes-III, Instrument Science Report STIS 2001-01R (Baltimore: STScI)
Stys, D., et al. 2003, this volume, 205
Valenti, J. A., Lindler, D., Bowers, C., Busko, I., Kim Quijano, J. 2002, ’2-D Algorithm for
Removing Scattered Light from STIS Echelle Data, Instrument Science Report STIS
2002-001 (Baltimore: STScI)
Valenti, et al. 2003, this volume, 209
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
Correcting STIS CCD Photometry for CTE Loss1
Paul Goudfrooij
Space Telescope Science Institute, Baltimore, MD 21218, USA
Randy A. Kimble
NASA Goddard Space Flight Center, Code 681, Greenbelt, MD 20771, USA
Abstract. We review the various on-orbit imaging and spectroscopic observations
that are being used to characterize the Charge Transfer Efficiency (CTE) of the
Charge-Coupled Device (CCD) of the Space Telescope Imaging Spectrograph (STIS)
aboard the Hubble Space Telescope. We parametrize the CTE-related loss for aperture photometry of point sources in terms of dependencies on X and Y positions,
the brightness of the source, the background level, and the time of observation. Our
parametrization of the CTE loss is able to correct point source photometry with STIS
to an accuracy similar to the Poisson noise associated with the source detection itself.
1.
Introduction
Astronomical observation was revolutionized more than two decades ago by charge-coupled
device (CCD) technology, due to a combination of generally linear response over a very large
dynamic range and high quantum efficiency. One shortcoming of CCDs, however, is the
imperfect transfer of charge from one pixel to the next. Charge Transfer Efficiency (CTE)
is the term commonly used to describe such charge loss, and it is quantified by the fraction
of charge successfully moved (clocked) between adjacent pixels. In practice it is often more
useful to use the term Charge Transfer Inefficiency (CTI = 1−CTE). The observational
effect of CTI is that a star whose induced charge has to traverse many pixels before being
read out appears to be fainter than the same star observed near the read-out amplifier.
Laboratory tests have shown that CTE loss of CCDs increases significantly when being
subjected to radiation damage (e.g., Janesick 1991). This is particularly relevant for spaceborne CCDs such as those aboard Hubble Space Telescope (HST ), where the cosmic ray
flux is significantly higher than on the ground. The purpose of the current paper is to
characterize the CTI of the CCD of the Space Telescope Imaging Spectrograph (STIS)
for point source photometry in terms of its dependencies on the X and Y positions, target
intensity, background counts, measurement aperture size, and elapsed on-orbit time. Earlier
on-orbit characterizations of the CTI of the STIS CCD have been reported by Gilliland,
Goudfrooij, & Kimble (1999) and Kimble, Goudfrooij, & Gilliland (2000). The current
paper uses two more years of on-orbit data, which provides a significantly more accurate
temporal dependence. Furthermore, we provide (for the first time) an algorithm to correct
STIS CCD imaging photometry for CTI.
The STIS CCD is a 1024 × 1024 pixel, backside-illuminated device with 21 µm × 21 µm
pixels. It was fabricated by Scientific Imaging Technology (SITe) with a coating process
that allows it to cover the 200–1000 nm wavelength range for STIS in a wide variety of
1
Based on observations with the NASA/ESA Hubble Space Telescope, obtained at the Space Telescope Science
Institute, which is operated by AURA, Inc., under NASA contract NAS5-26555.
105
106
Goudfrooij
Figure 1. Schematic architecture of the STIS CCD. The 1024 × 1024 pixel device
has two serial registers and four readout amplifiers. The nominal amplifier (amp D)
is at the top right.
imaging and spectroscopic modes. Key features of the STIS CCD architecture are shown
schematically in Figure 1. Two serial registers are available. A read-out amplifier is located
at all four corners, each with an independent analog signal processing chain. The full image
can be read out through any one of the four amplifiers, or through two- and four-amplifier
combinations. By default, science exposures employ full-frame readout through amplifier
‘D’, which features the lowest read-out noise. Further technical details regarding the STIS
CCD in particular is provided in Kimble et al. (1994), while background information on the
design of STIS in general can be found in Woodgate et al. (1998).
This paper is organized as follows. We first address CTE degradation. Section 2
describes methods used to monitor the CTI: One method using standard, internal dark
exposures, and two methods designed to quantify the CTI appropriately to observations of
point sources in sparse fields for spectroscopic and imaging modes. We derive functional
dependences of the CTI on source and background counts, X and Y position on the CCD,
and elapsed on-orbit time in Section 3. Finally, Section 4 summarizes these results and
describes an upcoming method to apply the CTI correction to photometric data tables
derived from STIS images.
2.
2.1.
Monitoring the CTE
Cosmic Ray Tails
An elegant method of monitoring CTI using the average profiles of cosmic rays observed
in standard dark current measurements (and hence not requiring any valuable pointed,
“external” telescope time) has been developed by Riess, Biretta, & Casertano (1999; see
also Riess, this volume, p. 47). The method works as follows. While cosmic rays typically
produce charge in more than one pixel, their induced charge distribution should statistically
(i.e., averaged over the whole CCD) be symmetric about their highest-count pixel, without
any preferred angular orientation. Hence, any systematic asymmetry in the cosmic ray
profiles in the clocking direction of the CCD is a measure of the CTI (through charge
Correcting STIS CCD Photometry for CTE Loss
107
Figure 2. CTI increase with on-orbit time, as measured by the excess signal in
the trailing vs. the leading pixels of cosmic ray events detected in standard dark
frames. Plotted is the amplitude of that excess signal after 512 transfers (representative for the center of the CCD). Note the much stronger CTI in the parallel
clocking direction vs. the serial one. The gaps in time indicate extended periods
during which STIS was in safe mode (zero-gyro mode and Servicing Mission 3A
around 2.8 on-orbit years; Side-1 failure around 4.3 years; Servicing Mission 3B
near 5 years). Data on serial CTI are only plotted when the fit was converging.
trapping and subsequent release). Referenced to the highest-count pixel of the cosmic ray
event, one measures the excess signal in the trailing pixels relative to that in the leading
pixels. Averaging the results over thousands of cosmic rays in dark frames, a significant
trailing charge excess is found which grows linearly with distance from the readout amplifier,
a clear signature of CTI origin. Figure 2 shows the growth of the cosmic ray tails with onorbit time, showing (i) the steady growth of the parallel CTI since STIS was placed into
HST, and (ii) the serial CTI.
Since dark frames are taken daily with the STIS CCD, this method is excellent for
providing a finely time-sampled measure of one aspect of CTE performance. However,
it does not provide an adequate measure of the dependence of CTI on signal level, and
only charge lost beyond a short tail is being measured. Charge trapping with longer time
constants is measured using methods described below, which provide measures that are
directly applicable to typical imaging and spectroscopic observations with the STIS CCD.
2.2.
Internal Sparse Field Test
A novel test method, which we designate the “internal sparse field” test, was developed
by the STIS Instrument Definition Team for both ground calibration and in-flight use. It
quantifies two key aspects of CTE effects on spectroscopic measurements: (i) The amount
of charge lost outside a standard extraction aperture, and (ii) the amount of centroid shift
experienced by the charge that remains within that extraction aperture.
The test utilizes the ability of the STIS CCD and its associated electronics to read out
the image with any amplifier, i.e., by clocking the accumulated charge in either direction
for both parallel and serial directions. A sequence of nominally identical exposures is taken,
108
Goudfrooij
Figure 3.
Representative images used for the parallel version of the “internal
sparse field” CTE test. At each of the five positions along the CCD columns, a
sequence of exposures is taken, alternating between amplifiers ‘D’ and ‘B’. Systematic variation of the relative intensities observed by the two amplifiers as a
function of position reveals the CTE effect.
alternating the readout between amplifiers on either side of the CCD (e.g., amps ‘B’ and ‘D’
for measuring parallel CTI). After correcting for (small) gain differences in the two readout
amplifier chains, the observed ratio of the fluxes measured by the two amps can be fit to a
simple CTE model of constant fractional charge loss per pixel transfer. By inspecting the
dependence of the observed flux ratio (e.g., ‘amp B’/‘amp D’) on the source position on the
CCD, it can be confirmed that what is measured is indeed consistent with being due to a
charge transfer effect.
A key virtue of this method is that neither a correction for flat-field response nonuniformity is required, nor an a-priori knowledge of the source flux (as long as the input
source is stable during the alternating exposures). It should be noted that what is being
measured is actually a sum of the charge transfer inefficiencies for the two different clocking
directions. However, given the identical clocking voltages and waveforms and with the
expected symmetry of the radiation damage effects, we believe the assumption that the
CTI is equal in the two different directions is a reasonable one.
The implementation of this “internal”1 version of the sparse field test is as follows.
Using an onboard tungsten lamp, the image of a narrow slit is projected at five positions
along the CCD columns. At each position, a sequence of exposures is taken, alternating
between the ‘B’ and ‘D’ amplifiers for readout. An illustration of such an exposure sequence
is depicted in Figure 3. For each exposure, the average flux per column within a 7-row
extraction aperture (which is the default extraction size for long slit STIS spectra of point
sources, cf. Leitherer & Bohlin 1997) as well as the centroid of the image profile within
those 7 rows are calculated. The alternating exposure sequence allows one to separate CTE
effects from flux variations produced by warmup of the tungsten lamp. As the slit image
extends across hundreds of columns, high statistical precision on CTE performance can be
obtained even at low signal levels per column.
Although these data are taken in undispersed (imaging) mode, the illumination is
representative for typical spectroscopic observations (as the dispersion direction of STIS
CCD spectral modes is along rows). The slit image has a narrow profile (2-pixel FWHM),
similar to a point source spectrum. The CTI resulting from this test is “worst-case,”
since there is essentially no background intensity (“sky”) to provide filling of charge traps
in the CCD array. Early results from this test were reported in Kimble et al. (2000); a
comprehensive update will be forthcoming (Goudfrooij et al. 2003, in preparation).
1
“Internal” in this context means that the necessary observations can be performed during Earth occultations,
hence not requiring any valuable “external” HST observing time.
Correcting STIS CCD Photometry for CTE Loss
2.3.
109
External Sparse Field Test
Similar sparse-field CTE tests using “external” astronomical data have also been carried
out in flight, on an annual basis (since 1999). These calibration programs (HST Program
ID’s to date have been 8415, 8854, and 8911) has utilized both imaging and spectroscopy
modes. The two observing modes are discussed separately below.
Imaging Test Series of imaging data have been acquired once a year (since 1999) on a
sparse field in the outskirts of the Galactic globular cluster NGC 6752, a field containing
several hundred stars spanning a large range of intrinsic brightness. Every visit of the
field consisted of 3 HST orbits, in which several exposures were taken using two different
exposure times (20 s and 100 s per exposure). Several repeat exposures were taken at both
exposure times, alternating again between opposing readout amplifiers.
The image field is “sparse” in the sense that there are not many stars per CCD row or
column. We deliberately choose a portion of NGC 6752 to ensure this, as it is well known
that the CTE-induced loss in crowded fields are significantly ameliorated (due to trap filling)
relative to the effects on isolated point sources, while the latter is what we intend to measure
here. To allow an assessment of the effect of a varying sky background level, we took the
data in the so-called Continuously Visible Zone (CVZ) of HST , in which the bright Earth
comes closer than usual to the telescope pointing direction. The varying amount of scattered
light from the bright Earth allows one to obtain a varying “sky” background during the
CVZ orbits, and hence to obtain CTI measurements at a suitable range of sky background
levels.
Spectroscopy Test Series of spectroscopic data have also been acquired once a year since
1999, as part of the same calibration proposals as the aforementioned imaging CTE tests.
The exposure setup is very similar to that used for the imaging exposures mentioned in the
previous section (3 orbits per HST visit, cycling through three exposure times, alternating
between the two different amplifiers). Slitless spectroscopy is performed of a sparse field
within NGC 346, a young star cluster in in the Small Magellanic Cloud, again located within
the CVZ. The G430L grating is used, which covers the wavelength region 2900–5700 Å at
a dispersion of 2.73 Å/pixel.
The sky background of this field features an ionized gas cloud (H ii region). Due to the
spectral energy distribution of H ii regions within the wavelength coverage of the G430L
grating (e.g., the three strong emission lines [O ii] λ3727, Hβ λ4861 and [O iii] λ5007), the
“sky” background spectrum of these (slitless) data features three relatively constant flux
levels along the dispersion direction. This aspect of this dataset allows one to average the
star spectra over a suitably large number of columns, and hence increase the S/N ratio of
the measurements, while the sky background (and hence its charge trap-filling effect) stays
relatively constant. Results on the CTI in spectroscopic mode will be presented in Ralph
Bohlin’s contribution to this volume (p. 115) as well as in Goudfrooij et al. (2003). We
concentrate on the imaging results in the remainder of this paper.
3.
CTI Analysis for Aperture Photometry of Point Sources
The images were first sorted into groups with a given exposure time and background level.
Images in each group were then averaged together into cosmic-ray-rejected images, using
tasks basic2d and ocrreject in the stis package of stsdas. Aperture photometry was
then performed using the daophot-ii package (Stetson 1987) as implemented within iraf
using fixed-size apertures.
Representative results on the parallel CTI for a short-exposure imaging dataset acquired in 1999 are shown in Figure 4 in which the observed flux ratio (amp D/amp B) vs.
distance from amp B is plotted for two different ranges of stellar flux level per exposure.
110
Goudfrooij
Figure 4.
Relative fluxes as a function of position on the CCD, measured by
amplifiers ‘D’ and ‘B’ for an image acquired in the external sparse field test. The
best-fitting line in each panel has a slope equal to 2 × CTI. Object flux ranges,
sky background value, aperture size and fitted CTI values (per pixel) are shown
in each panel.
The expected CTE behavior is clearly seen, with the closer readout amplifier systematically
measuring a higher stellar flux than the more distant amplifier. In the CTE model we have
been considering (i.e., a constant fractional charge loss per pixel transfer), the predicted flux
ratio is a straight line with a slope equal to 2× CTI. It is clear that the CTI decreases with
increasing signal level. Note also that the charge loss incurred for parallel clocking through
the image area of the CCD is quite substantial for signal levels of a few hundred electrons
or less. Serial CTIs were also determined, and found to be negligible for all purposes (i.e.,
orders of magnitude smaller, and consistent with zero given the uncertainties). In what
follows, CTI is equated to parallel CTI. The default gain = 1 setting (i.e., 1.0 e− /ADU)
is used throughout.
To evaluate the dependence of the measured CTI on aperture size, measurements were
made through three popular aperture sizes (diameters of 5, 7, and 11 pixels). The result is
depicted in Figure 5. Fortunately, we don’t find any significant difference in CTI among the
three apertures used. i.e., one can perform small-aperture photometry (with its increased
S/N relative to larger apertures) without incurring a larger CTI.
The dependence of CTI on the sky background, as derived from the 1999 visit of the
external imaging sparse field test, is depicted in Figure 6. A particularly striking result from
these measurements is the marked decrease in the CTI values for data with increasing sky
background. Furthermore, the slope of log CTI vs. log background decreases systematically
with increasing signal level. This suggests that the sky background fills traps in the bottoms
of the potential wells of the CCD, mostly affecting the transfer of small charge packets.
This substantial benefit of (only modest) sky background is good news for most STIS CCD
imaging observations (which typically have longer exposure times than those used for these
tests), in the sense that most exposures will not suffer from the large CTI values experienced
during the low-background tests reported here.
3.1.
Functional Form for CTI
In evaluating a suitable functional form to characterize the CTI of the STIS CCD, we
considered the following. First, CTI values measured for a given combination of signal
and sky background levels show a time dependence that is consistent with linear (see
Correcting STIS CCD Photometry for CTE Loss
Figure 5.
Similar to Figure 4, but the three panels now show results for three different aperture radii at fixed signal level and background. Note that the measured
CTI is the same for each aperture (within the errors).
Figure 6.
Log CTI as a function of log background for six different flux levels (in
October 1999). The flux levels, the slopes of the best-fitting line, and the latter’s
uncertainties (in parentheses) are mentioned in each panel. Note the systematic
decrease of the CTI dependence on the sky background with increasing signal level.
111
112
Goudfrooij
http://www.stsci.edu/hst/stis/proposing/phase2/cy11 update.pdf), as was found
earlier for WFPC2 data (Whitmore et al. 1999; Dolphin 2000). Furthermore, a glance at
panel (a) of Figure 7 demonstrates that the logarithm of CTI scales roughly linearly with
the logarithm of signal level (for a given background level), i.e., CTI ∝ exp(−a ln(counts)),
while Figure 6 suggests a functional form similar to CTI ∝ exp(−a [sky/counts]b ). We
attempted to fit the CTI values with a combination of those two functional forms as well
as other (additional) terms. After extensive experimentation, the following functional form
produced the best fits to the data:
bck
CTI = (1 + a yr) × b exp (−c lcts) × d exp (−e lbck) + f exp −g
counts
where
h (1)
yr ≡ (MJD − 51831)/365.25
lcts ≡ ln(counts) − 8.5
bck ≡ sqrt(sky2 )
lbck ≡ ln(sqrt(sky2 + 1)) − 2
The constants 51831, 8.5, and 2 were roughly the averages of the corresponding parameters
in the data, and were included to provide numerical stablity as well as to produce independent coefficients (a through h). The sqrt(sky2 ) term in the “bck” and “lbck” parameters
was introduced to avoid logarithms of negative values. The parameter a was determined
by defining a set of signal and sky background levels that were in common between the
data from all epochs, and measuring the fractional CTI increase per year for all those sets.
Parameter a was then defined to be the weighted mean fractional increase of CTI per year
(weighted by the inverse variance of the fractional CTI increase for each set). Initial estimates of the values of parameters b through h and their uncertainties were made using
bootstrap tests. A robust fit parameter was then minimized using a non-linear minimization routine from Numerical Recipes (Press et al. 1992). The resulting best-fit values of the
parameters in eq. (1) are listed in Table 1.
Table 1.
Coeff.
a
b
c
d
e
f
g
h
Best-fit Values of Coefficients in CTI Functional Form
Value
0.11 ± 0.03
(9.32 ± 0.09) 10−5
0.37 ± 0.01
0.23 ± 0.02
0.60 ± 0.05
0.48 ± 0.01
1.80 ± 0.10
0.40 ± 0.04
Description
Time dependence
CTI normalization
Flux (count) level dependence
Normalization for background dependence
Background level dependence
Normalization for background/flux ratio dependence
Background/flux ratio dependence
Power of background/flux ratio
The quality of this parametrization of the CTI correction is depicted in Figure 7 (panels
b – d), separately for each observation epoch. Quantitatively, the CTI parametrization
formula yields a correction that is accurate within 7% for any data point. To put this in
perspective, an observation of a typical faint object with a signal of 100 e− and a background
of 2 e− per pixel at the center of the CCD underwent a CTE loss of ∼ 17% in September
2001 (cf. Figure 7d). The CTI parametrization corrects this loss to a photometric accuracy
of 1.2%, which is similar to the uncertainty due to Poisson noise for that object.
Correcting STIS CCD Photometry for CTE Loss
Figure 7. Panel (a): CTI vs. object signal for the September 1999 dataset. Panels
(b)–(d): CTI vs. parameter “lcts” (see eq. 1), separately for each observation epoch
(1999 through 2001). The percentage charge loss at row 500 of the CCD is shown
on the right side of panels (b) and (d). The lines in panels (b)–(d) depict the bestfit CTI parametrization as discussed in Section 3.1. Symbols and line types are
associated with specific sky background levels, as depicted in the legend of each
panel. The CTI parametrization fits the data to within 7% (max. error), leading
to a photometric accuracy of <
∼ 1% after applying the correction.
113
114
3.2.
Goudfrooij
Applying the CTI Correction Formula
Once the CTI has been determined for a given object by applying Eq. (1) using the coefficients in Table 1, the actual CTI correction to the observed counts is:
Corrected Counts =
Observed Counts
1 − CTI ×
1024
YBIN
− YCEN
where YBIN is the binning factor in the Y direction (data header keyword binaxis2) and
YCEN is the observed Y coordinate of the object in question.
4.
Concluding Remarks
We have reviewed the methods used to monitor the evolution of the CTI of the STIS CCD,
using both internal and external exposures which provide measures that are directly applicable to typical imaging and spectroscopic observations with the STIS CCD. We analyzed
the imaging datasets observed through the fall of 2001 to derive a functional form for the
CTI correction in a semi-empirical fashion. After applying this CTI correction formula to
observed data, systematic residuals stay within 1%.
In the near future, we will perform a similar characterization of the CTE loss occurring
for spectroscopic observations of point sources. CTE correction formulae for STIS CCD
observations will be incorporated in a stsdas task within the stis package. In case of
imaging photometry, we will provide an option to use either ascii or stsdas tables (e.g.,
those supplied by the daophot package) as input files to the task.
As always, STIS observers will be informed of STIS calibration updates by email,
through the Space Telescope Analysis Newsletter which is also available through the “Document Archive” section of the STIS website at http://www.stsci.edu/hst/stis.
Acknowledgments. We appreciated discussions with Ron Gilliland, Adam Riess, and
Brad Whitmore.
References
Dolphin, A. E. 2000, PASP, 112, 1397
Gilliland, R. L., Goudfrooij, P., & Kimble, R. A. 1999, PASP, 111, 1009
Goudfrooij, P., Kimble, R. A., Gilliland, R. L., & Potter, M. 2003, in preparation
Janesick, J., Soli, G., Elliot, T., & Collins, S., 1991 SPIE Electronic Imaging and Technology
Conference on Solid State Optical Sensors II, 1147
Kimble, R. A., Brown, L., Fowler, W. B., Woodgate, B. E., Yagelowich, J. J., et al. 1994,
Proc. SPIE, 2282, p. 169
Kimble, R. A., Goudfrooij, P., & Gilliland, R. L., 2000, Proc. SPIE, 4013, p. 532
Leitherer, C. & Bohlin, R. C. 1999, Instrument Science Report STIS 97-13 (Baltimore:
STScI), available through http://www.stsci.edu/hst/stis
Press, W. H., Flannery, B. P., Teukolsky, S. A., & Vetterling, W. T. 1992,Numerical Recipes
in Fortran (Cambridge: Cambridge University Press)
Riess, A., Biretta, J., & Casertano, S. 1999, Instrument Science Report WFPC2 99-04
(Baltimore: STScI), available through http://www.stsci.edu/instruments/wfpc2
Stetson, P. B. 1987, PASP, 99, 191
Whitmore, B. C., Heyer, I., & Casertano, S. 1999, PASP, 111, 1559
Woodgate, B. E., Kimble, R. A., Bowers, C. W., Kraemer, S., Kaiser, M. E., et al. 1998,
PASP, 110, 1183
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
STIS Flux Calibration
R. Bohlin
Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218
Abstract. The low dispersion STIS spectrophotometric flux calibration must account for all instrumental non-linearities and all changes in sensitivity with time and
temperature. An empirical algorithm for CCD Charge Transfer Efficiency (CTE) is
presented along with the wavelength dependent sensitivity changes. Precise STIS
spectrophotometry includes corrections for CTE losses as large as 20% for faint signals and low sky, for MAMA non-linearities of ∼ 2%, and for loss in system throughput that currently reaches a maximum ∼ 15% at 1625 Å with an average loss of
more than 10% for G140L spectra on the FUV-MAMA. Newly available LTE and
NLTE models from I. Hubeny’s TLUSTY code are compared to the old flux distributions for the Koester/Finley LTE models of the four primary flux standards.
The uncertainty in the corrections and in the flux standards are all small enough,
so that flux distributions relative to 5500 Å in the V -band can be measured in the
photometric 52 × 2 arcsec slit to a precision of ∼ 1% over much of the 1150–10,000 Å
STIS wavelength coverage. The NLTE model flux distributions are probably correct
within ∼1% to 2.5 µ in the IR; and a model for the total uncertainty is presented.
1.
Introduction
The SNAP (SuperNova/Acceleration Probe) mission to determine the dark energy equation
of state parameters has motivated attempts to improve the precision of spectrophotometric
standards to an accuracy of ∼ 1% in the relative flux. In order to achieve the prime science
goals, the SNAP program requires ∼ 1% accuracy in the relative flux calibration over its
0.4–2 µ wavelength range. The HST primary standard stars used for the absolute flux
calibration of STIS and NICMOS are directly relevant to the SNAP program.
In simple terms, the sensitivity of a spectrometer as a function of wavelength is used
to measure the flux F of an observed science target or a secondary stellar flux standard by
F (λ) = C(λ)/S(λ) ,
(1)
where C is the observed count rate. To determine the sensitivity S, the most straightforward
method is to observe a standard star with a known flux distribution Fstd
S = Cstd /Fstd .
(2)
The best primary stellar flux standards from 0.1–3 µ are the set of four unreddened, pure
hydrogen white dwarf (WD) stars G191B2B, GD153, GD71, and HZ43 (Bohlin 2000), while
STIS secondary standards are presented by Bohlin, Dickinson, & Calzetti (2001). The temperature and gravity of the primary standards are determined from fits to the Balmer line
profiles (e.g., Finley, Koester, & Basri 1997); and model atmosphere calculations determine
the relative flux distributions (e.g., Barstow et al. 2001). Precise V -band photometry relative to Vega (Landolt 1992 & 1999, private comm.) sets the absolute flux scale of these
four primary standards. The uncertainty in the absolute Vega flux distribution of Hayes
(1985) combined with uncertainties in the normalization to the Landolt V magnitudes is
115
116
Bohlin
dash-MAMAs, solid-CCD
STIS Sensitivity Change in 5 years: 1997.38-2002.38
1.00
0.95
0.90
2000
3000
3.04
Wavelength (A)
4000
5000
Figure 1.
Change of sensitivity for the five low dispersion modes after five years
of on-orbit operations: dotted lines—MAMA modes G140L and G230L, solid
lines—CCD modes G230LB, G430L, and G750L corrected for CTE losses. All
of the G750L sensitivity changes are assumed to be due to CTE loss.
∼ 2% at V , while the uncertainties in C(λ) relative to 5500 Å in the V -band are discussed
in Section 2. Uncertainties in the calculated model flux distributions relative to 5500 Å are
quantified in Section 3; and Section 4 has examples of actual uncertainties achieved in the
measured STIS flux distributions of secondary flux standards.
2.
Uncertainties in Measured Count Rates
The count rate C in Eq. 1–2 is for an observation with all of the instrumental signatures
removed. These signatures include flat fielding, background removal, stray light, flagging of
artifacts, fringing, operational changes, wavelength calibration, temperature effects, changes
of sensitivity with time, and non-linear response of the detector. Because the effect of a
slowly varying flat field in the dispersion direction is accounted in the sensitivity S calibration, the only benefit of a flat field calibration is the removal of the pixel-to-pixel sensitivity
fluctuations, as long as the spectrum is always located at the same location on the detector.
Precise background subtraction is important for the faintest stars and is complicated by
geocoronal emission lines of Ly-α at 1216 Å and of OI at 1300 Å in the case of HST. For
STIS spectra in the wide 52 × 2 photometric slit, stray light from the wings of the PSF fills
in absorption lines. Beyond ∼ 6620 Å, fringing due to reflective interference in the CCD
substrate becomes increasingly important as the sensitivity drops; and at 1 µ the fringing
correction limits the repeatability to ∼ 1%, even for bright stars. In addition to the absence
of atmospheric effects, observations from space generally benefit from the constancy of operational modes. In the case of STIS, the same wavelength hits close to the same pixel on
the detector, year after year, so that slow temporal changes can be easily tracked. However,
two STIS low dispersion modes have had major adjustments: the G140L default aperture
moved from 3 arcsec above detector center to 3 arcsec below, while for the CCD modes, a
new aperture at row 900 is available and reduces CTE losses by a factor of 5. Precise wavelength calibration is important in wavelength regions with steep sensitivity gradients. For
the G140L mode on the FUV-MAMA there is a temperature dependence of the sensitivity
STIS Flux Calibration
117
Figure 2. Ratios of the first (bottom), second (middle), and most recent third
(top) to the average of all 31 observations of the bright CCD monitor star
AGK+81D266 with NO correction for any change in sensitivity with time or for
any CTE losses. The increasing effect of CTE losses with time are evident from
the ratios > 1 in the bottom panel and < 1 in the top panel. There are small deviations from unity at the shorter wavelengths where the signal peaks at ∼ 40, 000
electrons and larger deviations as the signal drops continuously to ∼ 1000 electrons
at 1 µ. The increasing magnitude of the non-linearity at lower counting levels is a
signature of CTE losses.
of 0.3%/C. The two most important limitations on the photometric precision of the STIS
count rate corrections are: a) the observed scatter about the mean changes of the sensitivity
with time and b) the uncertainty in the non-linear correction for CTE losses (see below).
The change in sensitivity as a function of wavelength and time has been reported
by Bohlin (1999) and by Stys, Walborn, & Sahu (2002). Figure 1 shows the wavelength
dependence of the sensitivity loss after five years in orbit for the five low dispersion STIS
modes. Within a mode, the changes are continuous functions of wavelength; and the two
modes, G230L on the NUV-MAMA and G230LB on the CCD, show the same changes to
within ∼ 0.5%. The four discontinuous jumps from one mode to the next may be the result
of blaze angle shifts in the first order gratings that are caused by shrinkage of the epoxy
substrate of the replica grating rulings, as suggested by Bowers (this volume, p. 127) for
the echelle modes.
Figure 2 illustrates the effects of Charge Transfer Efficiency (CTE) losses for the G750L
mode. The monitoring observations of AGK+81D266 are divided into three epochs and
compared to the average spectrum of all 31 observations of AGK+81D266 since launch.
While the ratio of the middle third is near unity, the early 11 observations in the lower
panel are higher than the average; and the 9 spectra obtained since 2001 July in the top
Bohlin
Charge Transfer Inefficiency = (1-CTE) at CCD Center
118
STIS CTI at 2000.6
B=0.5
0.100
B=2
0.010
0.001
101
102
103
104
105
106
Electrons per Column in 7-px Extraction
Figure 3.
CTI(G,2000.6) at 2000.6, i.e. 3.44 years after launch at the center of
the CCD. The abscissa is G, the total gross electrons recorded in the default 7-px
high extraction box. The dashed line is a fit to the measurements (triangles) of
Kimble (2001, private comm.) at 3.17 and 3.71 years after launch and is relevant
to an image with zero background from a single readout of the CCD. The measurements have been divided by small amounts, 0.92 and 1.08 at 3.17 and 3.71 years,
respectively, to correct the data points to the mean time of 3.44 years, so that
the scatter within the pairs of points at each of the six electron levels is indicative
of the uncertainty. Heavy solid lines: The CTI for background levels of 0.5 and
2 electron/px that are typical of gain 1.
Figure 4.
As in Figure 2 AFTER correction for CTE losses per Equation 1 and
Figure 3. Residuals exceed 0.5% only for the last 100 Å of wavelength coverage
beyond 10,100 Å.
STIS Flux Calibration
119
Figure 5.
Ratios of Hubeny to Koester-Finley pure hydrogen models in LTE for
the temperature and gravities indicated in each of the four panel titles. Differences
approach 1% only in the 10 Å bin at the center of H-α and H-β for the two hottest
stars. Compare with Fig. 1 of Bohlin (2000), which makes the same comparison
before the TLUSTY code upgrades.
panel are below unity everywhere. The assumption that the behavior observed in Figure 2
is attributed entirely to CTE losses and not to optical sensitivity degradation is justified
by the trend toward unity, i.e. no loss, at long wavelengths in Figure 1 and by the trend to
larger losses toward longer wavelengths in Figure 2. CTE losses increase as the G750L signal
decreases toward longer wavelengths. The triangles in Figure 3 illustrate the measured CTE
losses at a mean date of 2000.6, 3.4 years after launch, for a mean background level of zero
(R. Kimble 2001, private comm.). The heavy solid lines are the modeled losses at sky
background levels of 0.5 and 2 electrons per pixel and are a best compromise correction for
the AGK+81D266 data and for some short exposure observations of LDS749B and G191B2B
(Bohlin 2002). The zero background fit to the data triangles in Figure 3 is the dashed line
CTI(G, t = 2000.6), while the preliminary Charge Transfer Inefficiency CT I = (1 − CT E)
at time t in decimal years and rolloff for sky background B is
0.20
CT I(B, G, t) = CT I(G, t)e−2.2(B/G)
0.20
= CT I(G, 2000.6){(t − 2000.6)0.242 + 1}e−2.2(B/G)
,
(3)
where G is the gross signal in electrons in the seven pixel extraction height and where
the CTI is assumed to linearly increase from zero at the 1997.16 launch date. The CTI
increases linearly with the number of CCD rows from the readout amp, while the dashed
line, CTI(G,2000.6), in Figure 3 is for the center of the CCD at row 512. The constants
2.2 and 0.20 in Equation 1 produce the best fit correction to the deviations from unity in
Figure 2; and the ratios for these corrected observations of AGK+81D266 are illustrated in
Figure 4, where the deviations from unity are < 0.5% below 1 µ.
120
Bohlin
Figure 6.
As in Figure 5, except with Hubeny NLTE models in the numerator.
Differences approach 2% only in the 10 Å bin at the center of H-α and H-β for the
two hottest stars. Compare with Fig. 2 of Bohlin (2000), which makes the same
comparison before the TLUSTY code upgrades.
3.
Uncertainties in the Model Flux Distributions
The model Teff and log g are determined by fitting the observed Balmer line profiles (e.g.
Finley, Koester, & Basri 1997). These fits have an internal consistency that implies an
uncertainty of much less than 1% in the shape of model flux distributions from 0.1 to
3 µ. Since Bohlin (2000) reported inconsistencies between LTE calculations with the same
temperature and gravity for the simplest pure hydrogen case, the model atmosphere codes
have been upgraded (Barstow et al. 2001). Figure 5 demonstrates agreement within one
percent in the optical between the new Hubeny and old Koester models for the four primary
standards, even in the 10 Å band at the Balmer line centers. Furthermore, the Hubeny
NLTE model line profiles agree with the Koester/Finley profiles to 2% at line center, as
shown in Figure 6. Therefore, the Hubeny NLTE models are adopted as the best estimates
of the true stellar flux distributions, because the physics is more realistic and because the
Balmer line profiles are in agreement with the models and observations used by Finley to
derive the temperature and gravity. The model flux distributions are placed on an absolute
flux scale by normalizing to the V band magnitudes measured by Landolt with a precision
of a few milli-magnitudes (Bohlin 2000). In the top panel for the hottest star G191B2B,
the difference in the continuum slope starts to be evident and becomes larger in the IR.
In the IR, Figure 7 illustrates differences between the LTE and NLTE models of up
to 4% at 3 µ for the hottest star. The NLTE model for G191B2B resolves the discrepancy
(BDC 2001) between the standard star flux ratio and the measured NICMOS flux ratio for
G191B2B/P330E, as shown in Figure 8. On average, the data points lie closest to the heavy
solid line for the NLTE model.
STIS Flux Calibration
Figure 7.
As in Figure 6, except that the wavelength range is 1–3 µ. There is a
difference between LTE and NLTE of ∼ 4% at 3 µ for the hottest star G191B2B.
Figure 8.
The data points with error bars are the observed NICMOS count rate
ratios for G191B2B/P330E divided by the ratio of the standard star fluxes, using
the old LTE model for G191B2B. The systematic trend for the data points to lie
above unity means that either the standard model for G191B2B is too faint or
the IR spectrum of the solar analog P330E is too bright. The heavy solid line
is the change in the predicted ratio of the stellar observations based on the new
NLTE model for G191B2B. The NLTE model predicts a ratio in better agreement
with the bulk of the observations. The dotted line illustrates the minor difference
between the new LTE models of Hubeny and the old LTE Koester standard.
121
122
4.
4.1.
Bohlin
Can Relative Fluxes Be Measured with a One Percent Accuracy?
Uncertainties in the Observations
The STIS observational uncertainties are dominated by the repeatability of observations of
the same non-variable star and by uncertainty in the correction for CTE losses in the CCD
modes, as long as the counting statistical uncertainty is negligible.
Repeatability: After fitting the changes in sensitivity as a function of time and wavelength,
the residuals measure the lack of perfect repeatability for the 57 FUV and 59 NUV monitoring observations of GRW+70D5824 with the MAMAs and for the 31–35 observations of
AGK+81D266 with the CCD. The repeatability depends on wavelength and on the bandpass bin size as summarized in Table 1. Some of the narrower bins have a comparable
or better repeatability than the broad bands. Therefore, the correlation length for the
fluctuations must be rather long, because the scatter does not “average out” with broader
binning in wavelength. The observed narrower bandpass repeatabilities are corrected for
the small effects of counting statistics and contribute to the modeled uncertainty as a linear
interpolation between the measured wavelengths.
Table 1.
Repeatability in broad bands.
MODE
a
Bandpass (Å)
one sigma (%)
G140L
350
50
0.54
0.46–.79a
G230L
1000
100
0.18
0.19–.59a
G230LB
1000
100
0.27
0.27–.39a
G430L
2000
200
0.32
0.20–.56a
G750L
2800
400
0.15
0.13–.91a
A small correction has been made for the contribution from counting statistics.
CTE Correction: Figure 9 shows some measured errors in the CTE correction and a fit to
these points as a function of the number of electrons in the 7 pixel high standard extraction.
This uncertainty affects the three low dispersion modes that utilize the CCD detector.
4.2.
Uncertainties in the Model Flux Distributions
The formal uncertainty of ∼ 1000 K in the Teff determined from the Balmer line fits implies
an uncertainty in the model fluxes relative to the V band of much less than 1%. In practice
after normalizing the four models to their V magnitudes, there is as much as 1% scatter
in the relative measured vs. model fluxes at 2000 Å and below. The agreement of the flux
distributions from the two independent LTE codes is excellent below 1 µ, as illustrated in
Figure 5. In the continuum, there are differences of 1% at the convergence of the Brackett
lines around 1.6 µ for GD71 and at 3 µ for G191B2B, as shown in Figure 10. The NICMOS
observations of G191B2B vs. P330E in Figure 8 support an uncertainty of ∼ 1% for the
STIS Flux Calibration
Figure 9.
Measured errors in the CTE correction for various cases (triangles
with error bars). The solid line is a fit to these data points and is used to model
uncertainties in observed CCD spectrophotometry as a function of signal strength
in electrons for point sources.
Figure 10. As in Figure 5 for the 1–3 µ region. The big dips at the locations
of the B-β, B-δ, and B-γ are caused by the omission of these features in the old
Koester models. In the continuum, differences between these LTE calculations
approach ∼ 1% around 1.6 µ for GD71 and at 3 µ for G191B2B.
123
124
Bohlin
Figure 11. One percent of the stellar flux for HS2027+0651 (squiggly line), the
total uncertainty relative to V (heavy smooth solid lines), and the three main
components of the total uncertainty (light lines): repeatability (solid), model flux
distribution (dotted), and the CTE correction (dashed). The total uncertainty
for these data obtained about a year after launch exceeds 1% of the flux beyond
9300 Å, mainly because of poor repeatability combined with some CTE uncertainty.
adopted NLTE models beyond 1.5 µ. In summary, the model for the uncertainty in the STIS
sensitivity and in measured secondary flux standards due to possible errors in the adopted
flux distributions of the four prime standards is 1% below 2000 Å, decreasing linearly to
zero in the normalization region at 5500 Å, and increasing to 1% at 1.5 µ and beyond.
4.3.
Achieved Uncertainties for Secondary Standards
Figures 11–13 are examples of measured flux distributions and their uncertainties. The 2%
grey uncertainty from the absolute flux uncertainty of Vega at V is not included; and a
wavelength bin size sufficient to make the counting statistical error 1% (i.e. 10, 000
counts) is assumed. One percent of the measured flux is plotted and compared to the
component and total uncertainties. For the faint standard HS2027+0651 in Figure 11, the
observations were obtained early in the mission, so that the CTE error (dashed line) is
minimal and the total uncertainty (heavy solid line) exceeds 1% of the flux only beyond
9300 Å. Figure 12 is for another faint standard LDS749B obtained about five years postlaunch, where the larger uncertainty in the CTE correction causes > 1% uncertainty at the
short wavelength end of G430L at 3000–3300 Å and at the long wavelength end of G750L
beyond 8600 Å. Because the repeatability error (thin solid line) is assumed to scale as the
square root of the number of co-added observations, the repeatability error for the single
G230L observation of LDS749B at 1900 Å is twice as large in % as for the co-addition
of four spectra of HS2027+0651. In Figure 13, the recent single STIS observation of the
bright Sloan standard BD+17D4708 with the three CCD modes has an uncertainty of < 1%,
except at wavelengths below 2100 Å, where this F star is faint.
STIS Flux Calibration
Figure 12. As in Figure 11 for LDS749B observed after five years of on-orbit
radiation damage to the CCD. Uncertainty in the current preliminary CTE correction limits the precision in two regions of low sensitivity: 3000–3300 Å on G430L
and beyond 8600 Å for G750L.
Figure 13. As in Figure 11 for the Sloan standard BD+17D4708. Only at the
faint short wavelengths of G230LB do uncertainties of this bright (in the visible)
secondary standard exceed one sigma = ∼ 1%. Only one recent observation has
been utilized.
125
126
Bohlin
Can fluxes relative to V over the 0.1 to 3 µ region be determined to 1%? If the
Balmer line profile analyses give the correct stellar temperature and gravities and the model
atmosphere calculations correctly represent the physics, then the answer is yes, in some
cases. The best flux distributions are measured for bright stars or early in the mission to
avoid the uncertainties in correcting for CTE losses in the CCD detectors. (See the collection
of HST standards at http://www.stsci.edu/instruments/observatory/cdbs/calspec.html.)
At the longer wavelengths, the CCD fringing conspires with low sensitivity and large CTE
losses to produce larger uncertainties. In summary, uncertainties as small as 1% can be
achieved at some wavelengths, at some times, and for some ranges of stellar brightness.
Don’t forget that sigma here is only ONE sigma, not THREE and that the uncertainty in a
comparison of two objects is the combination in quadrature of the sigmas of the separately
measured fluxes.
References
Barstow, M. A., Holberg, J. B., Hubeny, I., Good, S. A., Levan, A. J., & Meru, F. 2001,
MNRAS, 328, 211
Bohlin, R. 1999, Instrument Science Report STIS 99-07 (Baltimore: STScI)
Bohlin, R. C. 2000, AJ, 120, 437
Bohlin, R. 2002, Instrument Science Report STIS 02-xx (Baltimore: STScI) in preparation
Bohlin, R. C., Dickinson, M. E., and Calzetti, D. 2001, AJ, 122, 2118 (BDC)
Finley, D. S., Koester, D., & Basri, G. 1997, ApJ, 488, 375
Hayes, D. S. 1985, in Calibration of Fundamental Stellar Quantities, Proc. of IAU Symposium No. 111, ed. D. S. Hayes. L. E. Pasinetti, A. G. Davis Philip, (Reidel,
Dordrecht), p. 225
Landolt, A. U. 1992, AJ, 104, 340
Stys, D., Walborn, N., & Sahu, K. 2002, Instrument Science Report STIS 2002-002 (Baltimore: STScI)
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
STIS Echelle Blaze Shift Correction1
C. Bowers and D. Lindler2
Laboratory for Astronomy and Solar Physics, Code 681, NASA’s Goddard Space
Flight Center, Greenbelt, MD, 20771
Abstract. Planned offsets of the STIS Mode Select Mechanism (MSM) result in
changes to the nominal calibration curves, particularly noticeable in the echelle
modes. The spectral wave calibration exposures (wavecals) obtained with each observation can be used to predict a simple, linear offset of the nominal calibration
curve to be used for each MSM shift. In addition, a time dependent variation has
been detected which is attributed to small changes in the grating itself. An algorithm has been developed which applies the offsets necessary to correct both the time
dependent and MSM shift effects for the echelle modes.
1.
Introduction
Ultraviolet spectra acquired with the Space Telescope Imaging Spectrograph (STIS) have
been periodically shifted in position at the UV MAMA detectors by a small amount to more
uniformly age the UV MAMA detectors. This was done by moving the Mode Select Mechanism (MSM) slightly from its nominal orientation for each mode. However, it was observed
that these motions caused small errors in the echelle calibration, particularly noticeable in
the overlap between orders. Re-calibration for each offset position was possible but time
consuming and inefficient. We thought it might instead be possible to correct for this calibration error by using the wavelength calibration spectra acquired with each observation,
to indicate the offset, and shift the initially acquired calibration curve accordingly.
2.
Magnitude and Cause of The Blaze Shift Effect
Figure 1 shows a portion of an echelle stellar spectrum acquired in STIS E230H mode following an MSM shift from the nominal setting. About six adjacent orders of the spectrum near
2575 Å are presented in the figure. The calibration error introduced is approximately linear
over each order, causing the slanted appearance. The resulting flux mis-match is about
10% at the overlap regions where the same spectral bandpass is measured simultaneously
in adjacent orders.
The calibration error is due to the change in the direction of light incident on the echelle
gratings when the MSM orientation is changed. Changing the light incident angle at the
echelle, causes two changes in the detected spectrum: the spectrum itself shifts position at
the detector, and the grating blaze, or grating efficiency curve, shifts. However these two
shifts are by different amounts. This relative shift between the echelle spectrum and the
grating blaze function is illustrated in Figure 2. In the upper panel, a spectrum consisting
1
2
Based upon observations with the NASA/ESA Hubble Space Telescope, obtained at the Space Telescope
Science Institute, which is operated by the Association of Universities for Research in Astronomy, Inc.,
under NASA contract, NAS 5-26555.
Sigma Space Corporation
127
128
Bowers & Lindler
Figure 1.
The spectrum of a star near 2575A with the E230H echelle, following
a change of the MSM orientation. Six spectral orders are shown. A systematic
calibration error, approximately linear with wavelength over each order has been
introduced by the MSM change.
of several orders (m, m+1, m+2) is illustrated with two distinct emission lines (small slit
symbols) shown. The grating blaze function is illustrated by the gray, trapezoidal region.
Peak grating efficiency is indicated by the bright, central region and the free spectral range
by the two diagonal, dashed lines. The lower panel illustrates the changes in this pattern
following a change of direction of light incident onto the echelle. The spectrum shifts
(indicated by the spectrum offset) and the blaze function shifts (indicated by the blaze
function offset), however the magnitude and direction of these two shifts are not equal.
The overall result is that the relative position of spectral lines with respect to the blaze
efficiency curve has changed. Calibration using a blaze function which does not account for
this relative shift between the spectrum and blaze function will result in the error observed.
3.
Correcting for Echelle Blaze Shift
The blaze function angular shift (dβblz) due to a change in the direction of incident light
on the echelle(dαgrt) in the dispersion direction is
dβblz = −dαgrt
(1)
From the wavecal observations, we can determine the change of the exit angles of light
from the echelle gratings in both the dispersion dβgrt and cross dispersion directions dφgrt.
Using the general grating equations (Namioka, 1959), these can be related to the change of
dispersion direction input angle dαgrt as
sin αgrt + sin βgrt =
mλ
σ cos φgrt
φgrt = −φgrt
dαgrt = −
cos βgrt
sin αgrt + sin βgrt
dβgrt + (
) tan φgrtdφ = −dβblz
cos αgrt
cos αgrt
(2)
(3)
(4)
The change of blaze angle is then in general a function of the exit angles in both the
dispersion (dβgrt) and cross dispersion (dφgrt) directions for out-of-plane grating mounts,
STIS Echelle Blaze Shift Correction
Figure 2. The changes at the detector in spectrum location and grating blaze
function due to a change of incident angle on an echelle grating are illustrated.
The top panel shows several spectral orders and the position of a few spectral
features by the small slit symbols. The grating efficiency curve (blaze) is the gray,
trapezoidal region, centered on the detector. After an MSM change, the spectrum
and blaze are seen to shift by different amounts, causing the relative efficiency of
spectral features to change.
129
130
Bowers & Lindler
Figure 3.
The blaze position as a function of dispersion direction (X) spectral
offset for a set of stellar observations in mode E230H. The blaze shifts about
60 pixels due to occasional variation of the MSM orientation. The results of
estimating the blaze position using only dispersion (single parameter model) or
both dispersion and cross dispersion data (two parameter model) are shown.
i.e., those for which φ = 0. For the STIS echelles the out-of-plane angle is small but not
negligible particularly for motion near the cross dispersion direction.
We thus tried to fit the blaze shift (∆Xblz) with a two parameter function, linear in
the dispersion (∆Xsp ) and cross dispersion (∆Ysp ) directions:
∆Xblz = A1 ∆Xsp + A2 ∆Ysp
(5)
A series of observations of mode E230H were selected for an initial test of correlating
observed blaze function shift with spectrum shift as determined from the accompanying
wavecal spectra. The spectra selected were all stellar with good S/N and few features and
were acquired over a period spanning about 1500 days. Relative spectral shifts in dispersion
(X) and cross dispersion (Y ) directions were determined for each selected spectrum from
the wavecal spectra. Blaze shifts were determined by shifting the echelle ripple pattern (this
is the pattern shown in Figure 1) until the overlap regions were coincident.
Figure 3 shows the relative blaze position as a function of the spectrum offset in the
dispersion (X) direction as the filled circles. The blaze function was seen to shift by about
sixty pixels throughout this series of observations. The correlation between blaze and spectral shifts is evident; the dashed line (single parameter model) shows the best, linear fit
between these quantities. The separation between the dashed line and the data points indicates the error which would still remain if only this single parameter model were used.
The greatest error occurs at the point with the unusually low blaze position of pixel 380,
with a residual error of about 20 pixels. Fitting the blaze position as a linear function of
the spectral shift in both dispersion (X) and cross dispersion (Y ) yields the results shown
in Figure 3 by the unfilled diamond symbols. The improvement compared to the single
parameter fit is evident. The distribution of errors is still somewhat large, with a standard
deviation of 7.5 pixels.
Examining the details of the remaining errors shows that the largest discrepancies occurred for spectra acquired at substantially different times. Figure 4 shows the difference
STIS Echelle Blaze Shift Correction
131
Figure 4. The difference between measured and expected blaze position using
the two parameter model (Figure 3) as a function of observation time. The remaining fitting error is clearly correlated with observing time approximately linearly.
between the measured blaze position and the two parameter model fit as a function of relative observation time. A correlation which is approximately linear is evident; the measured
blaze appears to have shifted about 25 pixels (out of a total in the data of sixty) over
the 1500-day period from which these observations were drawn. Including a linear time
dependence for the blaze as a third parameter produces the fit shown in Figure 5 (the three
parameter model). The improvement compared to the two parameter fit is clear with the
standard deviation between measured and fit blaze positions reduced to four pixels. Using
such a fit, the blaze position can be well estimated from the spectral shift (both X and
Y ) and the time of observation. But what is changing with time? We will return to this
question in Section 5.
4.
Implementation of the Correction Algorithm
To provide the most accurate data for blaze shift correction, all non-proprietary echelle
observations of sources with a continuum over the period from launch to December, 2001
were selected. The spectral shifts, in both dispersion and cross dispersion directions, were
determined from the accompanying wavecal images. The blaze shift of each spectrum was
determined by shifting the echelle ripple pattern until the overlap regions are coincident.
Finally a three parameter, linear fit of blaze position as a function of spectral location (X
Table 1.
STIS Echelle Mode Blaze Shift Model Parameters
Mode
E140M
E140H
E230M
E230H
A1
−0.30
−0.66
0.10
1.49
A2
0.01
−0.11
−0.15
−0.31
A3
0.008
−0.021
−0.002
−0.017
132
Bowers & Lindler
Figure 5. The blaze function measured (filled circles) and fit (open diamonds)
using a model linear in dispersion (X), cross dispersion (Y ) and time. Comparison
with Figure 3 shows the significant improvement achieved by including observation
time in the model.
and Y ) and time was produced for each echelle mode. The fit coefficients are presented
in Table 1. To apply to a spectrum, the wave calibration image is used to determine the
X and Y spectral shifts and the observation date provides the time value. The sensitivity
curve is shifted accordingly and then applied to the data. This algorithm was implemented
in the STIS pipeline reduction package, calstis, version 2.13b.
Figure 6 shows the result of applying this model to the data set of Figure 1 for Mode
E230H. The overlap agreement is good to about 1.5%. Similar results are obtained with
spectra in all echelle modes. We conclude that this method, if the time dependent term
is included, can produce spectra well corrected for the MSM offsets and time variations
observed so far.
The MSM offset procedure for the echelle modes was ended August 5, 2002, so corrections for this effect can be applied to all STIS echelle data accumulated to date. The
current incorporation of this algorithm will continue however, to apply a time dependent
correction, with the same slope as determined here, to all future observations.
5.
Time Dependent Variation of Blaze
A variation with time of the alignment of any optical components following the STIS entrance slit or the detector, could cause an apparent shift of the blaze function. But such an
instability would also cause a time dependent change in the location of the wavecal spectra.
We used a selected set of wavecal spectra to assess the temporal stability within STIS.
From a series of echelle observations in Mode E140H acquired over a period of about 1450
days (approximately contemporaneous with the E230H presented in Section 3) we extracted
those for which the MSM position repeated, that is the MSM was set to the nominal position
for the mode. The spectral locations, as determined by the wavecal spectra, in dispersion
(X) and cross dispersion (Y ) are illustrated in Figure 7. Over this four year period, the
position of the spectrum varied by no more than ±2 pixels in both directions, showing a
very high level of internal stability within STIS for the entire optical system following the
STIS Echelle Blaze Shift Correction
133
Figure 6. The same spectrum as Figure 1, now corrected using the wavecal
spectrum and results from the blaze shift algorithm for Mode E230H.
entrance slit. There may be a very small level of real drift amounting ≤ 4 pixels over the
four year period in the dispersion and cross dispersion directions but this is negligible when
compared with the blaze drift in this same measurement set illustrated in Figure 8. This
plot shows the relative measured blaze position over the four year observation period for
this same set of data which are nominally at the identical alignment. The blaze position
has shifted by over thirty pixels in this time.
The apparent shift of the blaze efficiency function with time could be caused by a
change of sensitivity across the detectors, approximately linear in the high dispersion direction. Such a change would not be wavelength dependent but position dependent; it would
vary similarly in each order (see Figure 1). Note however, from the results of Table 1,
that this time dependent sensitivity change would have to occur in both MAMA detectors
since we see shifts in both detectors. Such a change with time would also show up in all
other modes which utilize the MAMA detectors, including mode G230L for example. Any
changes in sensitivity in such a first order mode could be due to either a variation with
wavelength, the usual interpretation, or position on the detector. The dispersion direction
for G230L is the same as the high dispersion direction for echelle mode E230H. Repeated
sensitivity calibrations of mode G230L with wavelength are presented in Figure 9 of Instrument Science Report STIS 99-07 over the time period 1997.38–2000.38. The variation
with wavelength is not linear and is no more than +2% to −1% at any wavelength over this
time period. This data suggests that this detector is highly stable both positionally and
with wavelength. From these results we would expect to see echelle spectra taken over this
period to have order overlap errors no larger than these limits, provided the MSM was in the
same position. Figure 9 shows five orders from E230H of a calibration white dwarf taken at
1998.4 (upper curve) and 2001.9 (lower curve). The slit and MSM position was identical for
both observations and the spectral locations were within 3.6 pixels (cross dispersion) and
0.5 pixels (dispersion direction). Both spectra have been approximately normalized and the
earlier spectrum has been offset for clarity. The calibration and order overlap is very good
for the initial spectrum but 3.5 years later the systematic calibration error is about 8%, not
consistent with the measured sensitivity stability. The blaze shift effect does not seem to
be due to any change in detector sensitivity.
For the shorter wavelength detector, similar stability tests with Mode E140L (Figure 8,
Bohlin) do show some variability with time. The variation with wavelength is not monotonic; interpreting this as a possible positional sensitivity error would require more detailed
modeling to understand the effect on the echelle blaze curve. However the variation shown
is likely to be due to low level contamination and thus be a true wavelength dependent
134
Bowers & Lindler
Figure 7.
The relative positions of spectra obtained with the E140H echelle over
a four year period, with the MSM positioned to the same orientation. The internal
temporal stability of this STIS mode is very high; any possible drift is no greater
than 4 pixels over this period.
Figure 8.
The temporal stability of the blaze function, from the same data set
of repeated MSM orientations used in Figure 6. Over this four year period, the
blaze function has shifted about 30 pixels while the spectrum shifted no more than
4 pixels in the dispersion direction.
STIS Echelle Blaze Shift Correction
135
1.4
Normalized Flux
1.2
1.0
0.8
0.6
2320
2330
2340
Wavelength
2350
2360
Figure 9. A portion of normalized, E230H spectra of a white dwarf calibration
star from 1998.4 (upper, offset for clarity) and 2001.9 (lower) taken with the same
slit and same nominal MSM positions. The initial, consistent calibration is in error
by about 8% after 3.5 years, though the detector sensitivity has varied by no more
than 2%.
effect and not a positional sensitivity variation at all. The mirror coatings for STIS modes
in this wavelength range were tailored to have the least sensitivity to contaminants near
1216 Å and should be most sensitive near 1600Å. Mode G140L has decreased least near
1300 Å and most near 1600 Å similar to expectations.
Without other possibilities, it appears that the echelle blaze itself must be changing in
time, that is the grating groove angle is slowly changing. The rate of blaze shift for each
mode is listed in Table 2; it is very small, measured in either pixels or in tilt of the grating
grooves (arcseconds/year). The indicated rate of change of Mode E230M is within the
measurement errors, but the rates for the other three modes appear real. Without the high
degree of stability of HST and STIS such small changes would be difficult to detect. All
four echelles are replicated from master rulings. One possible mechanism for blaze change
is very slight shrinkage with time of the epoxy used in replication, though we note that the
measured rates of change are not well correlated with groove depth as might be expected
in a simple model.
Table 2.
STIS Echelle Mode Blaze Shift Rate
Mode
E140M
E140H
E230M
E230H
Blz shift [pixels/yr]
2.9
7.7
0.7
6.2
Blaze tilt change [/yr]
5.9
15.4
1.5
12.5
136
6.
Bowers & Lindler
Summary and Recommendations
The calibration error introduced by shifting the MSM from its nominal orientation can be
well corrected by using the original sensitivity curve for each echelle mode, shifted according
to the spectral shifts determined from the associated wavecal spectra and using a linear,
time dependent term. Linear fits to data collected over a 4-year period provide the necessary
coefficients for the algorithm, presently incorporated in the pipeline reduction procedure,
calstis. Shifting the MSM for the echelle modes has been halted as unnecessary so that
application of the shift terms is now incorporated only for archival data. However the time
dependent term will be continue to be applied for future reductions. As such, the time
dependence of the blaze function should continue to be monitored for any changes from
the simple, linear dependence used in the current model. The cause of this time dependent
term appears to be a change in the gratings themselves.
Acknowledgments. We would like to thank Jeff Valenti and the Space Telescope
Science Institute STIS Team for their help and support of this work some of which was
performed under purchase order 40095. We would also like to thank Ted Gull for several
useful discussions about this work.
References
Bohlin, R. 1999, Instrument Science Report STIS 99-07 (Baltimore: STScI)
http://www.stsci.edu/hst/stis/documents/isrs
Namioka, T. 1959, JOSA, 49, 446
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
Coronagraphic Imaging with HST and STIS1,2
C. A. Grady
NOAO, Eureka Scientific and GSFC, Code 681, NASA’s GSFC, Greenbelt, MD
20771; cgrady@echelle.gsfc.nasa.gov
C. Proffitt
Space Telescope Science Institute, 3700 San Martin Dr., Baltimore, MD 21218
E. Malumuth
Science Systems and Applications, Inc., Lanham, MD 20706
B. E. Woodgate, T. R. Gull, C. W. Bowers, S. R. Heap, and R. A. Kimble
Code 681, Laboratory for Astronomy & Solar Physics, NASA’s GSFC, Greenbelt,
MD 20771; members of the Space Telescope Imaging Spectrograph Investigation
Definition Team
D. Lindler
Sigma Research and Engineering, Lanham, MD 20706
P. Plait
Department of Physics and Astronomy, Sonoma State University, Rohnert Park,
CA 94928
Abstract. Revealing faint circumstellar nebulosity and faint stellar or substellar
companions to bright stars typically requires use of techniques for rejecting the direct,
scattered, and diffracted light of the star. One such technique is Lyot coronagraphy.
We summarize the performance of the white-light coronagraphic capability of the
Space Telescope Imaging Spectrograph, on board the Hubble Space Telescope.
1.
Introduction
As part of its optical imaging capabilities, the Space Telescope Imaging Spectrograph (STIS)
is equipped with an opaque focal plane mask to occult the star, a sub-aperture, circular, pupil plane mask, and an unfiltered CCD providing a simple white-light coronagraph
(Woodgate et al. 1998; Kimble et al. 1998). Fig. 1 shows a simplified version of the STIS
optical path for coronagraphic observations (after Heap, et al. 2000). The bandpass of the
coronagraph is CCD-limited to 0.2–1.0 µm with λeff ≈ 5875 Å. Since 1997, the STIS coronagraph has been used to image reflection nebulosity, protoplanetary disks, emission line
1
2
Based on observations made with the NASA/ESA Hubble Space Telescope, obtained at the Space Telescope
Science Institute, which is operated by the Association of Universities for Research in Astronomy, Inc., under
NASA Contract NAS5-26555.
Data included in this study come in part from the STIS IDT protoplanetary disk key project.
137
138
Grady et al.
Figure 1.
Simplified depiction of the optical path for STIS coronagraphic imagery, after Heap (2000).
nebulae including Herbig-Haro objects, and candidate stellar companions associated with
stars spanning 0.34 ≤ V ≤ 15 (Heap et al. 2000; Grady et al. 1999, 2000, 2001a,b, 2002;
Schneider 2001; Mouillet et al. 2001; Danks et al. 2001). In this paper we summarize the
operation, calibration, and performance of the STIS coronagraph.
2.
Coronagraphic Observations
Coronagraphic observations with STIS are carried out with the star placed under one of
two orthogonal wedges, or under a 3.0 × 10 bar. While the wedges vary smoothly in
diameter from 0.5”-3.0” over their 50” length, a limited set of coronagraphic “apertures”
has been defined to simplify planning for coronagraphic observations. These apertures are
at locations where the wedges are 1.0, 1.8, 2.0, 2.5, and 2.8 wide (fig 2). The vertical
wedge (wedge A) is oriented along the STIS CCD axis 2, the same direction as the long
slits for spectroscopic observations, and parallel to the CCD charge transfer direction. This
ensures that charge saturation from the stellar point spread function (PSF) at the edge of
the wedge does not contaminate the bulk of the coronagraphic image. The majority of STIS
coronagraphic observations have been made with this wedge.
3.
Coronagraphic Data Reduction
Coronagraphic observations made with STIS typically consist of a suite of short exposures
which are grouped using the CR-SPLIT optional parameter in order to avoid saturation
in the vicinity of the coronagraphic wedge. This grouping facilitates cosmic-ray rejection
by median filtering the images in each observation data-cube prior to coaddition. The
standard CCD image reduction then follows with overscan bias subtraction, conversion to
count rates, flat-fielding, dark image subtraction, bad pixel flagging and hot pixel repair.
This reduction is carried out both for science target and calibration star observations, with
data obtained at different spacecraft orientations reduced separately. To ensure the largest
achievable dynamic range in coronagraphic data, the observations are typically obtained
with GAIN = 4, which introduces a low-level video noise with a characteristic scale of 10–30
pixels (0.5–1.5) and amplitude 0.5–1 DN. The pattern is more conspicuous in observations
made since mid-2001, using the STIS side 2 electronics (Brown 2001). At this point in
time, no attempt has been made to Fourier filter the coronagraphic observations, since the
available algorithms are designed for full CCD observations and can introduce ringing into
the images if there are high contrast, sharp structures (e.g., the wedges) in the images.
Coronagraphic Imaging with HST and STIS
139
A2.8
Bar10
A2.5
A2.0
A1.8
B1.0
B1.8 B2.0
B2.5
B2.8
A1.0
Figure 2. The STIS coronagraphic wedge structure with the “aperture” locations defined by the STScI indicated with the stellar diffraction spikes.
4.
Comparison of Coronagraphic and Direct Imaging Data
The goal of coronagraphic observations is to detect faint emission near a bright star. STIS
direct light images taken under these conditions are saturated near the location of the star,
and along the charge transfer direction, and suffer from the presence of multiple, bright,
window ghosts (Fig. 3). Use of the coronagraphic wedge enables longer exposures without
saturation, and prevents formation of conspicuous window ghosts. As noted by Heap et al.
(2000), the STIS Lyot stop is a circular aperture passing the central 77% of the beam area,
without apodization of the diffraction spikes. All STIS images are taken with the Lyot stop
in the optical path, and have PSF wings which are reduced by a factor of ≈ 2 compared to
models for the optical telescope assembly. Occulting the star with the coronagraphic wedge
reduces the brightness of the PSF by an additional factor of 3–5 near the star and up to an
order of magnitude at distances r ≥ 10 , but only slightly depresses the diffraction spikes,
principally by reducing scattering of longer wavelength photons within the STIS CCD.
The change in the prominence of the diffraction spikes relative to the azimuthally
symmetric part of the PSF is not the only change between direct-light and coronagraphic
imagery with STIS. Other bright features in the STIS PSF such as the “stool legs” flanking
wedge A and the “tuft” seen above wedge B in coronagraphic data obtained with that
occulting bar (see Heap et al. 2000), are present in both data, but are more conspicuous
in coronagraphic data. The coronagraphic PSF is sufficiently different from the direct-light
PSF that direct light data cannot be used in the reduction of STIS coronagraphic data. Fig. 5
shows the contrast as a function of distance from the star for direct light and coronagraphic
observations. Far from the star, the suppression of the PSF is similar for observations where
the star is placed under the coronagraphic wedge, or off the active detector area (Fig. 6).
5.
Color Dependence
The bandpass of the STIS coronagraph is the bandwidth of the unfiltered CCD, and spans
0.2–1.0 µm. The HST PSF is known to vary across this bandwidth, and thus it is no
surprise that STIS has prominent color effects in the coronagraphic PSF. In Fig. 4 we show
the PSF wings for a suite of 4.74 ≤ V ≤ 7.92 single stars spanning 0.06 ≤ (B − V ) ≤ 1.65.
The width of the observed PSF increases with increasing (B − V ), and is accompanied by a
progressive deepening of a dark ring at 1.5. Subtracting a scaled PSF which differs from a
140
Grady et al.
IDL
-2.16567
2.83433
Log scale
Figure 3. Comparison of direct light and coronagraphic imagery of the V =
8.2 Herbig Ae star HD 95881. upper) The direct light image is dominated by
saturation and bleeding, and window ghosts. lower) the coronagraphic image has
reduced PSF wings compared to the diffraction spikes and the high power “stool
legs” flanking wedge A.
science target by ∆(B − V ) ≥ 0.08 results in the appearance of a series of rings, resembling
the diffraction rings seen in narrower-band WFPC2 or ACS imagery. Whether the ring
pattern is seen as a positive or negative pattern depends upon whether the science target
is bluer or redder than the PSF star. The pattern is similar along the diffraction spikes,
and can be used to select comparison stars for a science target whose color is not known
a priori.
6.
Shape Dependence Upon Aperture Location
While the higher power features in STIS coronagraphic images do show some field dependence, the shape of the wings of the PSF is less dependent upon the aperture used for the
observations. Comparison of PSF star observations at different wedge locations (Fig. 5) indicates that the suppression of the wings of the PSF induced by occulting the star does not
depend strongly upon the wedge location, with wider wedges suppressing the inner portions
of the PSF, and leaving the outer PSF wings largely unaffected. The primary advantage of
using the coronagraphic wedge is to reduce the dynamic range of the image, or equivalently
to go deeper by exposing longer before the CCD saturates. At wedge A1.0, exposures can
extend a factor of ≈ 400 longer than for direct light images before reaching the CCD fullwell. The ability to integrate longer without saturation means that fewer detector readouts
are needed to build up the desired signal to noise far from the star, and hence that the read
noise + sky background in the final image is lower.
Coronagraphic Imaging with HST and STIS
Figure 4.
Dependence of the shape of the PSF wings on the stellar (B − V ) from
(B − V ) = 0.06 (black) through (B − V ) = 1.65 (red).
Figure 5. Comparison of the PSF radial profile for HR 4413 (V = 5.2) for direct
light images (black), at wedge A1.0 (blue), and at wedge A1.8 (orange).
141
142
Grady et al.
Figure 6. Suppression of the PSF wings is similar for coronagraphic images and
observations where the star is placed off the active detector area. Here data are
shown for 2 Eri in direct imaging mode (black), at wedge A2.0 (blue), 5 off the
detector (green), and the contrast achieved following PSF subtraction for the offthe-detector observations.
IDL
-2.31903
2.68097
Log scale
Figure 7.
PSF subtraction residuals for 2 successive observations of the V = 5.2
star HR 4413. The image shown here is 10 on a side. Within 2 of the star, the
radial “tendrils” dominate.
Coronagraphic Imaging with HST and STIS
143
Figure 8. Variation in PSF subtraction residuals during a single 3-orbit visit
to V = 5.2 HD 141653 relative to one of the observations from the 3rd orbit.
Residuals were computed in 1 arcsec2 boxes to the left and right of wedge A.
Orbit breaks are indicated in red. STIS shows a clear orbital phase dependence
which is quite different from the response of NICMOS.
7.
Stability
In addition to color matching of the PSF star to the science object, successful removal of the
PSF from coronagraphic observations hinges upon the extent to which the science object
and the PSF star were placed at the same position behind the coronagraphic wedge, and the
extent to which the HST and STIS are in the same alignment and focus condition in both
observations. Compared to ground-based observations, the HST point spread function
is remarkably stable, with the principal changes involving redistribution of light due to
thermally-driven changes in the primary-secondary mirror separation (see discussion in
Pavlovsky et al. 2001). When there is no color difference or difference in stellar placement
under the wedge, net images (star-PSF star) are dominated by 1–2 long radial features
due to differences in the dispersed speckles, which Schneider has termed “tendrils” (Fig. 7).
These features provide a structured, high contrast zone within 2 of the star, and are
minimized for short (e.g., bright occulted source) observations taken within a few minutes
of each other, but show systematic increases in amplitude with orbital phase (Fig. 8).
8.
Limiting Performance
For exo-planet searches and detection of faint nebulosity, one parameter of interest is the
contrast achieved after removal of the PSF via subtraction of a comparison star. In the
absence of color differences between the science target and the calibration star, at 2 from
the star we have achieved contrasts of 10−6 arcsec−2 relative to the total stellar flux. The
equivalent contrast for a point source, measured over the 0.1 ×0.1 HST resolution element
is 10−8 . This is achieved for back-to-back observations of the same V = 5.2 calibration star,
with no change in telescope pointing between observations, other than the FGS jitter.
The same measurement for data obtained on successive orbits with independent target
acquisitions for each orbit (separate visits) is approximately a factor of 2 worse, reflecting
changes in the placement of the star under the wedge. This performance is comparable
144
Grady et al.
Figure 9. Star-to-Nebulosity contrast for the STIS coronagraph at V = 5.2. The
narrow dot-dash line is the direct imaging radial profile, the narrow solid line is
the raw coronagraphic radial profile. The profile following PSF subtraction for
optimally positioned stars with no color mismatch and contemporaneous data is
shown in bold. The limiting contrast when the observations are made as separate
visits separated by months is shown in aqua. Contemporary, separate visits have
radial profiles which are intermediate.
Figure 10. Radial surface brightness profiles for the Herbig Ae stars coronagraphically imaged by STIS. Dashed = direct imaging profile, black = raw coronagraphic profile, blue = AB Aur (nebulosity at r ≥ 3 visible in WFPC/2 direct
imagery), aqua = HD 100546, orange = HD 163296. The debris disks β Pictoris
and HR 4796A are close to the raw coronagraphic data profile in surface brightness, while HD 141569 A is intermediate between AB Aur and HD 163296. The
dot-dashed profile is the limiting radial profile for the V = 7–8 Herbig Ae star
non-detections. The fainter T Tauri disks that have been similarly observed have
radial surface brightness closer to 10−3 arcsec−2 relative to their stars.
Coronagraphic Imaging with HST and STIS
145
to that seen when the spacecraft is rolled in the middle of an orbit (Fig. 9). This latter
finding differs significantly from the NICMOS experience, and reflects the fact that STIS
is located on the sunny side of the HST spacecraft, in a less thermally benign environment
than NICMOS or ACS.
In all cases far from the star the contrast relative to the star is limited by the combined
effects of the sky background, detector dark counts, and more importantly, the aggregate
effects of the detector read noise. Further exploration of the suitability of Fourier filtering
techniques developed for studies of galaxies is needed.
9.
Application to PMS Stars
To date, the principal application of the STIS coronagraphic capability has been to detect nebulosity associated with nearby, young stars. Objects that have been successfully
coronagraphically imaged by STIS span 15 ≤ V ≤ 0.34, and have circumstellar nebulosity,
LCS , which has LCS /L∗ = 10−3 to 10−5 arcsec−2 2 from the star (Fig. 10). These include
stars with nebulosity detectable in direct imaging with WFPC2 (e.g., TW Hya, Krist et al.
2000; AB Aur, Grady et al. 1999), as well as nebulosity detectable in the coronagraphic
data only following PSF subtraction. This latter case includes some intrinsically faint disks
(e.g., HD 163296 Grady et al. 2000) as well as objects with disks with steep radial surface
brightness profiles, e.g., following r −3 . Comparatively small (r ≤ 2 ) disks, even when
bright, present a more challenging case in that their angular extent is small compared to
the STIS wedge, located in a region subject to the breathing “tendrils.”
10.
The STIS Coronagraph in an Era of Multiple HST Coronagraphs
HST now has 3 working coronagraphs. Compared to NICMOS, STIS offers a larger field of
view, which can be important in determining where the nebulosity ends (e.g., the HD 100546
envelope, Grady et al. 2001), higher spatial resolution due to imaging at shorter wavelengths,
and comparable sensitivity limits. For detection of comparatively bright, point-source companions to occulted stars, NICMOS’s ability to roll in mid-orbit offers a more time-efficient
observing strategy. NICMOS coronagraphic imagery can be combined with filters to provide
some albedo and chemistry information.
Both ACS and STIS are optical coronagraphs. ACS offers better sampling of the
image, with pixels that are a factor of 4 smaller in area than the STIS CCD pixels, and
the use of filters. The sensitivity of ACS is within a factor of 2 of STIS’s performance
for extended objects (Pavlovsky et al. 2002). ACS will be the preferred coronagraph for
large-scale (r ≥ 1 ) and brighter nebulosity where the more limited throughput in each
filter is compensated for by the ability to carry out albedo studies. The limiting contrast
for faint nebulosity will depend, for ACS observations, upon the impact of the re-imaged
direct light on the data, and has yet to be determined. For smaller disks, such as even the
larger ones around classical T Tauri stars, the large ACS spot size will occult much of the
disk and also has the potential to impair recognition of the presence of an optically visible
disk. Given the small advantage ACS has in apodization, STIS may still be the preferred
instrument for initial surveys for disks and more extended nebulosity as a result of both
the broad bandpass and equal sensitivity to reflection nebulosity and emission-line nebulae
in one observation. STIS provides access to narrower occulter locations than ACS and may
be the preferred instrument for optical imaging of smaller circumstellar disks.
146
11.
Grady et al.
Recommendations for STIS Coronagraphic Observations
• The location on wedge A used for observations of HR 4796A and HD 141569A, where
the wedge has a diameter of 0.6, should be defined as a formal aperture, to facilitate
observation planning, and to minimize confusion in the archive with data taken at
wedge A1.0 (the current location such observations are filed under). To optimize
observations at this location, given the breathing “tendrils,” observers will need to
have dedicated PSF observations, so there will be no cost to the STScI, other than
maintaining the aperture location calibration.
• For optimal reduction of STIS coronagraphic data, where PSF star observations are
used, the PSF data should be as close a color match to the program star as feasible
(and certainly with ∆(B − V ) ≤ 0.08), and obtained on adjacent orbits.
• For STIS coronagraphic observations where the PSF of the star itself will be used in the
data reduction, experience with 2 Eri suggests that multiple spacecraft orientations,
with n ≥ 4, and preferably closer to 9–11 are needed to ensure a median PSF image
which is free of contamination by background galaxies. Fewer observations are needed
in the presence of bright nebulosity (see Heap et al. 2000 for β Pic).
• The practice for coronagraphic observations of bright objects has been to read out only
part of the detector, to minimize detector read overheads during the orbit. Longer
observations, such as are appropriate for 10 ≤ V ≤ 15 T Tauri stars do not benefit
appreciably from this strategy. For such observations, we recommend reading the full
detector, both to give a better view of the environment of the star, and to permit use
of filtering techniques to reduce the read noise.
Acknowledgments. This study has made use of calibration observations obtained as
part of proposals 7088, 8037 8491, 8896, GO-9241, 8925, 8419, GO-8842, GO-9037, and
GO-9136 and parallel observations of HD 95881 obtained on GO-8796. Support for the
analysis under proposal HST-AR-9224 was provided by NASA through a grant from the
Space Space Telescope Science Institute, which is operated by the Association of Universities
for Research in Astronomy, Inc., under NASA contract NAS 5-26555. Support for the
STIS IDT was provided by NASA Guaranteed Time Observer (GTO) funding to the STIS
Science Team in response to NASA A/O OSSA -4-84 through the Hubble Space Telescope
Project at Goddard Space Flight Center. CAG was supported through transfer of funds to
The National Optical Astronomy Observatories. NOAO is operated by the Association of
Universities for Research in Astronomy (AURA), Inc., under cooperative agreement with
the National Science Foundation. Data analysis facilities were provided by the Laboratory
for Astronomy & Solar Physics, at NASA’s GSFC.
References
Brown, T. M. 2001, “STIS CCD Read Noise During Side-2 Operations,” Instrument Science
Report STIS 2001-005 (Baltimore: STScI)
Grady, C. A., Woodgate, B., Bruhweiler, F. C., Boggess, A., Plait, P., Lindler, D. J.,
Clampin, M., & Kalas, P. 1999, ApJ 523, L151.
Grady, C. A., Devine, D., Woodgate, B., Kimble, R., Bruhweiler, F. C., Boggess, A., Linsky,
J. L., Plait, P., Clampin, M., & Kalas, P. 2000, ApJ 544, 895.
Grady, C. A., et al. 2001, AJ 122, 3396.
Heap, S. R., Lindler, D. J., Lanz, T. M., Cornett, R. H., Hubeny, I., Maran, S. P., &
Woodgate, B. 2000, ApJ 539, 435.
Coronagraphic Imaging with HST and STIS
147
Kimble, R. A., et al. 1998, ApJ 492, L83
Krist, J. E., Stapelfeldt, K. R., Ménard, F., Padgett, D. L., & Burrows, C. J. 2000, ApJ
538, 793.
Kuhn, J. R., Potter, D., & Parise, B. 2001, ApJ 553, L189.
Leitherer, C., et al. 2001, STIS Instrument Handbook, V. 5.1 at
http://www.stsci.edu/instruments/stis
Lyot, M. B. 1939, MNRAS 99, 580.
Malbet, F., Yu, J. W., & Shao, M. 1995, PASP, 107, 386.
Mouillet, D., Lagrange, A. M., Augereau, J. C., & Ménard, F. 2001, A&A 372, L61.
Pavlovsky, C., et al. 2001, ACS Instrument Handbook, V. 2.1 at
http://www.stsci.edu/instruments/acs
Schneider, G. 2001, AAS 198, 8301.
Schultz, A. B., Hart, H. M., Kochte, M., Bruhweiler, F., DiSanti, M. A., Miskey, C., Cheng,
K.-P., Reinhard, K., & Schneider, G. 2001, AAS 198, 7707
Stapelfeldt, K. R., et al. 1999, ApJ 516, L95.
Woodgate, B. E., et al. 1998, PASP 110, 1183
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
The STIS CCD Spectroscopic Line Spread Functions1
T. Gull, D. Lindler,2 D. Tennant,3 C. Bowers, C. Grady,4 R. S. Hill,5 and
E. Malumuth5
Laboratory for Astronomy and Solar Physics, Code 681, NASA’s Goddard Space
Flight Center, Greenbelt, MD, 20771
Abstract. We characterize the spectroscopic line spread functions of the CCD
modes for high contrast objects. Our goal is to develop tools that accurately extract
spectroscopic information of faint, point or extended sources in the vicinity of bright,
point sources at separations approaching the realizable angular limits of HST with
STIS. Diffracted and scattered light due to the HST optics, and scattered light effects
within the STIS are addressed. Filter fringing, CCD fringing, window reflections, and
scattering within the detector and other effects are noted. We have obtained spectra
of several reference stars, used for flux calibration or for coronagraphic standards,
that have spectral distributions ranging from very red to very blue. Spectra of each
star were recorded with the star in the aperture and with the star blocked by either
the F1 or F2 fiducial. Plots of the detected starlight along the spatial axis of the
aperture are provided for four stars. With the star in the aperture, the line spread
function is quite noticeable. Placing the star behind one of the fiducials cuts the
scattered light and the diffracted light is detectable even out to 10000 Å. When the
star is placed behind either fiducial, the scattered and diffracted light components,
at three arcseconds displacement from the star, are below 10−6 the peak of the star
at wavelengths below 6000 Å; at the same angular distance, scattered light does
contaminate the background longward of 6000 Å up to a level of 10−5 .
1.
Introduction
The distinctive advantages of Hubble Space Telescope (HST ) are near-diffraction-limited
imaging performance and access to the ultraviolet. The Space Telescope Imaging Spectrograph (STIS) takes advantage of the near-diffraction-limited capability of HST and provides
spectral dispersions ranging from R 500 and 10,000 from 1175–10,000 Å and 30, 000
to 180,000 from 1175 to 3200 Å. The optical design and detector performance of STIS was
carefully matched to science problems that the STIS Instrument Development Team (IDT)
realized could be addressed with high angular resolution and selected spectral dispersions.
We designed the detector formats to utilize the angular resolution of the primary optics.
For the CCD modes (1650–10,000 Å, R 500 and 10,000), the pixel sampling is 0.0504. In
1
2
3
4
5
Based upon observations with the NASA/ESA Hubble Space Telescope, obtained at the Space Telescope
Science Institute, which is operated by the Association of Universities for Research in Astronomy, Inc.,
under NASA contract, NAS 5-26555.
Advanced Computer Concepts
Naval Academy
National Optical Astronomy Observatory
Science Systems Applications, Inc.
148
The STIS CCD Spectroscopic Line Spread Functions
149
keeping with the philosophy of developing the second generation of instruments for HST,
state-of-the-art detector technology was pushed to obtain the best detectors possible for
space observations and numerous spectral modes were installed to provide a range of resolving powers. Late in testing of the CCD modes, we realized that increased transparency
in the near red of the silicon bulk material led to increased internal scatter within the detector and support substrate. This red scatter would complicate spectroscopy, direct imagery
and especially coronagraphic imagery done with the STIS CCD.
Success of STIS is measurable in many ways. With each cycle of competition for HST
observational time, many successful proposals use the STIS. Already, key discoveries include
measurements of black hole masses in the nuclei of many galaxies, and the cores of globular
clusters, of the spectroscopic transit of a planet across the surface of a distant star, of the
lack of planets in globular clusters, of measurements of the Gunn-Peterson effect and of the
Lyman alpha forests, of the first ultraviolet spectra of gamma ray bursters, and of nebular
structures in very close vicinity to bright stars. The HST /STIS has broken many barriers
to ground-based spectroscopy, yet data reduction and analysis continues to be challenging
when we attempt to pull out weak, extended structures close to a bright central source. As
we have learned more and more about the performance of STIS, we have felt encouraged
to push the limits of its capabilities. In this discussion, we present some measures of the
line spread function for the CCD spectroscopic modes as a function of wavelength. In the
future, we hope that software will be developed to enable all users to take full advantage
of the remarkable rejection that STIS provides off axis. More importantly, we hope that
this information on the realized performance of STIS will provide insights for improved
instrument performance of future ground-based and especially space-based instruments.
For the HST /STIS user, many observations can be accomplished routinely. If two
objects are at the classical separation (one full at half maximum separation), then the data
reduction/analysis is relatively straightforward. Here we address optical performance that
must be taken into account when the relative intensities are > 20. With the potential
of reaching statistical S/N > 20, large contrast factors can be addressed. In short this
discussion is in the very important application when a very low extended source, or even a
faint point source, is detectable near a significantly brighter, point source.
2.
Examples of a STIS CCD High Contrast Observations
We start with a spectrum of a K0 star (HD 181204) dispersed by the G750L grating from
5000 to 10,000 Å (Figure 1). The top and middle spectra display the same spectrum with
relative flux scales of 100. The bottom spectrum is of the same star behind the F1 (0.5)
fiducial1 which blocks the core light by nearly four orders of magnitude. The grey scale
for the bottom spectrum is 1/300th that of the top spectrum. Longward of 7000 Å, the
silicon structure of the CCD absorbs less radiation, and the light is reflected within the chip
structures. Diffuse scattering becomes increasingly apparent with wavelength and spreads
across the CCD. The CCD in the near-red behaves much like a Fabry-Perot interferometer,
and develops wavelength-dependent fringes in response to the dispersed light. Properly
executed flat fields can be used to correct the fringed response for objects positioned within
the aperture. Recently, Malumuth et al. (2002) developed a calibration scheme for objective
1
The STIS has a aperture wheel that allows for a selection of optimized apertures to fit the desired scientific
observation. An internal calibration system (WAVECAL) feeds light from a Pt(Cr) lamp to provide reference
wavelengths for wavelength and velocity measures. Positional information is defined by two fiducials (F1,
which is 0. 5 wide, and F2, which is 0. 8 wide) on each long aperture. The aperture wheel encoding permits
very precise placement of the apertures, sufficiently accurate in position, that a stellar image can be blocked
by rotating the aperture into position. The fiducial tests in this paper were performed with the 52 × 0. 2
and the 52 × 0. 1 apertures.
150
Gull, et al.
Figure 1.
Spectra of HD 181204 (MIII) from 5000 to 10,000 Å through the 52 ×
0. 2 STIS aperture. The top and middle spectra are the same spectrum, but with
a grey scale change of 102 . The bottom spectrum is with the star placed behind
the F1 fiducial (0.5 wide), displayed with a grey scale 1/300th that of the top
spectrum. The HST diffracted light pattern is now visible.
grating spectra of objects in the STIS CCD parallel modes (see also Malumuth et al. 2003).
The fringes seen in the bottom image of Figure 1 are not CCD-induced fringes, but the
classical fringes due to diffraction of the telescope optics. In a perfect spectrograph, these
fringes would be the limiting factor in observing faint sources in the vicinity.
The collapsed Line Spread Function (LSF) is plotted in Figure 2 for the same spectra
in Figure 1. The upper trace is of the unobstructed star LSF sampled through the 52 × 0.2
aperture. The middle trace is of the same star, HD 181204, placed behind the F1 fiducial
(0.5 wide) and the bottom trace is with the star placed behind the F2 fiducial (0.8 wide).
The fiducial cuts the wings of the LSF by 102 . We note that this spectral LSF is a linear
approximation of a much more complex function. In the stellar spectrum spread horizontally
across the two-dimensional CCD, each wavelength has a scattering function that spreads
in two dimensions. The narrow aperture at the entrance to the STIS cuts the PSF of
HST to a thin slice that is then modified by each optical element. In the ideal situation,
this modification would simply be the spectral separation of the input light. However each
optical element can contribute scatter or modulation to the light, creating a PSF response
that is wavelength dependent. We find that the major effect is light scattering in the CCD
detector itself and that the scattering becomes more pronounced to the red part of the
spectrum. However to get reasonable S/N measures of the LSF, we had to collapse the data
in the spectral direction to measure the LSF along the cross-dispersion, the angular or the
spatial direction. The reader is cautioned that a wavelength LSF is needed, especially at
wavelengths longward of 7000 Å. Between 3000 and 7000 Å, an averaged LSF, scaled with
wavelength due to HST optics diffraction effects likely will suffice.
Scattered light is a major problem for faint, extended or complex structures along the
aperture. Eta Carinae, with the Homunculus (a bi-lobed nebula plus disk system of ejecta) is
one such complex structure with scales well matched to the angular resolution of HST. Most
of the nebula is a reflection nebula shell with some ionized metals, but neutral hydrogen.
Close to the star is a series of emission structures, including a Little Homunculus (Ishibashi,
et al. 2002). Figure 3 shows an extracted portion of three spectra including Balmer alpha
(G750M, centered on 6768A). Eta Carinae is centered within the 52× 0.1 aperture in the
lower spectrum. In the logarithmic stretch across five decades, a faint ghost is noticeable
The STIS CCD Spectroscopic Line Spread Functions
151
Figure 2. Collapsed Line Spread Function (LSF) for G750L (5000 to 10,000 Å) of
the HD 181204 (MIII) (Figure 1). The upper trace is the LSF for the unobstructed
spectrum. The middle trace is the LSF for the star placed behind the F1 (0.5
wide). The bottom curve is the LSF for the star placed behind the F2 (0.8 wide).
to the lower right of the bright, broadened Balmer alpha emission profile of the star. Faint
nebular emission consisting of narrow Balmer alpha and [N II], [S II] emission lines can be
seen, but the brightness is only slightly above the video pickup, with 4DN amplitude.
The middle and top spectra were recorded with the 52× 0.2 aperture, but with the F2
(0.8width) fiducial rotated to block the star. The middle spectrum (10 seconds CR-SPLIT)
brings out the nebular emission lines well above the video pickup, and the scattered starlight
continuum can be seen, including the broad Balmer alpha emission, complete with a nebular
absorption line. Some video pickup is noticeable. The top spectrum (150 seconds CRSPLIT) goes very deep, but some of the scattered Balmer alpha emission has saturated at
the edge of the fiducial and is bleeding into the nebular portion of the spectrum. Closer
inspection of this spectrum pulls out over twenty unique narrow nebular emission lines, two
of which are [Sr II] detected for the first time (Zethson, et al. 2000).
3.
Known Limitations
As with any spectrograph, each optical element alters the output. Some changes are not
desirable, whether they are diffraction effects, scattered light, faint reflections at every
transmitting surface, stray light, or detector performance. The challenge is to anticipate
these problems and to minimize deleterious effects on the product. STIS is no exception, and
as shown in this paper, we are pushing the instrument capabilities to the realizable limits.
Indeed one major reason for preparing this paper is to document the realized instrument
capabilities and to sensitize designers of future spectrographs to shortcomings that must be
overcome if astronomers wish to study complex systems with even higher contrasts.
Each first order grating has a blocking filter to ensure no second order (blue) leakage
contaminates first order spectra at the red end. Each blocking filter is attached directly to
the grating mount. Collimated light passes through these filters both in the incident and
diffracted beam. A small modulation is detectable especially in the spectra of calibration
emission lamps where the f/ratio of the optical system is very large and the intrinsic line
widths are much narrower than the resolving power of the spectrograph (R 500–10,000).
For astronomical continuum sources the modulation is significantly less than a percent and
for emission line sources, the modulation would be a few percent for intrinsic linewidths
152
Gull, et al.
Figure 3. Spectra of Eta Car and Very Nearby Ejecta from 6400 to 7000 Å,
plotted with a dynamic log scale of 105 . Bottom: Eta Carina in the 52 × 0.1
aperture (exposure 0.3 s). The broad, diffuse structure to the right and just below
the Balmer alpha P-Cygni emission line is a ghost image due to reflections off the
CCD detector window. Middle: Eta Car behind the F2 (0.8) fiducial ( 10 second
CR-SPLIT). Top: 150 second, CR-SPLIT=2 exposure. The stellar Balmer alpha
spills over the fiducial into the extended nebula. Each spectrum extends −2 to+2
from Eta Car. The ghost, seen when the star is within the aperture, is greatly
decreased by the fiducials as the scattered light is suppressed.
approximately 20 km s−1 . Spectroscopy of NGC 7009 yielded a measured peak-to-peak
modulation of 3% for the ratio of [O III] 5007 Å/4959 Å (Rubin et al. 2001).
The ghost image (Figure 3) is of much greater concern. Direct images (Figure 4) show
a faint double ring to one side of each bright star image. The position of these rings moves
relative to the star image. R. S. Hill (63 and 65) analyzed many direct and spectroscopic
images recorded by STIS (Figures 4, 6 and 7). The integrated flux of the double ring is a
few percent of the total flux of the star. Figure 8 shows the path traced by the principal ray
impinging upon the CCD. The CCD surface reflects the focused image, at f/48, back to the
fused silica housing window (necessary to prevent contamination accumulating on the CCD
surface!). Reflections from both window surfaces result in the two out of focus, displaced
images next to each star in Figure 4. Hill’s report (2000) demonstrates that the rings
on direct images are well behaved. Lines, extending through each stellar and associated
pair of ghost images, come to a common region on the detector. Figure 5 reproduces
his measures of the positional displacements. The dispersed spectrum of a star, shown in
Figure 6, demonstrates that the double rings build up a shoulder offset below the stellar
spectrum. The integrated amplitude, which changes with wavelength, is a few percent of
the total stellar flux in the red, but nearly negligible in the 2000–3000 Å spectral region as
an antireflection coating was place on the window for the blue portion of the spectrum.
Quite a different response is noted for WAVECAL spectra (Figure 7). The double ring
phenomenon is not present, but a series of fringes in the approximate positions, expected
for the CCD window ghosts, are seen. A spectrum of the brightest portion of the Orion
nebula was recorded of H alpha and [N II] lines. No fringes were observed at the ghost
positions. Twelve sub-exposures were combined by CALSTIS using a cosmic-ray rejection
algorithm. Detection of the ghost is marginal as the relative brightness per pixel is 10−4
that of the peak emission line brightness. However, WAVECAL exposures provide fringe
The STIS CCD Spectroscopic Line Spread Functions
Figure 4.
Broad-Band Direct Image Recorded by the STIS CCD. Each stellar
image has two ghost rings due to reflections off of the CCD detector housing
window. Note that the reflections relative to the stellar position move about a
point due to the pupil plane being a significant distance behind the detector.
Centering of Ghost 1
Centering of Ghost 2
1000
1000
800
800
600
600
400
400
200
200
0
0
200 400 600 800 1000
0
0
200 400 600 800 1000
Figure 5. Centering of the Two Ghosts within the CCD Detector Format. A
line, drawn through the stellar image and the centers of one of the ghosts, extends
to a common region on the CCD format. With the spectrum dispersed across in
row 511, the ghost is BELOW the spectrum. Were the spectrum placed in row
100, the ghost would appear ABOVE the spectrum.
153
154
Gull, et al.
Figure 6.
Stellar Spectrum with the Ghost. The ghost flux is a few percent of
the stellar flux, but extends over a number of pixels. Typical amplitudes per pixel
are a few parts in 104 of the peak stellar amplitude.
Figure 7. Sample Calibration Spectra and Simulations. Prelaunch external and
internal lamp spectra demonstrate that the ghost position shifts with respect to
the line position along the dispersion and that the HITM ghosts are fringed. The
HITM optical beam is nearly collimated relative to the HST f/24 optical beam.
The STIS CCD Spectroscopic Line Spread Functions
155
Figure 8. Optical Paths of the Incident Light and the Reflected Light. The
ghost images are displaced from the primary spot due to the pupil plane being at
a significant distance behind the detector surface.
amplitudes 10−3 of the peak emission line brightnesses. We realized that WAVECALs
were done through a different, highly collimated, optical train, before the STIS aperture.
Lifetime of a mechanism in a space instrument is always a concern. The STIS shutter
moves in and out of the input light path to prevent light contamination by STIS of any
other instrument operating in parallel. The backside of the shutter has a mirror that feeds
light into the instrument for LAMP calibrations. To minimize the movement of the shutter,
an alternate light path feeds light from the calibration optical train through a permanently
mounted mirror and a hole in the second relay mirror (the relay mirrors correct for optical
aberrations of HST ), and to illuminate the STIS aperture. This mode, labeled HITM (hole
in the mirror ... what else!), is within the shadow of the HST secondary. Thus the HITM
mode feeds light with a very large f/ratio beam into STIS. The lamp intrinsic line widths are
far narrower than the spectral resolving power of the first order gratings. The CCD-reflected
light is reflected back by the two surfaces of the detector window and is modulated. We do
not detect modulation of continuum sources, internal or external, because the wavelength
variation is continuous. Hence we see rings instead of fringes. We confirmed this by a
comparison of a LAMP spectrum to a HITM spectrum. The LAMP mode overfills the
f/24 beam of the STIS collimator. The LAMP spectrum for the Pt(Cr) lamp is noticeably
less that the HITM spectrum. As nearly all astronomical sources have intrinsic line widths
(thermal, turbulence, etc.), the fringing is negligible.
4.
Spectroscopic LSF Observations
To characterize the faint scattering characteristics of STIS, spectroscopic LSF measurements were done of stars placed within a long aperture and then blocked by a fiducial.
Indeed, we realized that characterization of coronagraphic imagery (see Grady, et al. 2003)
could be enhanced by these spectroscopic measurements. The HST point spread function
for a star is a weighted function of the spectral distribution. For a well-behaved panchromatic detector, the HST /STIS response would be expected to be spectral distribution of a
known star weighted by the diffraction-limited HST PSF and sampled by the selected STIS
entrance aperture. However, the STIS CCD has a very significant scattering component
156
Gull, et al.
longward of 7000 Å as demonstrated in Figures 1 and 3. Our desire to quantify the detector
scattering component and to understand the effects on observations led to requesting stellar
measurements performed with proposal 8844.
Four stars, point-like with respect to HST ’s angular resolving power, were selected.
Other criteria included no measurable infrared excess (suggesting possible dust and gas
surrounding the star); reasonable exposure times to record the scattered light passed by
the STIS aperture fiducials F1 (0.5) and F2 (0.8); not be in a crowded field; and good
accessibility, namely close to the orbital pole. STIS has an aperture wheel and a grating
wheel. The selected aperture and grating are rotated into the light path. Internal calibration
spectra are required to determine the spectral position on the detector. A WAVECAL is
recorded along with each astronomical observation. In line identifications of the Weigelt
blob B and D, located close to Eta Carinae, Zethson (2000) found that the measured,
velocity-corrected wavelengths were accurate to about one/fifth of a STIS CCD pixel, based
upon measured versus laboratory wavelengths of about 2000 emission lines. Position along
any long aperture is referenced by two fiducials within each aperture. As the aperture wheel
is very finely encoded, the F1 or the F2 fiducial can be precisely placed in front of the stellar
image. Currently only F2 for the 52 × 0. 2 aperture is supported for routine observations.
However other fiducials have been used. In proposal 8844, we used the F1 and F2 fiducials
for both the 52 × 0. 2 and the 52 × 0. 1 apertures with excellent success.
The most pressing criterion was to accomplish these observations in a reasonable number of orbits. We chose to use all three low dispersion gratings: G750L (5000 to 10,000 Å),
G430L (3000 to 5000 Å) and G230LB (1600 to 3100 Å). As there are approximately forty
primary settings for the M gratings, only the most frequently used M-mode settings were
tested: G750M (8561 Å and 6768 Å), G430M (4961 Å and 3936 Å) and G230MB (2836 Å).
Most grating settings were used with the A0V star, HD 141653, as the exposure times were
reasonable and much could be learned throughout the spectrum. Only a subset of grating
settings were used in the blue for the BD+75D325 (WD) and in the red for HD 115617
(GV) and HD 181204(MIII). The observed combinations are listed in Table 1. The WD,
BD+75D325, was selected because it is a primary standard used in monitoring the sensitivity of STIS. The other three stars were selected because they had proved to be excellent
reference stars for STIS coronagraphic observations (Grady, 2003) and because they spanned
the spectral types from MIII to GV to A0V. Initially, we thought that the reference PSF
of stars of intermediate spectral types could be modelled by a weighted combination of the
PSF’s measured for these stars. However, the best reference PSF’s are those of similar
stars taken during orbits immediately before or after the stars of interest because thermally
induced collimation and focus changes are of greater impact.
Table 1.
Star
BD+75D325
HD141653
”
”
HD115617
HD181204
a
Stars Observed in Proposal 8844 for this Study
SpT
WD
A0V
”
”
GV
MIII
APER
52 × 0.2
52 × 0.2
52 × 0.2
52 × 0.1
52 × 0.2
52 × 0.2
G750L
–
A,F1,F2
–
–
A,F1,F2
A,F1,F2
G750M
–
A,F1 6768
A,F1,F2 8561
A,F1,F2 8561
A,F1,F2 6768
A,F1,F2 6768
G430L
A,F1,F2
A,F1,F2
–
–
A,F1,F2
A,F1,F2
G430M
A,F1 4961
–
A,F1,F2 3936
A,F1,F2 4961
–
–
G230LB
A,F1,F2
A,F1,F2
–
–
A,F1,F2
A
G230MB
A 2836
–
–
–
–
–
For GXXXL(B) grating tests, the grating settings are the default central wavelengths. For GXXXM(B)the
grating settings are listed as the central wavelength in Å. A = star in slit. F1 = star behind fiducial F1
(0.5/arcsec). F2 = star behind fiducial F2 (0.8/arcsec).
Exposure times, which are not listed, were selected to keep the recorded spectra significantly below the 32,000 DN levels with the CCD GAIN = 4. Where the fluxes were
The STIS CCD Spectroscopic Line Spread Functions
157
Figure 9. Comparison of LSF for G750L for HD 181204 (MIII, upper right),
HD 115617 (GV, lower left), and HD 141653 (A0, lower right). Upper left is the
LSF for G750M for HD 181204 (MIII) centered at 6768 Å.
significantly lower, GAIN = 1 was used to keep above video pickup levels close to the star.
As the desired dynamic range is quite large, we rapidly reach flux levels affected by detector
noise, cosmic ray events and even bias shifts across the CCD columns. Indeed some LSF
plots exhibit a pronounced asymmetry from below the star in the spatial direction to above
the star. This is a known bias shift problem. Correcting for it is not a simple matter. We
chose to not correct for the shift as a means of cautioning the observer that it is there.
The line spread functions are presented in Figures 2, 9–13. These are averaged LSFs.
Each spectrum is precisely aligned using WAVECAL lamp spectra and the trace of the
stellar spectrum, or the illuminated edges of the fiducial, when the star is blocked by the
fiducial, F1 or F2. We caution the reader that these LSFs are collapsed along the spectral
dispersion (re-sampled row) direction. For the G750L spectra, the average is taken from
5000 Å to 10,000 Å, and is heavily weighted by scatter beyond 7000 Å. For contrasts
up to 103 , the current data is sufficient, but a wavelength-dependent LSF from 5000 to
10,000 Å will have to be modelled as the measurements do not have sufficient S/N for
complete measure. For Figure 2 and 9–13, the flux along the spatial (cross-dispersion) axis
is logarithmic ranging from 100 to 10−8 . Each CCD pixel subtends 0.0504. These plots
extend from approximately 10 below to 10 above the star, which is centered near row 512.
The top trace is of the star centered in the 52 × 0. 2 aperture and is normalized to the
total measured flux. The LSF drops slowly, but relatively symmetrically. At 10 distance
from the star, the detected flux is ∼ 10−4 of the total flux. The central trace is the LSF
with the star placed behind F1. The flux at the position of the star drops by 104 , and
the off-axis scattered light drops by a factor of 102 . The bottom trace is the LSF with the
star placed behind F2. The scattered light is not affected significantly in the innermost few
arcseconds, but at ∼ 10 , the scattered light is decreased by a factor of two compared to
the scattered light from the F1 fiducial measurement.
The left-hand shoulder is the ghost that we discussed above (Figures 3–6). Placing the
star behind either fiducial drops this shoulder as expected. However, the light that passes
the fiducial, diffracted and scattered light from the telescope, also produces ghosts. As the
signal/noise is excellent, we can see the fainter ghost due to the left edge of the fiducial for
158
Gull, et al.
Figure 10. Comparison of LSFs for HD 114653 (GV) Upper left: G750L Lower
left: G750M (8561 Å) Upper right: G750M (6768 Å) with 52 × 0.2 aperture.
Lower right: G750M (6768 Å) with 52× 0.1 aperture.
F1 and even for F2. The ghost for the right hand portion of the light passing the fiducial
contributes to filling in the area subtended by the fiducial and to the flux on the left hand
side of the fiducial. If the star were in row 100, the ghost would shift above the spectrum.
In the following section, we will intercompare LSFs for different grating combinations
with spectral type. The top curve is always the star in aperture; the middle curve is F1;
and the bottom curve is F2. These fiducials were moved in and out without re-centering
the star between observations. As demonstrated by the symmetry of the stellar core and
the scattered light on both sides of the fiducials, the relative offsets are excellent. Indeed,
success of proposal 8844 is good evidence to extend support to all four fiducials.
5.
LSF Intercomparisons
The LSFs (Figure 9) for the three coronagraphic reference stars are collapsed from 5000
to 10,000 Å. They vary because of the CCD response to the very different stellar spectral
distributions. The AOV star LSF (lower right) drops off faster due to the blue spectral
distribution. Indeed, this LSF, weighted towards 5000 Å relative to 10,000 Å, is similar to
the spectral LSF measured by the G750M grating (upper left) centered at 6768 Å.
Figure 10 compares LSFs for HD 114653 (A0V) for G750L and selected G750M settings.
The G750M (8561 Å) LSF (lower left) is very similar to the G750L LSF (upper left). The
G750M (6768 Å) LSF (upper right) wings drop off faster for the F1 fiducial than the LSF
centered on 8561 Å. The G750M (6768 Å) grating setting was observed with both apertures.
Little difference, other than S/N, was noted.
The G430L LSFs for all four stars are plotted in Figure 11. As expected, scattering
drops off faster than for the G750L LSFs, since the telescope optics diffraction pattern is
sharper and the CCD silicon layer is optically much thicker. Visible-wavelength photons do
not penetrate very deep into the silicon layer. By contrast, the CCD is nearly transparent
to photons near 10,000 Å and the diffraction pattern is twice as large. The G430M LSFs
produced by the four stars are quite similar: the diffraction pattern is close to the core of
The STIS CCD Spectroscopic Line Spread Functions
Figure 11.
159
G430L comparisons for all four stars.
the star; the ghost is easily seen on the left shoulder. Knowledge of the bias level is now a
significant problem as demonstrated by the asymmetry of scatter above and below the star.
Figure 12 plots three measures of the G430M LSF in comparison to the G430L LSF.
S/N is an issue as scattered/diffracted flux drops down to levels marginally detectable with
32000 DN encoding range. Determining the bias level is also a problem. BD+75D325 LSF
measures are strongly limited by detector background. The ghost signal appears stronger
for the G430M (4961 Å) setting than for the 3936 Å setting due to the antireflection coating
applied to the window for the blue optimization.
While several LSFs were measured for the G230LB and G230MB gratings, they differed
only in S/N. Figure 13 shows the best LSF measures for BD+75D325. The ghost on the
shoulder is significantly weaker than for spectra to the red. Scatter measurements in all three
LSFs are limited by the finite DN range integration time. The three LSFs match so well
that we are tempted to combine all three. The fiducials in the G230LB and G230MB modes
primarily attenuate the core of the telescope PSF. The STIS CCD does not contribute a
significant scatter component as the blue photons are absorbed at the surface of the silicon.
Figure 14 brings together the contribution to the background as a function of wavelength for an angular displacement offset by three arcseconds from the star in the +Y direction. The three curves are for the star in the aperture, behind the F1 fiducial and then
behind the F2 fiducial. Compared to the total normalized flux, the scattered/diffracted
starlight contribution is below a few parts in 105 from 2000 to 6000 Å. Longward of 6000 Å
the scattered light component becomes noticeable, primarily due to scatter in the STIS
CCD. It climbs to a level of ∼ 10−3 at 10,000 Å. With the star placed behind either fiducial, the scattered component drops 100-fold at 10,000 Å and is below 10−6 shortward of
7000 Å. The rise in scattered light shortward of 3000 Å is consistent with measured HST
light scatter due to roughness of the mirror surface. Finally we wish to point out that
the LSF measures in this discussion are really a linear approximation to a scattering phenomenon that must radially transmit through the CCD chip in a fairly random pattern.
This pattern will depend on the details of the CCD fabrication: how thick and uniform
is each layer of etched circuitry and the uniformity of sensitivity. From Malumuth et al.
(2002) we learned that the CCD chip is wedge-shaped and that the sensitivity fringes are
distorted by this apparent shape. A proper model of the detected scattered light would also
160
Gull, et al.
Figure 12.
BD75+325.
100
G430L and G430M comparisons for A0V HD141653 and WD
G230LB 2375 BD+75D325
10-2
51195: 52X0.2
51197: 52X0.2F1
51199: 52X0.2F2
10-4
10-6
10-8
300
Figure 13.
400
500
starplot/51195.ps
600
G230L LSF comparisons for WD BD75+325.
700
The STIS CCD Spectroscopic Line Spread Functions
161
Figure 14.
The Measured Scattered Light Component as a Function of Wavelength at a Position Offset 3 from the Star in the cross dispersion direction on
the CCD. Abscissa: alog10(flux) relative to the total flux. Top curve: scattered
light with the star centered within the STIS 52 × 0. 2 aperture; bottom curves:
measures of the scattered flux with the star is positioned behind F1 or F2.
have to take into account the smoked support glass that supports the CCD chip within the
housing, necessary to help the CCD survive during the launch vibrations.
6.
Conclusions
This discussion describes the instrumental scatter of STIS in combination with HST. We
observed four stars ranging from a WD to MIII in spectral type, working in coordination
with coronagraphic observers. Future work is intended to define a tool for subtracting the
ghost feature, and possibly defining a wavelength-dependent LSF.
Acknowledgments. We are grateful to the Space Telescope Science Institute STIS
Team, their contribution of observing time and positive encouragement in reducing and
interpreting the data. One of us (Don Tennant), as a volunteer summer student from the
United States Naval Academy, provided much of the data reduction used in this paper. Mr.
Keith Feggans assisted in preparation of the figures for this publication.
References
Grady, C. A. 2003, this volume, 137
Hill, R. S. 2000–2001, STIS Postlaunch Quick Look Reports, No. 63 and 65
http://hires.gsfc.nasa.gov/stis/postcal/quick reports./quick reports.html
Ishibashi, K., et al. 2002, AJ, submitted
Malumuth, E., et al. 2002, PASP, accepted
Malumuth, E., et al. 2003, this volume, 197
Rubin, R., et al. 2001, MNRAS, 334, 777
Zethson, T. 2000, PhD. Dissertation, Lund University, Sweden
Zethson, T., Gull, T., et al. 2000 AJ, 12, 34
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
FOS Post-Operational Archive and STIS Calibration Enhancement
M. R. Rosa,1 A. Alexov, P. D. Bristow, F. Kerber
Space Telescope European Coordinating Facility, European Southern Observatory,
D-85748 Garching, Germany
Abstract. The first part of this report summarizes the scientific quality enhancement for FOS/BL data—the Post-Operational Archive project. Part two describes
the status and planning for the follow up, a calibration enhancement for the 66,095
data sets obtained with STIS between 1997 and 2001 when it was operated using
“Side 1 electronics.”
1.
Introduction
When in 1999 the Memorandum of Understanding (MoU) between NASA and ESA about
funding and specific contributions towards the continuation of the multi-agency HST project
had to be renewed, it became clear that there was little room for hardware contributions
from ESA’s side. The agencies agreed that the additional ESA contributions will instead
concentrate on the “users end,” for our purposes here in particular the science data archives
and the science data calibration status.
As a result, the “Instrument Physical Modeling Group” was created at the ST-ECF
with 3 additional staff from ESA HST funds. In October 1999 we could start with the first
project, the “Post-Operational Archive for FOS,” concluded at year’s end in 2001. Since
January 2002 we are now fully engaged in the “STIS Calibration Enhancement” (STIS/CE)
project. In the following status, results and plans for these two projects, implemented under
NASA/ESA MoU, are present with emphasis on the science end user’s point of view.
2.
Scopes and Interfaces
The prime responsibility for HST science operations rests with the STScI, including the
dissemination and archival of data in a form useful for the science end user. It has always
been a key objective for these activities at STScI to ensure a very high standard of calibration
on the data. This high quality default calibration was (and still is) to be available not weeks
or years after the actual observation took place, but right from the moment the raw data
are received on the ground and subjected to the pipeline for the first time.
At the occasion of the first HST calibration workshop in the new millennium it should
be noted that the staff at STScI, with the support from IDTs and others, has done a
very good job on maintaining the high standards for the “instantaneous” calibration, even
under the adverse conditions of responsibility for operations, planning and dissemination
of products in almost real time. Naturally, this goal could only be reached by investing in
an 80/20-like fashion, that is, assuring a good level for almost all data at all times from all
frequently used modes, rather than concentrating on a few specific or difficult modes at a
1
Affiliated to the Space Telescope Division, Research and Space Science Department, Directorate of Science,
European Space Agency
162
FOS Post-Operational Archive and STIS Calibration Enhancement
163
particular epoch. Also, the operational requirement to make available calibration solutions
“momentarily” did not help in understanding long term trends.
The de-commissioning of the first generation of HST instruments, in particular the
FOS, offered the ideal case to take a break and to view the entire set of data, calibration
observations and science exposures as well, as a coherent set. On the firm basis of an
existing, usually quite well documented calibration pipeline system the key objectives for
POA/FOS project were then initially defined by
1. A comprehensive review of the calibration (documentation, calibration data, suitable
science data in comparison to expectations)
2. Analysis of the trends, isolating slowly varying instrumental effects from HST environmental ones (long baseline for trending)
3. Selection of the calibration areas with best ratio of scientific value per effort for follow
up
4. Implementation of superior solutions into the pipeline
5. Release of re-calibrated data and documentation, user support
Ultimately the recipients of the products should be the scientific users world wide. And,
while STScI on behalf of NASA is formally the primary customer, it can also act efficiently
as a redistribution center both ways, products and user inquiries. In order to maximize
the benefits and to minimize the overhead close collaboration and exchange of information
between ST-ECF and STScI were agreed upon as well and set up as required. Reporting is
annually to the ST-ECF Users Committee and at the ESA/ESO Annual Review of the STECF on this side of the Atlantic. Progress reports are usually also included in the regular
meetings of the STScI Users Committee.
3.
A New Paradigm
From the outset it was clear that our group with a nominal power of 2.5 FTE would never
be able to repeat (even partially) the work of many that had covered the calibration of an
instrument like FOS as members of the IDT, as participants in the laboratory testing before
launch, as instrument scientists and archive specialists at STScI. In order to improve the
calibration of the entire set of archived data of the FOS at a scientifically significant level
we therefore had to choose a paradigm quite different from the canonical, empirical one.
Our approach then rests on the idea that a large (well above 90 percent) fraction of the
calibration relations is very well described by functions that are either directly derived from
first principles or well known laws from physics textbooks. The description of dispersion
relations through trigonometric functions from physical optics, with meaningful parameters,
rather than by fitting of unspecific low order polynomials may serve as an example.
The procedure then is to replace a multitude of isolated empirical corrections (instrumental signatures) available for the various modes, environmental conditions and epochs
by a physically correct chain of transformations that are motivated by insight into the engineering and physics of the instrument, its optical setup and its detectors. Early successes
in that direction include for example the scattered light model for the FOS, described in
the Appendix C (M. Rosa) of the FOS Instrument Handbook, (Keyes et at. 1995) and in
Rosa (1994).
4.
Post-Operational Archive for the FOS
The review and subsequent improvement of the calibration of the FOS archival data started
in 1999 and was concluded with the final release of the upgraded pipeline software (ver-
164
Rosa, et al.
sion 1.2.1), necessary reference data and on-line documentation in November 2001 (see
http://www.stecf.org/poa/fos/, and Alexov et al. 2002). The data actually corrected concern all spectrophotometric observations FOS/BL mode. Excluded are FOS/BL spectropolarimetry and FOS/RD all modes (see also the last subsection further down).
4.1.
Wavelength Calibration—Zero Points
In agreement with STScI we focused in 1999 on known and suspected issues of the wavelength calibration. User reports and the analysis of data obtained annually for the check
of the dispersion solutions pointed towards problems with the wavelength calibration in
all modes. The findings included inconsistencies of radial velocity measurements between
repeated exposures, between different wavelength ranges, between HST FOS data and observations from the ground and even within the wavelength range visible in certain high
resolution modes.
Even without the user driven suspicion towards the dispersion solutions and zero points
it was clear to us that a close review of the wavelength calibration and its stability over the
entire FOS lifetime was of high interest. This because the dispersion solution reference data
(polynomial parameters) employed by the calfos pipeline had remained unchanged since
1992. Indications of large, albeit random-like wavelength zero point shifts from trending
analysis while FOS was still operational had always been explained with the so-called “filtergrating-wheel un-repeatability.”
We undertook a mass analysis of all internal wavelength calibration lamp observations
available, dedicated ones planned by FOS instrument scientists and those taken by GOs in
connection with science exposures on targets. In total we had some 1800 such observations
spread over the entire lifetime of FOS (1990 to 1996) at our disposal. The raw data were
cross-correlated with those that had formed the basis for the dispersion coefficients installed
in CDBS.
Soon it became clear that there were rather large uncertainties in the zero points of
the wavelength scale. On the blue channel modes there were very well defined almost linear
trends with time amounting to an offset of 4 pixels for data taken in late 1996. On the red
channel modes the situation was much more complicated, presenting a seemingly random
scatter −2 to +6 pixels throughout the entire lifetime (see Rosa et al. 1998).
In the end the in-famous geomagnetic image motion problem (GIMP) noted in September 1990, and the attempts to correct it on-board after April 1994 were found to be the
primary source of the scatter in the wavelength scale zero-points. The long term steady increase in zero point deviation for the FOS/BL data could be identified with the side-effects
of the repeated adjustment of the Digicon image focusing parameter YBASE during operational life, and with the general increase of the HST aft-shroud temperature. Details can
be found in the POA/FOS Technical Reports #1 and #2 (Rosa et al., 2002a, b) available
from the above mentioned web site.
Correction for the blue side FOS data was achieved using a model that combined the
detector physics (Digicon magneto-electrical focusing) with highly accurate models of the
HST orbit (from NORAD) and the geomagnetic field (from GSFC). This code segment was
implemented into a POA/FOS modification of calfos (poa calfos) released in August 2000.
4.2.
Wavelength Calibration—Dispersion Relations
After correction of the GIMP/YBASE/Temperature related drifts and shifts of the wavelength zero points (really the location of the photocathode image on the diodes), it was
possible to analyze the dispersion relations and their variations in shape for the different
epochs. Not surprisingly—the FOS was built according to high standards—it was found
that the shapes did NOT vary at all. The alleged variation of dispersion solutions had in
fact been the zero point shifts masked by the use of unspecific low order polynomials.
FOS Post-Operational Archive and STIS Calibration Enhancement
165
Figure 1. Residuals of the calfos dispersion relations (different shades are different high resolution modes). Although most of the data points are within the
nominal error limit of a pixel (= 1/4 diode indicated by horizontal bars), there is a
clear structure common to all modes—actually the S-Distortion not fully matched
by the 3rd order polynomials. In addition, several lines of the original line lists
are either blends or have bad identifications.
During the review of the dispersion solutions we also found that the scarcity of useful,
unblended lines in almost all of the FOS wavelength ranges combined with the use of low
order polynomial solutions led to unphysical behavior of those solutions at the wavelength
range limits (run-away).
Accordingly, STPOA poa calfos release version 1.1 in May 2001 included an entirely new
wavelength calibration module based on a physical optics model of the FOS spectrograph.
This model is a derivative of a generic spectrograph model described by Ballester and Rosa
(1997), and contains only engineering parameters such as focal lengths, grating constants
and configuration angles. The solution implemented for the FOS/BL high resolution modes
represents the best set of such parameters for those components that are common between all
modes (collimator, overall configuration, camera). It also includes the effects of S-Distortion
in the Digicon detector. For each of the high resolution modes the internal accuracy of the
solutions are a factor 5 to 10 improved over the original polynomial solutions.
4.3.
Update of Flats and Dark Correction
Subsequently the entire suite of flat fields needed to be redone, because the FOS/BL raw
data were now adjusted in zero point location. We used the occasion to investigate areas of
improvement for the flat fields, but concluded that any additional improvement was visible
only for exposures of exceptional total length and signal to noise ratios. The only set of
such exposures in the FOS/BL archive are, however, the standard star observations made
for the purpose of deriving flat fields.
A noticeable improvement was possible for the dark correction module of the pipeline.
As noted earlier (Rosa 1993), the scaling of darks in the calfos pipeline was insufficient at the
extremes of geomagnetic latitudes traveled by HST. Since the new pipeline poa calfos had
already been augmented with a module that predicted HST orbital location and geomagnetic
parameters to a much higher accuracy than previously calfos, we were able to scale the darks
using a physical connection with the energetic particle density through the so-called Shellparameter L. This new dark scaling and correction of an error in the scaling algorithm in
166
Rosa, et al.
Figure 2.
Residuals from the FOS Dispersion model, same total number of lines
used. Improvement by a factor of 5 to 10 is mostly due to the generic inclusion
of the S-Distortion and the ability to select against blends or misidentifications
without significant change of the shape of the solution.
the pipeline itself were implemented into the final release of STPOA in late 2001 (Bristow
et al. 2002b).
Additional dependencies of the dark rate on solar cycle or day-night differences could
be seen in the data, but the large spread of actual dark measurements at any given geomagnetic location and epoch did not warrant any further modeling. In any event, the residual
uncertainty in dark scaling is only relevant for the faintest of all targets. However, for such
observations the only trustworthy dark measurements could have been obtained only by
simulating a near-simultaneous sky or dark observation through use of a 50/50 cycle rapid
beam switching in the Digicon—actually never performed.
4.4.
And What about FOS/RD Data...
The entire suite of successful improvements made for the FOS/BL data archive rests on
the ability to correct the GIMP/YBASE/temperature induced zero point shifts of the raw
data in the first place. In the case of the blue side FOS Digicon the magnetic shielding was
near spec, so that the GIMP (and its, albeit wrong, on-board compensation) were small
compared to the other two sources of error.
However, the magnetic shielding of the red side detector was so poor (factor 8 less than
on the blue side), that the GIM and the subsequent on-board compensation are significant
nominally 4 pixel max each. The seemingly stochastic variation of the zero-points on the
red side from −2 to +6 pixel exceeded this range about twice (see Rosa et al. 1998).
The suspicion, that the on-board GIM compensation was sometimes out of phase is—
very probably—the answer. Probably, because our deep investigation into the convoluted
sequence of scheduling software, telemetry up, execution of the on-board GIM segment and
telemetry down did lead nowhere. Actually it led into a US Government storage facility
near Bowie, MD, where some 10,000 reels of 16 inch magtape from VAX resided—to be
destroyed within a month’s time (that was in June 2000). More details on this sad outcome
can be found in Bristow et al. (2002c).
In principle the code we produced for FOS/BL can be used to analyze some 1000
wavelength calibration data from FOS/RD and to investigate whether a suitable and meaningful scaling of the parameters might explain the zero point shifts in a homogeneous and
FOS Post-Operational Archive and STIS Calibration Enhancement
167
Figure 3. MWG halo absorption line velocities measured in data from two successive orbits as calibrated in the original FOS archive. Note the unphysical slope
(wavelength dependency), and offset relative to the values expected from 21 cm
line data bracketed by the dashed horizontal lines.
comprehensive manner. Our peers and the internal review of plans between STECF and
STScI however felt that we should now concentrate on the 66,095 data sets from STIS side 1
instead (see below)—and this is why there will likely never come a POA-like improvement
to the FOS/RD part of the archive.
4.5.
Science Verification of POA/FOS
Finally, the improved calibration of the FOS/BL data was verified on science data in two
ways. The consistency of the wavelength calibration across several adjacent wavelength
ranges was shown using the emission lines of several LMC planetary nebulae (Kerber et al.
2002).
The full impact of the POA/FOS correction of zero-points and dispersion solutions
can be judged from a “before and after” comparison shown in Figures 3 and 4. Galactic
halo absorption lines seen against the background of quasar continua have been measured
in data sets calibrated with both, the standard calfos and with the new poa calfos pipelines
respectively. Shown are data from two successive orbits denoted by symbols of different
grey-scale. The horizontal lines indicate the range of radial velocities expected from the
line of sight velocity distribution in the H2 column. The old calibration resulted in a
velocity distribution that was (a) not commensurate with the 21 cm data, (b) displaying
an unphysical correlation with wavelength, and (c) not even consistent from one orbit to
the next. Surprisingly, a reasonably strong absorption line at 140 nm, if identified as a
prevalent feature seen in many lines of sight through the halo, seems to be the only one
matching up with expectation.
The measurements of the same data sets, now from the POA/FOS archive, dramatically
demonstrate the science potential for archival research with the improved FOS data. The
measurements are now in agreement with expectation, from one orbit to the next as well,
and independent from wavelength. In addition, the “mystery” line can now clearly identified
168
Rosa, et al.
Figure 4. Same raw data as Figure 3, now calibrated with the POA/FOS
pipeline. The feature at 140 nm is definitely not a MWG halo absorption, but
Lyα from the quasar host galaxy at the redshift indicated by the vertical line.
with Lyα absorption in the quasar host galaxy at the expected redshift (indicated by the
vertical line in Figure 4).
4.6.
User Support
Upon delivery and release of STPOA v1.2.1 (August 2001), the STECF assumed full responsibility for the FOS user support world wide. Users entering the STScI HST help pages
are transparently routed through to http://www.stecf.org/poa/fos/, and mail links are set
to our POA help desk at ecf-poa@eso.org.
5.
STIS Calibration Enhancement—STIS/CE
Since January 2002 we are engaged in a follow-up project that will bring to the STIS data all
the insight gained with implementing the new paradigm for FOS. Since STIS is not a postoperational instrument yet, extra care has to be taken to not interfere with the operational
pipeline and calibration data base. Hence, the STIS Calibration Enhancement (STIS/CE)
project has as its objective the archive of STIS data obtained between the commissioning
after Servicing Mission 3 (March 1997) and the switching of electronics to Side 2 (July
2001). This STIS Side 1 archive comprises 66,095 data sets.
Between January and March 2002 we collected and digested a very near complete pile
of relevant STIS documentation from STScI, from the IDT archives at GSFC and from the
manufacturer (Ball Aerospace). In April we discussed the resulting 4-year-plan with the
STScI and presented it to both the STECF and the STScI Users Committees. The plan is
a phased approach to the key areas of STIS calibration for which a significant improvement
of science data can be expected from our paradigm of reducing calibration to understanding
an instrument and list environment in physical terms.
FOS Post-Operational Archive and STIS Calibration Enhancement
5.1.
169
Preparations
Phase 1 (2002) foresaw the consolidation of a documentary archive, additional hardware
installation to mass process the 66,095 data sets repeatedly, and to archive raw and processed data in an on-line storage for analysis. By mid-2002 we had installed a 5 TB raid
array on a SUN Fire 280R machine with two 900 MHz CPUs, supported by the over-night
processing power of additional 4 SUN Blade 150 Workstations. By late 2002 we have now
in place the complete raw data archive of all STIS Side 1 data processed by STIS OTFR
Opus v 1.4.2. This will serve as the reference against which improvements are verified (to
be updated as necessary).
5.2.
Geometry
As was the case with FOS, paramount to this approach is a firm basis of the instrumental
geometry for all epochs of data taking, both for imaging and for spectral long slit or echelle
data. The pipeline ultimately will contain a module that comprises geometric distortion,
2D spectrometric imaging and wavelength assignment in a coherent fashion. Phase 2 (2002–
2003) calls for the implementation of the 2D spectrograph model (Ballester & Rosa, 1997),
which has been working successfully already for the UVES Echelle spectrograph on the ESO
VLT. Testing of the analysis part of the code has been very successful by June 2002 already
for both, the FUV and the NUV echelle modes. As an example for the E140H FUV mode
we predict the location wavelength calibration lines to better than 0.3 pixel (chi-square for
400 lines) from first principles across the entire frame (2 by 2 K).
In support of this Phase 2 (geometry and wavelengths) several calibration lamps (vintage STIS flight spares and newly produced) have been observed with the UV vacuum
spectrograph at NIST to obtain highly accurate laboratory list of the Pt, Cr and Ne spectra as observed in orbit. This is necessary because the laboratory list currently in use by
the default STIS pipeline are not based on any observation of a STIS vintage lamp and, in
particular do not have a single entry for the abundant lines of Cr.
Also in support of the geometry phase we have begun to build up a physical CCD
readout model that will be able to correct the raw data for the effects of CTE deficiencies.
It is a well known fact that the loss and redistribution of charge during readout is not only
a nuisance for photometric applications, but also leads to geometric distortion of the data
in an illumination/scene dependent manner. The CCD readout model will serve also during
Phase 4 (flux calibration).
5.3.
Trending and Orbital Environment
Phase 3 (mid-2003 to mid-2004) will make use of the STIS internal geometry corrected data
to assess the impact of orbital environments (MSM repeatability, aft-shroud temperature
variations, breathing of the telescope, etc.). The analysis of the data will also have an impact
on the predictability of dark frames, hot pixels and other items related to photometry.
5.4.
Flux Calibration
After conclusion of Phase 3 the data processed by the STIS/CE Side 1 pipeline will be ready
to establish a new system for flux calibration. Improvements of the geometric stability
and predictability will be used to address again the optimal extraction of spectral data,
this time aided by the predictions from the 2D echelle or long slit model—particularly
beneficial for faint targets or deep absorption troughs. Phase 4 then will include the review
of flux calibration items (e.g., flats, darks, extraction, inverse sensitivity, throughput and
vignetting).
170
Rosa, et al.
5.5.
And what about STIS Side 2...
Between the STIS project at STScI and the STIS/CE project we have a close collaboration
and information exchange, supported and regulated by several agreed upon documents. In
order to protect the operational STIS pipeline at STScI from the experimental versions of
our STIS/CE pipeline there will not be a direct port of modules and builds into calstis.
However, the collaboration stipulates to exchange all insight gained into STIS calibration and functioning at the earliest possible date. For example it will certainly be possible
to upgrade the operational calstis to the improved geometry and wavelength calibration
once these modules have been tested extensively for STIS/CE of Side 1 data, if the STScI
STIS project opts to do so. In any case it will be necessary to evaluate the situation—i.e.
the evolutionary stage of both pipelines, STIS Side 1 ce calstis and STIS Side 2 calstis—at
around 2004 and to discuss possibilities for mergers for the benefit of all STIS users.
References
Alexov, A., Bristow P. D., Kerber, F., & Rosa, M. R. 2002, Re-processing of HST FOS
Data, POA/FOS Technical Report 2002-08, STECF
Ballester, P. & Rosa, M. R. 1997, A&A Rev., 126, 563
Bristow, P. D., Alexov, A., Kerber, F., & Rosa, M. R. 2002b, POA Investigation of FOS
Dark Correction, POA/FOS Technical Report 2002-07, STECF
Bristow, P. D., Alexov, A., Kerber, F., & Rosa, M. R. 2002c, Tracking FOS on-board GIMP
in commanding, AEDP telemetry and header contents, POA/FOS Technical Report
2002-02, STECF
Kerber, F., Alexov, A., Bristow P. D., & Rosa, M. R. 2002, POA/FOS Science Verification,
POA/FOS Technical Report 2002-06, STECF
Keyes, C. D., Koratkar, A. P., Dahlem, M., Hayes, J., Christensen, J., & Martin, S. 1995,
FOS Instrument Handbook, v.6.0, (Baltimore: STScI)
Rosa, M. R. 1993, in Calibrating Hubble Space Telescope, J. C. Blades and S. J. Osmer
(Baltimore: STScI), p. 190
Rosa, M. R. 1994, The FOS Scattered Light Model Software, CAL/FOS-127, (Baltimore:
STScI)
Rosa, M. R., Alexov, A., Bristow, P. D., & Kerber, F. 2002a, GIMP and YBASE Induced
Zero-Point Shifts in FOS Data, POA/FOS Technical Report 2002-01, STECF
Rosa, M. R., Alexov, A., Bristow, P. D., & Kerber, F. 2002b, Physical Model FOS Dispersion
Relations, POA/FOS Technical Report 2002-04, STECF
Rosa, M. R., Kerber, F., & Keyes, C. D. 1998, Zero-Points of FOS Wavelength Scales,
CAL/FOS-149, (Baltimore: STScI)
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
Accuracy and Precision of Measuring Emission Line Velocities with
the Space Telescope Imaging Spectrograph
Thomas R. Ayres
University of Colorado (CASA), Boulder, CO, 80309
Abstract. I describe some of the issues connected with measurements of emission
line velocities in STIS spectra, primarily in the key E140M band. These issues are
important not only in studies of the magnetodynamics of stellar outer atmospheres,
but also to gain insight into ways of bootstrapping calibrations onto solar FUV
instruments, which typically have avoided internal wavecal lamps (like those flown
in all of the HST spectrometers) in favor of using in situ spectral ”standards” (such
as the average velocity of weak chromospheric emission lines to set the zero point
offset). I address the issue of accuracy by comparing apparent emission line radial
velocities, as measured by STIS in the FUV, with high-quality optical measurements
of photospheric spectra, for a large sample of late-type stars. I address the issue of
precision by conducting a series of numerical experiments to simulate Gaussian line
fitting in the presence of Poisson noise. I also discuss generalization of these principles
to the next generation HST spectrometer, the Cosmic Origins Spectrograph.
1.
Introduction
I discuss some of the issues connected with measuring emission line velocities in HST STIS
spectra, focussing on the widely used E140M band (1150–1710 Å). These issues are important not only in studies of the magnetodynamics of late-type stellar outer atmospheres, but
also to gain insight into ways of bootstrapping calibrations onto solar FUV instruments,
which for cost reasons typically have avoided internal wavecal lamps (like those flown in
all of the HST spectrometers) in favor of using in situ spectral “standards” (such as the
average velocity of weak chromospheric emission lines) to set the zero point offset of the
wavelength scale. I am conducting these studies as part of the development of an extensive
catalog of STIS ultraviolet spectra of late-type (“cool”) stars: COOLCAT.
Table 1.
Star
Name
ζ Dor
χ1 Ori
κ Cet
τ Cet
ξ Boo A
2 Eri
AU Mic
AD Leo
EV Lac
STIS Velocities: DWARFs
Sp Typ
Lum Cl
F7 V
G0 V
G5 V
G8 V
G8 V
K2 V
M0 V
M3.5 V
M3.5 V
V
(mag)
+4.72
+4.41
+4.83
+3.50
+4.55
+3.73
+8.61
+9.43
+10.06
υrad
∆υ
υSTIS
−1
←− (km s ) −→
+ 0.5 ± 0.2
−2.0
+2.5
−15.6 ± 0.3 −13.5
−2.1
+18.4 ± 0.1 +19.9
−1.5
−17.1 ± 0.1 −16.4
−0.7
+ 1.5 ± 0.2
+3.0
−1.5
+18.0 ± 0.2 +15.5
+2.5
− 4.4 ± 0.2
+1.2
−5.6
+12.6 ± 0.2 +10.8
+1.8
+ 0.5 ± 0.6
−1.5
+2.0
< ∆υ > ± 1 σ =
171
−0.1 ± 2.1
NOTES
dMe star
dMe star
dMe star
dG + dK, only
172
Ayres
Figure 1. Sample measurements, using a semi-autonomous Gaussian fitting algorithm, of selected C I emissions in an E140M spectrum of the K dwarf 2 Eridani.
Table 2.
Star
Name
υ Peg
31 Com
35 Cnc
HR 9024
24 UMa
µ Vel
ι Cap
β Cet
α Boo
α Tau
α TrA
β Aqr
β Cam
2 Gem
STIS Velocities: GIANTs and SUPERGIANTs
Sp Typ
Lum Cl
F8 III
G0 III
G0 III
G1 III
G4 III
G5 III
G8 III
K0 III
K1.5 III
K5 III
K2 II
G0 Ib
G1 Ib
G8 Ib
V
(mag)
+4.40
+4.94
+6.58
+5.90
+4.57
+2.72
+4.30
+2.04
−0.04
+0.85
+1.92
+2.91
+4.03
+3.02
υrad
∆υ
υSTIS
←− (km s−1 ) −→
− 6.2 ± 1.1 −11.1
+4.9
+ 1.1 ± 1.8
−1.4
+2.5
+42.4 ± 2.3 +36.0
+6.4
− 2.6 ± 0.3
+0.7
−3.3
−26.8 ± 0.3 −27.2
+0.4
+ 6.8 ± 0.3
+6.2
+0.6
+12.1 ± 0.2 +11.5
+0.6
+13.4 ± 0.2 +13.0
+0.4
− 4.3 ± 0.2
−5.2
+0.9
+54.2 ± 0.2 +54.3
−0.1
− 3.4 ± 0.3
−3.3
−0.1
+ 7.1 ± 0.3
+6.5
+0.6
− 0.8 ± 0.6
−1.7
+0.9
+ 6.8 ± 1.0
+9.9
−3.1
< ∆υ > ± 1 σ =
+0.4 ± 0.4
NOTES
high
high
high
high
υ sin i
υ sin i
υ sin i
υ sin i
broad-lined
broad-lined
broad-lined
narrow-lined, only
STIS Emission Line Velocities
173
Figure 2. Summary of velocity measurements in the sample. The target features
all are narrow chromospheric lines. The solar-type star α Cen A shows that such
features fall to better than 1 km s−1 of the expected stellar radial velocity (in this
case based on the well-determined orbit of the α Cen system).
2.
Accuracy
I addressed the issue of accuracy by comparing apparent emission line velocities, as measured
by STIS in the FUV, with optical determinations of radial velocities, for a sample of nearly
thirty late-type stars. Figure 1 depicts the semi-autonomous measuring procedure for a
group of C I lines in a representative K-type dwarf. Figure 2 summarizes average emissionline velocities of the targets, based on selected narrow low-excitation chromospheric emission
lines, ostensibly free of blends and optical depth effects. Tables 1 and 2 compare the STIS
velocities (±1 s.e. [standard error] of the mean) with υrad s from SIMBAD (the radial velocity
material, unfortunately, is somewhat inhomogeneous) for the two dozen or so single (or wide
binary) stars of the sample. The absolute velocity accuracy of a typical STIS pointing on a
bright late-type star, based on the standard deviations, appears to be better than ±2 km s−1 ,
with an uncertain contribution due to the optical υrad s themselves.
174
Ayres
Figure 3.
Simulation of Gaussian fitting process for emission features governed
by photon statistics (on a negligible background, in this case). The distribution
functions (lower three panels) were obtained from the results of 105 , or so, trials at
each of several S/N levels (top panel illustrates representative line profiles). “x” is
the wavelength displacement in pixels, and “stn100 ” is a measure of the S/N relative
to the N = 100 counts case. The “normalized” quantities on the abscissa allow
a scale-independent comparison of the distribution functions for different FWHM
cases. Here, a FWHM= 2 pix simulation is shown. The dot-dashed horizontal
lines mark the probability of a (two-sided) 1 σ deviation. The lower horizontal
lines indicate 2 σ (darker) and 3 σ (lighter) deviations.
STIS Emission Line Velocities
3.
175
Precision
I addressed the issue of precision by examining the internal consistency of the emission line
measurements within the spectrum of a given star: see, again, Figures 1 and 2, Tables 1
and 2. The internal consistency of line positions appears to be extremely good, limited
largely by the photon statistics of the measurements themselves. I routinely am seeing
sub-km s−1 standard deviations in typical cool star emission spectra.
These two exercises rely upon reasonable assessments of the errors incurred in fitting
narrow emission lines with, say, a least-squares Gaussian algorithm. I have re-examined
this question by conducting a series of numerical experiments to simulate Gaussian line
fitting in the presence of Poisson noise: see Figure 3. These simulations lead to a series of
scaling laws to describe estimated 1 σ , 2 σ, and 3 σ two-sided confidence intervals for the
key Gaussian parameters—centroid wavelength λ0 , full-width at half maximum intensity
FWHM, and integrated line flux fL —as a function of the S/N of the flux measurement (i.e.,
√
N for counting statistics). Complete results will appear in a future publication.
Acknowledgments. Supported by HST archive research grant AR–09550.01–A.
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
Modelling Charge Transfer on the STIS CCD
P. Bristow, A. Alexov, F. Kerber and M. Rosa
Space Telescope European Co-ordinating Facility, ESO, Karl-Schwarzschild-Str. 2,
D-85748, Garching bei München
Abstract. The Calibration Enhancement effort for the Space Telescope Imaging
Spectrograph (STIS) aims to improve data calibration via the application of physical
modelling techniques. We describe here a model of the Charge Transfer process
during read-out of modern Charge Coupled Devices, and its application to data
from STIS. The model draws upon previous investigations of this process and, in
particular, the trapping and emission model developed by Robert Philbrick of Ball
Aerospace.
Early comparison to calibration data is encouraging. Essentially, a physical
description of the STIS CCD combined with the physics of known defects in the
silicon lattice expected to arise in a hostile radiation environment, is enough to yield
results which approximately match real data. Uncertainties remain, however, in the
details of the model and the physical description of STIS.
1.
Introduction
The STIS CCD is known to have a steadily declining Charge Transfer Efficiency, CTE, which
is normal for a CCD in a hostile radiation environment (Cawley et al. 2001). Consequently,
charge is lost or deferred as it is read out to the on-chip amplifier from the pixel where it
was collected during the exposure.
The CCDs in HST instruments represent a rather unusual problem:
• HST orbits in a hostile radiation environment
• By contrast, ground based astronomical CCDs are not subject to bombardment from
high energy radiation
• Expensive servicing mission are required to replace HST hardware
• CCDs used in medical imaging, X-ray crystallography, etc., can be replaced at relatively low cost when they degrade
• HST continues to acquire data from its CCDs which is, in all other senses, of the
highest quality.
• Few other space based missions have enjoyed this longevity
Consequently, the HST CCDs provide an incentive for an accurate correction of CCD
readout effects which does not arise in other CCD applications.
The simplest correction applicable to data suffering significant effects of poor CTE is to
apply an empirically derived function relating pixel position, epoch and signal level to lost
charge (Goudfrooij 2002). However, in reality charge trapping and deferral lead to more
complex effects which can only be well understood by considering the transfer of charge
through each pixel and potential trapping site. We discuss here an attempt to model the
transfer of every electron through every pixel electrode on the STIS CCD during read out.
176
Modelling Charge Transfer on the STIS CCD
2.
177
The Model
The model began as a “toy model” to help us understand the processes involved. Encouraged by the ease with which CTE phenomena were qualitatively reproduced by this model,
we sought to replace our initial ad-hoc trapping and emission simulation with a more accurate representation of the physical processes involved. We were greatly assisted in this by
Rob Philbrick of Ball Aerospace who had developed such a model for the Kepler mission,
which will use similar CCDs to STIS (though provided by STA and E2V Technologies). Indeed Philbrick used STIS calibration data in his testing. This provided us with an emission
and trapping model derived from the physics of known bulk state traps.
This modelling approach is by no means unique, however we bring together a number
of features:
• Emission and capture is via the known bulk state traps summarized in the table:
Name/Description
P-V (Si-E)/Phosphorus Vacancy Complex
O-V (Si-A)/Oxygen Vacancy complex
V-V/Divacancy
Energy
(MeV)
0.44
0.168
0.3
Capture Timescale
τcn (µs) at -83◦ C
0.65
0.91
0.91
Emission Timescale
τcn (µs) at -83◦ C
1.3E+06
0.065
80
• The density of the traps is determined from the estimated Non-ionizing Energy Loss
(NIEL) experienced by STIS appropriate for the length of time on orbit (Philbrick
2001, Robbins 2000).
• Time scales for tapping and emission for each trap type are related to their energy
levels and the operating temperature.
• The state of all traps is tracked throughout the readout simulation. The initial distribution on the chip may be specified.
• The charge loss for each transfer is a function of the charge packet and the instantaneous state of the traps. There is no assumption of constant fractional charge loss.
• Electrons are treated as indivisible entities and capture and emission is modelled in a
Monte-Carlo fashion (e.g., there is no floating point averaging assumption).
• Transfer of charge packets between every electrode is modelled.
• Output includes fits images of CTE degraded data, difference images and statistics of
the capture and emission history of traps under every electrode.
• Other detectors suffering poor CTE may be modelled simply by specifying appropriate
CCD parameters.
The implementation of these features is described in detail in Bristow & Alexov (2002).
3.
Early Results
The Sparse Field technique makes use of the ability of the STIS CCD to read out to
registers on opposite sides of the array. A sequence of nominally identical exposures is
taken, alternating the readout between the registers. The observed differences between the
results obtained from the two registers can be fit to a simple CTE model. Calibration data
have been obtained at STScI for two tests based upon this idea.
178
Bristow, et al.
The internal sparse field test uses slit images from a flat field lamp placed at various
positions across the detector and read out through opposite registers (below these are the
registers corresponding to the B amplifier and D amplifier). The ratio of the line of the
signal readout from the slit image from the two amplifiers allows the calculation of the CTE
(Kimble et al. 2001). The external sparse field test makes use of exposures of the outer
regions of globular clusters, once again read out through opposite registers.
We take advantage of the plentiful internal sparse field data available from the STIS
archive. These datasets contain slit images of varying intensity and varying distances from
the amp used for readout. We choose a dataset in which the slit image is very close to the
readout register so that CTE effects are likely to be small. This assumption is not critical;
all that matters is that we have a slit image to use as our reference. If it has some small
distortion to its shape due to real CTE effects this will be effectively divided out in the
analysis below. We then use this as our input slit image and simulate the read out as if this
line was placed at varying distances from the register. The background is also that from
the real data.
In this way we are able to produce plots equivalent to those of Kimble et al. (2000). To
B
, we divide the
calculate Kimble et al.’s “Amplifier B signal ÷ Amplifier D signal,” Dsignal
signal
signal from a read out over y rows by another from a read out over 1024 − y rows. Of course
our results will necessarily have the property
Bsignal
Dsignal
Bsignal
Dsignal
(y) =
Bsignal
Dsignal
(1024 − y)
−1
and in
(y = 512) = 1.0. Here, y is the number of rows from the slit image to the
particular
register of the B amplifier. Indeed, deviations from this in the real results indicate either
non-uniformity in the trap distribution or a difference in behavior of the readout amps.
Figure 1 top and bottom show our results plotted along with Kimble et al.’s Figures 5a
and c respectively. For the lower signal level (top) the agreement is not bad for higher
values of y, especially given that the loss over 512 rows is almost identical. As discussed
above, the data points for y = 512 and below from Kimble et al. are anomalous. Also in
good agreement is the centroid shift for this data, measured at 0.38 pixels in the simulation
results compared to ∼ 0.35 (read from Figure 6b of Kimble et al.) for the real data.
For the higher signal level (bottom) the simulation results appear to underestimate the
effect of degraded CTE. The centroid shift in the simulation output is also about 40% lower
than in the real data.
We have not chosen to include error bars in the simulated results. This is because
uncertainty arises, at least in part, from the validity of the model itself, which is difficult
to quantify and in any case this is what we are trying to ascertain with this test. Uncertainty also stems from the physical parameters of the STIS CCD. If this uncertainty
was represented by error bars then the dominant contribution would come from our lack
of knowledge of the exact dimensions and effectiveness of the STIS mini-channel (this is
discussed in somewhat more detail by Bristow & Alexov 2000).
A more accurate fit to the real data is always possible by arbitrarily fixing trap densities
and trapping and emission constants, but we have plotted the values for the theoretical
values of these parameters only.
4.
Conclusions
We have developed a simulation of the CCD read out process that attempts to reproduce
quantitatively the effects of poor CTE in realistic astronomical data. This approach is by
no means unique, but we attempt to build upon recent progress in the understanding of
the underlying processes. The simulation has been customized to the STIS CCD, but is, in
principle, portable to other space based instruments. Moreover, the model is adaptable to
differing operating conditions, illumination patterns and levels of radiation damage.
Modelling Charge Transfer on the STIS CCD
179
2
1.5
1
Calibration Data
CTE Model
Low Signal
0.5
0
200
400
600
800
1000
800
1000
Rows from Amp B
1.05
1
Calibration Data
CTE Model
High Signal
0.95
0
200
400
600
Rows from Amp B
Figure 1.
Top: Signal per column for 60e- (cf. Kimble et al. 2000, Figure 5a).
Bottom: Signal per column for 3400e- (cf. Kimble et al. 2000, Figure 5c)
The simulation gives an insight into the processes involved and serves as a useful tool for
better understanding CTE. It reproduces qualitatively, known degraded CTE phenomena.
Preliminary quantitative results are encouraging. Without any empirical tweaking of the
models physical parameters we are able to get an approximate match to various values of
CTE measured in orbit. These comparisons now need to be extended to a wider range of
data, CTE measurement techniques and radiation damage levels. Simultaneously we intend
to improve the physical model by reducing the uncertainty in some parameters and better
understanding some processes. It will be interesting to see if this brings the results into
closer agreement with the data.
Acknowledgments. We would like to thank Rob Philbrick for taking the time to
answer our many questions in great detail and Paul Goudfrooij for useful comments.
References
Bristow, P. & Alexov 2002, Instrument Science Report: CE-STIS-2002-01
(http://www.stecf.org/poa/pdf/ccd sim isr2.pdf)
Cawley, L., Goudfrooij, P., & Whitmore, B. 2001, Instrument Science Report WFC3 2001-05
(Baltimore: STScI)
Goudfrooij, P. 2002, STIS Instrument Science Report No. TBD, STScI, in preparation.
Kimble, R. A., Goudfrooij, P., & Gilliland, R. L., 2000, in Proc. SPIE Vol. 4013, 532
Philbrick, R. H. 2001, Modelling the Impact of Pre-flushing on CTE in Proton Irradiated
CCD-based Detectors, Ball Aerospace & Technologies Corp.
Robbins, M. 2000, The Radiation Damage Performance of Marconi CCDs? Marconi Applied
Technologies, Technical Note: S&C906/424, 17 Feb 2000 (available on request)
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
STIS Status after the Switch to Side 2
Thomas M. Brown and James E. Davies
Space Telescope Science Institute, Baltimore, MD 21218
Abstract.
Since July 2001, STIS has been operating on its secondary (Side-2) electronics,
due to the failure of the primary (Side-1) system. The change to Side 2 has required
new calibration work. The dark rate of the STIS CCD varies since the switch to
Side 2, as it depends on the temperature of the CCD (which cannot be regulated
precisely using Side-2 electronics). We find that the dark rate is a linear function of
the housing temperature for pixels at a given dark rate, but the slope of this relation
varies for pixels with different dark rates. Scaling of the darks as a function of the
temperature has been incorporated into the STIS pipeline. An additional feature of
the switch to Side-2 is that the STIS CCD read noise has increased by 1 e− sec−1
for all four amplifiers when using a gain of 1. This increased read noise is due to
electronic pick-up pattern noise (on Side 1 the noise was primarily white noise).
Although an algorithm exists for filtering this additional pattern noise, it will not be
incorporated into the STIS pipeline.
1.
Introduction
The Space Telescope Imaging Spectrograph (STIS) was launched in 1997 with two sets of
redundant electronics, but a unique set of detectors. The sets of electronics are referred
to as “Side 1” and “Side 2.” STIS ran on its Side-1 electronics until May of 2001, at
which point the instrument safed due to catastrophic failure of these electronics. After an
extended period of testing, it was determined that the Side-1 electronics were unrecoverable;
they are also not designed for repair during the servicing missions. Thus, STIS operations
were resumed with the Side-2 electronics in July of 2001. STIS performance did not change
significantly in the switch from Side 1 to Side 2, except that the CCD read noise has
moderately increased, and the loss of CCD temperature control produces a variable dark
rate. We summarize these changes here; they are discussed more fully by Brown (2001a,
2001b).
2.
Dark Rate Variations with Temperature
On Side 1, a temperature sensor mounted on the CCD carrier provided closed-loop control
of the current provided to the thermoelectric cooler (TEC), thus ensuring a stable detector
temperature at the commanded set point (−83o C). Side 2 does not have a functioning
temperature sensor, so the TEC is run at a constant current. Thus, under Side-2 operations,
the CCD temperature varies with that of the spacecraft environment. Although no sensor
is available to measure the temperature of the CCD itself, there is a sensor for the CCD
housing temperature, which should track closely with the detector temperature under Side-2
operations (but not for Side-1). This housing temperature is reported in STIS CCD science
headers under the keyword OCCDHTAV; note that it is far hotter than the detector itself
(the housing is typically near 18o C while the detector runs near −83o C).
180
Rate relative to rate at 18oC
STIS Status after the Switch to Side 2
181
1.2 log(rate) =−2.50
log(rate) =−2.25
log(rate) =−2.00
log(rate) =−1.75
log(rate) =−1.50
1.1
1.0
0.9
slope =0.047
slope =0.056
slope =0.074
slope =0.087
slope =0.088
1.2 log(rate) =−1.25
log(rate) =−1.00
log(rate) =−0.75
log(rate) =−0.50
log(rate) =−0.25
1.1
1.0
0.9
slope =0.086
slope =0.080
slope =0.075
slope =0.072
slope =0.068
1.2 log(rate) = 0.00
log(rate) = 0.25
log(rate) = 0.50
log(rate) = 0.75
log(rate) = 1.00
1.1
1.0
0.9
slope =0.063
slope =0.062
slope =0.059
slope =0.057
slope =0.055
16 17 18 19
16 17 18 19
16 17 18 19 o 16 17 18 19
16 17 18 19
CCD Housing Temperature ( C)
Figure 1. The change in dark rate with temperature, for pixels at different rates.
The change in dark rate is linear for pixels at a given rate, but the slope of this
linear relation depends on the rate in question.
Because the dark rate for the STIS CCD is strongly temperature dependent, the dark
rate is now much more variable in CCD observations (compared to observations on Side-1).
The relation is linear for pixels at a given dark rate (Figure 1), but the increase with
temperature varies from ∼ 5%–9% for every o C, depending upon the rate of a particular
pixel. Pixels with low dark rates will vary by ∼ 5% per o C, pixels with moderately high
dark rates will vary by ∼ 9% per o C, and the pixels with the highest dark rates will vary
by ∼ 6% per o C.
This variable dark rate presents a problem for calibration, as the temperature can vary
between one science exposure and the next, and between one dark exposure and the next.
For accurate subtraction of the dark current, the temperature dependence of the dark rate
must be included in the pipeline processing of the raw data.
To this end, a new header keyword OCCDHTAV giving the CCD housing temperature
has been added to the STIS science headers. A simplified scheme that assumes a dark
rate variation of 7% per o C for all pixels has been implemented into the STSDAS calstis
software and the archive On The Fly Reprocessing (OTFR) software. Calibration dark files
are created so that they reproduce the dark at a housing temperature of 18o C, and then
are scaled up or down to match the temperature of a given science frame prior to dark
subtraction. The new dark subtraction algorithm for processing Side-2 CCD data came
into use in January 2002. Note that more complicated schemes for subtracting the dark
current (which use the pixel-to-pixel variation in dark rate with temperature) show little
improvement over this simplified scheme, mainly due to the poor statistics in characterizing
individual pixels in a given time period.
Side-1 data retrieved from the archive should have a negative value for OCCDHTAV
(indicating to the software that it should not be used for scaling the dark subtraction).
Unfortunately, when the temperature-dependent dark subtraction was implemented, a bug
in the software used only the year of the observations to determine if the data were taken
on Side 1 or Side 2, and thus the first 5 months of 2001 were treated as Side-2 data and
given a positive (and meaningless) OCCDHTAV value. This bug was fixed with the archive
software update of September 2002. Thus, any observations taken from January to May in
2001, but retrieved from the archive in January to August of 2002, will have an incorrect
dark subtraction, and should be re-retrieved from the archive with OTFR.
3.
Increased Read Noise
CCD data taken using Side 2 show an increase in effective read noise of approximately
1 e− pix−1 for gain = 1 and 0.2 e− pix−1 for gain = 4. This elevated read noise manifests
182
Brown & Davies
section of bias frame on Side 2
Power spectrum of bias frame
580
0.4
0.3
Magnitude
y pixel
560
540
520
0.2
0.1
500
480
260
280
300
320
x pixel
340
0.0
16.0
16.2 16.4 16.6 16.8
Frequency (kHz)
17.0
Figure 2.
Left panel: A section of a raw CCD bias frame. Notice the herringbone pattern noise. Right panel: A Fourier transform of the 1-D time series (as it
was read out in the Side-2 electronics) of the same bias frame. There is a peak in
the power spectrum at 16.65 kHz, which corresponds to a horizontal pattern with
2.73 pixel spacing on the CCD.
itself as a herring-bone pattern that can be seen easily on a short exposure such as a raw
bias frame (Figure 2). When the 2-D image is converted to a 1-D time series (using the
timing intervals for clocking out the CCD), a Fourier transform of the series indicates that
the read noise pattern is temporally correlated. In this example, there is a peak in the power
spectrum at about 16.65 kHz. Pipeline-processed CCD data also show this pattern noise.
In fact, the pattern noise is usually much more apparent in processed, cosmic-ray-rejected
images than it is in the raw, unprocessed images.
The frequency of the pattern noise is typically 15.5–18 kHz. There is a correlation
between the pattern noise frequency and the CCD housing temperature, but the correlation
is too loose to be useful for predicting the frequency accurately enough to assist with
filtering (see below). Timing frequencies in this range correspond to a spatial frequency of
approximately 3 pixels on the detector (horizontally, in the direction of serial clocking).
STIS CCD images and spectral images can sometimes be filtered by interpolating the
power spectrum to remove the peak from the pattern noise. We have provided an IDL
script to analyze the pattern noise and attempt filtering, at
ftp://ftp.stsci.edu/pub/instruments/stis/stisnoise.pro.
This noise removal will not be added to the archive pipeline; the current procedure requires
careful tuning of the filter position and width in Fourier space, which requires user interaction and evaluation of the results. Also, the filter often introduces undesirable artifacts
into the data, which must be weighed against the advantages of filtering on a case-by-case
basis. As the width of the filter is decreased, the artifacts decrease, but at some point some
of the pattern noise escapes the filter (due to the frequency wandering across the image).
The best filter width is typically about 20 Hz. Note that our software uses a fairly primitive
filtering method; we would appreciate any feedback on superior filtering techniques that
have been shown to work with the STIS pattern noise.
Figures 3 and 4 show two examples of data filtered with the IDL routine provided: the
crowded star field of 47 Tuc and the diffuse galaxy UGC 2847. These examples represent
the two extremes encountered in filtering. Images crowded with point sources tend to suffer
the most artifacts, while images with diffuse sources tend to filter well.
Acknowledgments. We are grateful to L. Dressel, R. Allen, P. Goudfrooij, R. Kimble,
and T. Gull for their insight and useful discussions.
STIS Status after the Switch to Side 2
no filter
183
filter 16.40 − 16.44 khz
y pixel
550
500
450
800
850
x pixel
900
800
850
x pixel
900
Figure 3.
A crowded star field in 47 Tuc, before (left) and after (right) filtering.
This is an example of data that cannot be filtered with the current algorithm.
The power spectrum did not reveal the pattern noise frequency, and a narrow
filter introduces artifacts around the bright stars.
no filter
filter 16401 − 16421 Hz
y pixel
550
500
450
475
525
x pixel
575
475
525
x pixel
575
Figure 4.
Left panel: An image section of UGC 2847. The pattern noise is very
striking in this short exposure, and makes it difficult to see the galactic structure.
Right panel: Application of a narrow filter completely removes the pattern noise
without introducing artifacts. Filtering works best in sparse, faint, diffuse images.
References
Brown, T. M. 2001a, “Temperature Dependence of the STIS CCD Dark Rate During Side-2
Operations”, Instrument Science Report STIS 2001-03 (Baltimore: STScI)
Brown, T. M. 2001b, “STIS CCD Read Noise During Side-2 Operations”, Instrument Science Report STIS 2001-05 (Baltimore: STScI)
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
Optimal Extraction with Sub-sampled Line-Spread Functions
Nicholas R. Collins1
Science Systems and Applications, Incorporated, 5900 Princess Garden Parkway,
Suite 300, Lanham, MD 20706
Theodore Gull and Chuck Bowers
NASA’s GSFC, Code 681, Greenbelt, MD 20771
Don Lindler1
Sigma Space Corporation, 9801 Greenbelt Road, Lanham, MD 20706
Abstract. STIS long-slit medium resolution spectra reduced in CALSTIS extendedsource mode with narrow extraction heights (GWIDTH = 3 pixels) show photometric uncertainties of ±3% relative to point-source extractions. These uncertainties
are introduced through interpolation in the spectral image rectification processing
stage, and are correlated with the number of pixel crossings the spectral profile core
encounters in the spatial direction.
The line-spread-function may be determined as a function of pixel crossingposition from calibration data sub-sampled in the spatial direction. This line spread
function will be applied to science data to perform optimal extractions and pointsource de-blending. Wavelength and breathing effects will be studied. Viability of
the method to de-convolve extended source “blobs” will be investigated.
1.
Introduction
In the course of analyzing rectified long-slit spectral observations of nebular targets such as
η-Carinae, or of crowded stellar fields, it may be desirable to use a small spectral extraction
height to avoid contamination from neighboring sources. Such extractions suffer from ±3%
systematic uncertainties introduced by the interpolation step during the spectral image
rectification.
We investigate the viability of an optimal extraction procedure that uses the crossdispersion profile of the STIS long-slit data to improve the extraction of isolated sources
and to de-blend neighboring sources.
This work is in progress. To date, we have used a standard star cross-dispersion profile
to perform an optimal extraction of the same and other isolated standard stars. We have
tested the viability of de-blending point sources by creating a model data set with two
point-source spectra of known intensity and relative position. Work remains to be done on
improving the point-source de-blending solution and on extending the algorithm to handle
resolved sources. Limitations of the method with respect to deriving a cross-dispersion profile from one observation to perform the optimal extraction of another observation include
the MSM-repeatability and the focus changes induced by HST “breathing.”
1
NASA’s GSFC, Code 681, Greenbelt, MD 20771
184
Optimal Extraction with Sub-sampled Line-Spread Functions
2.
185
Observations
The observations used in this analysis were obtained under proposal 8844, “Deep Spatial/Spectral PSF Calibration of STIS,” Charles Proffitt, principle investigator. We have
used the subset of un-occulted, 52 × 0.2 long-slit medium resolution mode observations of
the following stars: HD 181204 (M0, variable-irregular), HD 115617 (G5 V), HD 141653
(A2 IV), BD+75D325 (O5pvar).
3.
The Pixel-Crossing Problem
As a spectral image is converted to an extended source image, the interpolation introduces
artifacts in the data. These artifacts are apparent at the ±3% level in spectral extractions
that use extraction heights less than the number of pixel crossings.
Figure 1 shows comparisons of 3-pixel extractions from spectral images rectified using
three different interpolation methods (linear, quadratic, and cubic) against a standard pointsource extraction (no interpolation, extraction height = 7 pixels). The observed (unrectified)
spectrum of HD 181204 (observation ID O68M03090) crosses 7.5 CCD rows. Note that the
cubic and quadratic methods yield the same results at the ±1% level. The relative pixelcrossing positions are shown at the bottom. The cycles in that plot are clearly correlated
with the variations in the ratio plots. The spectrum was sub-sampled by a factor of four in
the cross-dispersion direction and by two in the dispersion direction. The grating used was
G750M.
4.
4.1.
Optimal Extraction Using the Cross-Dispersion Profile
Long-Slit Cross-Dispersion Profile Derivation
A cross-dispersion profile may be derived as a function of pixel-crossing position from a
rectified, background-subtracted observation. The columns for a particular pixel-crossing
cycle (see bottom plot in Figure 1) are extracted from a rectified spectrum. The signal
to noise for each column in the cycle is increased by averaging it together with its two
neighboring columns. For each end column, the average is performed with its neighboring
inside column. A record of the relative pixel-crossing position for each column is made.
If the cross-dispersion profile array contains artifacts, it is rejected and the procedure is
repeated for a different pixel-crossing cycle.
4.2.
Optimal Extraction Method
The optimal extraction is performed column-by-column across the background-subtracted
rectified spectral image. Each column in the observation is matched to a column in the
cross-dispersion profile array by its pixel crossing value. Next, the relative offset between
the profile and the data in the cross dispersion is determined (at this point manually). The
optimal extraction algorithm described by Horne (1986) uses a linear regression method.
The generalization of the single-source extraction to multiple-source extraction is made
using the multiple linear regression method described by Bevington and Robinson (1992).
y(xi ) =
k
(ak × fk (xi ))
(1)
m=1
where xi is a column element, y(xi ) is the fitted function at xi , m is the number of spectra
to fit, fk is the cross-dispersion profile co-aligned with the kth spectrum, ak is the parameter
of the fit, or the optimal extraction for the kth spectrum is the row of ak in each column
(wavelength).
186
Collins, et al.
Figure 1.
Comparison of extended-source (3-pixel) and point-source (7-pixel)
spectral extractions. Top to bottom: All extractions, extended-source (linear interpolation) vs. point-source, extended-source (quadratic interpolation) vs. pointsource, extended-source comparison (quadratic vs. linear interpolation), extendedsource comparison (cubic vs. quadratic), relative pixel position vs. wavelength.
The row matrix a may be determined as follows: a = β × α−1 , where βk ≡
yi × fk (xi )] and αlk ≡ i [σi−2 × fl (xi ) × fk (xi )]
5.
−2
i [σi
×
Optimal Extraction: Single Point-Source Results
Figure 2 shows the optimal extraction result for observation O6M03090 (HD 181204) using a cross dispersion profile derived from observation O6M02080 (HD 115617). The top
plot shows both the standard 7-pixel point-source extraction from the unrectified spectrum
and the optimal extraction from the rectified (cubic interpolation) spectrum. The middle
plot shows the ratio of the two extractions, and the bottom plot shows the relative rowpixel position. The cross-dispersion profile was created from the pixel-crossing cycle in the
O6M02080 data between 6600 Å and 6700 Å. The overall residuals are very low (< 2%)
with a ∼ 5% residual at the position of Hα 6563 Å absorption. Correcting for the 10% offset
seen in the (middle) ratio plot would require deriving a sensitivity curve using the optimal
extraction method. Using the cross-dispersion profile array from O6M02080 to perform
an optimal extraction on the O6M02080 spectrum itself (not shown) also produces small
residuals (except at Hα) but with a slight (∼ 2%) drop from the blue end to the red end of
the spectrum.
Deriving a profile array from O68M03090 for optimal extraction is not as successful,
perhaps due to absorption features in the pixel crossing cycle between 6600 Å and 6700 Å,
although no obvious artifacts are present in the profile array. When applying this array to
Optimal Extraction with Sub-sampled Line-Spread Functions
187
Figure 2.
Optimal extraction of observation O68M03090 using the crossdispersion profile array from observation 068M02080.
the O68M03090 data itself, 2% column-to-column residuals are produced, and a overall 4%
variation is observed from the blue to the red. Applying this profile array to O68M02080
produces 5% features in the residuals plot that correspond to the absorption features in
O69M03090 between 6600 Å and 6700 Å. Also a 5% drop from the blue end of the spectrum
to the red end is observed in the residuals.
6.
6.1.
Application to Simulated Point-Source Data
Creating the Simulated Blended Point-Source Spectra
Two test data-sets were created to monitor the effectiveness of point-source de-blending.
Both used the standard star spectrum from observation O68M03090. In each case the
rectified, background-subtracted spectrum was shifted, scaled and added to itself to produce
an artificial spectral image with two spectra. The fits to the test data sets were made
using the second complete pixel-crossing cycle from the short wavelength end of observation
O68M02080. The first test set used an offset of 5 pixels and a scale factor of 5. The second
test set used an offset of 25 pixels and a scale factor of 7.
6.2.
Simulated Data: Extraction/Deblending Results
The extractions for both components of test data set 1 (5-pixel offset, scale factor=5) show
large column-to-column residuals (∼ 5%) and much larger variations (∼ 20%) from the blue
end of the spectrum to the red end when compared to the standard 7-pixel point-source
extraction. The ratio plot of one component to the other has an average value of 2 with
20% variations. The average value should be 5, the input value of the scale factor for this
test data set.
The extractions for both components of test data set 2 (25-pixel offset, scale factor = 7)
also show large column-to-column residuals (∼ 3%), but smaller variations across the whole
spectrum. When compared to the standard point-source extraction, the residuals for the
unscaled, unshifted component drop by 5% across the spectrum and show 2% features that
are correlated with the relative pixel-crossing position. The scaled and shifted component
188
Collins, et al.
residuals compared to the point source extraction is flat within the column-to-column 2%
noise except for a 2% discontinuity coincident with the pixel-crossing cycle that begins on
the blue side at ∼ 6600 Å, and a 2% variation coincident with the pixel crossing cycle at
the red end of the spectrum (> 6900 Å).
7.
Conclusions/Directions
For isolated spectra, results comparable to the standard point-source extraction can be
obtained using an optimal extraction method that relies upon the actual cross-dispersion
profile instead of an analytic function. De-blending multiple spectra will clearly involve
more work, particularly for very closely blended sources. More work remains to be done
in verifying how robust the method is for handling isolated spectra, and in understanding
how features in the spectrum, such as absorption and emission lines, affect the line-spreadfunction template. Future work will also include multiple source extractions of real data.
References
Bevington, P. R. and Robinson, D. K., 1992, “Data Reduction and Error Analysis for the
Physical Sciences,” (New York: McGraw-Hill)
Horne, K., 1986, PASP 98, 609
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
Recent Improvements to STIS Pipeline Calibration
Rosa I. Diaz-Miller, Jessica Kim Quijano, Jeff A. Valenti, Charles R. Proffitt,
Kailash C. Sahu, Ralph C. Bohlin, Thomas M. Brown
Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218
Don Lindler
Computer Science Corporation, Inc.
Abstract. In the last few months a number of improvements to the STIS pipeline
calibration have been developed and implemented, which include the following.
We have released new low order flat files for use with the G140L observations.
These flats should reduce uncertainties of the extracted flux with position from 12%
to 2%.
To better reflect the change with time in the overall shape of the NUV MAMA
dark current, new dark reference files were created for different epochs. To further improve the dark subtraction, these darks are also scaled using an improved
algorithm, which takes into account long term changes in the behavior of the NUV
MAMA dark current.
Additional improvements which have been implemented are described in the
posters by Stys et al., Valenti et al., Davies et al. and Lindler et al. Future improvements include background smoothing for low signal spectroscopic data, and updating
the Pixel-to-Pixel flat library and the current CCD bad pixel table.
1.
New Low Order Flat Image File for G140L
As part of STIS calibration program 7937, observations of the spectrophotometric standard
star GD 71 were taken at 20 different positions along the length of the 52×2 slit, and showed
significant variation in sensitivity with position. These discrepancies in the extracted flux
relative to a point source at the standard position are as large as 12%. To correct for this,
new LFLATs (lfl files) were prepared using the 7937 observations. Use of these new lfl files
should reduce these discrepancies to 2% or less, except when the spectrum falls across the
shadow of the FUV MAMA’s repeller wire.
Prior to March 15, 1999, G140L spectra were shifted 3 above the repeller wire, while
after this date the default position was changed to 3 below the repeller wire. Different
sensitivity calibrations are applied before and after this date to correct for the differences
between the two positions. One new lfl file was prepared for each of these epochs and
normalized to unity near the appropriate standard extraction position.
Figure 1 (left) shows the ratio of the extracted count rates from the 7937 observations of
GD 71 to a reference spectrum, both with and without the application of this new LFLAT.
The same star was also observed at a more limited number of positions as part of program
7097. The 7097 data were not used to constrain the LFLAT, and while the scatter is larger
than for the 7937 data, the improvement is still significant (Figure 1, right).
189
190
Diaz-Miller, et al.
Figure 1.
Count rate comparison after using new G140L LFLAT (solid lines) and
without it (dashed lines). GD 71 observations: program 7930 (left) and program
7097 (right).
2.
New MAMA Dark Reference Files
The NUV MAMA dark current is dominated by phosphorescence from impurities in the
detector window. Cosmic rays excite electrons in these impurities into metastable states,
which can be collisionally excited to unstable states that then decay by emitting UV photons.
This process can be modeled by the equation:
rate = constant ∗ number of excited states ∗ exp(−δE/kT)
(1)
In addition to the variations in the overall level of the dark current which have been
discussed elsewhere (Ferguson & Baum 1999), there has been a subtle change over time in
the relative intensities of this dark current at different places on the NUV MAMA detector.
To better reflect this change in the overall shape of the NUV MAMA dark, four new dark
files were created; each intended for use with NUV MAMA data during a given time period.
On the other hand, the original NUV dark scaling formula used in the pipeline assumed that the number of excited states could be well approximated by a constant value,
and adopted fixed coefficients for this formula. However, long term changes in the mean
operating temperature of the detector can significantly change the equilibrium population
of excited states. In addition, it is observed that the dark current at the lowest detector
temperatures does not drop as low as the original formula would predict. This may indicate that there are other sources of dark current with different time scales and thermal
characteristics.
To better model the dark current, formula (1) was changed to allow the overall normalization to vary slowly over time to represent the long term evolution of the number of
excited metastable states, and to impose a lower limit to the temperature used in this formula. This minimum temperature was also allowed to vary with time. These coefficients, as
well as other coefficients in the equation which are currently left fixed, are read into Calstis
from the TDC table, a new STIS reference file type, which contains the adopted coefficients
as a function of time.
Dark monitor observations were used to calibrate these coefficients, and it was found
that using the modified formula substantially reduces the discrepancy between observed
and predicted rates. The NUV dark current predicted by the revised formula should be
within 5% of the correct value about 85% percent of the time, and within 10% about 95%
Recent Improvements to STIS Pipeline Calibration
191
Figure 2. Comparison observed and predicted NUV-MAMA dark rate using the
old (top) and the new (bottom) NUV dark scaling formula.
of the time (see Figure 2). To maintain this accuracy as the mean thermal condition of the
detector changes, periodic updates to the TDC table will be necessary.
3.
Smoothing the Background in STIS Spectroscopic Data
For low signal data, it is advantageous to smooth the background since such a smoothing
can decrease the background noise level, and hence improve the quality of the results. In
addition to the smoothing that is currently being done in the pipeline in the cross-dispersion
direction, we are implementing additional smoothing in the dispersion direction by fitting a
polynomial of order 3 to the background. In this procedure, we exclude the region occupied
by the geo-coronal lines in the fit (in the UV-region), and for these regions use the simple
row-averaged background instead.
Figure 4 shows the results before and after the background smoothing algorithm is
applied: the results are slightly better after the smoothing. The figure at the bottom shows
the ratio of the flux before and after smoothing, which is close to 1 as expected.
4.
Updating the Pixel-to-Pixel Flat Library
We have created new p-flats (pixel-to-pixel, high-frequency flats) for the STIS MAMAs.
These p-flats combine the data from four years of lamp exposures to achieve a signal-tonoise ratio of 200 per low-resolution pixel (compared with a signal-to-noise ratio of 100 in
the extant p-flats). Details on the creation of these P-flats can be found in Brown & Davies
(2002).
192
Diaz-Miller, et al.
Figure 3. Top: Spectrum before and after the background smoothing. Bottom:
Ratio of the flux before and after smoothing.
5.
MAMA Bad Pixel Table
The MAMA bad pixel mask (in the Data Quality extension) will be updated to reflect the
true occulted regions on each detector (notably the horizontal repeller shadow across the
middle of the far-UV detector and the corners of the near-UV detector). Currently, the DQ
extension reflects the bad pixels in the G140M and G230M modes, which is not appropriate
for the other spectroscopic and imaging modes. Furthermore, the bad pixel mask changes
somewhat with time due to changes in the Mode Select Mechanism position over the course
of the mission. Thus, the new Data Quality flags will be mode- and time-dependent.
References
Brown, T. M. & Davies, J. E. 2002, Technical Instrument Report STIS 2002-03 (Baltimore:
STScI)
Ferguson, H. & Baum, S. 1999, Instrument Science Report STIS 99-02 (Baltimore: STScI)
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
Autofilet.pro: An Improved Method for Automated Removal of
Herring-bone Pattern Noise from CCD Data
Rolf A. Jansen
Arizona State University, Dept. of Physics & Astronomy, Tempe, AZ 85287-1504
Nicholas Collins
SSAI, Raytheon
Rogier A. Windhorst
Arizona State University, Dept. of Physics & Astronomy, Tempe, AZ 85287-1504
Abstract. We present an improved method for the automatic removal of the highly
variable pattern-noise that was introduced in HST /STIS CCD data when it was
switched to its redundant (“Side-2”) electronics in July 2001. While mainly a cosmetic nuisance for work on bright objects, this “herring-bone” noise severely limits
the sensitivity at optical wavelengths for projects that aim to push STIS to its design limits. We build on the Fourier filtering technique described by Brown (2001)
and present a method to automatically find and remove the power associated with
the noise patterns in frequency space, while avoiding the introduction of ringing
(aliasing) around genuine astronomical signal—in particular around stellar images,
spectroscopic (emission) lines, and cosmic ray hits. We implement this method as an
IDL procedure and show several applications. Details of the method will be discussed
in Jansen et al. 2003.
1.
Introduction
After STIS operations were resumed in July 2001 using the redundant “Side-2” electronics
(following the failure of the primary electronics on May 16, 2001), the read-noise of the
CCD detector had noticeably increased due to a superposed and variable “herring-bone”
pattern-noise. For most work the pattern will be a mere nuisance, but projects aiming to
detect signals so faint as to push STIS to its design limits will be severely affected. One such
project of interest to the authors is program #9066, which aims to detect the exceedingly
faint spectroscopic signal left behind in the general extragalactic background due to the
reionization of Hydrogen at z ≥ 6 (e.g., Baltz, Gnedin, & Silk 1998) using deep STIS/CCD
parallel exposures. In order for this program to be successful, we needed to develop a
reliable automated method for removal of the pattern-noise.
Brown (2001) presented a method to filter the pattern noise by noting that the sequential charge shift and read-out allows one to convert a CCD image into a time-series.
That time-series may be Fourier transformed to the frequency domain, where the frequencies responsible for the noise pattern may be suppressed via various methods. His method
works well in images where few bright and/or spatially very concentrated (sharp) features
are present, but requires manual definition of the frequency limits of the filter. If the filter is chosen too wide, or if many genuine high-frequency non-periodic signals (e.g., stars,
spectral lines, cosmic ray events) are present, ringing may occur (see, e.g., Brown 2001, his
Figures 1b and 6b).
Here we present a method that builds on the work by Brown (2001) and which mitigates
both these issues.
193
194
2.
Jansen, et al.
Strategy
The problem of automatically and robustly finding the frequencies that correspond to the
herring-bone pattern is greatly reduced if we first model and subtract the genuine science
signal. The residuals image, ideally, only contains shot-noise, read-noise and the herringbone pattern. In practice, there are systematic residuals of genuine features in the data
as well, but the contrast of the herring-bone pattern is much higher than in the original
science image. This means that in the frequency domain we will be able to blindly run a peak
finding routine with much relaxed constraints on the frequency interval (or alternatively on
much poorer data—e.g., very long spectroscopic exposures that are riddled with cosmic ray
hits) and still correctly find, fit, and filter out the pattern frequencies. Also, since most of
the genuine signal has been removed prior to constructing the power spectrum, the problem
of ringing has effectively been avoided.
Second, instead of setting the power at all frequencies corresponding to the noisepattern to zero, or suppressing it using a multiplicative windowing function, we opt to
substitute the power at the affected frequencies by white noise at a level and amplitude that
matches the “background” power in two intervals that bracket the affected frequencies. This
is less likely to introduce artefacts due to the absence of power at frequencies that should
have some, or aliasing that may result when many adjacent frequencies have the same power.
The resulting modified power spectrum may be inverse Fourier transformed and converted to a 2-D image, which by adding back in the fitted “data model” produces a patternsubtracted science frame.
3.
Autofilet.pro—Getting Rid of them Herring Bones
The optimized Fourier filtering method briefly outlined above was implemented in an IDL
procedure, Autofilet.pro, available from the authors. Details of the routine and results
of its application will be presented elsewhere (Jansen et al. 2003). Two real-valued (32bits per pixel) FITS format images are output for every science extension in a raw FITS
image (usually 16-bits per pixel), one containing only the herring-bone pattern, the other
containing the pattern-subtracted science image. To remove the herring-bone pattern from
a science image that is part of a multi-layer image set (e.g., [SCI,ERR,DQ] for STIS CCD
data), the herring-bone image may simply be subtracted from the appropriate science extension, as long as both the arithmetic and the output pixel format are real-valued. An
example clarifying the procedure and results is given in Figures 1 and 2.
Although written for the removal of the variable pattern-noise in HST /STIS CCD data
taken after July 2001, Autofilet contains place holders for adaptation to CCD data from
other telescopes and instruments that display similar pick-up noise.
Acknowledgments. The data shown are from HST parallel program #9066, “Closing
in on the Hydrogen Reionization Edge of the Universe at z < 7.2 with Deep STIS/CCD
Parallels,” which aims to detect a signal so faint that it necessitated the project on which
we report here. We acknowledge support from NASA grants GO-08260.* and GO-09066.*.
We thank Bruce Woodgate for getting us started. We would not have had the same success
without the work by Thomas M. Brown.
References
Baltz, E. A., Gnedin, N. Y., & Silk, J. 1998, ApJ, 493, L1
Brown, T. M. 2001, Instrument Science Report STIS 2001-005 (Baltimore: STScI)
Jansen, R. A., Collins, N., & Windhorst, R. A. 2003, PASP (in prep.)
Automated Removal of Pattern Noise from CCD Data
Figure 1.
(a) Section of a raw CCD bias frame (‘o6dc9b040’), taken with
HST /STIS on 2001 July 23, after operation of STIS had been resumed with
its redundant (“Side-2”) electronics. This section displays different features: a
herring-bone noise pattern is seen, as well as several (vertical) columns where the
bias level and noise differ slightly from the mean, and three regions affected by
cosmic ray hits;
(b) a model of the image section shown in (a) containing most of the signal (as fitted to the image lines and columns) and also all pixels deviating from the mean by
more than 3 σ or by more than 0.5 σ when adjacent to pixels deviating more than
3 σ. The difference of the original image section and this model, i.e., the residuals
image, is converted to a time-series and Fourier transformed to frequency space;
(c) image of the herring-bone pattern inferred from the peak in the power spectrum of the residuals image (see also Figure 2). We fit the center and width of that
peak, and replace all signal in the Fourier transformed time-series within ±3 σ of
the peak by white noise matching the noise in the two intervals located 4–7 σ away
from the peak. The result is inverse Fourier transformed and converted back into
a 2-D image.
The resulting pattern-subtracted bias image is shown in (d ). The remaining noise
closely resembles white noise with rms ∼4.2 e− . Note that there is no “ringing”
around the bright regions affected by cosmic ray hits.
195
196
Jansen, et al.
Figure 2. (a) Portion of the power spectrum centered on the frequencies responsible for the herring-bone pattern in the BIAS frame displayed in Figure 1(a).
The power spectrum is generated by first converting the 2-D residuals image [Figure 1(a) minus Figure 1(b)] into a time-series followed by Fourier transformation
of that time-series. After finding the peak frequency (for this image 16.155 kHz),
an estimate of the width of the peak is obtained by fitting a Gaussian function
to the power spectrum. The finite width of the peak results from the (erratic)
drift in frequency of the noise pattern during the time it takes to read the CCD.
All power within ±3 σ of the peak frequency (solid, blue, vertical lines) is then
replaced by white noise that matches the noise in the two bracketing regions in
frequency located 4–7 σ away from the peak (dotted, blue, vertical lines). The
resulting modified power spectrum is inverse Fourier transformed and converted
to a 2-D image, which by adding back in the fitted “data model” [Figure 1(b)]
produces the pattern-subtracted bias frame [Figure 1(d )].
(b) Distribution of pixel values in the raw BIAS frame of Figure 1(a) (dotted, green)
and in the pattern subtracted BIAS frame of Figure 1(d ) (solid, red ). Whereas the
noise in the raw BIAS frame is distinctly non-Gaussian near the mean pixel value
and has a σ ∼ 5.5 e− , after subtraction of the inferred herring-bone pattern of
Figure 1(c) the noise is well described by random Gaussian noise with a standard
deviation σ = 4.20 e− . Autofilet therefore successfully reproduces—perhaps even
slightly improves upon—the nominal “Side-1” CCD read noise observed prior to
July 2001. The
√ inferred amplitude−of the herring-bone pattern (assuming Gaussian
statistics) is 5.52 − 4.22 ∼ 3.6 e .
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
Removing Fringes from STIS Slitless Spectra and WFC3 CCD
Images
E. M. Malumuth,1 R. S. Hill,1 T. Gull, B. E. Woodgate, C. W. Bowers, R. A.
Kimble, D. Lindler,2 R. J. Hill,1 E. S. Cheng, D. A. Cottingham,3 Y. Wen,1 and
S. D. Johnson3
Laboratory for Astronomy and Solar Physics, Code 681, NASA’s Goddard Space
Flight Center, Greenbelt, MD 20771
Abstract. We have developed a model that allows us to defringe slitless 2-dimensional
spectra taken with the Space Telescope Imaging Spectrograph (STIS). An IDL tool
has been developed which allows the user to defringe any spectrum obtained with
the G750L grating on STIS. This technique has been employed to model the fringing
on Wide Field Camera 3 (WFC3) flight candidate CCDs.
1.
Introduction
Interference fringes are an annoying fact of life for many astronomical CCD detectors; the
Space Telescope Imaging Spectrograph (STIS) and Wide Field Camera 3 (WFC3) CCDs
are no exception. Fringes are caused by the interference of the incident and internally
reflected beams within the thin layers of the CCD. The interference can be constructive or
destructive within the CCD detection layer, leading to strong variations of the detection
efficiency as a function of wavelength and local CCD thickness.
Fringing is not significant at wavelengths below ∼ 7000 Å where the absorption path
length in silicon is less than the thickness of the CCD detection layer, but it becomes a
serious issue at near-infrared wavelengths.
As reported previously (Malumuth et al. 2000, Malumuth et al. 2002), STIS CCD
fringing can be modeled as an instance of multilayer thin-film interference using the Fresnel
equations and the formalism of Windt (1998). The modeling requires a detailed knowledge
of the CCD’s physical structure (i.e., how many and what materials make up the stack). We
then solve for the physical parameters (thickness and interfacial roughness) of each layer for
each pixel. For the STIS CCD, we found that we could keep all of the parameters constant
except for the thickness of the detection layer of the CCD.
In this paper we concentrate on how to use the results of the model fitting to defringe
STIS slitless spectra. We also show that the same formalism and data obtained in the
Goddard Space Flight Center’s (GSFC) Detector characterization Laboratory (DCL) is
being used to model the fringing of the WFC3 flight candidate CCDs.
1
2
3
Science Systems and Applications, Inc.
Sigma Space Corp.
Global Science and Technology, Inc.
197
198
Malumuth, et al.
Figure 1.
2.
A image, spectral image pair shown in SLWIDGET.
STIS Slitless Spectra
STIS slitless spectra are usually obtained as image, spectral image pairs. Figure 1 is an
example of a slitless spectral image pair taken as part of the STIS parallel program. The
image, spectral image pair is shown in the IDL widget SLWIDGET (Lindler 1998) which
can be used to extract the spectrum of any object in the field. The image is needed so that
we will know where in the field of view the object was located when the spectral image was
taken. This is necessary to determine the wavelength scale.
3.
Using The Model
The core of the model is a program that solves the Fresnel equations for a given pixel. The
IDL procedure (LAYERS) is based on the formalism of Windt (1998) and is a function of
the wavelength on the pixel, the number of layers in the CCD, the index of refraction and
absorption coefficient of each layer at that wavelength, the thickness of each layer, and the
“roughness” of the boundary between each layer (expressed as a diffusion distance). The
function returns the transmission and the absorptance for that pixel at that wavelength.
The fringe amplitude at that wavelength for the given pixel is the normalized absorptance.
Therefore, to defringe a given spectrum we need to know the following.
• The wavelength of light hitting each pixel used in the spectral extraction.
• The structure of the CCD, including the composition and thickness of each layer in
the CCD.
• The index of refraction and absorption coefficient of each material in the CCD as a
function of wavelength.
Removing Fringes from STIS and WFC3 Data
199
Figure 2. Three fringe models for the spectrum of the star shown in Figure 1.
The top line is for a wavelength shift of −10 Å, while the bottom line is for a shift
of +10 Å.
The details of how the structure of the CCD at each pixel was derived, as well as
an adjustment to the index of refraction of silicon as a function of wavelength have been
previously reported by Malumuth et al. (2002).
Given the above information a model of the fringing for an object anywhere in the
field of view can be computed. The resulting model for a given set of pixels is a sensitive
function of the wavelength of the light on that pixel. Thus, a wavelength error in the
spectral extraction of only a few angstroms can result in a poor fringe correction. Figure 2
shows how a shift of ±10 Å effects the fringe location.
3.1.
STIS Spectra Defringing Tool
An IDL widget tool (see Figure 3) brings all of the elements together to allow the user to
defringe an extracted spectrum.
The user reads in the extracted spectrum in the form of a structure contained in a
FITS table. The structure should include the wavelength vector, the flux vector, and the
CCD location where the extraction was done. This maps the wavelength to pixel location.
The widget reads in the tables of the optical properties (index of refraction and absorption coefficient as a function of wavelength) of the materials in the CCD, the thickness
vector, the roughness vector, and the map of the thickness of the detection layer as a
function of pixel location.
The widget will normalize the spectrum by fitting a spline to the spectrum and dividing
by the result. The normalized spectrum is displayed in the upper right window.
The user presses the “Find Wavelength” button. The widget will calculate 11 model
fringe patterns for the center row of the extraction using a slightly different wavelength
scale for each. The user may control the wavelength scales by changing the “Wavelength
Offset” and “Wavelength Step Size.” The defaults are 0 Å for the offset and 2 Å for the
step size. At each step the calculated fringe pattern is plotted on the normalized spectrum
as a blue line (seen in Figure 3 as a thin black line). The normalized spectrum is divided
by this “model” fringe pattern and the result is displayed in the lower window, and the
signal to noise ratio is calculated between 8300 Å and 10000 Å. The S/N ratio is plotted
200
Malumuth, et al.
Figure 3. The IDL STIS Spectra Defringing Tool. The top plot shows the
normalized spectra, with fringes, of the star shown in Figure 1. The thick line
is the data, the thin line is the best fit model. The bottom right plot shows the
resulting defringed spectra.
as a function of offset in the small window on the left. After the eleventh wavelength offset
position is finished a gaussian is fit to the S/N vs. offset plot to find the best offset.
A fringe model is calculated for the best fit wavelength offset and divided into the flux
calibrated spectrum and displayed in the lower window. If the user wishes, a full fringe
model using all of the rows in the extraction may be calculated at this time by pressing the
“Full Fringe Extraction” button. The defringed spectrum can be saved in a fits table using
the “save” button.
4.
WFC3 Fringing
The WFC3 instrument being developed for the Hubble Space Telescope will have a UV/VIS
channel which will include two 2051 × 4096 pixel thin backside illuminated CCDs similar
to the STIS CCD. The techniques developed for modeling the STIS CCD are being applied
to the flight candidate CCDs for WFC3, in the Goddard Space Flight Center’s Detector
Characterization Lab.
References
Malumuth, E. M., et al. 2000, A Model for Removing Fringes from STIS Slitless Spectra,
BAAS, 197, 1204
Malumuth, E. M., et al. 2002, Removing the Fringes from STIS Slitless Spectra, PASP, in
press
Malumuth, E. M., et al. 2002, Model of Fringing in the WFC3 CCDs, SPIE, in press
Windt, D. L. 1998, Computers in Physics, 12, 360
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
Absolute Flux Calibration of STIS Imaging Modes
Charles R. Proffitt,1,2 James Davies, Thomas M. Brown, Bahram Mobasher
Space Telescope Science Institute, Baltimore, MD 21218
Abstract.
The absolute flux calibration of STIS imaging photometry presents a number of
unique challenges. The very wide wavelength coverage of most STIS imaging modes
leads to significant color dependence in both the throughputs and the aperture corrections, complicating the determination of detector sensitivity. For CCD imaging
modes, these difficulties are further complicated by the very broad scattered light
halo at long wavelengths. For MAMA imaging modes, it is also necessary to take the
time and wavelength dependent sensitivity changes of the detectors into account. We
present deep imaging observations of a number of stars with well measured spectral
energy distributions. These data have been used to derive improved color dependent
aperture corrections for all STIS imaging configurations, and to revise the wavelength
dependent detector sensitivities. These new aperture corrections and sensitivity revisions should allow absolute flux calibration of imaging observations with better than
5% accuracy for most STIS imaging modes.
1.
Introduction
STIS has a small number of imaging modes for each detector, many of which have very
broad bandpasses. Aperture corrections can be very large and strongly dependent on source
color. Because STIS is also a well calibrated spectrograph, detailed wavelength dependent
filter throughputs have been measured by taking slitless spectra through each filter. There
remain, however, uncertainties in the overall wavelength dependent system throughput.
Our strategy is to calibrate the detector throughputs and encircled energy curves as
a function of wavelength by obtaining very deep images of stars of a variety of colors and
with well measured spectral energy distributions.
2.
CCD Imaging Throughputs
Early STIS F28X50LP CCD images of several stars measured count rates 28% lower than
expected, while 50CCD images of one hot star showed about the expected count rate. The
initial response to this discrepancy was to lower F28X50LP throughput by 28%. However,
this solution is not consistent with imaging results for cooler stars (e.g., Houdashelt, Wyse,
& Gilmore 2001). Spectra taken through the F28X50LP filter also shows the throughput
to be close to prelaunch estimates. The correct solution is to lower the long wavelength
throughput of all STIS CCD modes and to properly measure the strongly color dependent
aperture corrections.
1
2
Science Programs, Computer Sciences Corporation
Catholic University of America Institute for Astrophysics and Computational Science
201
202
Proffitt et al.
As part of STIS/CAL programs 8422, 8844, and 8924, we contained contemporaneous
deep STIS 50CCD and F28X50LP images as well as STIS CCD spectra of a number of
stars (Table 1) of a variety of colors. The deep images were used to derive color-dependent
aperture corrections (Table 2), while the measured spectral energy distributions were used
to adjust the CCD throughput curve (Figure 1) to obtain good agreement between predicted
and measured imaging count rates (Figure 2).
In Table 1 we show the magnitudes and colors calculated for the CCD calibration
stars using their observed spectral energy distributions and the currently adopted STIS
component throughputs. Table 2 shows the measured aperture corrections for these stars.
Table 1. Calculated Magnitudes and Colors of CCD Imaging Calibration Stars
in STMAG and VEGAMAG Systems
Star
GRW +70◦ 5824
WD 310-688
CPD -60◦ 7585
CPD -35◦ 9181 A
BD -11◦3759
STMAG
F28X50LP 50CCD-LP
13.747
−0.899
12.361
−0.759
10.574
−0.196
10.328
+0.074
9.994
+0.325
50CCD
12.848
11.603
10.378
10.402
10.318
V
12.716
11.358
10.041
10.202
11.263
U BV Rc
U −B B −V
−0.772 −0.057
−0.624 +0.041
−0.017 +0.417
+0.833 +0.917
+1.241 +1.476
V − Rc
−0.100
−0.079
+0.275
+0.583
+1.274
Table 2. Measured Aperture Corrections for 50CCD and F28X50LP Images As
a Function of the Aperture Radius in Pixels
Aperture Radius in CCD Pixels (0.05071)
5
7
10
15
19.7
Star
2
3
4
GRW +70◦ 5824
WD 310−688
CPD −60◦ 7585
CPD −35◦ 9181 A
BD −11◦ 3759
−0.464
−0.461
−0.523
−0.585
−0.806
−0.283
−0.276
−0.314
−0.352
−0.524
−0.219
−0.216
−0.242
−0.282
−0.379
50CCD
−0.183 −0.137
−0.194 −0.138
−0.210 −0.169
−0.244 −0.193
−0.328 −0.280
−0.102
−0.105
−0.130
−0.152
−0.235
−0.074
−0.076
−0.104
−0.120
−0.197
GRW +70◦ 5824
WD 310−688
CPD −60◦ 7585
CPD −35◦ 9181 A
BD −11◦ 3759
−0.621
−0.643
−0.632
−0.662
−0.813
−0.344
−0.363
−0.375
−0.397
−0.535
−0.245
−0.262
−0.288
−0.299
−0.392
F28X50LP
−0.214 −0.166
−0.228 −0.184
−0.259 −0.215
−0.273 −0.225
−0.346 −0.300
−0.124
−0.140
−0.158
−0.182
−0.251
−0.094
−0.109
−0.127
−0.148
−0.212
39.4
59.2
−0.057
−0.060
−0.088
−0.102
−0.173
−0.021
−0.022
−0.043
−0.053
−0.104
−0.009
−0.010
−0.024
−0.029
−0.064
−0.078
−0.091
−0.110
−0.126
−0.187
−0.036
−0.047
−0.059
−0.074
−0.109
−0.021
−0.026
−0.035
−0.039
−0.068
Measurements and predictions (Table 3) are now in good agreement for stars of all
colors (Figure 1).
3.
MAMA Imaging Throughputs
Bright object limits prevent the usual WD flux standards from being observed with many
MAMA imaging modes. Time dependent sensitivity changes are also important, especially
for FUV MAMA imaging modes.
To calibrate the MAMA imaging throughput we use a strategy similar to that used
for the CCD modes. For the narrow and some intermediate band MAMA filters we use
STIS images and spectra of a number of WD standards from Bohlin, Dickinson, & Calzetti
(2001). To calibrate the broad-band and intermediate imaging modes we use hot HB stars
in the globular cluster NGC 6681 that were measured in STIS/CAL programs 8842 and
9631, and the star UIT1 in NGC 2808 (images are from STIS/CAL program 8511 and
spectra from GO program 7436).
Absolute Flux Calibration of STIS Imaging Modes
Figure 1.
shown.
The old (solid line) and revised (dashed line) CCD throughputs are
Figure 2.
Here we compare the predicted and observed count rates for 50CCD
(squares) and F28X50LP (circles) imaging of our standard stars for assuming the
old (filled symbol) and the revised CCD throughputs (open symbols). For all
calculated F28X50LP magnitudes in this figure, the revised F28X50LP curve is
assumed.
203
204
Proffitt et al.
Figure 3. The solid line shows the throughput curve for NUV MAMA imaging
modes adopted in July 1999. The broken curve shows our suggested revision.
The throughput curves shown here have not yet been multiplied by the OTA
throughput.
Wavelength dependant MAMA aperture corrections are based on the encircled energy
curves from the deep imaging observations of individual stars.
For most detector/filter combinations, observations agree with predictions based on
previously tabulated throughput curves to better than 5%. FUV F25QTZ observations
show unusually large scatter, possibly due to long wavelength throughput variations across
FUV MAMA. NUV MAMA F25CN182 observations are somewhat too bright compared
with predictions. We have proposed a revision to the short wavelength throughput curve
for NUV MAMA imaging modes (Figure 3) that alleviates this problem, and modestly
improves the consistency of results for a number of other filters.
Table 3. Aperture Corrections for Individual MAMA Imaging Modes Calculated
Assuming a Source Spectrum with Fλ =constant.
Detector/filter
NUV F25CIII
NUV F25CN182
NUV F25ND3
NUV F25SRF2
NUV F25QTZ
NUV F25CN270
NUV F25MGII
FUV F25LYA
FUV 25MAMA
FUV F25ND3
FUV F25SRF2
FUV F25QTZ
Aperture Radius
3
5
−0.680 −0.441
−0.664 −0.431
−0.561 −0.365
−0.578 −0.375
−0.572 −0.371
−0.512 −0.332
−0.485 −0.316
−1.030 −0.651
−0.855 −0.556
−0.851 −0.554
−0.785 −0.515
−0.752 −0.494
in MAMA Pixels (0.025)
10
15
20
−0.287 −0.194 −0.130
−0.279 −0.186 −0.123
−0.223 −0.141 −0.093
−0.232 −0.149 −0.099
−0.230 −0.147 −0.099
−0.198 −0.121 −0.082
−0.182 −0.110 −0.073
−0.381 −0.227 −0.127
−0.340 −0.214 −0.122
−0.339 −0.214 −0.120
−0.323 −0.210 −0.120
−0.321 −0.218 −0.132
References
Bohlin, R. C., Dickinson, M. E., & Calzetti, D. 2001, AJ, 122, 211
Houdashelt, Mark L., Wyse, Rosemary F. G., & Gilmore, G. 2001, PASP, 113, 49
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
Sensitivity Monitor Report for the STIS First-Order Modes
D. J. Stys, N. R. Walborn, I. Busko, P. Goudfrooij, C. Proffitt, K. Sahu
Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218
Abstract. The analysis of the STIS Sensitivity Monitor observations from 1997
through 2002 continues to show sensitivity trends correlated with time for all firstorder modes, as well as temperature dependence in the FUV. The wavelengthaveraged rate of sensitivity loss for the MAMA low-resolution (L) modes is nearly
2%/yr for G140L and about 1.5%/yr for G230L; the observed trends in the CCD
modes are dominated by charge transfer efficiency (CTE) loss. Selected wavelength
settings of the medium-resolution (M) gratings G140M, G230M, G230MB, G430M,
and G750M have also been followed by this monitoring program. The lower exposure
levels in the CCD M-mode observations result in significantly larger effects from CTE
losses, than is the case for the L-mode observations. In general, the sensitivity losses
are found to be wavelength dependent. The limited MAMA M-mode wavelength
coverage is consistent with the same sensitivity trends observed in the L modes at
the corresponding wavelengths. On this basis, the STIS pipeline processing software
is currently being revised to correct the extracted fluxes for these sensitivity changes
in both the MAMA L and M modes. The CCD sensitivity losses due to CTE depend
on the signal and background exposure levels as well as detector location, and so
require tailored corrections.
1.
Observations
A spectroscopic flux standard, either the white dwarf GRW+70D5824 or the subdwarf
AGK+81D266, is monitored with each STIS grating to detect changes in sensitivity due
to contamination or other causes. Observations included in this report are from the STIS
Sensitivity Monitor calibration program. The data are from Cycles 7–10 through May 2002.
Each observation uses the 52 × 2 slit and is processed with on-the-fly reprocessing (OTFR).
The MAMA L modes have been monitored monthly while the M mode measurements were
taken at two-month intervals. The CCD L modes were monitored every three months while
the CCD M modes were checked every six months. CCD observations are CR-SPLIT = 2
and GAIN = 1. G750L and G750M observations have contemporaneous ‘fringe-flat’ exposures for fringe removal. The normal mode select mechanism (MSM) shifting is disabled for
these monitoring observations in order to minimize variations due to spatial displacements
of the spectra. Detailed information regarding observing strategy can be found in Walborn
and Bohlin (1998).
2.
Analysis and Results
Prior to analysis, the FUV MAMA fluxes are corrected for a slight (0.25%/?C) temperature
dependence. Thereafter, each STIS mode is examined for any variation in sensitivity with
time. The earliest observation for each mode is defined as the reference. The sum of
the NET counts for each observation is divided by the sum of the NET counts for the
reference. These relative sensitivities are plotted versus time and are fitted with linear
205
206
Stys, et al.
Figure 1.
G140L Relative Sensitivity vs. Time
Figure 2.
G230L Relative Sensitivity vs. Time
Sensitivity Monitor Report for the STIS First-Order Modes
Figure 3.
207
G750L Relative Sensitivity vs. Time
segments (as seen in Figures 1–3). The slope, i.e., percent-per-year change in sensitivity
and 1σ uncertainty in the fits are printed at the bottom of each plot. The 1σ rms(%) of
the data residuals from the linear fit is also provided (SIGMA). The tables below sumarize
the wavelength-averaged results for the MAMA and CCD L-modes.
Prior to analysis, the FUV MAMA fluxes are corrected for a slight (0.25%/C) temperature dependence. Thereafter, each STIS mode is examined for any variation in sensitivity
with time. The earliest observation for each mode is defined as the reference. The sum
of the NET counts for each observation is divided by the sum of the NET counts for the
reference. These relative sensitivities are plotted versus time and are fitted with linear
segments (as seen in Figures 1–3). The slope, i.e., percent-per-year change in sensitivity
and 1σ uncertainty in the fits are printed at the bottom of each plot. The 1σ rms(%) of
the data residuals from the linear fit is also provided (SIGMA). The tables below sumarize
the wavelength-averaged results for the MAMA and CCD L-modes. Smith et al. (2002)
discussed the entire analysis procedure for this program while Stys and Walborn (2001)
reports the latest sensitivity trends.
3.
Pipeline Correction
Corrections for the time dependences of all MAMA first-order configurations have been
incorporated into the STScI data-reduction pipeline as of mid-2002, so that all OTFR
retrievals of such data will have corrected fluxes regardless of their acquisition epoch. The
CCD sensitivity losses due to CTE depend on the exposure levels and detector location, and
so require tailored corrections. Sensitivity-monitor data for the echelle and MAMA imaging
configurations are currently under investigation; they appear to show analogous effects,
and pipeline corrections for them will become available at a later time. The SYNPHOT
208
Stys, et al.
Table 1.
MAMA Time-Dependent Sensitivity Trends
Mode
G140L
G140M
G140M
G230L
G230L
G230M
G230M
Table 2.
Epoch
<
>
<
>
1998.7
1998.7
1998.7
1998.7
λ-Range (Å)
1300–1500
1150–1190
1542–1592
2200–2600
2200–2600
2775–2860
2775–2860
%/yr
−2.02
−2.52
−2.17
1.66
−1.50
1.40
−0.90
+/−
0.07
0.12
0.08
0.17
0.04
0.23
0.05
SIGMA
0.66
0.67
0.52
0.17
0.25
0.12
0.22
CCD Time-Dependent Sensitivity Trends
Mode
G230LB
G230LB
G230MB
G230MB
G430L
G430M
G750L
G750M
Epoch
< 1998.7
> 1998.7
< 1998.7
> 1998.7
λ-Range (Å)
2000–3000
2000–3000
2340–2490
2340–2490
3100–5500
3050–3300
5600–7000
7000–7500
%/yr
0.79
−1.59
0.23
−2.23
−0.39
−1.20
−0.28
−1.07
+/−
0.20
0.08
0.49
0.14
0.07
0.06
0.05
0.10
SIGMA
0.16
0.34
0.29
0.43
0.46
0.28
0.30
0.51
package has been updated to model the time dependent sensitivities for both spectroscopic
and imaging modes.
References
Walborn, N. & Bohlin, R. 1998, Instrument Science Report STIS 98-27 (Baltimore: STScI)
Bohlin, R. 1999, Instrument Science Report STIS 1999-07 (Baltimore: STScI)
Smith, T. E., Stys, D., Walborn, N., & Bohlin, R. 2000, Instruments Science Report STIS
2000-03 (Baltimore: STScI)
Stys, D. & Walborn, N. 2001, Instrument Science Report STIS 2001-01R (Baltimore: STScI)
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
2-D Algorithm for Removing STIS Echelle Scattered Light
Jeff Valenti, Ivo Busko, and Jessica Kim Quijano
Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218
Don Lindler
Advanced Computer Concepts, Inc.
Chuck Bowers
Goddard Space Flight Center, Greenbelt, MD 20771
Abstract. We provide excerpts from Instrument Science Report STIS 2002-01
(Valenti et al. 2002), which describes in more detail a 2-D algorithm for removing
scattered light from STIS echelle spectra.
1.
Introduction
Ideally, a spectrograph should yield a one-to-one mapping between detector pixel and
monochromatic source intensity. In practice, background, scattered light, and finite resolution contaminate the monochromatic signal in each pixel. Background subtraction and
scattered light removal typically precede spectral extraction. Bias and dark subtraction
removes the component of background that is independent of exposure level, leaving only
the source spectrum and a component due to scattered light. For echelle spectrographs,
1-D linear interpolation of the minimum intensity between echelle orders provides a simple
model of the scattered light beneath each order. Originally, this basic scattered light model
was the only choice in the IRAF task x1d (McGrath et al. 1999), which is often used to
extract echelle spectra obtained with the Space Telescope Imaging Spectrograph (STIS).
Beginning with CALSTIS version 2.9 (installed in the archive pipeline on 2000 Dec 21 and
released as part of STSDAS version 2.3 on 2001 June 12), x1d also includes a new 2-D scattered light model (algorithm = sc2d) that supplements the original 1-D model (algorithm =
unweighted). The sc2d algorithm was developed by Lindler & Bowers (2001), implemented
in CALSTIS by Busko, and tested by Valenti.
Several authors have suggested simple enhancements of the basic 1-D algorithm. For
example, Howk & Sembach (2000) inferred the background beneath each order by fitting 1-D
polynomials to an extended region around the minima between echelle orders. Alternatively,
scattered light may be decomposed into a local 1-D component that scales with counts
detected in the two immediately adjacent orders and a global 2-D polynomial component
(e.g., Gehren & Ponz 1986). The formalism developed to interpret echelle data from the
Goddard High-Resolution Spectrograph includes components that scale with total counts in
an order and counts detected in each extracted wavelength bin (Cardelli et al. 1993). These
1-D components correspond to scatter by the echelle and the cross-disperser, respectively. In
contrast to the models described above, the sc2d algorithm iteratively builds an empirical 2D description of scattered light from 1-D extracted spectra and known scattering properties
of the telescope and spectrograph.
209
210
Valenti, et al.
800
Dataset: o4qx04010
HD 303308 (O3V)
Order 334
E140H
1.5
Counts in Column 500
Flux (10−11 erg s−1 cm−2 Å−1)
2.0
1.0
0.5
2−D Algorithm
1−D Algorithm
0.0
1259
1260
1261
Wavelength (Å)
1262
600
400
200
0
600
650
Row Number
700
Figure 1. Left: These interstellar absorption lines should have zero flux in the
line cores, but extraction with a 1-D scattered light model yields negative fluxes.
Right: In this cut across echelle orders, the deep absorption line in the order at
row 640 drops below the 1-D scattered light model (smooth horizontal line).
2.
Empirical Motivation
Development of a background subtraction algorithm (Lindler & Bowers 2001) with a 2-D
scattered light model was motivated by unrealistic negative fluxes in saturated cores of interstellar absorption lines punctuating spectra of continuum sources. Figure 1a demonstrates
the problem with extracted spectra obtained by subtracting either 1-D or 2-D estimates
of the scattered light background. With the traditional 1-D background subtraction algorithm, the saturated line core is systematically below zero by 9.0 ± 0.4% of the neighboring
continuum flux. This must be an artifact of inadequate background subtraction. With the
new 2-D scattered light model described here, the line core is only 1.0 ± 0.4% below zero,
indicating significant improvement.
Figure 1b presents a cut at column 500 through the subimage in Figure 2a. Echelle
orders 330–338 are spaced nearly uniformly, except that order 334 is missing because of the
strong interstellar line extracted in Figure 1a. The unweighted algorithm in x1d estimates
the background beneath each order by linearly interpolating the mean interorder light along
columns. Order 334 (near row 640) is systematically below the 1-D background estimate
(smooth horizontal line), indicating the need for a better background estimate.
Figure 2a shows a portion of a cross-dispersed echelle spectrum obtained in the FUV
with the STIS E140H grating. The continuum of HD 303308 (black horizontal bands) is
cut by interstellar absorption lines (white gaps) which should have zero signal in the final
extracted spectrum. The image in Figure 2a has been clipped at 6% of the peak to highlight
the behavior of the background. Note that absorption line cores are fainter (whiter) than
the “background” between echelle orders. In this case, linear interpolation of interorder
light along columns does not provide a good estimate of the background beneath an order.
Lindler & Bowers (2001) developed a 2-D algorithm which provides a better estimate of
the background everywhere on the detector. Figure 2b shows the resulting 2-D background
estimate for the subimage shown in the left panel. The region around the strongest interstellar line complex is highlighted (dashed box) in both images. The background between
orders is brighter (blacker) than the general background beneath each order, which in turn
is higher than the faint background (white patches) beneath strong absorption lines.
700
700
650
650
Row
Row
STIS Echelle Scattered Light Correction
600
211
600
Clipped
at 6%
of Peak
0
200
400
600
Column
800
1000
0
200
400
600
Column
800
1000
Figure 2.
Left: Counts in deep absorption lines are below (less dark) than the
background between orders. A vertical line indicates the cut used in Figure 1.
Right: The 2-D scattered light model is lower (less dark) in the cores of deep
absorption lines. A rectangle outlines the same regions as in the left panel.
3.
Algorithm
In the Lindler and Bowers (2001) algorithm, a flat-fielded image is fitted with a 2-D model
constructed in each iteration by folding the best current estimate of the extracted spectrum through a semi-empirical simulation of STIS optical properties. For STIS data, selfconsistency between the model image and the extracted spectrum is achieved after three
iterations. A 2-D scattered light model is then constructed considering only the echelle
scatter outside an 11 pixel wide vertical window centered on each order. This 2-D scattered
light model is subtracted from the original image, and the final spectrum is obtained using
standard 1-D extraction. See Valenti et al. (2002) for details.
Scattered light from the echelle gratings is a significant contribution to STIS scattered
light and is the main reason why a 1-D background model does not accurately reflect
the scattered light beneath an order. Figure 3a shows echelle scatter functions for three
orders of the E140M grating. A majority of the light is concentrated in the central pixel,
but integrated light in the wings of the scattering function can be significant. At the
shortest wavelengths, 37% of the light is scattered more than 15 pixels from the nominal
position. As indicated in the table inset, scattered light increases dramatically at the short
wavelength end of the FUV bandpass, presumably because the wavelength of incident light
is becoming comparable to the size of irregularities on the reflection grating surface. At
visible wavelengths, echelle scatter would be greatly diminished, reducing the need for a
2-D scattered light algorithm.
4.
STSDAS Implementation
The 2-D algorithm was first implemented by Lindler in IDL, taking advantage of the software
and database environment maintained for the STIS Instrument Definition Team. Busko
incorporated the algorithm into the existing x1d task in the IRAF package STSDAS. The
x1d implementation is used in the archive pipeline and is supported for general use by the
STIS community. Both the IDL and IRAF implementations have tasks named CALSTIS
212
Valenti, et al.
1
2
0.1
0.01
Order
λ
(Å)
128
104
84
1154
1424
1758
Central >15
Pixel Pixels
60%
75%
76%
37%
24%
16%
10−3
Order
128
104
84
10−4
−400
−200
0
200
Pixel Offset
400
Fractional Error in Line Core (%)
Echelle Scattering Function
E140M
2−D Algorithm
0
−2
−4
−6
E140H E230H
−8
1−D Algorithm
−10
−12
1500
2000
Wavelength (Å)
2500
Figure 3.
Left: Echelle scatter is modelled as a sharp core with broad wings.
At short wavelengths, 1/3 of the light is scattered by more than 15 pixels. Right:
The cores of numerous deep absorption lines yield errors, relative to the local
continuum and as a function of wavelength, for 1-D and 2-D models.
that drive end-to-end processing of STIS data, so an additional descriptor is required to
distinguish between the two implementations (e.g., “the sc2d algorithm in the x1d task”
uniquely specifies the IRAF implementation). The sc2d algorithm in x1d first appeared
in CALSTIS version 2.9, which was installed in the archive pipeline on 2000 Dec 21 and
released as part of STSDAS version 2.3 on 2001 June 12.
The 2-D algorithm requires a component level description of STIS optical properties,
which Lindler and Bowers bundled into a variety of reference files. Implementation of the
sc2d algorithm in x1d required the creation of 7 new reference file types, beyond those
required for the original unweighted algorithm. Scattered light reference files used by the
sc2d algorithm in x1d have content identical to the original Lindler reference files, but
organization and FITS structure have been modified to match STSDAS conventions.
Figure 3b shows measured errors for strong absorption interstellar line cores, as a
function of wavelength and echelle grating. Although the four echelle gratings could in
principle have significantly different surface roughness, 1-D extraction errors for all the
echelle gratings have a similar dependence on wavelength. The factor of four increase in
error from 1400 to 1100 Å, despite only a factor of two increase in echelle scatter over the
same interval (table inset in Figure 3a), may simply be due to the decrease in order spacing
for bluer orders. This same effect could also account for the larger errors in E230H spectra
at 1700 Å, relative to E140H spectra.
References
Cardelli, J. A., Ebbets, D. C., & Savage, B. D. 1993, ApJ, 413, 401
Gehren, T. & Ponz, D. 1986, A&A, 168, 386
Howk, J. C. & Sembach, K. R. 2000, AJ, 119, 2481
Lindler, D. & Bowers, C. 2001, BAAS, 197.1202
McGrath, M. A., Busko, I., & Hodge, P. 1999, Instrument Science Report STIS 1999-03
(Baltimore: STScI)
Valenti, J. A., et al. 2002, Instrument Science Report STIS 2002-01 (Baltimore: STScI)
Part 3. NICMOS
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
NICMOS Status
A. B. Schultz,1 D. Calzetti, S. Arribas,2 T. Böker,2 M. Dickinson, S. Malhotra
Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218
L. Mazzuca
NASA’s Goddard Space Flight Center, Greenbelt, MD 20771
B. Mobasher,2 K. Noll, E. Roye, M. Sosey, T. Wiklind,2 and C. Xu
Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218
Abstract. The Near Infrared Camera and Multi-Object Spectrometer (NICMOS)
is operational with improved scientific performance relative to the Cycle 7 & 7N. It
has been performing routine science observations since June 12, 2002. The NICMOS
detectors are operating at a higher temperature (77.1 ± 0.08 K (3σ)) than during
Cycles 7 & 7N (∼62 K). The instrument temperature has been fairly stable (±0.08 K
at the detector), which allows stable optics and instrument’s characteristics. Due to
the higher operating temperature, the detector QE is higher, which results in ∼0.23
magnitude fainter detection limit in H for a 3σ detection of a point source in a 3,000
second integration.
1.
Introduction
The Near Infrared Camera and Multi-Object Spectrometer (NICMOS) was installed on the
Hubble Space Telescope (HST ) in February 1997 during the Second HST Servicing Mission
(SM2). The on-orbit life time of NICMOS was shortened due to a thermal short. The
nitrogen ice cryogen was exhausted on January 4, 1999, and data taking was suspended on
January 11, 1999. The NICMOS detectors warmed up to temperatures around 260 K and
were not available for scientific observations for about three years. A mechanical cooler,
the NICMOS Cooling System (NCS), using a reverse-Brayton cycle (Cheng et al. 1998)
was installed on March 8, 2002 during the HST Servicing Mission 3B (SM3B). The NCS
activation was on March 18, 2002, followed shortly by the start of the NICMOS cool down.
Subsequently, a set point temperature resulting in a detector temperature of 77.1 K was
reached, and NICMOS data taking resumed.
The NICMOS has three infrared cameras (NIC1, NIC2, NIC3) with different focal
ratios (f/80, f/45.7, f/17.2). The NIC1 PSF is critically sampled at 1.0 µm, the NIC2 PSF
is critically sampled at 1.75 µm, and the NIC3 PSF is under sampled. Camera operations
are independent and non-confocal. The three cameras have adjacent, but not spatially
contiguous, fields-of-view (FOVs). Each camera has a 256 × 256 pixel HgCdTe focal plane
array (NICMOS 3 detector architecture). With the dewar “anomaly,” the detectors were
moved forward toward the dewar face plate. This resulted in non-confocal imaging at the
detectors. However, NIC1 and NIC2 can successfully be used in parallel to each other
1
2
Science Programs, Computer Sciences Corporation
On assignment from the Space Telescope Division of the European Space Agency (ESA)
215
216
Schultz, et al.
at an intermediate focus position for most applications. The focus interface for NIC3 is
beyond the adjustable range of the Pupil Alignment Mechanism (PAM). However, NIC3
can still be used for science observations at the best available focus as the slight defocus
relative to optimal focus induces only a modest (∼15%) reduction of the PSF peak energy.
The characteristics of the NICMOS cameras after the installation of NCS are presented in
Table 1. More details about the new capabilities of NICMOS can be found in the NICMOS
Instrument Handbook for Cycle 12 (Malhotra et al. 2002).
Table 1. Characteristics of NICMOS Cameras. The read noise is measured
per read pair, the difference of two reads. The dark current is the linear signal
remaining after removing the amp-glow and shading contribution from the total
“dark” signal. The DQE values (fraction of incident photons detected) for postNCS temperature of 77.1 K scaled from the ground-based values.
Characteristics
Pixel size
Field of View
read-noise
dark current
ADGAIN
DQE (1.6 µm)
Camera 1
0.043
11× 11
∼30 e− /pixel
0.145 e− /sec/pixel
5.4 e− /DN
0.426
Camera 2
0.075
19.2× 19.2
∼30 e− /pixel
0.110 e− /sec/pixel
5.4 e− /DN
0.442
Camera 3
0.2
51.2× 51.2
∼30 e− /pixel
0.202 e− /sec/pixel
6.5 e− /DN
0.414
NICMOS provides direct infrared imaging in broad, medium, and narrow-band filters
from 0.8 to 2.5 µm as well as three special observational modes; i.e., coronagraphic imaging,
broad-band imaging polarimetry, and slit-less grism spectroscopy. These special NICMOS
capabilities enable fundamental investigation into the nature of a wide variety of objects.
Accompanying papers describe in detail each of these modes.
2.
NICMOS Cooling System (NCS)
The NICMOS Cooling System (NCS) consists of three major subsystems: a cryocooler
which provides the mechanical cooling; a Capillary Pumped Loop (CPL) which transports
the heat dissipated by the cryocooler to an external radiator; and a circulation loop which
transports heat from the inside of the NICMOS dewar to the cryocooler via a heat exchanger.
Cold neon gas is circulated to cool the detectors. The NCS was activated on March 18,
2002, and reached the operational detector temperature of 77.1 K four weeks later. The
time line of the NICMOS detector cool down is presented in Figure 1.
A number of test programs were executed to verify the stability and repeatability of the
NCS control law. Results to date indicate that the NCS has good control of the temperature
and that there is a margin in the compressor speed to compensate for extra heat due to
changes in spacecraft attitude, operation of the NICMOS and other science instruments in
the Aft Shroud, and any changes due to seasonal conditions. An in-depth discussion of the
NICMOS cool down can be found in the accompanying paper Böker et al. (2003).
3.
NICMOS Optical Stability
The optimal focus for each NICMOS camera was determined on a
ing Cycles 7 & 7N with the last of the pre-NCS focus measurements
uary 4, 1999 (Suchkov et al. 1998). These results can be found on
page (http://www.stsci.edu/hst/nicmos/performance/focus). Images
regular basis durperformed on Janthe NICMOS web
of the star cluster
NICMOS Status
217
260
220
Temperature (K)
200
180
160
NICMOS switched on
NICMOS switched off
240
140
120
100
Target Temperature
80
60
0
Figure 1.
5
10
15
20
Time (days)
25
30
35
Time line of the NICMOS detector cool down.
NGC3603 (NIC/CAL program 8977) were obtained on May 3, 2002 to determine the locations of the NICMOS detectors with respect to the HST f/24 input beam. The data were
independently analyzed using recovery of Zernike polynomial coefficients by phase retrieval
and minimization of PSF core flux density dispersal by encircled energy. Comparisons made
between phase retrieval and encircled energy measurements are in good agreement.
Results from the analysis indicated that the PAM positions, that affect best foci at the
detectors, appeared to have moved marginally in the negative direction for NIC1 (PAM1)
and NIC2 (PAM2) and in the positive direction for NIC3 (PAM3). The movement of NIC3
is towards its best focus. Thus, NIC3 focus is somewhat better than during Cycles 7 &
7N. The new PAM positions for all three cameras were uplinked to the telescope on May 9,
2002 together with the new intermediate NIC1 & NIC2 (PAMI) focus. No focus adjustments
were implemented for NIC3 as the camera’s optimal focus is located beyond the physical
range of the PAM. No change to the coronagraphic focus (PAMC) was recommended at
that time. The pre-NCS and current PAM positions are presented in Table 2. A monitor
program, executed monthly for the first six months of Cycle 11 and bi-monthly afterwards,
shows that the camera foci are relatively stable with a maximum excursion of ∼0.5 mm of
PAM movement from the default set points. A plot of the post-NCS focus history compared
to the focus history for Cycles 7 & 7N is presented in Figure 2. Details about the focus
determination can be found in the accompanying paper by Roye and Schultz (2003).
Table 2.
The recommended nominal focus positions in mm of PAM motion. No
breathing correction has been applied.
PAM1
2.36
1.90
PAM2
0.69
0.20
PAM3
−9.50
−9.50
PAMI
1.75
1.22
PAMC
2.69
2.69
Date
pre-NCS
post-NCS
218
Schultz, et al.
Figure 2. Post-NCS NICMOS focus history results compared to Cycles 7 &
7N. The x -axis represents the number of days since January 1, 1997 and y-axis
represents the position of the PAM in mm.
A small amount of coma, observed in the post-NCS NICMOS and essentially absent
in the Cycles 7 & 7N observations, has been corrected. The absence of such coma in the
pre-NCS observations implies that a small amount of induced differential shear has occurred
since January 1999. The presence of a small amount of coma did not affect the determination
of the “best focus” by either of the two employed analysis methods discussed above. Coma
corrections were determined separately for all three cameras. A set of data called a “tilt
grid” was obtained which is a series of images taken at a variety of PAM tilt positions
around the default position. New optimal PAM tilt setting for NIC1 were uplinked to HST
on May 16, 2002. The correction successfully alleviated the observed NIC1 coma. New
settings for NIC2 were uplinked on September 29, 2002, successfully alleviating all coma.
No update was required for NIC3. Figure 3 presents the NIC1 and NIC2 PSFs before and
after PAM tilt correction for coma. The NIC3 PSF is also shown for comparison. Details
about the coma determination can be found in the accompanying paper Roye and Schultz
(2003).
4.
Sensitivity Limits
The sensitivity of the NICMOS detectors is checked during each HST observing cycle by
observing standard stars in each filter, the solar analog P330E and the white dwarf G19B2B. In addition, a photometric monitor program is executed regularly during each cycle
NICMOS Status
219
Figure 3. NIC1 and NIC2 PSFs before and after PAM tilt correction for coma.
NIC3 PSF is shown for comparison.
to observe standard stars in a selection of filters. These observations are used to determine
the NICMOS photometric calibration and to determine the photometric stability of all
three NICMOS detectors. The Cycles 7 & 7N photometric calibration data showed that
the absolute photometry is accurate to 6% (1-σ uncertainty), and the temporal and spatial
relative photometry is within 2% as well.
A check of the NICMOS photometric calibration was performed during June 2002
(SM3/NIC 8986). The solar analog P330E was observed in all three cameras with every
broad, medium, and narrow band filters. Preliminary data reduction indicates an increase in
the post-NCS photometric sensitivity of 20–70% depending on wavelength; this is consistent
with the observed increase of the camera’s DQE due to the higher operating temperature
(see accompanying paper by Böker et al. 2003). Figure 4 shows the ratio of post-NCS
photometric calibration to the Cycle 7 calibration. Further details about the photometric
calibration can be found in the accompanying paper by Dickinson et al. (2003). The higher
sensitivity can be quantified as a ∼0.23 mag fainter 3 σ detection limit at H for a faint
source with 3,000 sec exposure time. The new sensitivity limits are presented in Table 3
below.
Table 3.
NICMOS Sensitivity Limits. Cycle 11 magnitude limits to achieve
S/N = 3 in 3600 sec for point source (80% encircled energy). AB magnitudes
are at the standard NICMOS filter bands, where AB magnitudes are defined as
AB = −2.5 log(F (ν)) − 48.57.
JAB
HAB
KAB
NIC1
24.6
24.1
—
NIC2
25.2
24.6
21.8
NIC3
25.2
24.7
21.4
220
Schultz, et al.
Figure 4.
5.
Post-NCS photometric calibration ratioed against Cycle 7 calibration.
NICMOS Pipeline and User Tools
A suite of software tools have been developed to help the NICMOS observer calibrate, reduce, analyze, and to determine the quality of his data. Some of these tools are old favorites,
while others are relatively new since Cycles 7 & 7N. Most importantly for those submitting
Cycle 12 proposals is the updated Exposure Time Calculator (ETC). The ETC is set for
the Cycle 11 operating temperature of 77.1 K. And for those retrieving NICMOS observations from the HST Archive, On-The-Fly-Reprocessing (OTFR) is now implemented for all
NICMOS data retrieved from the Archive. Of note are two tools to create “Temperature
Dependent Darks” and “Temperature Dependent Flat fields and Color Dependent Flats.”
A summary of all available NICMOS software tools useful for data reduction and analysis along with a short discussion on the most recent updates to the calibration pipeline
software (CALNICA and CALNICB) can be found in the accompanying paper by Sosey
(2003).
An “old” problem with NICMOS observations is the persistence (residual image) induced by cosmic rays from passages through the South Atlantic Anomaly (SAA). In order
to help alleviate the problem, the following steps have been taken:
1. Pairs of ACCUM darks are obtained with each camera after SAA passages to provide
an “imprint” of the persistence image (post-SAA darks).
2. Four new keywords have been added to the headers of the science data (SAA EXIT,
SAA TIME, SAA DARK, SAACRMAP). The filenames of the post-SAA darks closest
in time to the respective observation are written into the header keywords.
Finally, a persistence-removal method using the post-SAA darks is currently being investigated, together with the feasibility of an “automatic” routine to alleviate the persistence in
science images.
NICMOS Status
Table 4.
The NICMOS Cycle 11 Calibration Goals
Calibration
Dark current and shading
Flat fields
Photometry
PSF and Focus
Astrometry
Coronagraphic PSF
Grism wavelength calibration
Grism photometric calibration
Polarimetry
NIC3 intrapixel sensitivity
High S/N capability character
6.
221
Accuracy
4-5% on MULTIACCUM sequences
1% broad band, 3% narrow-band
6% absolute, 2% relative and stability
1 mm
0.5% plate scale, 0.1 to FGS frame
0.013 in positing in hole
0.005 µm
∼30% absolute and relative
1%
1-5%
S/N∼10,000
NICMOS Calibration Plan
A recalibration of the NICMOS primary capabilities has been initiated following the installation of the NCS and the subsequent cool down of the detectors. The NICMOS Cycle 11
calibration activities cover a period of 13 months. They complement the SMOV3B and the
Cycle 10 interim calibration programs. These activities pursue the following major objectives: i) to monitor detector properties, ii) to provide data to test the model for generating
darks, iii) to provide high S/N flat fields; iv) to investigate the intra-pixel sensitivity effects
for NIC2 and NIC3, and v) to calibrate the three special observing modes; i.e., coronagraphy, polarimetry, and grism spectroscopy. Table 4 presents a list of the Cycle 11 calibration
goals. Further details about the NICMOS Cycle 11 Calibration Plan can be found in the
accompanying paper by Arribas et al. (2003).
Acknowledgments. We are grateful to all of the NICMOS presenters at the HST
Calibration Workshop for sending copies of their papers and figures. Special thanks goes to
Rodger Thompson, Marcia Rieke, Glenn Schneider, and Dean Hines (University of Arizona),
and to Wolfram Freudling (ST-ECF).
References
Arribas, S., et al. 2003, this volume, 263
Böker, T., et al. 2003, this volume, 222
Cheng, E. S., Smith, R. C., Jedrichm N. M., Gibbon, J. A., Cottingham, D. A., Swift,
W. L., & Dame, R. E. 1998, SPIE Proc., 3356, 1149
Dickinson, M., et al. 2003, this volume, 232
Hines, D. C., Schmidt, G. D., & Schneider, G. 2000, PASP, 112, 983
Malhotra, S., et al. 2002, NICMOS Instrument Handbook for Cycle 12, Version 5.0, (Baltimore: STScI) url: http://www.stsci.edu/hst/nicmos
Mazzuca, L. & Hines, D. C. 1999, “User’s Guide to Polarimetric Imaging Tools,” Instrument
Science Report NICMOS-99-004 (Baltimore: STScI)
Roye, E. and Schultz, A. 2003, this volume, 267
Sosey, M. 2003, this volume, 275
Suchkov, A., Bergeron, L., & Galas, G. 1998, “NICMOS Focus Monitoring,” Instrument
Science Report NICMOS-98-004 (Baltimore: STScI)
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
NICMOS Detector Performance in the NCS Era
Torsten Böker,1 Louis E. Bergeron, Lisa Mazzuca,2 Megan Sosey, and Chun Xu
Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218
Abstract. After a three-year hiatus following the exhaustion of its solid nitrogen
coolant, NICMOS was revived with the installation of the NICMOS Cooling System
(NCS) during the HST Servicing Mission 3B in March 2002. NICMOS now operates
at about 77.1 K, some 15 K warmer than during its initial operating period. In this
paper, we briefly describe the on-orbit performance of the NCS. In addition, we use
results from the early NICMOS calibration program to characterize the impact of
the higher operating temperature on the behavior of the NICMOS detectors, with
a focus on those parameters that are relevant to the scientific performance of the
“new” NICMOS.
1.
Introduction
The Near Infrared and Multi-Object Spectrometer (NICMOS) provides the Hubble Space
Telescope (HST ) with its only means to study the universe at infrared wavelengths. Installed
on HST during the second Servicing Mission in February 1997, NICMOS suffered from
a shortened lifetime because of a thermal anomaly that led to an increased sublimation
rate of the solid nitrogen coolant used to maintain a detector temperature of ≈ 61 K.
Following the nitrogen exhaustion in January 1999, the NICMOS instrument warmed up to
temperatures around 260 K, much too high for scientifically useful observations. NICMOS
thus lay dormant for about three years, awaiting the installation of the NICMOS Cooling
System (NCS), a mechanical cooler using a closed-loop reverse-Brayton cycle (Cheng et al.
1998).
Since the NICMOS detectors show a number of subtle effects that are sensitive to
temperature, both the value of the operating temperature and its stability are crucial parameters for the scientific performance of NICMOS. Prior to the NCS on-orbit installation,
the evaluation of the thermal performance of the NICMOS/NCS system had to rely on models, because for obvious reasons, the NICMOS dewar was not available for ground testing.
Therefore, various aspects of the NCS performance remained rather uncertain, including
the parasitic heat load that the NCS had to overcome and its ability to react to environmental changes during the orbital (and seasonal) cycle of HST. However, the results from
the early NICMOS calibration program and the NCS telemetry during the first few months
of on-orbit operation indicate that all is well with the revived NICMOS. In what follows, we
describe the stable and efficient performance of the cryocooler, and the impact of the higher
operating temperature on detector parameters such as dark current, quantum efficiency,
and readout noise.
1
2
On assignment from the Space Telescope Division of the European Space Agency (ESA).
NASA/GSFC, Code 681, Greenbelt, MD 20771
222
NICMOS Detector Performance
Figure 1. Thermal history of the NCS. Top: weighted average of the Neon inlet
and outlet temperature sensors. Bottom: Camera 1 mounting cup sensor which
closely traces the actual detector temperature.
223
224
2.
2.1.
Böker, et al.
NICMOS/NCS Performance in 2002
The Cooling System Performance
Much effort had been spent over the last few years to understand the thermal performance
of the NICMOS/NCS system. The latest pre-launch models had predicted a cooldown time
of about 10 days. However, it became clear very soon that the NICMOS dewar was cooling
much slower than expected, which triggered frequent revisions to the SMOV timeline, not
only for NICMOS, but also for the other HST instruments. What made matters worse, early
extrapolations of the cooldown profile indicated that the target temperature of around 77 K
for the NICMOS detectors might not be reached. A number of options to increase the NCS
cooling capacity or to reduce the parasitic heat load were discussed in a flurry of status
meetings. The more drastic of these proposals included disabling the safety heaters that
provide leakage protection of the cryogenic Neon lines.
Finally, a decision was made to temporarily safe the NICMOS instrument in order to
reduce the heat load from its electronic boxes. A consequence of this decision was the loss of
all telemetry data from within NICMOS, and the interruption of the dark current monitoring
program which was supposed to provide early indications of the NICMOS performance in
the NCS era. However, the NICMOS safing resulted in an accelerated cooldown and, after
about 4 weeks of continued cooling, the Neon gas inside the NCS circulator loop finally
reached the target temperature of 72 K. NICMOS was switched on again, and the dark
current measurements resumed while the system was stabilizing.
Since the start of NCS operations, STScI has continously monitored the system performance. The thermal history of a few key temperature sensors until mid-November 2002
is summarized in Figure 1. The plots show that over the first 6 months of operation, the
NCS has maintained the NICMOS detectors to within 0.1 K of their target temperature.
The slight increase in the average detector temperature over the last month is probably
a reflection of the hotter season for HST, as the earth currently is closer to the sun and
hence the mean temperature of the HST aft shroud is slightly increased. STScI is currently
investigating whether this trend is significant enough to warrant adjustments to the NCS
control law.
Because the NICMOS detectors react sensitively to temperature variations, the superb
stability of the cooling system is extremely positive news for the scientific performance of
NICMOS. In what follows, we discuss in some more detail the characteristics of the NICMOS
detectors in the NCS era.
2.2.
The Early NICMOS Calibration Program
NICMOS datasets consist of a series of non-destructive detector readouts, with varying
time intervals (∆-times) between reads. The observer can choose from a number of predefined sequences that are designed to optimize the dynamic range for a variety of science
projects. For details about the readout sequences, we refer to the NICMOS instrument
handbook at http://www.stsci.edu/hst/nicmos/documents/handbooks/v5/ . A number of
proposals using different readout sequences were executed early in the SMOV process to
assess the NCS performance and to obtain essential information about the NICMOS health
and detector performance. Table 1 summarizes the programs that were used to derive the
results presented in this paper. To a large extent, the data analysis procedures follow those
of the NICMOS warm-up monitoring program after the cryogen depletion in early 1999.
The analysis has been discussed in detail in Böker et al. (2001), and hence will not be
repeated here.
2.3.
Detective Quantum Efficiency
The detective quantum efficiency (DQE) of the NICMOS detectors changes as a function
of temperature, in the sense that higher operating temperatures result in higher sensitivity.
NICMOS Detector Performance
Table 1.
225
Summary of Early NICMOS SMOV Programs
Program #
8944
8945
8975
8985
Purpose
Filter wheel functional
Dark current monitor
Readnoise & Shading
DQE
Filter
All
Blank
Blank
All
Readout Sequence
ACCUM
SPARS64
SCAMRR & STEP256
many
Figure 2. NICMOS DQE: comparison between post-SM3B (at operating temperature of 77 K) and 1997/1998 (62 K) eras.
This is one of the reasons why the re-instated NICMOS under NCS control was expected
to be more efficient than during the solid cryogen era—at least for some science programs.
This section quantifies the gain in DQE at the new operating temperature.
Relative changes of the NICMOS DQE can be measured from “flat-field” exposures
generated from a pair of “lamp off” and “lamp on” exposures. Both are exposures of the
(random) sky through a particular filter, but one has the additional signal from a flat field
calibration lamp. Differencing these two exposures then leaves the true flat-field response.
The countrate in such an image is a direct (albeit relative) measure of the DQE. The DQE
increase of the three NICMOS cameras between 77 K and 62 K as a function of wavelength
is presented in Figure 2.
From the DQE monitoring program during the 1999 instrument warmup, we were able
to construct a model that predicts the DQE as a function of wavelength and temperature.
This model—together with dark current predictions—provided the basis for the sensitivity
calculations in the NICMOS exposure time calculator (ETC), a widely used web tool for
NICMOS users. With the new post-SM3B data, we are now able to test the accuracy of
226
Böker, et al.
Figure 3.
Comparison of NICMOS DQE as measured during April 2002 to predictions based on data from the 1999 instrument warmup.
this model, and to verify that the predicted exposure times for all NICMOS science projects
are correct. In Figure 3, we compare the new DQE measurements through the NICMOS
filter set to the model predictions. The plot demonstrates that the new data agree with the
model predictions to within a few percent, which means that earlier sensitivity calculations
are indeed accurate for the revived NICMOS.
Because the flat-field lamps inside NICMOS have not been calibrated in an absolute
sense, these data can not be used to determine the absolute DQE of the NICMOS detectors.
This is a notoriously difficult problem, which typically relies on observations of standard
stars with a well-known energy distribution over the NICMOS wavelength range. This
calibration program is currently still under analysis. For the purpose of this paper, we
show in Figure 4 the absolute DQE of the three NICMOS detectors—measured during
NICMOS ground-testing at ≈ 60 K—compared to the same curve scaled by the linear fits
to the relative DQE increase from Figure 2. This gives another impression of the sensitivity
improvement under the NCS.
2.4.
Read-out Noise
Each NICMOS detector has four independent readout amplifiers, each of which reads a
128 × 128 pixel quadrant. The noise associated with the amplification process, commonly
referred to as read-out noise, has been shown to be independent of temperature (Böker
et al. 2001). However, it is important to answer whether the three-year hiatus in space has
in any way degraded the performance of the NICMOS electronics.
Subtracting the first two reads of a SPARS64 sequence eliminates all effects of bias
variations or shading. The effective integration time of this difference image is only 0.3 s,
too short for the linear dark current signal to become important. Therefore, the RMS
deviation of the pixel values across the detector array is an accurate representation of the
NICMOS Detector Performance
227
Figure 4.
Approximate improvement of absolute DQE of the NICMOS detectors. Here, the pre-launch measurements have been scaled by the relative DQE
increase as shown in Figure 2.
intrinsic read-out noise of the detectors. From a series of about 400 such first-difference
images, we determined the average RMS pixel-to-pixel variation across the three NICMOS
detector arrays. The results, which are summarized in Table 2, indicate that the NICMOS
noise levels are not in any way degraded from the pre-NCS values.
Table 2.
Read-out Noise of NICMOS Detectors
Camera
NIC1
NIC2
NIC3
2.5.
RMS
[DN]
5.1±0.15
5.1±0.15
4.6±0.15
Gain
5.4
5.4
6.5
Readnoise
[e− /readpair]
27.5±0.8
27.5±0.8
29.9±0.9
Linear Dark Current
The linear dark current (i.e., that component of the signal in a “dark” exposure that accumulates linearly with time) of the NICMOS detectors had been the subject of much debate
and speculation over the last few years. The reason for this was an anomalously high dark
current in the temperature range between 78 and 85 K that was observed during the 1999
warmup (Böker et al. 2001). The elevated dark current would have compromised NICMOS
sensitivity (if it had to be operated in this temperature regime), and hence the question of
whether or not the anomaly would be present after the cooldown was an important one.
228
Böker, et al.
60.18
105
Dark Current (e-/s)
104
71.46
Temperature (K)
87.96
114.3
163.3
0.0114
0.0087
1/T (Inverse Temperature)
0.0061
CAM 1
EOL Data
SM3B Data
103
102
101
100
0.147
10-1
105
CAM 2
Dark Current (e-/s)
104
103
102
101
100
0.114
10-1
105
CAM 3
Dark Current (e-/s)
104
103
102
101
100
0.166
10-1
0.0166
0.0140
Figure 5.
Linear dark current of all three NICMOS cameras as a function of
temperature. The open symbols are from the 1999 instrument warm-up, while the
closed symbols mark the new data obtained during and after the NCS cooldown.
Note the good agreement in the exponential increase towards high temperatures,
and the fact that the mean dark current at the new operating temperature of 77 K
(plotted in e− s−1 ) is significantly lower than during the warm-up.
NICMOS Detector Performance
229
The safing of the NICMOS instrument prevented dark current measurements for the
better part of the cooldown. However, some non-saturated datasets were taken before the
NICMOS safing, and the measurements resumed as soon as the NCS had reached its target
temperature. These new measurements are shown in Figure 5, compared to the results of the
1999 warm-up monitor. It is clear that while the exponential increase at high temperatures
is fully consistent between the two datasets, the anomalously high dark current around 82 K
is absent in the new data.
One minor concern is the fact that all three NICMOS arrays show an increased number
of “hot” pixels, i.e., pixels with higher-than-average linear dark current. This is not entirely
unexpected because the warmup monitor did show a similar behavior, although it was
unclear at the time whether this was a transition effect that would disappear once the
detectors stabilize in temperature. The new data show that the hot pixels are indeed a
genuine feature of the new operating temperature. However, they can be fully corrected for
by dithering NICMOS observations.
2.6.
Amplifier Glow
Amplifier glow is a well-known feature of NICMOS-3 arrays. It manifests itself as a spatially
variable, but highly repeatable signal component in every detector read-out. The signal is
highest in the corners of the array, i.e., closest to the read-out amplifiers, and gets fainter
towards the center of the array. Typical values for the amplifier glow are 2 DN/read in
the center of the array, and up to 15 DN/read in the corners, independent of detector
temperature. The signal is extremely repeatable and can be well modeled and removed
during pipeline calibration.
The amplifier glow is measured by subtracting the first two reads in a STEP64 sequence
which are only 0.3 s apart, thus making the signal contribution from the linear dark current
negligible. In agreement with expectations, the amount and structure of amplifier glow is
unchanged compared to NICMOS data obtained in 1997/1998.
2.7.
Reset Level and Self-Calibration
The reset level (i.e., the count level immediately following a detector reset) of the NICMOS
detectors is extremely sensitive to temperature. With proper calibration, this fact can be
exploited to use the reset level as a detector thermometer. In the absence of other reliable
information, e.g. from diode sensors, this method can provide a means to correct for various
temperature-dependent effects.
Prior to the on-orbit testing of the NCS, estimates of its performance remained highly
uncertain, both in terms of final operating temperature and stability. In addition, there was
a real possibility that the operating temperature would fall above 78 K in which case the
internal temperature sensors would become unusable due to the limits of their analog-todigital converters. For these reasons, the STScI NICMOS group spent a significant amount
of work in order to use the reset levels to essentially enable self-calibration of NICMOS
data.
However, the surprising stability of the NCS, and the fact that at least some diode
sensors within NICMOS are still within their usable range, allow consideration of an alternative strategy for NICMOS calibration. Given temperature variations of less than 0.1 K
RMS since the start of active NCS control in mid-April, the NICMOS calibration pipeline
could possibly rely on reference files taken at the actual operating temperature, at least for
the near future. This possibility is currently under study. However, should the temperature
stability of the NCS degrade significantly (RMS > 0.1 K), data self-calibration via the reset
levels is the only viable option for the NICMOS pipeline.
230
Böker, et al.
2.8.
Saturation Levels and Dynamic Range
The saturation level of a given detector pixel is defined by the amount of charge “loaded”
onto it during the detector reset (i.e., the well depth). The reset voltages of the NICMOS detectors and—to a much smaller degree—the capacitance of the pixel both vary
with temperature, and hence the pixel saturation levels will also depend on the operating
temperature. The flat field exposures taken after the cooldown allow us to measure the saturation levels at the new NICMOS operating temperature of 77K. The data suggest that for
cameras 1 and 2, the average pixel saturates around 25,000 DN (or ≈ 135, 000e−s−1 ) which
is about 15% lower than during the nitrogen period. For camera 3, the reduction is only
about 7%. This rather small loss in dynamic range of NICMOS data can be compensated
for by a proper choice of readout sequences, so that NICMOS will be able to execute the
same wide range of science projects as before.
2.9.
Cosmic Ray Persistence
The NICMOS detector are susceptible to image persistence from a bright source, for example a luminous star or a cosmic ray hit. Persistence signal from cosmic ray hits is a
problem especially for exposures taken soon after an HST passage through the South Atlantic Anomaly (SAA), a region in the Earth’s atmosphere with higher than average cosmic
ray incidence. The random spatial distribution of the cosmic ray “afterglows” is effectively
an additional component to the dark current. It increases the noise level in the image and
therefore limits faint source detections.
Some of the dark current monitoring data were taken soon after an SAA passage. Our
analysis of these datasets indicates that the persistence signal decays exponentially over a
period of about 30 min. This is the same timescale as in the solid nitrogen period, and hence
new NICMOS datasets will be affected by cosmic rays in much the same way as the old ones
did. However, for the upcoming NICMOS science program, STScI plans to automatically
schedule a pair of dark exposures immediately following each SAA passage in order to
provide a map of the persistent cosmic ray afterglow. Experiments using 1998 NICMOS
data have shown that it is possible to scale and subtract such “post-SAA” darks from
subsequent science exposures, and thus to significantly reduce the impact of CR persistence.
A more detailed advisory on how to use the “post-SAA” darks will be distributed soon.
2.10.
Detector Cosmetics
All NICMOS detector arrays have a small number of particulates on their surfaces. These
are believed to be flecks of black paint scraped off the baffles during the mechanical deformation that led to the accelerated cryogen depletion. There was some concern that the
warmup and subsequent cooldown with their associated mechanical motions could have
produced an increased number of these particles. Fortunately, these concerns are not confirmed. Except for one additional particle in the lower right corner of camera 1, no new
contaminants are apparent in the SMOV data. Moreover, the number of “bad” pixels (i.e.,
pixels with significantly degraded responsivity) is roughly the same as before, so that the
cosmetic appearance of the NICMOS detectors is as good as in the pre-NCS era.
3.
Summary
We have described various aspects of the detector performance of the reinstated NICMOS
instrument. All parameters are in line with expectations, and will enable a “better-thanever” scientific performance of NICMOS. In particular, the DQE increase agrees very well
with models derived from the 1999 warmup, and the linear dark current shows no sign of
anomalously high levels. Other parameters such as readnoise, amplifier glow, or detector
cosmetics are unchanged compared to the “old” NICMOS. Together with the stable and
NICMOS Detector Performance
231
quiet performance of the cooling system, HST and NICMOS are in great shape for new,
exciting scientific discoveries.
Acknowledgments. It is impossible to name all individuals who have helped to make
the idea of a NICMOS revival become a reality. We are indebted to the NCS team at GSFC
for making the cooling system a success, to the GSFC thermal group for their continued
assistance in modeling the NICMOS/NCS system, and to the NICMOS group at UofA, in
particular G. Schneider and R. Thompson, for invaluable help in the design and analysis of
the NICMOS verification and calibration program.
References
Böker, T., et al. 2001, PASP, 113, 859
Cheng, E. S., Smith, R. C., Jedrich, N. M., Gibbon, J. A., Cottingham, D. A., Swift, W. L.,
& Dame, R. E. 1998, SPIE Proc., 3356, 1149
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
NICMOS Photometric Calibration
Mark Dickinson, Megan Sosey
Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218
Marcia Rieke
Steward Observatory, Tucson, AZ
Ralph Bohlin, Daniela Calzetti
Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218
Abstract. We review procedures and measurements used to calibrate the photometric zeropoints of the HST Near Infrared Camera and Multiobject Spectrograph
(NICMOS). New spectrophotometric models of solar analog and white dwarf standard stars have been calibrated and tested against ground-based photometry, as well
as against one another using NICMOS observations through many filters. These
are then used to derive the bandpass-averaged flux densities of these stars through
the NICMOS filters. We describe the characteristics of the on-orbit standard star
observations made in Cycle 7, and the procedures used to correct finite aperture
measurements to total count rates, and finally to derive the NICMOS zeropoints.
We conclude with preliminary measurements of the increase in delivered quantum
efficiency (DQE) at the warmer operating temperatures for Cycle 11.
1.
Introduction
The photometric calibration of NICMOS is challenging for several reasons. During the first
lifetime of NICMOS, in Cycles 7 and 7N, our knowledge of the camera, detectors, and
of the means to process and calibrate the data evolved substantially. This affected the
quality of early NICMOS photometric calibrations. There are still limits to the quality and
reproducibility of NICMOS photometry due to certain instrumental effects such as intrapixel
sensitivity, as discussed below. Second, flux calibration in the near-infrared is challenging
in general, due to the lack of absolutely calibrated spectrophotometry for primary standard
stars. Moreover, ground-based calibrations are restricted to measurements in the J, H
and K atmospheric windows. NICMOS filters have no such limitation, and indeed do
not correspond to standard, ground-based bandpasses, adding further complication to the
process of photometric calibration.
Here we review the procedures used to calibrate NICMOS photometry in Cycle 7, when
the instrument was operating at T ≈ 61.5 K. In era of the NICMOS Cooling System (NCS),
the instrument is significantly warmer, T = 77.1 K. The detector DQE has changed (increased) substantially, and the photometric calibration constants have changed as well. At
the time of this writing, the recalibration of NICMOS photometry for Cycle 11 is underway,
and we present only very preliminary results.
232
NICMOS Photometric Calibration
2.
233
NICMOS Standard Stars
In order to calibrate NICMOS photometry, we must observe standard stars for which we
believe we understand the flux distribution with wavelength. No direct spectrophotometric
observations of standard stars with continuous wavelength coverage are available over the
near-infrared wavelengths covered by NICMOS. Instead, we choose stars for which we believe we understand the expected spectrophotometry, and then check and calibrate this with
ground-based broad-band photometry. Two types of stars have been used: solar analogs,
and hydrogen white dwarfs. The near-IR spectral energy distribution of the Sun is known
to reasonably good accuracy (Colina, Bohlin & Castelli 1996), and for a good solar analog we may scale the solar spectrum to match the JHK ground-based photometry of the
standard star and adopt it as our model. Hydrogen white dwarf stars have comparatively
simple atmospheres which have been accurately modeled, and are commonly used for HST
calibration at UV and optical wavelengths. These models have been extended to infrared
wavelengths, and then as with the solar analogs, they can be scaled to match ground-based
JHK photometry of a particular standard star.
The NICMOS photometric calibration programs in Cycle 7 observed several different
standard stars. The solar analog P330E and the white dwarf G191B2B were the primary
stars, observed through all filters in all cameras. P330E was also the main target for
the NICMOS photometric monitoring program, and thus has many repeated observations
through a limited set of filters over the instrument lifetime. Two other “ordinary” standards
were observed through a limited subset of NICMOS filters: the solar analog P177D, and
the white dwarf GD153. Finally, three red stars, BRI0021, CSKD−12, and OPH−S1, were
observed to assess filter red leaks and color transformations. We will not discuss those red
stars further at this time.
Ground-based JHK photometry of the solar analogs and the white dwarf G191B2B
was calibrated by Persson et al. (1998, and priv. comm.), primarily from the Las Campanas
Observatory. We will refer to those observations as photometry on the “LCO system,”
which was nominally calibrated to the “CIT system” of Elias et al. (1982). The other
white dwarfs are UKIRT faint standard stars (Hawarden et al. 2001), and we have derived
empirical color transformations to place those observations onto the LCO system.
2.1.
Absolute Flux Calibration for the LCO System
Campins, Rieke & Lebofsky (1985) derived an absolute flux calibration for the Arizona
infrared photometric system using the solar analog method.1 Their Table IV summarizes
their derived absolute flux measurements for α Lyr at J and K from Blackwell et al. (1983),
who compared stellar photometry to an absolutely calibrated terrestrial source, and at J,
H and K from the Campins et al. solar analog method. They adopt an average value for
each bandpass, which we use as our assumed flux calibration for the Arizona system.
The effective wavelengths and bandwidths of the Arizona photometric passbands are
not identical to those of the LCO filters (see Figure 1). Therefore we must correct the
Campins et al. flux zeropoints measurements to the LCO bandpass system. To do this, we
have used a Kurucz model atmosphere for α Lyr. We have not assumed that the absolute
flux of this model is correct, nor even that its shape accurately represents that of Vega over
a long wavelength baseline, but have only used it to transfer the empirical flux calibration
from Campins et al. to the LCO bandpasses:
fν (CRL) ×
1
fν (model)LCO
= fν (LCO, transferred),
fν (model)Az
In the Arizona system, α Lyr has J = H = K = 0.02, whereas in the CIT system it is defined to have
mag = 0.
234
Dickinson, et al.
Figure 1.
Comparison of the J, H, K, and Ks bandpasses used in the standard
star photometry of Persson et al. (1998) (the “LCO system”) with tapered boxcars
representing the bandpasses of the Arizona system, for which Campins et al. (1985)
have presented absolute flux calibration. The effects of atmospheric absorption on
the net bandpass functions are also illustrated.
where fν (CRL) is the Campins et al. flux density measurement, and fν (model)X indicates the bandpass-averaged flux density computed using the α Lyr model and filter system
X. In this way, we assume that (1) our bandpass functions for the two systems are accurately determined, and (2) the Vega model is relatively accurate over the wavelength
interval between the two “versions” of a given bandpass.
Given the dimensionless system throughput (filter transmission times DQE) for the
LCO filter system, it is straightforward to integrate the spectrophotometric models through
the LCO bandpasses to derive bandpass-averaged flux densities. Compare these to the
ground–based photometry, however, requires an absolute flux calibration for the LCO photometric system. The LCO system is calibrated to the CIT standards of Elias et al. (1982).
The CIT system is nominally calibrated so that α Lyrae has zero magnitude at JHK. We
assume here that this is correct, and thus require an absolute flux calibration for α Lyrae
through the LCO filters.
3.
Spectrophotometric Data and Models
In this way, we have JHK photometry in the LCO system for the four basic NICMOS standards, as well as Ks photometry for the solar analogs. Bohlin, Calzetti & Dickinson (2001)
obtained high-quality STIS spectrophotometry in the wavelength range 1150–10200 Å for
HST standard stars including P330E and G191B2B. They have made a detailed comparison
to stellar atmosphere models in terms of effective temperature, reddening, etc. They use
the solar spectrum and white dwarf atmosphere models to extend the spectrophotometry
to near-infrared wavelengths.
We integrated the spectrophotometric models through the LCO infrared bandpasses to
generate synthetic photometry. Figure 2 shows the relative magnitude differences between
these models and the observed JHK and, for some objects, Ks , photometry. (The points for
NICMOS Photometric Calibration
235
Figure 2.
Magnitude differences between ground-based photometry of NICMOS
standard stars and synthetic photometry derived by integrating the model spectra
of Bohlin et al. (2001) through the nominal NICMOS bandpasses. The panel at
upper left shows the result of integrating the Vega model atmosphere retrieved
from CDBS (Calibration Database and Operations) at STScI, compared to an
assumed m = 0.
“Vega” compare the CDBS model Vega spectrum to the definition m = 0. The agreement
is not particularly good, suggesting that the absolute flux calibration of the Vega model in
the near-infrared is not particularly accurate.) For most of the NICMOS standards, the
agreement at J and K is well within the uncertainties (taken as the quadrature sum of
the ground-based standard star photometry and the quoted systematic uncertainty of the
Campins et al. JHK flux zeropoints). At H-band, all of the stars fall systematically below
zero offset, although in general by something only slightly larger than the uncertainty in
the Campins et al. calibration. Since this is seen for both the solar analog and white dwarf
model spectra, and for stars with both LCO and UKIRT photometry, we are inclined to
believe that the error is mainly in the Campins et al. flux calibration for the H-band, which
also required the largest color correction between the Arizona and LCO/CIT systems. We
therefore adopt the Bohlin et al. model spectra for NICMOS flux calibration.
4.
NICMOS Cycle 7 Standard Star Observations
The NICMOS standards P330E and G191B2B were observed in various Cycle 7 calibration
programs. For some filters there were a large number of independent observations, while
for others there were only a few (as few as three) exposures taken during Cycle 7. The data
were processed using the most up-to-date calibration reference files (including temperaturedependent dark frames) and pipeline software, together with some post-processing to remove
the NICMOS “pedestal” offset and residual bias “shading.” Photometry for all stars was
236
Dickinson, et al.
Figure 3.
Examples of NICMOS Cycle 7 on-orbit measurements of the count
rate from the standard star P330E, in two filters from NICMOS camera 1. F110W
(left) was observed many times throughout Cycle 7 as part of the photometric
monitoring program. Other filters, such as F166N (right) were observed only
once, with three dither positions.
measured in a set of standard apertures, with diameters 1 , 1 , and 2 for cameras 1, 2
and 3, respectively. Examples of the photometry for two filters in camera 1, showing cases
with many and few individual exposures, are shown in Figure 3. Occasional outlier points
were rejected—these generally occurred when the star fell onto some bad pixels, “grot,” or
the unstable central column or row of the detector. Unfortunately, when only 3 exposures
were obtained, this sometimes led to a “2-out-of-3” vote on the mean count rate.
The photometric scatter is acceptably small for most filters in cameras 1 and 2, although
for some of the narrow band filters the signal-to-noise ratio of the individual exposures was
disappointingly low. For camera 3, the scatter was appreciably larger, especially at short
wavelengths. This is due to the effects of intrapixel sensitivity—the NICMOS detector
responsivity varies within the area of a pixel, and the highly undersampled PSF of the short
wavelength NIC3 observations results in significant count rate variations for a point source
depending on where it was centered within the pixel. This can be seen in the left panel of
Figure 4, which shows NIC3 photometry for F110W. At longer wavelengths, the diffraction
limit broadens the PSF and the effects of undersampling and intrapixel sensitivity variations
are substantially smaller (right panel of Figure 4). The RMS of the count rate measurements
(after outlier rejection) in camera 3 is plotted versus wavelength in Figure√5. For filters
with a large number N of exposures, this RMS averages down by roughly N , providing
acceptable mean count rates. However, some NIC3 filters have only a few observations,
leading to large uncertainties in the final calibrations.
Aperture corrections to total count rates were computed using Tiny Tim (Krist &
Hook 2001) model PSFs, generated for each camera and filter combination. These were
“photometered” using the same aperture sizes and background annuli to account for PSF
spill-over into the regions used for background subtraction. The adopted aperture corrections are shown in Figure 6, and depend strongly on wavelength as the diffraction limit
varies. Early Cycle 7 photometric calibration adopted overly simplistic, constant aperture
corrections for all filters, leading to significant errors in the derived zeropoints.
NICMOS Photometric Calibration
Figure 4.
More NICMOS Cycle 7 standard star count rate measurements, this
time for camera 3. The effects of intrapixel sensitivity variations introduce a large
scatter in the measured count rates for undersampled point spread functions, such
as the F110W data shown at right. This is reduced at longer wavelengths, where
the broader diffraction limit reduces the impact of undersampling.
Figure 5.
The RMS of individual standard star measurements for NICMOS
camera 3 filters, plotted as a function of wavelength. The reduced scatter at
longer wavelengths is readily apparent. (Circles: P330E; Squares: G191B2B)
237
238
Dickinson, et al.
Figure 6.
Aperture corrections from measured to total count rates, derived from
Tiny Tim PSF models, plotted against filter effective wavelength. The horizontal
lines show the constant aperture corrections that were adopted for preliminary
NICMOS photometric calibrations early in Cycle 7.
5.
Cycle 7 Zeropoints
NICMOS photometric calibration is encoded in the image headers (and PHOTTAB reference files) with the keywords PHOTFNU and PHOTFLAM, representing the bandpassaveraged flux densities (in fν and fλ , respectively) corresponding to a count rate of 1 ADU/
second. These are computed by integrating the standard star spectrophotometric models
through the NICMOS bandpass functions (including all optical and detector throughput
elements), and dividing by the measured, aperture-corrected mean standard star count
rates.
With earlier calibrations, there was a significant discrepancy between zeropoint results
for the solar analog P330E and the white dwarf G191B2B. With the new spectrophotometric
models, the agreement is generally very good (see Figure 7)—within the measurement errors
for many filters, generally < 3%, and rarely > 5%. The agreement is also excellent for the
ground-based JHK photometry, also shown on Figure 7; here, any uncertainties in the
Campins et al. absolute flux calibrations divide out, showing that the two stars genuinely
agree with one another to good precision. The agreement is best when we use recent
non-LTE model atmospheres with metals to represent the white dwarf (Hubeny, private
communication), rather than earlier, pure hydrogen, LTE models, and we have therefore
adopted these for the NICMOS calibrations.
The final NICMOS photometric calibration constants were taken to be the unweighted
mean of values derived for P330E and G191B2B, in order to average over the systematic
uncertainties of the spectrophotometric models for these two stars. The derived Cycle 7
calibration constants are given in the NICMOS Data Handbook.
NICMOS Photometric Calibration
239
Figure 7.
Ratio of photometric calibration constants (PHOTFNU) derived for
two different standard stars: the solar analog P330E, and the white dwarf
G191B2B. The calibrations agree within 3% for most NICMOS filters and well
within the errors for the ground-based photometry, giving confidence that the
spectrophotometric models of these two very different stars are accurate.
6.
Cycle 11 Recalibration
The NICMOS operating temperature under control of the NCS is substantially warmer than
it was in Cycle 7, necessitating a complete recalibration of the photometric zeropoints for
Cycle 11. P330E and G191B2B have been reobserved, and care was taken to increase the
number of dither positions, especially for camera 3, where intrapixel sensitivity variations
are most important. As of this writing, a preliminary analysis of the standard star data has
been carried out by Marcia Rieke of the NICMOS IDT. Early results, showing the ratio of
total throughput for Cycle 11 compared to Cycle 7, are shown in Figure 8. As anticipated,
the detector DQE is substantially higher at the warmer operating temperature, particularly
at shorter wavelengths; the gain is ∼ 60% for F110W, ∼ 35% for F160W, and ∼ 20% for
F222M. This increase in DQE results in improved signal-to-noise ratios for most observations, more than offsetting the increased noise from dark current. Final analysis of the new
NICMOS zeropoints has awaited the recalibration of other aspects of NICMOS instrument
and detector performance, particularly flat fields, darks, and nonlinearity corrections. When
these are complete (and as of this writing this is nearly the case), the standard star observations will be reprocessed and reanalyzed, and final photometric keywords will be derived
and placed into the PHOTTAB table used by the OPUS data processing pipeline. Until
that time, NICMOS data retrieved from the STScI archive will have incorrect header information for PHOTFNU and PHOTFLAM; observers should watch the STScI NICMOS web
pages and the Space Telescope Analysis Newsletter for updates as they become available.
240
Dickinson, et al.
Figure 8. Preliminary comparison of NICMOS sensitivity in Cycle 11 to that
from Cycle 7, based on an early analysis of Cycle 11 photometric calibration data
for P330E, carried out by Marcia Rieke.
References
Blackwell, D. M., Leggett, S. K., Petford, A. D., Mountain, C. M., & Selby, M. J. 1983,
MNRAS, 205, 897
Bohlin, R. C., Dickinson, M., & Calzetti, D. 2001, AJ, 122, 2118
Campins, H., Rieke, G. H., & Lebofsky, M. J. 1985, AJ, 90, 896
Colina, L., Bohlin, R. C., & Castelli, F. 1996, AJ, 112, 307
Elias, J. H., Frogel, J. A., Matthews, K., & Neugebauer, G. 1982, AJ, 87, 1029
Hawarden, T. G., Leggett, S. K., Letawsky, M. B., Ballantyne, D. R., & Casali, M. M. 2001,
MNRAS, 325, 563
Krist, J. & Hook, R., 2001, The Tiny Tim User’s Guide (Baltimore: STScI)
Persson, S. E., Murphy, D. C., Krzeminski, W., Roth, M., & Rieke, M. 1998, AJ, 116, 2475
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
NICMOS Grism Calibrations
Rodger I. Thompson
Steward Observatory and Department of Astronomy, University of Arizona,
Tucson, AZ 85721
Wolfram Freudling
Space Telescope-European Coordinating Facility, Garching, Germany
Abstract. Installation of the NICMOS cryocooler has restored NICMOS to operational status and has necessitated a recalibration of its primary capabilities including
grism observations. This paper describes the results of the recalibration. Most properties of the grisms are essentially unchanged. The principal change has been the
different quantum efficiency versus wavelength properties of the detectors. The quantum efficiency is higher at all wavelengths, particularly so at the shorter wavelengths.
1.
Introduction
The Near Infrared Camera and Multi-Object Spectrometer (NICMOS) contains three grisms
for multi-object infrared spectroscopy with the Hubble Space Telescope (HST ). A grism is
a combination of grating and prism that produces a dispersed spectrum with a selected
wavelength at essentially the same position as the image of the object in the field of view.
The selected wavelength is determined by a combination of the dispersions of the grating
and prism. In the case of the NICMOS grisms an interference filter was coated on the front
face of the grism to prevent mixing of orders. The front face of the grism is perpendicular
to the optical axis of the incoming light and the grating is ruled on the back side of the
grism. The ruled faces of the grating grooves are parallel to the front face to provide the
proper blaze for the first order. Depending on the location of the object in the image, the
zero, first, and second orders can appear in the field of view.
The grisms are in the Camera 3 filter wheel. When in position they are very near the
cold pupil of the camera. The optical beam passes through the grism at f/17 as opposed to
the usual parallel beam. The effect of the convergent beam on the spectrum is negligible
given the large 0.2 arcsecond pixels in camera 3. There are three grisms, G096, G141, and
G206 with undeviated wavelengths of 0.94, 1.40, and 2.04 microns.
On March 8, 2002 the NICMOS Cooling System (NCS) was installed on the HST
to restore NICMOS observational capabilities. The cooler established the final detector
(mounting cup sensor) set point temperature on May 10, 2002 at 77.1 degrees Kelvin and
the vapor cooled shield at 112 degrees Kelvin. The temperature of the vapor cooled shield
will be approximately the temperature of the NICMOS grisms that reside in the camera 3
filter wheel which is thermally tied to the vapor cooled shield. Both of these temperatures
are different from the operation in Cycle 7, requiring a recalibration of the grisms. The
primary difference in operation is due to the different operating temperature of the NICMOS detectors. The quantum efficiency versus wavelength of each camera 3 pixel must be
redetermined to accurately reduce grism spectra. The warmer grism temperature can result
in a slightly different dispersion solution. Mechanical distortion created by the instrument
warming and subsequent recooling can produce geometric effects such as a change in the
tilt of the spectrum on the detector.
241
242
Thompson and Freudling
Figure 1.
2.
Image and Spectrum Offset.
Observations
Proposal 8991 contains all of the grism calibration observations. The solar analog star
P330E provides a flux calibrated source to determine the grism-filter efficiency and geometric
factors. The compact emission line object Hubble 12 provides multiple spectral lines in each
grism for the calculation of the dispersion relation. The NICMOS detectors have significant
differences in sensitivity over the area of the detector and within the area of a single pixel.
To mitigate this effect images and spectra of each of the two objects were taken at 15
different positions in the camera 3 field of view. The entire first order spectrum is visible
at all positions and the zero or second orders are visible at many of the positions.
3.
Data Analysis
The basic steps in the data analysis are decomposition of the multiaccum integration into
the individual image readouts minus the first read to remove kTC noise, subtraction of
the appropriate dark individual readouts for the same multiaccum pattern, correction for
linearity, removal of cosmic rays, flat fielding if it is an image but not if it is a spectrum and
then correction of known bad pixels. All of this is done with IDL procedures developed by
the first author.
4.
Geometric Parameters
The geometric parameters for the grisms include the dispersion relation in terms of wavelength per pixel, the angle of the dispersed spectrum relative to the axes of the detector, the
offset of the spectrum from the image position and the wavelength of the image position.
4.1.
Offset of the Spectrum from the Image and Spectrum Slope
Taking the dispersion direction as the x axis, the image location is offset in the y direction
from the spectrum as shown in Figure 1, which is a combination of the F160W image and
the G141 spectrum for one of the dither pattern positions. The offset of the spectrum from
the image position for the 15 P330E image and spectra was measured for all three grisms
by the following process.
The centroids of the F160W image positions were measured with the IDL based procedure Image Display Paradigm (IDP3) developed by the IDT. Note that a check of this
procedure relative to several different IRAF procedures revealed significant differences in
results on the order of 0.5 pixels. The results from the IRAF procedures were found to vary
by the same order of magnitude relative to each other. Because of this the results reported
in this report differ from the results used at the Space Telescope–European Coordinating
Facility (ST-ECF) for HST in Garching, Germany. For a reason not known at the present
time the largest differences occur in the IRAF routine IMCENT.
NICMOS Grism Calibration
243
The spectral positions were determined by finding the peak signal position in y for
each column of pixels intersecting the first order by gaussian fitting and then finding the
best least squares linear fit to those values. In practice this fit always intersected the zero
or second order if they were present in the image. The offset was taken as the difference in
y pixel values between the centroid of the image and the fit to the spectrum at the x value
of the centroid. The slope of the fit also gave the slope of the spectrum relative to the x axis
of the detector in y pixels per x pixel. The angle of the spectrum is taken as positive for a
counter clockwise rotation. The values of the parameters are given in Table 1. The quoted
standard deviations are only from the statistics of the 15 fits.
Table 1.
Grism
G096
G141
G206
4.2.
NCS Era Grism Offset and Slope Parameters
Offset
−4.3
−6.7
−2.0
Offset std dev
0.4
0.4
0.4
Slope
0.054
0.013
0.023
Slope std dev
0.003
0.003
0.003
Angle in Degrees
3.1
0.73
1.3
Grism Dispersion Relations
The line spectra of Hubble 12 defined the spectral dispersion relation for each grism. The
IDL procedure for determining the dispersion assumes that the dispersion relation is linear.
This assumption was checked against the observed lines and appears to be accurate within
the spectral resolution of the grisms. The procedure takes three prominent and widely
spaced lines in each spectrum and finds the best fit using the IDL function POLY FIT.
Table 2 gives the vacuum wavelengths of the lines used in the analysis. The dispersion
solution is for vacuum wavelengths.
Table 2.
Wavelengths in Microns of the Lines Used in the Dispersion Solution
Grism
G096
G141
G206
Line 1
1.0832
1.8756
2.16609
Line 2
1.00507
1.28216
2.0581
Line 3
.953461
1.0832
1.8756
Note that adjacent wavelength grisms have a line in common to make sure the dispersion
solutions match across grism boundaries. An interactive program displays each of the 15
Hubble 12 spectra for a given grism and asks the operator to mark the position of the
3 lines used for the dispersion solution. The line center is found by fitting a Gaussian to
the line using the IDL function GAUSSFIT. The center wavelength (the constant) and the
dispersion (the slope) are the averages of the 15 measurements and the standard deviation
is calculated from the scatter of the measurement values by the IDL function MOMENT.
Table 3 shows the results of the measurements. Note that the standard deviation is from the
measurements only and does not reflect any possible systematic errors. Only the first order
dispersion was measured. The second order should be 1/2 of the dispersion of the first order
giving spectra at twice the resolution of the first order. Table 4 gives the regions of the zero,
first, and second orders relative to the image position which marks the center wavelength.
None of the positions of the spectra were far enough offset to see the long wavelength end
of the second order, therefore those positions are upper limits on the extent. Note that the
second order fades toward the long wavelength end as the angle of dispersion get further
from the blaze angle.
244
Thompson and Freudling
Table 3.
Grism
G096
G141
G206
Table 4.
Dispersion Solutions for the First Order of the NICMOS Grisms
Center Wavelength
(microns)
0.9415
1.396
2.039
Center Wavelength
Std. Dev
0.002
0.001
0.001
Dispersion
(microns/pixel)
−0.00552
−0.008016
−0.011353
Dispersion
Std. Dev.
0.0001
5 × 10−6
4 × 10−5
Pixel Positions for the Grism Orders
Grism
G096
G141
G206
Zero
+163–+167
+165–+170
+166–+173
First
−46–+27
−65–+40
−43–+58
Second
−110–< −184
−91–< −184
−57–< −184
Using the dispersion solutions listed in Table 3 the wavelengths of the observed spectrum of Hubble 12 were calculated. Figure 2 through Figure 4 show the observed grism
spectrum with the higher resolution ground based spectra overplotted. The ground based
spectra in the J, H, and K bands are from Luhman & Rieke 1996 and the shorter wavelength spectra are from Rudy et al. 1993. Since the Rudy data was not available in digitized
form, narrow triangular lines were just placed at the wavelengths of the observed lines in
the published line list, adjusted to vacuum wavelengths.
5.
Efficiency Calibration
The second part of the grism calibration is a determination of the efficiency. The observed
signal for a pixel in the spectrum is given by
S(λ) = F (λ)GE(λ)F E(λ)QE(λ) ,
(1)
where F (λ) is the object flux in photons per second, GE(λ) is the grism efficiency,
F E(λ) is the filter efficiency and QE(λ) is the quantum efficiency of the pixel. Since the
filter is applied directly to the front face of the grism it is combined with the grism efficiency
for a total grism efficiency such that
G(λ) = GE(λ)F E(λ)
5.1.
(2)
Input Spectrum
The solar analog P330-E provides the known input flux and spectrum F (λ). The detailed
solar spectrum (Wallace, Hinkle & Livingston 1993, Livingston & Wallace 1991) is matched
to the fluxes and slopes of ground based spectra of P330-E taken by Marcia Rieke (unpublished) and the NICMOS calibrated fluxes of P330-E. This detailed spectrum is then
convolved with the NICMOS Camera 3 PSF and grism spectral resolution to provide the
input spectrum.
5.2.
Pixel Quantum Efficiency
The remaining task is determination of the quantum efficiency of each pixel in the Camera 3
array as a function of wavelength. This is done using the Camera 3 narrow band filter flat
NICMOS Grism Calibration
Figure 2.
G096 Spectrum of Hubble 12.
Figure 3.
G141 Spectrum of Hubble 12.
245
246
Thompson and Freudling
Figure 4.
G206 Spectrum of Hubble 12.
fields. The flat fields created by the IDT have a median of 1 and are multiplicative, i.e.,
the image is flattened by multiplying by the flat field. The procedures for creating the flat
fields are covered in Section 5.3.
The median response in Janskys per ADUs/sec for each flat field has been measured
in the photometric recalibration of NICMOS. This data has been analyzed by Marcia Rieke
and Table 5 gives the median response for the Camera 3 narrow band filters.
Table 5.
Camera 3 Narrow Band Filter Responses
Filter
F108N
F113N
F164N
F166N
F187N
F190N
F196N
F200N
F212N
F215N
Janskys/(ADU/sec)
1.01E-04
8.57E-05
3.99E-05
4.28E-05
3.86E-05
3.89E-05
3.65E-05
3.58E-05
3.65E-05
4.03E-05
For each pixel the median responses are divided by the flat field value to get the response
of the pixel at the narrow band wavelengths. The responses are then fit with a quadratic
function to provide the response of the pixel with wavelength by an IDL procedure.
This procedure does not fit the long wavelength roll off of the detector since none of
the narrow band filters fall at those wavelengths and the broad band filters are too broad
to measure it. The only narrow band data that measure the roll off are from the laboratory
testing done at the University of Arizona before flight. The G206 grism is the only grism
that includes the roll off. For that wavelength region the measure preflight roll off is matched
to the current response at 2.44 microns to provide the roll off. More accurate determination
NICMOS Grism Calibration
Figure 5.
247
The NICMOS Grism Efficiencies.
can be achieved by using the G206 spectra of P330-E but only for those pixels that happen
to lie in that part of the spectrum.
5.3.
Camera 3 Narrow Band Filter Flat Fields
All of the Camera 3 narrow band filter flat field observations used in this analysis come
from Proposal 9557. The reduction procedure for the individual observations is the same as
described in Section 3.0 with two exceptions. First the flat field correction step is obviously
skipped. The second exception has to do with the nature of the data. The observations in
all of the data sets were driven to extremely nonlinear regions and often to saturation. Since
the linearity corrections for the NCS era have not yet been developed it was not possible to
correct for nonlinearity. Instead, the cosmic ray correction procedure, which also finds the
signal rate, was limited to only the linear regions. This was done with the special cosmic
ray correction procedures that accept a parameter indicating how many reads to eliminate
from the analysis. This parameter was determined by individual examination of the signal
growth in each filter.
After the individual observations were reduced the final flat field was created by taking
the median of all of the images, taking the reciprocal of all of the values, and setting the
median of the result equal to 1.0. Flat fielding is then accomplished by multiplying an
image by the flat field.
5.4.
Derived Grism Efficiency Functions
The calculated grism efficiencies are shown in Figure 5 for each of the grisms. The high
frequency structures in the efficiency are due to the interference filters.
5.5.
Intrapixel Sensitivity
It is well known that the sensitivity of the NICMOS detector array pixels is not constant
across the face of the pixel. The sensitivity decreases toward the edges of the pixel. Although
not confirmed by experiment it is probable that the degree of sensitivity decrease varies
from pixel to pixel. There are various methods to correct for this problem depending on
the nature and signal to noise of the spectrum. The efficacy of the various methods has
been debated and it is not clear which method works the best. The spectra used in this
248
Thompson and Freudling
analysis have not been corrected by any of these methods. Instead we have dithered the
observations to 15 different positions and have used a final spectrum that is the median of
the 15 individual spectra to mitigate the effect of intrapixel sensitivity.
6.
Calibration Data
All of the NICMOS grism calibration results have been provided to the Space Telescope Science Institute (STScI) for inclusion in their data base. Inquiries about grism calibration or
data reduction should be directed to either STScI or the Space Telescope-European Coordinating Facility (ST-ECF). The ST-ECF maintains a NICMOS grism data reduction program
called NICMOSlook which can be obtained from the ST-ECF at http://stecf.org/nicmoslook.
The current version is 2.12.0. The distribution of NICMOSlook includes calibration data
compatible with the results presented here.
Acknowledgments. This work is supported in part by NASA grant NAG 5-10843.
This work utilized observations with the NASA/ESA Hubble Space Telescope, obtained at
the Space Telescope Science Institute, which is operated by the Association of Universities
for Research in Astronomy under NASA contract NAS5-26555.
References
Livingston, W. & Wallace, L. 1991, N. S. O. Technical Report #91-001, National Solar
Observatory, National Optical Astronomy Observatories, Tucson, AZ 85726
Luhman, K. L. & Rieke, G. H. 1996, ApJ, 461, 298
Rudy, R. J., Rossano, G. S., Erwin, P., Puetter, R. C., & Feibelman, W. A. 1993, AJ, 105,
1002
Wallace, L., Hinkle, K., & Livingston, W. 1993, N. S. O. Technical Report #93-001, National
Solar Observatory, National Optical Astronomy Observatories, Tucson, AZ 85726
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
Coronagraphy with NICMOS
Glenn Schneider
Steward Observatory, University of Arizona, Tucson, AZ 85721 USA
Abstract. The Near Infrared Camera and Multi-Object Spectrometer (NICMOS)
provides a coronagraphic imaging capability in camera 2. NICMOS PSF-subtracted
coronagraphy routinely results in per-pixel background rejections of ∼ 107 of an occulted target’s total flux density at an angular distance of 1 , thus providing a highcontrast lever for the detection of close sub-stellar companions. At 1.1 µm (with an
∼ 0. 1 spatial resolution), occulted starlight is typically reduced by a factor of 105
over a 2 -3 annulus, thereby enabling the detection and spatially resolved imaging of
low surface brightness material in circumstellar environments. Achieving these performance levels in inherently very high contrast fields while maintaining photometric
and astrometric fidelity is challenging and requires careful planning, reduction, calibration and post-processing of coronagraphic imaging data in the presence of residual
systematic artifacts. We discuss coronagraphic calibration/processing methodologies
developed by the NICMOS IDT (successfully applied to Cycle 7 and 11 data), with
recommendations for future observations in light of the ongoing re-verification of
NICMOS coronagraphy following SM3B.
1.
Introduction
The Hubble Space Telescope (HST ) provides a unique venue for high contrast imaging which
is further exploited by NICMOS with the incorporation of coronagraphic optics in its intermediate resolution camera (camera 2 with ∼ 76 mas square pixels). After internally correcting for the well-known spherical aberration in the HST primary mirror, NICMOS+HST
delivers diffraction limited images with Strehl ratios panchromatically exceeding 98% in the
obscured pupil. Moreover, the NICMOS+HST PSF is highly stable and repeatable, which
permits extremely effective and efficient PSF subtraction. Coronagraphic PSF subtraction
is enabled by the high degree of targeting precision afforded by the HST pointing control
system coupled with autonomous target location and acquisition logic in the NICMOS and
HST flight software (FSW). Coronagraphically occulted targets are typically positioned
“behind” the occulting spot to an accuracy of ∼ 8 mas and with a post-acquisition stability
of ∼ ±4 mas. Intra-orbit field rotation on sub-orbit timescales (by rolling the telescope
around the line-of-site to the target) permits the identification and rejection of residual optical artifacts. Such artifacts are rotationally invariant in the image plane of the detector,
whereas circumstellar features of astronomical origin are not.
The NICMOS detector’s multiple non-destructive readout mode (multiaccum) and 16bit (per read) digital data quantization is well-suited for the high contrast capabilities of
its coronagraph (Schneider et al. 1998). Typical multiaccum readout strategies permit a
sampling dynamic range of ∼ 20 stellar magnitudes in a single visibility period, key to the
detection of faint objects in the presence of bright ones.
In H -band, the NICMOS coronagraph reduces the background scattered and diffracted
energy from coronagraphically occulted targets by factors of ∼ 10 at the edge of the 0.3
radius occulted region, ∼ 4 at 0.5–1.5 and ∼ 2 beyond 2 . After coronagraphic PSF
subtraction (i.e., by rotating the field) the background light is further reduced by factors
249
250
Schneider
Radius (Pixels) from Hole Center
5
7
9
11
13
16
B
A
C
K
G
R
O
U
N
D
R
E
D
U
C
T
I
O
N
15
14
13
12
11
10
9
8
INTENSITY (AZIMUTHAL AVERAGE)
10 0
5
17
19
21
23
25
27
29
31
33
35
37
REDUCTION IN BACKGROUND FLUX FROM F160W PSF
w.r.t. central pixel:
Fcentral(H) = 11% Fstar
-1
10
Unocculted PSF
Coronagraph
Coronagraph & PSF Subtraction
10-2
1
pixel
10-3
10-4
10-5
7
6
15
10-6
0
Coronagraphic
Hole
Radius = 0.3"
0.075 0.15 0.225
0.3 0.375 0.45 0.525
0.6 0.675 0.75 0.825
0.9 0.975 1.05
ARCSECONDS
4
3
2
0.3 0.45 0.6 0.75 0.9 1.05 1.2 1.35 1.5 1.65 1.8 1.95 2.1 2.25 2.4 2.55 2.7 2.85
Radius (Arcsec) from Hole Center
Figure 1.
H -band coronagraphic stray-light rejection compared to direct imaging. Inset: Per-pixel background relative to central pixel flux density.
of ∼ 30–50 at < 1 , and > 2 orders of magnitude beyond. Together, the background light
reduction at 1 from an occulted star yields per-pixel background intensities ∼ 10−7 of the
star’s brightness (Figure 1). These performance levels enable the direct imaging of young
(< few ×107 years) extra-solar Jovian-mass planets (which decline in luminosity with age)
and circumstellar debris disks (with scattering fractions > 10−4 at 1 or 10−5 at 2).
These levels of performance were repeatably demonstrated in HST Cycle 7/7N and
reverified by the recently completed SMOV-3B recommissioning program. Presently, the
final on-orbit calibration data required to fully re-enable NICMOS coronagraphy with performance levels achieved in the instrument’s earlier incarnation have not yet been acquired.
However, it is clear from the recommissioning tests which have been completed that the
coronagraphic system is well within the tunable envelope which will fully enable NICMOS
coronagraphy for HST Cycle 12 at the same performance levels as demonstrated above.
2.
The Coronagraphic Field-Of-View (FOV)
The NICMOS coronagraph is in camera 2, providing 256 × 256 pixel imaging into an
∼ 19.5 × 19.3 FOV with 0.9% X:Y linear geometrical distortion (so images must be rectified before rotationally combined). The coronagraph is optimized for peak performance
for wavelengths at and shorter than H -band (1.6 µm), where the diffracted energy from
an unresolved point source in the first Airy ring of its diffraction pattern is fully contained
in the coronagraphic “hole” at the instrument’s first image plane. The detector’s FOV is
asymmetric with respect to the occulted target. The coronagraphic hole is projected onto
the detector image plane [+73, −45] pixels (or [+5. 6, −3.4]) from the [-X, +Y] corner of
the FOV. For two-roll single-orbit imaging (normally with a 30◦ maximum differential roll
due to spacecraft constraints) the total area surveyed around a target is 475 arcsec2 with
an overlap area of 280 arcsec2 .
Coronagraphy with NICMOS
Direct (TA) Images
∆ Orientation = 30°
Coronagraphic Images
∆ Orientation = 30°
251
Positive/Negative
Separation
Difference Image
Rotate About Hole
Center and Co-add
Resampled
Figure 2.
HD 102982 (G3V, H =6.9, Lowrance et al. 1998). Two-orientation TA,
coronagraphic and PSF-subtracted images, and difference image recombination.
Circles indicate size and, except for TAs, location of coronagraphic hole.
3.
Coronagraphic PSF Subtraction
In Figure 2 we illustrate the process of NICMOS coronagraphic PSF subtraction for
HD 102982, a star with a companion separated by 0. 9, and a companion:primary H -band
brightness ratio of 0.007. Here, following target acquisition (TA) exposures we obtained
coronagraphically occulted images of HD 102982 at two spacecraft orientation angles differing by 30◦ (11 min. total integration time at each orientation). The companion is easily
visible in both coronagraphic images, with the fixed speckle pattern in the PSF “wings” of
the primary greatly reduced in intensity with respect to direct imaging. In the difference
image, the background light from the primary all but disappears. To take advantage of
“rotational dithering” which results from image reorientation, we separate the positive and
inverted negative components of the difference image and recombine them, after rotational
re-registration, resampled onto a sub-pixel grid.1 The reconstructed image of HD 102982B
can be centroided to a precision of ∼ 3 mas, and its brightness measured with an internal
precision of ∼ 12 %.
The total integration time of 11 min. per image orientation was set by the spacecraft
and instrument overheads required to execute a two-roll observation in a single visibility
period (typically ∼ 52–54 min.); notably 12 min. for two guide star acquisitions, 11 min.
for the spacecraft rotation and 3 min. for two TAs. For most targets of scientific interest,
NICMOS coronagraphic observations are not photon or read noise limited, but rather are
limited by imperfections in PSF subtraction. In the absence of background light (i.e.,
sufficiently far from the occulted target), the detection floor for a total integration time of
22 min. was ∼ H = 22.5 in Cycle 7 (22.9 in Cycle 11; §7.1), though clearly the detection
floor is a function of the radial distance from the occulted star. We illustrate this in Figure 3
for a coronagraphic PSF-subtracted (difference) image of LHS 3003 (H = 9.3) taken the
same manner as HD 102982 but displayed to the noise floor limit.
1
Software for post-processing of calibrated coronagraphic images (IDP3 & DSPK; Schneider & Stobie 2002)
is electronically available at: http://nicmos.as.arizona.edu/software/idl-tools/toollist.cgi
252
Schneider
H = 21.9
ρ = 9".36
∆H = 12.6
TA Persistence
Ghost Images
H = 22.3
ρ = 13".34
∆H = 12.9
(faint galaxies)
Figure 3.
LHS 3003 (H = 9.3, M7V). Single visibility period, two-orientation
coronagraphic PSF-subtracted faint-object limits at “large” angular separations.
4.
Variability of the “invariant” PSF
While the structure of the PSF is highly repeatable, it is not perfectly so. These imperfections arise predominantly from four effects (Schneider et al. 2001). First, changes in the
structure of the HST PSF arise from variations in the thermal input to the HST Optical Telescope Assembly’s (OTA) secondary mirror. As the telescope cycles through orbit
day/night, or as it undergoes a major change in attitude or orientation with respect to the
Sun, or sub-solar point of the Earth through its orbital phase, the secondary mirror moves
by several microns along the optical axis. Though this “breathing” phenomenon (Bely 1993)
results in only a small mechanical displacement, it affects the scale and structure of the input PSF illuminating the edge of the coronagraphic hole, which itself acts as a scattering
surface. Second, the position of the Lyot stop (pupil plane cold mask) which contributes its
own near-field diffraction signature to the final image plane, “wiggles” (Krist et al. 1997)
by very small amounts, usually on multi-orbit timescales. Recent data suggest that the
amplitude of these motions may be reduced from HST Cycle 7/7N (consistent with having
been driven by differential stresses in the NICMOS dewar from mass redistribution of the
sublimating SN2 ). Third, the projected location of the coronagraphic hole onto the image
plane of the detector changes with time due to changes in metrology between the NICMOS
warm optics and the detector cold bench. The location of the hole is determined by the
FSW at the time of each TA, and targets to be occulted are placed with precisions as noted
in §1 with respect to the hole. However, changes in the hole and target positions, by even
small sub-pixel amounts, result in differential errors in flat-field calibrations and intra-pixel
responsivity. Finally, differential targeting errors on the order of a few milliarcseconds can
themselves lead to changes in the coronagraphic PSF structure.
Scale changes in the HST +NICMOS PSF ultimately affect the coronagraphic sensitivity and detectability dominated by uncompensated background light. Observations of
the target and reference PSF should be closely spaced in time, less than the one to several orbit thermal responsivity time constant of the OTA. For two-roll PSF subtractions,
where the target serves as its own reference PSF (e.g., for companion detection), this can
be accomplished in a single visibility period as previously described.
Coronagraphy with NICMOS
253
A
B
C
D
E
F
Figure 4. TWA6 (H = 6.9, K7V, Schneider & Silverstone 2002). Top: Cycle 7
(2.19 µJy counts−1 s−1 ), Bottom: Early Cycle 11 (1.59 µJy counts−1 s−1 ).
5.
Pushing the Limits—Companion Detection
We illustrate companion detectability limits in the presence of residual circumstellar starlight
first by representative example, and then quantitatively from a statistically significant sample of stars observed in Cycle 7/7N. Figure 4 shows NICMOS coronagraphic images of
TWA6 at two image orientations differing by 30◦ (panels A and B), and acquired in the
same manner as the two stars previously discussed. A faint point-like feature is noted even
before PSF subtraction in both images. Upon subtraction, a stellar-like (unresolved) object
200,000 times fainter than TWA6 (∆H = 13.2) appears at an angular distance of 2.5 from
the occulted star (panel C). The two independent point-like images, each of S/N ∼ 20,
exhibit first Airy-ring structures, core profiles identical to point sources, and are rotated
about the occulted target by the field re-orientation angle. This observation is presented as
a representative demonstration of capability. Putative companionship must be tested via a
differential proper motion measure from two epochs (Figure 4 and §7.1).
We assert the typical nature of the TWA6 observation after assessing the instrumental sensitivity to, and detection probabilities for, identifying nearby companions. Twoorientation PSF-subtracted coronagraphic observations of 50 stars of spectral types G-M,
obtained by the NICMOS IDT in HST Cycle 7/7N, were subjected to statistical analyses
of the radially dependent background noise and star implantation experiments (Schneider
& Silverstone, 2002). The systematic noise from imperfect PSF subtractions is azimuthally
non-isotropic, as are the sensitivities and detection limits. We find a 50% probability of
companion detection of ∆H (50%) = 9.7 ± 0.3 + 2.1 × ρ(separation), assuming azimuthally
random target placement but with a 30◦ rotation between two field orientations. The 1 σ
dispersion of 0.3 magnitudes over the whole sample arises from breathing excursions, target
centering errors, and color terms. This may also be cast in terms of the achievable S/N ratio,
which we arbitrary quantify at S/N = 25 (closely comparable to the TWA6 observation with
both rolls combined) as: ∆H (S/N = 25) = 8.1 ± 0.3 + 2.1 × ρ. These levels of performance
are highly repeatable for two-orientation PSF subtractions with observations completed in
a single visibility period. Specific experiments performed in the HST Cycle 7 SMOV 7052
program (Schneider & Lowrance 1997) found degradation in these performance levels of
254
Schneider
Figure 5.
5–10 Myr debris disks. A: TW Hya (Weinberger et al. 2002),
B: HD 141569A (Weinberger et al. 1999), C: HR 4796A (Schneider et al. 1999).
∼ 1.5 stellar magnitudes in ∆H for observations taken in sequential orbits rather than in
the same orbit, and hence such a multi-orbit observational strategy is contra-indicated.
6.
Disk Imaging
Observing light scattered by circumstellar debris has been observationally challenging because of the very high star-to-disk contrast ratios (i.e., low surface brightness disks very
close to bright stars). Proto and young circumstellar disks (< 1 Myr) with embedded or
obscured central stars may be (and have been) imaged directly. Older, centrally-cleared,
disks present a contrast challenge requiring PSF-subtracted coronagraphy (Schneider 2002).
Until NICMOS, the only debris system which had been seen in scattered light was the
bright, large, and nearly edge-on disk around β Pictoris. The advent of NICMOS coronagraphy gave rise to the first subsequent scattered light observations of debris disks with a
surprising diversity of morphologies and characteristics (Schneider et al. 2002). Dusty disks
with radial and hemispheric brightness anisotropies and complex morphologies, possibly indicative of interactions with unseen planetary mass companions, were detected and spatially
resolved (e.g., Figure 5). Transitional disks (around Herbig AeBe stars) of intermediate age
(1–5 Myrs) have been observed coronagraphically with HST /STIS (Grady, 2002), which
has been also used for follow-on observations of disks imaged earlier by NICMOS.
In imaging sub-stellar companions, which are spatially unresolved point sources, one
can use the host star as its own reference PSF. Circumstellar disks are diffuse spatially
extended structures, and the two-roll subtraction technique used for companion detection
cannot be exploited. Rather, a reference PSF from a different star must be obtained and
used as a surrogate to subtract out the residual instrumentally scattered and diffracted
starlight. When a reference PSF is required, a “nearby” star should be chosen to permit
a retargeting slew within the same visibility period whenever possible, and such that the
Sun angle of the spacecraft does not change significantly. This is sometimes difficult since
a PSF star should be at least as bright as the target and of similar spectral type (within
one spectral class) to obviate the dominance of color effects (Weinberger 1999). In the
worst case a (disk) target and its reference PSF should be observed in subsequent orbits,
and at the same orbit phases. To minimize image artifacts resulting from imperfect PSF
subtractions, and to sample regions of the disks which will invariably be corrupted by the
HST diffraction spikes, observations should be obtained at two or more spacecraft roll
angles. Image contrast is significantly enhanced for young exoplanet and brown dwarf
companion imaging in H -band as such objects are self-luminous with higher emissivities
at longer wavelength. Disks, however, are seen by the scattering of the central star’s light
by circumstellar grains, with color terms, typically, not too different from the stars which
Coronagraphy with NICMOS
255
they orbit. For disk surveys it is preferable to image at shorter wavelength, e.g., at 1.1 µm
(F110W filter), which takes advantage of both the intrinsically higher spatial resolution and
the reduced coronagraphically induced instrumental scatter (Schneider et al. 2001). Multiband imaging with NICMOS can help elucidate the nature of the circumstellar grains, and
indeed the camera 2 filter set contains spectral elements which are diagnostic of ices which
can mantle such grains at circumstellar distances beyond their sublimation temperatures.
7.
Performance Characterization for Cycle 11
With the resurrection of NICMOS in the era of the NICMOS Cooling System we have
characterized the performance capabilities of the coronagraph under the SMOV-3B test
program. In particular we have evaluated data acquired from the NICMOS TA (8983), Focus Verification (8979) and Initial (Part 1) Performance Characterization (8984) tests. The
TA and Focus tests were executed, per the SMOV plan, prior to planned final updates to
the NICMOS plate scale and aperture rotation used by the FSW. The initial performance
characterization was executed prior to the re-determination and update to the FSW targeting logic’s “low scatter point” in the coronagraphic system. Because the test and update
program is not yet completed we cannot present final Cycle 11 (and beyond) performance
metrics, though the system “as is” is performing close to the previous mark. Based upon the
data acquired at this state of the instrument’s recommissioning, and comparing them to the
parameter spaces explored during the SMOV-2 (Cycle 7) program, it is apparent that the
NICMOS coronagraphic performance capabilities will be fully restored at the completion of
the re-enabling activities (to be carried out under the Cycle 11 calibration plan).
7.1.
TWA6 Revisited
As a demonstration of restored capability, we re-observed TWA6 “out-of-the-box” as part
of the NICMOS TA testing. In Figure 4 we compare Cycle 7 (top) and Cycle 11 (bottom)
coronagraphic images to illustrate the high degree of repeatability of the NICMOS coronagraphic PSF after a more than three year suspension of activity. The point object seen
in the Cycle 7 observation is seen at nearly the same image contrast as in the Cycle 11
observation. The Cycle 7 and 11 images were acquired at slightly different absolute orientation angles, so the point object is at slightly different azimuthal angles at the two epochs
(and is co-incident with the occulted star’s (-X,-Y) diffraction spike in panel D). To first
order, the system performance is roughly comparable. Residual optical artifacts (i.e., radial
streaks extending to ∼ 2 ) are more prominent in the Cycle 11 PSF subtracted image. This
is a consequence of differential targeting errors between the two image orientations and
is a direct result of executing this test before later-planned on-board calibration updates.
This is fully within the envelope of correction of the planned updates and will be mitigated
following the completion of Part 2 of the coronagraphic performance characterization test.
In Cycle 11 (NCS at 77K), the pixel-to-pixel (read) noise is lower in amplitude, relative
to the photon flux from an occulted target (and field object) as compared to Cycle 7 (SN2
at 62K). This improvement arises from an increase in H -band quantum efficiency (QE) of
∼ 37% (Figure 4). For comparative purposes the Cycle 7 and 11 images have been stretched
to permit direct comparison in light of the net gain in imaging efficiency.
In both the Cycle 7 and Cycle 11 observations we were able to measure the positions
of the unresolved object to a precision of ∼ 11 mas with respect to TWA6. Over the 4 yr
temporal baseline, the angular distance decreased by 160 mas in the direction expected from
the proper motion of TWA6 itself, implicating a background object rather than a Jupitermass companion to this 10 Myr star (unfortunately). However, these two-epoch observations
demonstrate the ease with which such observations can be reliably and repeatably made.
256
Schneider
Figure 6. A: OTFR/CALNICA pipeline reduction with synthetic dark and library flats, B: CALNICA with custom-matched dark and linear-regime flat-field.
8.
Calibration
The performance levels discussed assume properly calibrated and cosmetically clean data.
Whether searching for companions or circumstellar disks, local and global deviations from
true photometric backgrounds must be corrected (zeroed) before PSF subtraction. Failure
to appropriately do so will result in: loss of sensitivity (against the residual background);
degraded detectabilty in PSF-subtracted images; photometric zero-point errors; and spatially non-uniform detection limits. With coronagraphic data, in particular, one must be
critical of pipeline processing.
Standard reference flats from STScI’s calibration database system contain a static
imprint of the coronagraphic hole though the position of the hole is known to be unstable at
the level of roughly a pixel. Before flat-fielding high-S/N reference flats should be augmented
with contemporaneous (S/N ∼ 130 combined) lamp flats, provided as part of the TA process,
otherwise very significant near-hole flat-field gradients may arise. We also suggest that
reference flats be constructed so as not to rely on assumed high-fidelity knowledge of the
per-pixel linearity transfer when approaching saturation. I.e., discard non-destructive reads
in raw flat-field frames with pixel values > 50–70% full-well when making reference flats.
The efficacy of using “synthetic” (decomposed model) darks (currently generated by
OTFR) vs. reference dark frames made directly from observed calibration frames is somewhat conjectural and may be data driven. Preliminary indications from Cycle 11 data are
that synthetic darks in many cases may suffice as the NICMOS detectors (which have temperature dependent dark currents) are now thermally more stable than in Cycle 7. Manual
(and sometime laborious) construction of dark reference files from Cycle 7 data which are
selected to match (1) the temperature of the detector at the time of the observation (as
reported in the SPT file), and (2) the relative time from the prior SAA exit, often reduce
photometric errors relative to the Cycle 7 model darks. Fortunately, two-roll per visibility observations are inherently SAA non-interruptible, so usually they are less subject to
degradation from cosmic-ray induced excess dark current decays from prior SAA crossings.
As an example, in Figure 6, we show the result of re-processing a raw NICMOS coronagraphic frame using OTFR/CALNICA compared to an IDL-based analog to CALNICA
using reference darks made from selected, observed, data, flats unreliant on linearity corrections, and Gaussian-weighted bad pixel replacement. Post-processing tools exist in the
IRAF/STSDAS environment to mitigate calibration errors (such as in the example shown),
Coronagraphy with NICMOS
A
B
C
D
257
E
Figure 7. A: Raw TA image, B: Shading model, C: Shading removed (and bad
pixels replaced), D: Bars model, E: Bars removed.
but they often do not work well in regions of high flux densities and large signal gradients
which fill the field (e.g., for bright coronagraphic targets). Experimentation is required, and
improvements over pipelined results can be had with this additional work.
9.
Mode-2 Target Acquisition Data
Though often under-appreciated, data which are taken (and delivered) from TAs are of fundamental importance in the calibration and interpretation of coronagraphic imaging data
which follow. TA images serve as astrometric anchors, and are required to accurately determine the placement of a target into the coronagraphic hole. Such determinations cannot
be made from coronagraphic images themselves. In HST Cycle 7 targeting information
from the spacecraft slew, which put the target into the coronagraphic hole, was reported in
the ancillary SPT file (first in raw engineering units, later in detector pixels). In Cycle 11
the OTFR pipeline has been updated to place this information in the FITS headers of the
RAW and CAL files themselves, but co-ordinates are still given in the FSW’s detector pixel
system, not in the science instrument aperture system (SIAS). For camera 2: SIAS[X,Y] =
256-FSW[Y,X]. The information provided through these files, however, uses the fixed aperture “constants” (scale and rotation) employed by the FSW, though these may change over
time. For highly accurate astrometry, the TA information reported in the FITS headers
may require updating to reflect the actual “plate constants” at the time of the observation.
Such information historically has been provided by STScI through a web-based interface.2
With the improved detector QE in Cycle 11, stellar PSF cores will not saturate at
the shortest (0.2s) TA exposure times for the recommended TA filters as follows: F160W:
H > 7.2; F165M: H > 6.5; F171M: H > 5.5; F187N: H > 4.0. For targets with H < 4,
Mode-1 TAs are needed. Autonomous acquisition of targets of H > 18 are prone to failure
as such targets are difficult to discriminate from hot pixel clusters by the on-board software.
Properly exposed (∼ 70% full-well) TA images can be used to establish an in-band
magnitude for the filter used. Additionally, “hole location” lamp-flat background images
are taken (as two 7s ACCUM images) and may be used to ascertain the H -band magnitude
of an unsaturated target, even if the TA was done in a different filter. Color transformation
from a TA or lamp-background filter to a different science filter may be estimated by using
STSDAS SYNPHOT task. These serendipitous photometric determinations may not be
optimal. Whenever possible, unocculted images of subsequently occulted targets should be
obtained as part of the planned imaging sequence to establish the PSF core photometry.
This is particularly necessary for disk imaging where the in-band flux ratios of the target
and reference PSF must be known to obtain a proper scaled PSF subtraction.
TA ACCUM mode images, F160W lamp flats and background images (which will
contain an image of the star to be acquired) are not calibrated by the OPUS pipeline.
2
http://www.stsci.edu/hst/nicmos/performance/platescale
258
Schneider
Dark current is usually not an issue for short TA images and lamp background frames
(but may be so for hot pixels), but the detector reset signature (aka “shading”) will bias
target centrations with horizontal field gradients, and introduce photometric errors amplified
through the flat-field. TA images are also subject to a readout anomaly known as “the
bars”, which though correctable, are not handled by the OPUS pipeline. TA images may
be calibrated by: (1) pixel-paired minimization to eliminate cosmic rays, (2) building a
source-clipped (or masked) column-median to characterize and build a 2D shading image to
be subtracted, (3) building a source-clipped row-median image to eliminate the bars upon
subtraction, (4) flat-fielding with a library (or contemporaneous TA) flat. Alternatively to
(1), multiaccum darks could be taken, but because of subtle differences in detector clocking
compared to TA/ACCUM mode the integration time should be increased by +0.025s.
10.
Summary
The SMOV-3B program has demonstrated a full return of NICMOS coronagraph science capabilities with stray-light rejection closely replicating Cycle 7 performance. Coronagraphic
detection limits, with and without PSF subtraction, will be fully restored after planned
updates to the FSW’s TA calibration constants have been made. System sensitivities have
increased (∼ 37% in H -band) due to the higher QE of the detector now operating at 77K.
Final refinements to performance metrics and calibration data await the completion of the
Cycle 11 calibration plan. NICMOS is ready to resume coronagraphic science in Cycle 12.
Acknowledgments. This work is supported by NASA grants NAG5-3042 and 10843.
We thank M. Silverstone and J. Beattie for a careful proof reading of this manuscript.
References
Bely, P. 1993, STScI Report (SESD-93-16)(Baltimore: STScI)
Grady, C. 2002, this volume, 137
Krist, J. E., et al. 1998, PASP, 110, 1046
Lowrance, P. J., et al. 1998, ESO Conf. Proc., 55, 96
Schneider, G. & Lowrance, P. 1997, SMOV/7052 Test Report,
http://nicmosis.as.arizona.edu/gschneid/7052 PS/
Schneider, G., Thompson, R. I., Smith, B. A., & Terrile, R. J. 1998, Proc. of SPIE, 3356,
222
Schneider, G., et al. 2001, AJ, 121, 525.
Schneider, G. & Stobie, E. 2002, ASP Conf. Ser. 281, in press
Schneider, G. 2002, in Hubble’s Science Legacy: Future Optical-UV Astronomy From Space,
ed. C. Blades, ASP Conf. Series, in press
Schneider, G. & Silverstone, M., 2002, Proc. of SPIE, 4860, in press
Schneider, G., Weinberger, A. J., Silverstone, & M. D. Cotera, A. S. 2002, in Debris Disks
and the Formation of Planets, ed. D. Bachman, ASP Conf. Series, in press
Weinberger, A. J., et al. 1999, ApJ, 522, L53
Weinberger, A. J., et al. 2002, AJ, 566, 409
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
Polarimetry with NICMOS
Dean C. Hines
Steward Observatory, University of Arizona, Tucson, AZ 85721
Abstract. The Near Infrared Camera and Multi-Object Spectrometer (NICMOS)
aboard the Hubble Space Telescope (HST ) incorporates optics in Cameras 1 and 2
(NIC1 and NIC2) that enable high spatial resolution imaging polarimetry at ∼1 &
2 µm, respectively. Thermal vacuum tests prior to installation revealed that the
three polarizing elements in each camera have unique (non-unity) polarizing efficiencies, and their primary axes are not oriented at the nominal 120◦ intervals. This
non-ideal system requires a reduction algorithm that differs from the standard approach used for ideal polarizers. The coefficients of the algorithm are derived from
the ground-based thermal vacuum results and from on-orbit observations of objects
of known polarization. The Cycle 7 and 7N calibration resulted in excellent imaging
polarimetry performance, capable of producing uncertainties in measured polarization as small as σp ≈ 1%. The Cycle 11 calibration plan includes a recharacterization
of the polarimetry capabilities. Herein I review the reduction algorithm, describe the
Cycle 11 calibration plan, and present preliminary results. The latter indicate that
once fully calibrated, NICMOS will provide polarimetry performance comparable to
(or better than) Cycle 7 and 7N. Combined with the polarimetry mode of the Advanced Camera for Surveys (ACS), HST provides high resolution imaging polarimetry from ∼0.2–2.1 µm. The further possibility of combining imaging polarimetry
with coronography in both instruments has the potential to greatly enhance high
contrast imaging.
1.
Preflight Thermal Vacuum Tests
The NICMOS polarimetry system was characterized on the ground during thermal vacuum
tests using a light source that fully illuminated the field of view with completely polarized
light and with position angles variable in 5◦ increments. The primary results of these
thermal vacuum tests include:
• Each polarizer in each camera has a unique polarizing efficiency,1 with POL120S
having the lowest at 2POL120S = 48%.
• Angular offsets between the polarizers within each filter wheel differ from their nominal
values of 120◦.
• Instrumental polarization caused by reflections off the mirrors in the optical train is
small (≤ 1%).
• The grisms act as partial linear polarizers, with G206 producing the largest variation
in intensity (∼5%) for completely polarized light. Because the grisms reside in the
1
Polarizer efficiency is defined as = (Spar − Sperp )/(Spar + Sperp ), where Spar and Sperp are the respective
measured signals for a polarizer oriented parallel and perpendicular to the position angle of a fully polarized
beam.
259
260
Hines
NIC3 filter wheel, they cannot be used with either the NIC1 or NIC2 polarizers and
are unsuitable for spectropolarimetry.
2.
The Algorithm for Reducing NICMOS Polarimetry Observations
The thermal vacuum results showed that the standard reduction algorithm would not work
for NICMOS data. Instead, a more general approach was required (Hines, Schmidt & Lytle
1997; Hines, Schmidt & Schneider 2000).
At any pixel in an image, the observed signal from a polarized source of total intensity
I and linear Stokes parameters Q and U measured through the kth polarizer oriented at
position angle ϕk is
(1)
Sk = Ak I + 2k (Bk Q + Ck U ) .
Here,
tk
(1 + lk ), Bk = Ak cos 2ϕk , Ck = Ak sin 2ϕk ,
(2)
2
2k is the polarizing efficiency, tk is the fraction of light transmitted through the polarizer for
a 100% polarized input aligned with the polarizer axis, and lk is the “leak”—the fraction of
light transmitted through the polarizer (exclusive of that involved in tk ) when the incident
beam is polarized perpendicular to the axis of the polarizer. These quantities are related
under the above definitions, 2k = (1 − lk )/(1 + lk ).
This treatment can be shown to be equivalent to other approaches, once appropriate
transformations are made (Mazzuca, Sparks, & Axon 1998; see also Sparks & Axon 1999).
The values of tk were determined initially by the filter manufacturer from witness samples, and were not accurately remeasured during thermal vacuum tests. However, on-orbit
observations of the unpolarized and polarized standard stars enables refinement of these
numbers. Adopted characteristics of the individual polarizers and algorithm coefficients
derived during and applicable to Cycle 7 and 7N are listed in Table 1 of Hines, Schmidt &
Schneider (2000), and are also available in the NICMOS instrument manual and the HST
Data Handbook .
After solving the system of equations (eq. 1) for the Stokes parameters at each pixel (I,
Q, U ), the percentage polarization (p) and position angle (θ) are calculated in the standard
way:
1
Q2 + U 2
−1 U
, θ = tan
.
(3)
p = 100% ×
I
2
Q
Ak =
Note that a 360◦ arctangent function is assumed.
This algorithm has been tested by the NICMOS Instrument Definition Team (IDT) and
by the Space Telescope Science Institute (STScI) on several data sets. An implementation
has been developed by the IDT, and integrated into a software package written in IDL.
The package is available through the STScI web site2 and is described by Mazzuca & Hines
(1999).
3.
The Cycle 11 Polarimetry Characterization Program
The higher, yet more stable, operating temperature provided by the NCS and the three
year dormancy of NICMOS may contribute to changes in the properties of the polarimetry
optics, especially the tk coefficients. Therefore, a program to re-characterize the polarimetry
optics in Cycle 11 has been developed. The core program has been outsourced by STScI
2
http://www.stsci.edu/instruments/nicmos/ISREPORTS/isr 99 004.pdf
Polarimetry with NICMOS
261
to the author (NIC/CAL 9644: Hines), while flat fields and observations of photometric
calibrators are maintained by the NICMOS team at STScI.
The basic design of the program follows the strategy undertaken in Cycle 7 and 7N,
relying on observations of polarized and unpolarized standard stars as well as the protoplanetary nebula CRL 2688 (Egg Nebula). The stars are observed at two epochs separated
in time such that the spacecraft roll differs by ∼ 135◦ . This removes the degeneracy in
position angle caused by the pseudo-vector nature of polarization. The Egg Nebula is only
observed at a single epoch as a direct comparison with observations from Cycle 7 and 7N
(ERO 7115: Hines; Sahai et al. 1998; Hines, Schmidt & Schneider 2000), and to evaluate
any gross discrepancies across the field of view.
Observations were obtained between UT 2002 September 2 and UT 2002 October 6.
The second epoch observations of the polarized and unpolarized standards stars are scheduled between UT 2003 April and UT 2003 July. All observations were obtained using
MULTIACCUM sequences and four position spiral dither pattern in each polarizer. The
dither pattern used N+1/2 (N ≥ 30) pixel offsets to improve sampling, avoid latent images
and mitigate residual instrument artifacts. This is also the recommended observing strategy
for all NICMOS polarimetry programs.
4.
Preliminary Results of Cycle 11 Polarimetry Characterization Program
The observations were processed through the CALNICA pipeline at STScI using the currently available reference files. The linearity corrections are not yet fully characterized and
potentially pose the largest uncertainty in the present results. The flat-fields and the polarimetry calibration images are both potentially affected (this should be corrected soon
and will be applied for the final analysis of the complete characterization data set).
Observations of the unpolarized standard star processed with the Cycle 7 and 7N
algorithm coefficients yield pNIC1 = 0.7% ± 0.2%, θNIC1 = 74◦ and pNIC2 = 0.7% ± 0.3%,
θNIC2 = 73◦ . This suggests that the system may have changed, but the uncertainties are
currently too large. The observations of the polarized standard star is also larger (∆p ≈ 2%)
in NIC1 compared with the measurements of Cycle 7 and 7N, which themselves were in
excellent agreement with ground-based measurements (Hines, Schmidt & Schneider 2000).
The second epoch observations will reduce the uncertainties in the null measurements, and
enable the coefficients to be refined.
Observations of the Egg Nebula were also analyzed with the Cycle 7 and 7N coefficients. As for the polarized standard star, the results for the Egg Nebula suggest that the
polarimetry system has changed slightly,3 again by about 2% in p%. The Cycle 11 preliminary results are shown in Figure 1. In addition to the traditional polarization vector plot,
Figure 1 also shows maps of perpendiculars to the polarization position angle as a function
of position in the Egg Nebula. Perpendicular maps show the direction of the illuminating
source relative to the last scattering surface as projected on the sky. The covergent points
in the Cycle 11 data are consistent with the Cycle 7 and 7N data (Weintraub et al. 2000),
and indicate no significant field-dependent anamolies in the polarimetry system.
5.
Summary
The Cycle 11 re-characterization plan is partially complete. The preliminary results indicate that the coefficients of the algorithm for deriving Stokes parameters from images
3
The polarization structure of the Egg is not expected to change over the 5 year period between observations
even though the object is known to show photometric variations.
262
Hines
(b)
(a)
100%
Polarization
NIC1
(c)
(d)
100%
Polarization
NIC2
Figure 1.
Cycle 11 polarimetry observations of the Egg Nebula (CRL 2688).
The left-side panels show the polarization maps, while the right-side panels show
the perpendiculars to the polarization position angles (see text).
taken through the NICMOS polarizers will require slight adjustment after all of the standard star observations have been completed in the Spring of 2003. Observations of the
Egg Nebula suggest that, once fully calibrated, NICMOS will provide polarimetry performance comparable to (or better than) Cycle 7 and 7N. Combined with the polarimetry
mode of the Advanced Camera for Surveys (ACS), HST provides high resolution imaging
polarimetry from ∼0.2–2.1 µm. The further possibility of combining imaging polarimetry
with coronography in both instruments has the potential to greatly enhance high contrast
imaging.
References
Hines, D. C., Schmidt, G. D., & Lytle, D. 1997, in The 1997 HST Calibration Workshop,
eds. S. Casertano et al. (Baltimore: STScI)
Hines, D. C., Schmidt, G. D., & Schneider, G. 2000, PASP, 112, 983
Mazzuca, L., Sparks, B., & Axon, D. 1998, Instrument Science Report NICMOS-98-017
(Baltimore: STScI)
Mazzuca, L. & Hines, D. C. 1999, Instrument Science Report NICMOS-99-004 (Baltimore:
STScI)
Sahai, R., Hines, D. C., Kastner, J. H., Weintraub, D. A., Trauger, J. T., Rieke, M. J.,
Thompson, R. I., & Schneider, G. 1998, ApJ, 492, L163
Sparks, W. B. & Axon, D. J. 1999, PASP, 111, 1298
Weintraub, D. A., Kastner, J. H., Hines, D. C., & Sahai, R. 2000, ApJ, 531, 401
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
NICMOS Cycle 10 and Cycle 11 Calibration Plans
S. Arribas,1 S. Malhotra, D. Calzetti, E. Bergeron, T. Boeker,1 M. Dickinson,
L. Mazzuca, B. Mobasher,1 K. Noll, E. Roye, A. Schultz, M. Sosey, T. Wiklind,1
C. Xu
Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218
Abstract. The NICMOS calibration activities performed after the completion of
the Servicing Mission and On-orbit Verification program SMOV-3b are described. In
particular, we present the generic objectives pursued with the Cycle 10 (interim) and
Cycle 11 calibration plans, the specific programs involved, and the accuracy goals
for Cycle 12.
1.
Introduction
After the successful completion of the 2002 HST Servicing Mission, NICMOS went through
an on-orbit verification phase as part of the SMOV-3b program. The NICMOS SMOV-3b
program was intended to demonstrate that the instrument was functioning as expected after
the installation of the NICMOS Cooling System (NCS). Although this program included
many calibration-related activities, it did not provide a full calibration of the science modes.
Full calibration of the instrument is being performed thanks to the programs included in
the calibration plans for each cycle. In the NCS era the first such plans were the Cycle 10
(interim) and the Cycle 11 (routine) calibration plans. Here we summarize the generic
objectives pursued with these plans, the specific proposals involved, and the accuracy goals
for Cycle 12.
2.
Objectives of the Cycle 10 (Interim) Calibration Plan
The Cycle 10 interim calibration plan (ICP) lasted approximately five months and pursued
the following objectives:
i) Calibration of the imaging mode for the three cameras and all the spectral elements.
The imaging mode is by far the most commonly used NICMOS science mode. During
Cycles 7 and 7N more than 80% of the exposures taken with NICMOS were in this mode.
A full calibration of this mode requires a number of individual activities (e.g., obtaining high
S/N darks and flats, optimizing the image quality, evaluating photometric stability, etc).
The ICP provided high S/N flats for all narrow band filters (SMOV program 8985 provided
wide, medium, and polarizer filter flats), improved the accuracy of the darks obtained during
SMOV, and allowed a detailed study of the image quality and photometric stability of the
instrument.
ii) Calibration of the spectroscopic mode. This science mode was the second most
used during Cycles 7 and 7N with about 5.6% of the total number of exposures. The
ICP provided the flats for the narrow band filters for NIC3 which, together with those
1
Affiliated with the Space Telescope Division, Research and Science Support Department of the European
Space Agency (ESA)
263
264
Arribas
obtained with program 8991 (SMOV), allowed the calibration of this mode (see Thompson
and Freudling 2003).
iii) Monitoring the main instrument properties. ICP included four monitor programs
(darks, flats, focus, photometry). These programs are considered key for understanding the
behavior and stability of the instrument after the installation of NCS. The dark program
(ID 9321), which provided the linear component of the dark current, the shading, and the
amplified glow, was executed weekly. The other three monitor programs were executed
monthly.
iv) Special calibrations. The ICP also included two special calibrations: 1) the gain
test was aimed at demonstrating the benefits of implementing a new gain value, and 2) high
quality ACCUM darks were needed to calibrate the cosmic ray persistence effects in postSouth Atlantic Anomaly (SAA) observations.
v) ICP also provided the data necessary to implement the Dark Generator and the
Flat Generator tools, which allow users to create synthetic darks and flats, respectively.
Table 1.
Cycle 10 (interim) and Cycle 11 (regular) Calibration Programs. Some
SMOV calibration-related programs are also included. Details on individual programs can be obtained via the HST -STScI web site at http://www.stsci.edu/hst
Activity title
Multiaccum Darks
ID (Cycle/Program)
9321 (C10), 9636 (C11)
Flats
Flats for NIC1 and NIC2
8995 (SMOV)
9327 (C10)
Flats for NIC3
Photometry Test
9557 (C10)
8996 (SMOV)
Aperture Location
8981 (SMOV)
Plate Scale
Grism Calibration
Polarimetric Calibration
Coronagraphic focus
8982
8991
9644
8979
Coronagraphic
Performance Assessment
Focus Stability
Photometric Stability
8984 (SMOV)
Flat Fields Stability
Dark Generator Test
SAA-CR Persistence Test
9326 (C10), 9640 (C11)
9641 (C11)
8987 (SMOV)
Accum Darks
9322 (C10)
Thermal Background
8989 (SMOV)
Intra Pixel Sensitivity
High S/N Capability
Characterization
Gain Test
9638 (C11)
9642 (C11)
Pupil Transfer Function
9643 (C11)
(SMOV)
(SMOV)
(C11)
(SMOV)
9323 (C10), 9637 (C11)
9325 (C10), 9639 (C11)
9324 (C10)
Comments
Monitor programs. Linear component of the
dark current, shading, ampified glow. Include
all the information needed for the Dark Generator tool
Flats for the broad and medium-band filters
Narrow band filter flats. Broad and medium
band filter flats were obtained during SMOV
Narrow filters
Photometric zero points for all the spectral
elements
Location of the NICMOS apertures in the V2V3 plane
Plate scale, field rotation, and field distortion
Recalibration of the spectroscopic mode
Recalibration of the polarimetric mode
Determination of optimum focus for
coronagraphy
Quantitative re-evaluation of the coronagraphic mode
Monitor programs
Monitor programs. Observations of P3003E
with selected broad filters
Monitor programs, using a few selected filters
To characterize the accuracy of this tool
Calibration for mitigating the effects of cosmic ray induced persistence after passage of
the SAA
Darks needed for the calibration of the Cosmic Ray Persistence
Characterization of the thermal background
at the NICMOS focal plane
For cameras 2 and 3
Characterization of temporal photometric
variations at very high S/N regime
Engineering test to analize the advantages of
a new gain value
To correct large scale flat-field residuals (contingency program)
NICMOS Calibration Plans
3.
265
Objectives of the Cycle 11 Calibration Plan
The objectives pursued with this plan are:
i) Monitor Programs: Similar to Cycle 10, an important objective during Cycle 11 has
been the monitoring of the main properties of the instrument. The involved programs are
a continuation of the corresponding Cycle 10 (and SMOV) programs. Preliminary analysis
of the data indicated good thermal stability and, therefore, the frequency of some of these
programs has been reduced with respect to the corresponding programs for Cycle 10.
ii) Intrapixel sensitivity: Data obtained during Cycles 7 and 7N demonstrated that
one of the major factors limiting the photometric accuracy was the non uniform intrapixel
sensitivity. This is especially true for NIC3, for which the PSF is more severely undersampled. Although this limitation may be overcome by dithering, this approach may be quite
demanding in terms of observing time. Calibrations of intrapixel sensitivity may result in
acceptable photometric accuracy without the need for excessive dithering.
iii) Multiaccum darks: In order to generate the dark reference files for all the multiaccum readout sequences an empirical model, the so called Dark Generator Tool , is used.
One of the goals of the present plan is to test the accuracy of such a model.
iv) Polarimetry mode: This calibration has been outsourced to Dr. Dean Hines (University of Arizona), and it is aimed at recalibrating the polarimetry mode in both Camera 1
and 2 (see Hines 2003).
v) High S/N Capability Characterization (PI, Ron Gilliland): The goal here is to
establish the temporal (differential) photometric accuracy in the very high S/N regime.
In Table 1 we list the individual programs with their corresponding ID numbers. The
reader may find further details via the HST- STScI web page at http://www.stsci.edu/hst/
and in Arribas et al. (2002 a,b) and Malhotra et al. (2002).
Table 2.
Summary of Cycle 12 Calibration Accuracy Goals
Attribute
Detector dark
Flat Fields
Photometry
PSF and Focus
Coronagraphic PSF
GRISM wavelength calibration
GRISM photometry
Polarimetry
Astrometry
4.
Accuracy
< 10 DN
1% broad-band
3% narrow-band
< 6% zero point (filter dependent)
2% relative over the FoV
maintained within 1 mm
for NIC1 and NIC2, 4 mm for NIC3
0.013 arcsec pointing in the hole
0.05 microns
30%
1%
0.5% plate scale
0.1 arcsec to FGS frame
Limiting Factors (Notes)
Temperature fluctuations
Color and temperature dependence
S/N
Absolute calibration, Photometric systems, Intrapixel effects
Breathing and OTA desorption
Centroid of target for zero point
determination
Intrapixel sensitivity
Residual Flat-Field errors
(After geometric distortion correction)
Calibration Accuracies for Cycle 12
In Table 2 we summarize the calibration accuracy goals for Cycle 12. The calibration
proposals executed during the SMOV phase, as well as the ones included in Cycles 10 and 11
calibration programs (see Table 1), were aimed at reproducing and possibly improving
the level of accuracy achieved during Cycle 7 and 7N. Although at the time of writing
this paper only a fraction of these programs have been completed, we do not foresee any
problems in meeting these goals. The actual performance of NICMOS is closely related to
266
Arribas
its temperature stability. The results obtained so far (after 7 months of NCS operations)
indicate very good stability (rms fluctuations ∼ 0.07 K) and, therefore, the accuracies
quoted in Table 2 should be reached.
Acknowledgments. Thanks are due to the members of the IDT group (Dean Hines,
Marcia Rieke, Glenn Schneider, and Rodger Thompson) who have contributed to the recalibration of NICMOS after the installation of NCS. The contribution by Wolfram Freudling
is also very much appreciated. Thanks are also due to Ron Gilliland who has proposed and
designed program 9642.
References
Arribas, S., et al. 2002 NICMOS Cycle 10 Interim Calibration Plan, NICMOS Instrument
Science Report 02-02 (Baltimore: STScI)
Arribas, S., et al. 2002 NICMOS Cycle 11 Calibration Plan, NICMOS Instrument Science
Report 02-03 (Baltimore: STScI)
Hines, D. 2003, this volume, 259
Malhotra, S., et al. 2002 NICMOS Instrument Handbook, Version 5.0, (Baltimore: STScI)
Thompson, R. & Freudling, W. 2003, this volume, 241
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
NCS NICMOS Focus and Coma Analysis
E. W. Roye
Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, Maryland,
21218
A. B. Schultz
Science Programs, Computer Sciences Corporation and the Space Telescope Science
Institute, 3700 San Martin Drive, Baltimore, Maryland, 21218
Abstract. After the installation of the NICMOS Cooling System (NCS), the NICMOS focus was measured and found to be comparable to that observed in Cycle 7.
The NIC1 and NIC2 focii have both moved about 1 mm (in PAM space) in the negative direction since Cycle 7, and the NIC3 focus has moved fractions of a mm (in
PAM space) in the positive direction since Cycle 7, bringing it closer to focus than
it was during that time. New optimal focus values were uplinked to HST on May 8,
2002, and subsequent NICMOS focus monitoring data have revealed that the focus
has remained relatively stable. In addition, coma measurements have been made for
all three cameras. A tilt grid was executed for NIC1 on May 10, 2002 revealing a
significant amount of coma. New settings for NIC1 optimal PAM tilt were uplinked
to HST on May 16, 2002. The correction successfully alleviated the observed NIC1
coma. NIC2 and NIC3 tilt grids executed on June 9, 2002. Smaller amounts of
coma were observed in these two cameras. New settings for NIC2 optimal PAM tilt
were uplinked to HST on September 29, 2002, successfully alleviating all coma. No
update was required for NIC3.
1.
Introduction
The NICMOS Pupil Alignment Mechanism (PAM) is used to adjust the focus of the NICMOS cameras. By moving the PAM back and forth along its axis, the focus can be measured
and adjusted. The focus data consist of a series of images in and out of focus passing through
best focus. The best focus can be subsequently determined from these data. The PAM is
also used to take out misalignments between the HST exit pupil and the NICMOS entrance
pupil. As the PAM is tilted, the relayed HST exit pupil image is translated relative to an
internal NICMOS pupil at the Field Offset Mechanism (FOM). The FOM also carries the
HST spherical aberration correction in its surface figure, so any misalignment between the
aberrated HST pupil and the FOM produces a wavefront error. This error shows up as
a pseudo-derivative of the spherical aberration, i.e., a field-independent coma that varies
linearly with the misalignment of the pupil. The PAM is tilted to correct for such misalignment. The amount of tilt is measured in steps where one step is about 7.8. (PAM
movement can only be adjusted by an integer number of steps.) Measurement of coma is
hence accomplished via a set of data called a “tilt grid,” in which a series of images are
taken at a variety of PAM tilt positions around the default position. The tilt position at
which coma is minimized can be derived from this dataset.
267
268
2.
Roye & Schultz
Observations
The southern open star cluster NGC 3603, also observed for the Cycle 7 focus monitor and
tilt grid programs, was observed for all the focus sweeps (program IDs: 8977, 9323, and
9637) and tilt grids (program IDs: 8977, 9323, and 9637). The focus sweeps in NIC1 and
NIC2 consisted of a series of 17 MULTIACCUM images obtained over a range of ±8 mm of
PAM travel in 1 mm increments. For NIC3, the focus sweep was conducted at PAM settings
of −0.5 mm to −9.5 mm of motion, and consisted of only ten MULTIACCUM images. The
NIC1 tilt grid observations consisted of a 9 point grid surrounding the then-current default
position, with one MULTIACCUM observation taken at each of the 9 grid points. The data
spanned the range of [−4,0,4] in both the x and y directions. The NIC2 tilt grid consisted
of a 13 point grid, with one MULTIACCUM observation taken at each position. The inner
nine points comprised a 3 × 3 ± 4 step tilt grid, and the outer four points comprised a
2 × 2 ± 8 step tilt grid.
The data were re-calibrated off line with calnica using model-generated, color-dependent
flats and specially made NCS darks along with all the other standard reference files.
3.
Focus Monitor Results
The first NCS focus sweep was executed for all three cameras on May 3, 2002. Phase retrieval and encircled energy methods were used to measure the best focus position. The
independent results agreed favorably with one another. Adjustments to the PAM positions
were implemented on May 9, 2002 for NIC1 and NIC2 (PAM1 and PAM2). No focus adjustments were implemented for NIC3 (PAM3) or for the NIC2 coronagraphic focus (PAMC).
Since the first focus sweep there have been five subsequent sweeps: June 5, July 22,
August 26, October 15, and October 28. The focus is relatively stable and fairly consistent
with Cycle 7 and 7N measurements. However both the NIC1 and NIC2 focii have moved
slightly in the negative direction (bringing them closer to 0 mm in PAM space). The NIC3
focus has moved slightly in the positive direction since Cycle 7. This is very fortunate, as it
has brought NIC3 closer to focus than it was for the duration of Cycle 7. However, since the
optimal NIC3 focus is still slightly beyond the range of motion of the PAM, no adjustment
was required for the NIC3 focus position. See Figure 1 for a plot of the NCS focus history
compared with the focus histories for Cycles 7 and 7N.
Slight variations in the measured focus positions are due to periodic variations in the
HST optics, also known as telescope breathing. The present data are not corrected for
breathing due to the lack of a robust model. The breathing model that was implemented
during Cycles 7 and 7N no longer applies to the NCS data. For more detailed information,
see Schultz et al. (2002).
4.
4.1.
Coma Analysis Results
Serendipitous Data Analysis
During the filter wheel test portion of the SM3b SMOV, serendipitous first light images were
taken with NIC1 and NIC2. These images revealed some coma in both cameras. A coarse,
temporary adjustment was made and uplinked to HST on May 9, 2002 to help diminish
the coma in both NIC1 and NIC2.
4.2.
NIC1 Coma Analysis
A nine point tilt grid for NIC1 executed on May 10, 2002. Four analyses of the data were
performed using two independent methods. The first coma-analysis method utilized phase
retrieval to measure the X- and Y -coma values at each of the tilt grid positions. These
NCS NICMOS Focus and Coma Analysis
269
Figure 1.
NCS NICMOS focus history results compared to Cycles 7 and 7N. The
x -axis represents the number of days since January 1, 1997 and y-axis represents
the position of the PAM in mm.
values were then fit to a model and the position at which coma is nulled was extracted
from the model. Three separate analyses of the tilt grid were performed using this method.
A fourth analysis utilized an independent method in which composite PSFs were created
by combining the same nine stars from each of the tilt grid images. Interpolated and
extrapolated model PSFs were then built in order to find the optimal grid position at which
flux was most symmetrically distributed and coma amplitude minimized. All results agreed
well. A recommendation for the final position of (+16, +14) was made and uplinked to
HST on May 16, 2002. NIC1 coma was successfully removed. Figure 2 shows the NIC1
PSF before and after the final tilt correction. For more detailed information, see Roye &
Krist (2002).
4.3.
NIC2 Coma Analysis
A thirteen point tilt grid for NIC2 executed on July 9, 2002. Two separate models were built
based on the measured phase retrieval X- and Y -coma values at each of the grid points. The
results agreed well, and a recommendation for the final position of (+15, +10) was made
and uplinked to HST on September 30, 2002. NIC2 coma was successfully eliminated.
270
Roye & Schultz
Figure 2. NIC1 and NIC2 PSFs before and after PAM tilt correction for coma.
NIC3 PSF is shown for comparison.
Figure 2 shows the NIC2 PSF before and after the final tilt adjustment. For more detailed
information, see Roye et al. (2002).
References
Roye, E. W., et al. 2002, ”SM3B Coma Measurement” Technical Instrument Report NICMOS 2002-002 (Baltimore: STScI)
Roye, E. W. & Krist, J. 2002, ”NIC2 Coma Measurement” Technical Instrument Report
NICMOS 2002-005 (Baltimore: STScI)
Schultz, A. B., et al. 2002, ”SM3B Focus Check” Technical Instrument Report NICMOS
2002-001 (Baltimore: STScI)
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
Combining NICMOS Parallel Observations
A. B. Schultz1 and H. Bushouse
Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218
Abstract. Two problems when working with NICMOS parallel observations are
combining the images into mosaics and removing the telescope thermal background
from the impacted filter images for wavelengths greater than 1.7 microns. We describe a useful technique to combine parallel observations into associations, which
then allows for automated mosaicing and background removal using the NICMOS
pipeline task calnicb. We demonstrate the technique using parallel NICMOS K-filter
images of a region near the galactic center.
1.
Introduction
The majority of NICMOS observations in the HST Archive are obtained in parallel to
other instruments onboard HST and not pointed observations of specific targets obtained
as part of a General Observer (GO) science program. Some of these parallel observations
are coordinated parallel observations associated with the primary science observations, while
others are pure parallel and are unrelated to the primary science observations. Many parallel
observations have no proprietary period and are available to the general science community
within a day or two of arrival on the ground.
In this report, we present a method to combine individual parallel observations into
an association so that calnicb can be used in creating a mosaic of the images. And, we
address the problem of removing the thermal background from observations obtained with
the thermally impacted filters (λ > 1.7 µm).
2.
The Data Set
As part of the Cycle 11 calibration program to determine the stability of the HST +NCS+
NICMOS thermal background (program ID: 9269), NIC3 F222M filter observations are
obtained in parallel to other HST instruments. It is this data set that we will use to
demonstrate how to group parallel observations into associations. In particular, a star field
near the galactic plane was observed with NIC3 in parallel to the prime STIS/CCD science
observations of W-SGR (program ID: 9105), a binary Cepheid variable (HD164975).
Any targeted NICMOS F222M filter observation will need a background observation,
preferably of equal exposure and of a blank field, to remove the thermal background of the
telescope from the data. We will use exposures of a sparse field from the extended set of
observations for program 9269, which were obtained close in time to the star field images,
as the background data set. For this example, eight exposures each of the star field and the
background will be used to produce a mosaic. Both data sets were calibrated using the standard STScI data pipeline program calnica. Calnica removes the instrumental signature
from the individual exposures. The calibration steps for NICMOS data are described in
1
Science Programs, Computer Sciences Corporation
271
272
Schultz & Bushouse
the NICMOS Instrument Handbook (Malhotra et al. 2002) and in the HST Data Handbook
(Mobasher et al. 2002).
In the HST ground system, a proposal exposure line that yields multiple exposures
will automatically trigger the creation of an association for those exposures, including the
creation of an association table that lists the names of all the individual exposures. The
association table is then used as input to calnicb. Exposures obtained individually do not
trigger the automatic creation of an association table, therefore it must be created by hand
in order to process those data through calnicb. Note that not just any random collection
of images can be processed with calnicb, as calnicb does not handle rotations amongst
images, only simple x/y (RA/Dec) translational shifts. Any rotation between images must
be removed before processing with calnicb.
2.1.
Science Header Keywords
A few keywords in the headers of calibrated files (ipppssoot cal.fits) need to be modified to
enable calnicb processing of the parallel observations. If images are only to be stacked with
no pattern, then only the NUMITER keyword needs to be modified to reflect the number
of images to be stacked.
For the example discussed here (target and background), five keywords need to be
modified. The keywords PATTERN1, P1 NPTS, PATTERN2, and P2 NPTS only need to
be set in the header of the first image listed in the association table, as they are assumed to
be constant in all images. The keywords PATTERN2 and P2 NPTS will most likely need
to be added to the header as they are omitted for single exposures. The PATTERN2 and
P2 NPTS keywords indicate that a secondary pattern was used; such as, a chop pattern.
The PATTSTEP keyword will need to be set to a unique value in each image header.
The pattern type (PATTERN1 and PATTERN2) should be an accepted NICMOS
pattern as defined in the Phase II instructions. For collections of images that contain both
target and background exposures use one of the “CHOP” pattern types; such as, “NICONE-CHOP”. In the case for just target images, one of the dither pattern types; such as,
“NIC-SPIRAL-DITH” will work. The PATTERN1 and PATTERN2 keywords are set to
the same value.
The value of the P1 NPTS keyword needs to be the total number of pattern positions
observed, while the P2 NPTS keyword value will depend upon the pattern selected. For this
example, the NIC-ONE-CHOP pattern was selected as the target and background exposures
will only be stacked. The P2 NPTS keyword needs to be set to a value of “2”.
The PATTSTEP keyword must be set to a monotonically increasing number, starting
with 1 for the first image in the pattern up through (P1 NPTS × P2 NPTS) for the last
image. The necessary pattern keywords and their respective values are shown below.
> hedit n8c2f8pvq_cal.fits[0] P2_NPTS 2 add+
>imhead n8c2f8pvq_cal.fits[0] l+
...
/ PATTERN KEYWORDS
PATTERN1=
P1_SHAPE=
P1_PURPS=
P1_NPTS =
P1_PSPAC=
P1_LSPAC=
P1_ANGLE=
P1_FRAME=
P1_ORINT=
P1_CENTR=
BKG_OFF =
PATTSTEP=
’NIC-ONE-CHOP’
’
’
’
’
8
0.000000
0.000000
0.000000
’
’no
’
’
/
/
/
/
/
/
’
/
0.000000 /
’
/
/
2 /
/ primary pattern type
primary pattern shape
primary pattern purpose
number of points in primary pattern
point spacing for primary pattern (arc-sec)
line spacing for primary pattern (arc-sec)
angle between sides of parallelogram patt (deg)
coordinate frame of primary pattern
orientation of pattern to coordinate frame (deg)
center pattern relative to pointing (yes/no)
pattern offset method (SAM or FOM)
position number of this point in the pattern
Combining NICMOS Parallel Observations
PATTERN2= ’NIC-ONE-CHOP’
P2_NPTS =
3.
273
2
The Association Table
An association table (i.e., ipppssoot asn.fits) is stored in a FITS file, in a FITS binary table
extension. It contains a list of the members in the association, relevant information on the
exposures (target or background), and the name of the output product (ipppssoot mos.fits).
For example, the association table n626s4020 asn.fits displayed below contains three rows,
consisting of the names of the two exposures and the output product name.
Column
Label
1
2
3
> tread n626s4020_asn.fits
1
2
___MEMNAME____ ___MEMTYPE____
N626S4DCQ
EXP-TARG
N626S4DFQ
EXP-TARG
N626S4020
PROD-TARG
3
MEMPRSNT
yes
yes
yes
The easiest way to create an association table for non-association exposures is to copy
an existing table and use the ttools package task tedit to edit the table entries. The task
tedit allows the user to add or delete rows and to edit individual row entries. There should
be a row for each exposure and a row for the output product. For the following example,
there are eight rows for the target and background images and two rows for output products.
Column
Label
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
> tedit n8c2f8010_asn.fits
1
2
___MEMNAME____ ___MEMTYPE____
N8C2F8PVQ
EXP-TARG
N8C2F8PWQ
EXP-TARG
N8C2F8PXQ
EXP-TARG
N8C2F8PYQ
EXP-TARG
N8C2F8PZQ
EXP-TARG
N8C2F8Q1Q
EXP-TARG
N8C2F8Q4Q
EXP-TARG
N8C2F8Q5Q
EXP-TARG
N8C2GLE8Q
EXP-BCK1
N8C2GLE9Q
EXP-BCK1
N8C2GLEAQ
EXP-BCK1
N8C2GLEBQ
EXP-BCK1
N8C2GLEEQ
EXP-BCK1
N8C2GLEFQ
EXP-BCK1
N8C2GLEIQ
EXP-BCK1
N8C2GLEJQ
EXP-BCK1
N8C2F8010
PROD-TARG
N8C2F8011
PROD-BCK1
3
MEMPRSNT
yes
yes
yes
yes
yes
yes
yes
yes
yes
yes
yes
yes
yes
yes
yes
yes
yes
yes
Calnicb will create separate output products for the star field and the background,
stacking and averaging the images since no dithering was performed. Calnicb also performs
background subtraction and source identification on the images in the association.
> calnicb n8c2f8010_asn.fits
274
Schultz & Bushouse
Figure 1.
NIC3 F222M filter parallel imaging of the galactic plane, (left) the
calibrated star field image and (right) the star field with the thermal background
subtracted. In practice, the field stars in the background images should be removed
before the subtraction. The total integration time was (8 × 128 =)1, 024 seconds.
The band along the bottom of the images, about ∼ 15–20 rows wide, is due to
vignetting by the FDA mask.
For this example, the resulting background sky image contained a few point sources
which were removed by median filtering before subtracting it from the star field image. The
mstools package task msarith was used to perform the subtraction in order to properly
propagate the data quality and error arrays of the multi-IMSET files.
> msarith n8c2f8010_mos.fits - n8c2f8011_mos.fits starfield_bck
4.
Discussion
The combined F222M filter image and the same image with the background subtracted are
presented in Figure 1. The parallel NICMOS images revealed a couple of dozen bright stars
in front of a densely packed background of faint stars, thus demonstrating the usefulness of
the above technique to combine parallel observations into a single association.
References
Malhotra, S., et al. 2002, “Near Infrared Camera and Multi-Object Spectrometer Instrument
Handbook,” Version 5.0, October 2002 (Baltimore: STScI)
Mobasher, B., et al. 2002, “HST Data Handbook for NICMOS,” Version 5.0, January 2002
(Baltimore: STScI)
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
NICMOS User Tools and Calibration Software Updates
M. Sosey
Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218
Abstract. This is a summary of the available internet software tools useful for
processing NICMOS datasets. A general overview of each web tool and its output is
provided along with a short discussion on the most recent updates to the calibration
pipeline software—CALNICA and CALNICB.
1.
Currently Available Internet Tools
All of the tools described below may be found under the Software Tools heading at the
NICMOS Instrument website—http://www.stsci.edu/instruments/nicmos.
1.1.
Temperature Dependent Dark Tool
This tool generates synthetic dark reference files which correspond to a specific temperature. The web form of this tool currently only creates darks which can be used to calibrate data taken between August 22, 1997 and January 4, 1999. Similar code has already
been implemented in CALNICA, in the NICMOS calibration pipeline software, to create
on-the-fly temperature dependent darks. The CALNICA code accommodates all cycles
of NICMOS observations. Further information on the synthetic darks can be found at
http://www.stsci.edu/hst/nicmos/tools/syndark.html and in Monroe (1999).
1.2.
SAA Crossing Calculator
This web tool details when the South Atlantic Anomaly (SAA) passages occurred for a given
time period or set of observations. It uses a set of Science Mission Schedule (SMS) calendar
files, which contain the SAA entrance and exit times. The NICMOS SAA avoidance model
is number 23. Every instrument/detector has its own model, hence allowing changes to a
model without unnecessarily affecting other instruments. At the moment SAA model 05
and model 23 share the same definition, however, this may change in the future.
1.3.
Imaging Exposure Time Calculator (ETC)
This tool estimates the integration time and Signal-to-Noise Ratio (SNR) on a point source
or extended object, derived selection limits and checks for saturation. For point sources,
the SNR calculation is for the brightest pixel. For extended sources, the signal is the number of photons detected in one pixel. Noise is always the noise in a single pixel (standard
deviation in a pixel due to photon statistics and instrumental noise). Proposers are strongly
recommended to use the ETC to estimate exposure times. The ETC provides the most accurate estimates with the most current information on the instrument performance. This
is achieved by updating the ETC reference tables with the most recent values of the instrumental characteristics. Information on the scientific verification of the NICMOS CGI ETC
can be found in Sosey (2001).
275
276
1.4.
Sosey
Attitude History Tool
This tool returns two plots, both histograms. The first is Solar Elongation versus Time,
and the second is the Off-Nominal Roll (in degrees) versus Time. The only input required
is the start time of the dataset(s); the returned graph will cover the previous and following
two days.
1.5.
Temperature Dependent Flatfield Tool and Color Dependent Flats
This tool generates temperature dependent flatfield reference files and is available for all
imaging filters. NICMOS is operating at a stable temperature of 77.1 K under NCS control
and there is no temperature dependent data available to create the flatfield model around
this value. Therefore, this tool is only relevant for data taken between January 1997 and
January 1999 (pre-NCS).
There are two scripts available on the NICMOS website (http://www.stsci.edu/hst/
nicmos/tools/colorflat intro.html) for creating color dependent flatfield images. As detailed
in Storrs (1999), the pipeline flatfield data for NICMOS is made with the use of on-board
flatfield lamps which have an intrinsically blue color over most of the NICMOS sensitivity
range. Sources with extreme colors do exist and broadband images of such sources may
require special treatment.
1.6.
Units Conversion Tool
The NICMOS Units Conversion Program is a tool for converting fluxes of astronomical
sources from units which are widely used in Astronomy (e.g., magnitude, ergs/sec/cm2 /angstrom) into Jansky (Jy), and vice versa. The Jy, or for the case of an extended source,
Jansky/arcsec2 , is the flux unit adopted in the NICMOS Handbook and used by the NICMOS ETC. The details of the FORTRAN program which handles the unit conversion are
explained in Skinner (1996). The report (and all other Instrument Science Reports) are
available under the Documentation section of the NICMOS WWW page. Modification to
the FORTRAN program were added by Daniela Calzetti (July 02, 1996 and April 09, 1997)
to handle power-law spectra and AB mags).
Unit conversion between magnitude and fluxes, and visa versa, requires information
on the magnitude, zero-point value and the bandpass central wavelength. The zero-point
magnitudes used in this program are from the CIT system, or in the case of the L1 band—
commonly known as L band—from the UKIRT system.
1.7.
Polarimetric Imaging Tools
A set of IDL programs exsist which may be used to produce polarization coefficient images
from observed data in NIC1 or NIC2. The images are derived using input from the user
and contain options for superimposing polarization vectors and contours on the intensity
image. A detailed user’s manual exsists in Mazzuca & Hines (1999). Further information
on methodologies for reducing polarimetric data can be found in Mazzuca et al. (1998).
2.
NICMOS Calibration Pipeline Updates
The following updates have been made to the NICMOS calibration pipeline software, and are
applied to versions after and including CALNICA v4.0 and CALNICB v2.5. STSDAS v3.0
is the current release, it contains the updates to the CALNIC software and can be down
loaded at STSDAS webpage served from the STScI web pages.
2.1.
CALNICA Updates
Version 4.0 now creates and applies temperature dependent darks for all datasets. This is
applicable to all data taken since August 22, 1997. The new code is based on the code used
NICMOS User Tools and Calibration Software Updates
277
in the Web based tool. The web tool currently only accommodates pre-NCS data, while
the CALNICA code has been modified to accept all NICMOS datasets. This is accomplished with the new tdd.fits reference files which contain tables of shading profiles and are
referenced by temperature—keyed off the new global header keyword TEMPFILE. Setting
TEMPFILE=‘N/A’ will force CALNICA to use the dark file referenced in DARKFILE.
Using the temperature depended dark code is the default action that CALNICA takes. If
you wish to use the new version of CALNICA on files already saved to a local disk, you
must add the keyword TEMPFILE to the header.
On-the-fly processing (OTFP) is now implemented for all NICMOS data which is retrieved from the archive. OTFP data should still always be checked for agreement with the
most recent reference file.
Following deep SAA passages, residual images of CR hits incurred while in the SAA are
still visible in long exposures following the passage. The spatially correlated nature of the
decaying signal means CALNICA cannot find and reject them. The random distribution
of these ‘persistent cosmic rays’ increases the noise in the image and limits faint source
detections. The time scale of the decaying signal is exponential and is detectable for ∼
30 minutes following a passage (the same time scale as that from persistent images caused
by extremely bright sources). Starting in Cycle 11, a pair of ACCUM mode NICMOS dark
exposures is scheduled after each SAA passage in order to provide a map of the persistent
cosmic ray afterglow at a time when it is strongest, and has just begun to decay. Post-SAA
darks are delivered for each affected dataset retrieved from the HST Archive. However,
the analysis for creating and using crmaps with these darks is still underway. New header
keywords have been added to the global extension of all science images and are listed in
Table 1.
Table 1.
New Header Keywords
Keyword
SAA EXIT
SAA DARK
SAA TIME
SAACRMAP
Description
Time of last exit from the SAA 23 contour
Association name for the post-SAA dark exposure
Seconds since last exit from SAA 23
SAA cosmic ray map file (still in development)
Software is being developed to measure the detector temperature from bias measurements. This method is at least as accurate as reading the temperature from the mounting
cup sensors. This method will be the most accurate temperature determination for camera 3, for which the mounting cup sensor has a limit of 76.28K, below the current operating
temperature of 77.1 K. For this reason all camera temperatures are now being specified by
the mounting cup 11 sensor (NDWTMP11) in the spt[1] header.
2.2.
CALNICB Updates
CALNICB has been updated to recognize and process the generic pattern types of SPIRAL
and LINE. These can be used to perform dithering and chopping at the same time, producing
target and background images. The software does not recognize the background images as
such, but they will now process cleanly through CALNICB.
References
Mazzuca, L., Sparks, B., & Axon, D. 1998, Instrument Science Report NICMOS 98-017
(Baltimore: STScI)
278
Sosey
Mazzuca, L. & Hines, D., 1999 Instrument Science Report NICMOS 99-004 (Baltimore:
STScI)
Monroe, B. 1999, Instrument Science Report NICMOS 99-010 (Baltimore: STScI)
Skinner, C. 1996, Instrument Science Report NICMOS 96-014 (Baltimore: STScI)
Sosey, M. 2001, Instrument Science Report NICMOS 2001-01 (Baltimore: STScI)
Storrs, A., Bergeron, E., & Holfeltz, S. 1999, Instrument Science Report NICMOS 99-002
(Baltimore: STScI)
Part 4. WFPC2
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
WFPC2 Status and Overview
B. C. Whitmore
Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218
Abstract. The current status of the Wide Field and Planetary Camera 2 is reviewed, with special emphasis on the five years since the previous HST Calibration
Workshop. The WFPC2 continues to work nearly flawlessly, with the only major
problem being a shutter anomaly in October, 2000, which put the camera offline for
approximately a month. The two servicing missions (SM3a and SM3b) did not affect
the long-term characteristics of the WFPC2. We also report on the status of basic WFPC2 characteristics (photometry, focus, and astrometry) and briefly mention
some of the recent WFPC2 projects.
1.
Highlights From the Past Five Years
It has been five years since the previous calibration workshop, and the report by Whitmore
(1997) on the status of the WFPC2. An earlier status report was provided by MacKenty
(1995). The current report will therefore concentrate on the past five years. WFPC2 has
been operating for almost nine years now (launch was December, 1993), and over 125,000
external science exposures have been taken. The instrument has worked almost flawlessly
during this period, with the only mechanical problem being a shutter anomaly in October,
2000, which put the WFPC2 offline for approximately one month. This problem was due to
the degradation of the LED-sensor assembly in the shutter mechanism combined with slight
mechanical misalignments. The problem was solved by increasing the length of time the
LED is on before reading the sensor. A detailed report is available in Casertano (2000a).
Two servicing missions have occurred during the past five years, Servicing Mission 3a
in December, 1999, and Servicing Mission 3b in March, 2002. In both cases, extensive post
servicing mission tests have shown that the WFPC2 characteristics (e.g., quantum efficiency,
point-spread-function, flatfields, ...) have not been affected. The only minor exception is a
temporary (approximately one month) increase in the UV contamination rate, presumably
due to additional contaminants originating from the shuttle and/or other newly installed
instruments during the servicing mission. Detailed reports covering these post servicing
mission tests are available in Casertano et al. (2000b) and Koekemoer et al. (2002a).
Table 1 shows how the filter usage has changed over the past five years. There are
two trends responsible for most of the changes. The first is the increased usage of the very
wide filters F300W, F450W, F606W, and F814W. This evolution was inspired by the Hubble
Deep Field, which employed these four filters. The second trend is for the increased usage of
the narrow-band filters, especially F656N (Hα), F658N (redshifted Hα), and F502N (OIII).
Some of the scientific highlights originating from WFPC2 during the past five years
have been the Hubble Deep Field South; untangling the nature of gamma gay bursts; and
observations of supernovae at z > 1 that support the existence of dark energy, to name just
a few. On the calibration front, the development of the DRIZZLE software (Fruchter &
Hook 1997), and formulae for the correction of Charge Transfer Efficiency loss (Whitmore,
Heyer, & Casertano 1999, Dolphin 2000) were important contributions.
When one reflects on the fact that the WFPC2 has received the lion’s share of observing
time during the past nine years, on arguably the most important telescope ever developed,
281
282
Whitmore
it would appear to be a fair statement to say that the WFPC2 camera may be the most
successful astronomical instrument ever built. The astronomical community owes a great
debt of gratitude to John Trauger and the Instrument Development Team that designed
and built the WFPC2.
Table 1.
Historical Usage of Filters since Launch
Filter
F814W
F606W
F555W
F300W
F450W
F675W
F702W
F439W
F656N
F336W
F170W
F547M
F850LP
F502N
F1042M
F673N
F160BW
F658N
F410M
F953N
F255W
F218W
F791W
F785LP
F631N
F467M
F622W
F588N
F380W
F487N
F122M
a
Dec/93–Sep/02a
(# of exposures)
18867
16560
11700
6300
3729
3316
2756
2304
2283
2219
1991
1903
1542
1165
1043
1042
966
934
926
921
919
755
588
503
355
336
237
200
198
191
181
rank
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
Dec/93–Aug/97a
(# of exposures)
7689
3033
5506
1381
680
1516
1755
1254
684
1225
1129
806
698
451
500
518
755
300
382
549
487
469
rank
1
3
2
6
14
5
4
7
13
8
9
10
12
21
18
17
11
24
22
15
19
20
Comments
NOTE CHANGE
NOTE CHANGE
NOTE CHANGE
NOTE CHANGE
NOTE CHANGE
NOTE CHANGE
NOTE CHANGE
NOTE CHANGE
External exposures only
2.
2.1.
The Basics
Photometry
The photometric stability over a short period of time for the major broadband filters continues to be very good, with rms scatter <1% (Figure 1). However, a slow (few percent
during the 9 years in orbit), linear decrease is present in the throughput for most filters.
This appears to be due to the degradation of Charge Transfer Efficiency (CTE) with time
(See Figure 4.6 of Koekemoer et al. 2002a).
WFPC2 Status and Overview
Figure 1.
Comparison of post-SM3b photometric monitoring observations (the
large stars) with historical trends. (Whitmore & Heyer 2002a, Koekemoer et al.
2002a).
283
284
Whitmore
Figure 2.
Growth of Y-CTE with time for faint stars (Heyer, 2001).
This degradation of CTE with time continues to be perhaps the most significant problem for the WFPC2. The existence of CTE loss was first discussed by Holtzman et al.
(1995b). They advocated using a 4% ramp to correct for the problem at the time. Whitmore et al. (1999) more fully characterized CTE loss, and showed that it is a function of
1) Y position on chip (parallel CTE), 2) X position on the chip (serial CTE), 3) target
brightness, and 4) background level. They also found that the problem was growing linearly with time (see Figure 2). Since then, several other studies have confirmed the effect,
and have added important contributions to the characterization of the CTE. In particular,
the recommended formula for correcting for CTE loss is currently that of Dolphin (2000 =
original paper, 2002 = WWW site with latest update). Other contributions of interest include studies of CTE residuals (Biretta & Mutchler 1997, Baggett et al. 2000), using cosmic
rays to measure CTE loss (Riess 1999), extended source CTE (Riess 2000), CTE monitoring
(Heyer 2001) and CTE at very low levels and a reexamination of the long-vs.-short anomaly
(Whitmore & Heyer 2002b). There is also a nice summary of CTE loss for various HST
instruments (Stiavelli et al. 2001).
The determination of the photometric zeropoints is also an important topic for many
projects. A comparison of zeropoint determinations from five different studies for the primary broad band filters is shown in Figure 3. The rms scatters for the five studies are
0.043 mag for F336W, 0.034 mag for F439W, 0.016 mag for F555W, and 0.018 mag for
F814W. We note that the most widely used filters, F555W and F814W, are particularly
good, with rms scatter less than 0.02 mag.
In the past, the zeropoints determined at STScI have been based on observations of our
monitoring star, GRW+70D5824, and the SYNPHOT (synthetic photometry) package in
STSDAS. We are now engaged in a project to determine zeropoints (and check Holtzman’s
transformation equations) based on comparisons with: Stetson (2002) standards, Saha et al.
(2002) standards, and a few Landolt (1992) standards. See Heyer et al. (2002) for details.
WFPC2 Status and Overview
285
Figure 3. Comparison between five different zeropoint determinations (Heyer,
Richardson, Whitmore & Lubin 2003).
We conclude this section by noting that although CTE loss has compromised photometric accuracy to some extent, for typical observations it is still possible to obtain absolute
photometric accuracies of a few percent with WFPC2.
2.2.
Focus
The frequency of focus moves required to stay within 2.5 microns of best focus (which
roughly equals the orbital variations induced by “breathing”) has decreased, due to the
slowing rate of OTA shrinkage caused by water desorption. Figure 4 shows the WFPC2 focus
position since launch. Note that the recent trend is slightly negative. A focus adjustment
was therefore made in November, 2002. The focus measurements are made on the PC, due
to its better spatial sampling. The other chips are slightly out of focus, since there is no
mechanism to adjust separately the focus for each chip. In particular, the worst focus is
on the WF3 chip, which is about 5 microns out of focus relative to the PC. For further
discussion and information on how focus affects photometry see Suchkov and Casertano
(1997).
286
2.3.
Whitmore
Astrometry
The ability to transform accurately from x, y instrument coordinates to RA, DEC coordinates on the sky is fundamental for a large fraction of WFPC2 observations. In addition,
the advent of dithering increases the need for very accurate relative astrometry, so that
blurring of the image due to a poor geometric solution is not introduced when recombining
the image.
A number of geometric distortion solutions were determined shortly after launch, including those by Holtzman et al. (1995a), Trauger et al. (1995), and Gilmozzi et al. (1995).
These had ≈5 mas residuals near the center and 10–15 mas residuals in the corners. Casertano & Wiggs (2001) developed a much better solution with uncertainties ≈1.2 mas on
the WF and 4 mas on the PC. Most recently, Anderson & King (2003) produced what is
currently the best solution with ≈1 mas rms per star. Kozhurina-Platais et al. (2003) are
extending this work done in the F555W filter to encompass other wavelengths.
3.
3.1.
Recent WFPC2 Projects
UV Contamination Rate Update
The UV throughput for WFPC2 degrades with time due to contaminants within the cameras
that freeze out on the faceplate. For example, the throughput in the F160W filter on the
PC declines by about a percent per day. The throughput can be recovered by heating
the WFPC2 and sublimating the contamination off the faceplate. Until recently, these
decontaminations were done roughly once per month.
A recent reexamination of the UV throughput as a function of time by McMaster &
Whitmore (2002) shows that the contamination rates have decreased by roughly 50% since
WFPC2 was launched. Hence, we now have a long enough baseline to determine temporal
trends, rather than simply measuring the rate for a given year. This allows us to average the
data, and provides a simpler, more accurate correction. The lower rate has also allowed us
to save telescope time by extending the time between decontaminations from 30 to 50 days.
3.2.
Pointing Accuracy
Pointing accuracy is less important for the WFPC2 than for most of the other instruments,
due to its wide field of view. However, there are instances where larger than normal excursions between the true and intended positions can cause problems. Hence, we try to
keep the observed offsets to ≈ 1 , so that the uncertainty due to guide star positions is
the dominant error. However, as the FGS-to-FGS alignments change with time (e.g., do to
desorption in the new FGS’s after they are launched), the observed offsets also change, and
depend on which of the three FGSs is “dominant.”
A recent examination of the pointing accuracy by Brammer, Whitmore & Koekemoer
(2003) showed an offset of ≈ 1.5 between the commanded and actual pointing positions for
WFPC2. While not particularly important for most WFPC2 users, this does increase the
uncertainty in determining absolute astrometric positions from the 1–2 values imposed by
uncertainties in guide star positions to a value of 2–3. A realignment of the FGS-to-FGS
positions in October, 2002, is expected to solve this problem. We note that for cases where
accurate astrometric accuracy is important, the users can remove this offset by determining
the positions of astrometric standards that are in the WFPC2 field of view, if they exist.
3.3.
Flatfield Update
High quality flatfields are critical for a variety of astronomical projects, especially those
where very accurate photometry is required. Improved flatfields were developed by Koekemoer, Biretta, and Mack (2002) and installed in the calibration pipeline in March, 2002.
Improvements included: 1) the first major revision since previous flats were produced (based
WFPC2 Status and Overview
287
Figure 4.
Measured focus position for WFPC2 since launch. A value of 0 microns
is the optimal focus position.
on 1994–1995 data), 2) inclusion of new dust spots as a function of time, 3) measurement
of the time dependence of large-scale features caused by long-term changes in the camera
geometry, 4) measurement of pixel-to-pixel structure to levels below ≈0.3% for the PC and
≈0.2% for the WF chips. Improved UV flatfields were also produced by Karkoschka &
Biretta (2001; installed in 2001), and Karkoschka (2003).
3.4.
CADC “Association” Images
The Canadian Astronomical Data Center (CADC), in conjunction with the European Coordinating Facility (ECF), have combined WFPC2 images from the archives to produce “association” images. These are currently being used as preview images for the HST archives.
The images will be available through the HST archive in the near future. For more information see the CADC WWW site at: http://cadcwww.hia.nrc.ca/wfpc2/
288
Whitmore
Figure 5. On the left are the raw positions of our standard star (GRW+70d5824)
using different guide stars. On the right are the “corrected” positions. See Brammer, Whitmore & Koekemoer (2003) for details.
4.
References and Resources
The WFPC2 has now been in orbit for almost nine years, and an extensive set of documentation has been developed. This includes both STScI developed documents, such as
the WFPC2 Instrument Handbook, and papers in the astronomical journals, such as the
two classic Holtzman papers (Holtzman et al. 1995a,b). The Data Analysis Library (formerly know as the WFPC2 Clearinghouse) provides easy access to both sets of documents,
although it has not been extensively updated as far as the external astronomical literature since 1997, when the high volume of WFPC2 related articles became overwhelming.
A history file is also maintained that provides a chronological record of various WFPC2
information (e.g., focus moves, decontamination dates, software changes, etc.).
The primary sources of WFPC2 documents are available via the WFPC2 WWW site
at: http://www.stsci.edu/instruments/wfpc2/wfpc2 top.html. These include:
• The WFPC2 Instrument Handbook–Version 7.0, October 2002 (Biretta, Lubin, et al.
2002)
• The HST Data Handbook for WFPC2–Version 4.0, January, 2002 (Baggett, McMaster, et al. 2002)
• The HST Dither Handbook–Version 2.0, January, 2002 (Koekemoer et al. 2002b)
• The WFPC2 Tutorial–Version 3.0, July 2002 (Gonzaga et al. 2002)
Detailed reports on specific topics are available at the WFPC2 WWW site in the form
of Instrument Science Reports (ISRs) and Technical Instrument Reports (TIRs, generally
for internal use but available on request). The following software tools are also available:
• Exposure Time Calculator
WFPC2 Status and Overview
289
• Linear Ramp Filter Calculator
• CTE Estimation Tool
• Polarization Calibration Tool
• Data Analysis Library (formerly WFPC2 Clearinghouse)
Other WWW sites of interest are:
• Andrew Dolphin’s WWW page (e.g., CTE correction formula):
http://www.noao.edu/staff/dolphin/wfpc2 calib/
• The WFPC2 “Metrics” page: http://www.stsci.edu/hst/metrics/SiUsage/WFPC2/
5.
Summary
The WFPC2 continues to work almost flawlessly, with no major mechanical, electrical, or
systems problems. The most serious problem that has occurred in its nine years of operation
is a shutter malfunction in October, 2000, that resulted in one month of downtime before
it was fixed. During most of this time the telescope has received the lion’s share of the
observing time (≈40%). The images from the camera have inspired both scientists and
the public with incredible pictures that include the Hubble Deep Field, the collision of
Comet Shoemaker-Levy with Jupiter, and the Eagle Nebula, to name just a few. The
degradation of Charge Transfer Efficiency (CTE) with time has been perhaps the most
important calibration issue, but for typical science exposures this still only amounts to a
loss of about 10% of the throughput. A correction formula (Dolphin 2002) is available for
point sources, and work continues on a tool to correct images on a pixel-by-pixel basis.
An extensive set of documentation is available, both via the WFPC2 WWW site and in
the astronomical community at large. WFPC2 will be replaced by Wide-Field Camera 3
(WFC3) during Servicing Mission 4, which is currently scheduled for February, 2005.
Acknowledgments. We wish to thank the entire STScI WFPC2 team, both past and
present, for their support of the WFPC2. It has truly been a pleasure working with this
dedicated group of people. Special thanks to Sylvia Baggett, who determined the statistics
shown in Table 1, and Matt Lallo and Russ Makidon, who produced the focus plot shown
in Figure 4. Finally, thanks to John Trauger, Jon Holtzman, and the IDT for building the
WFPC2 and producing the original characterization of the camera.
References
NOTE: Instrument Science Reports listed below are available at:
http://www.stsci.edu/instruments/wfpc2
Anderson, J. & King, I. 2003, PASP, in press
Baggett, S., et al. 2000, Instrument Science Report WFPC2 00-03 (Baltimore: STScI)
Baggett, S., McMaster, M., et al. 2002, HST Data Handbook for WFPC2 (Baltimore:
STScI)
Biretta, J. & Mutchler, M. 1997, Instrument Science Report WFPC2 97-05 (Baltimore:
STScI)
Biretta, J., Lubin, L. M., et al. 2002, WFPC2 Instrument Handbook (Baltimore: STScI)
Brammer, G., Whitmore, B. C., & Koekemoer, A. 2003, this volume, 329
Casertano, S., et al. 2000a, http://www.stsci.edu/instruments/wfpc2/wfpc2 resources.html
290
Whitmore
Casertano, S., et al. 2000b, Instrument Science Report WFPC2 00-02 (Baltimore: STScI)
Casertano, S. & Wiggs, M. 2001, Instrument Science Report WFPC2 01-10 (Baltimore:
STScI)
Dolphin, A. 2000, PASP, 112, 1397
Dolphin, A. 2002, http://www.noao.edu/staff/dolphin/wfpc2 calib
Fruchter, A. & Hook, R. N. 1997, SPIE, 3164, 120
Gonzaga, S., et al. 2002, WFPC2 Tutorial (Baltimore: STScI)
Gilmozzi, R., Ewald, S., & Kinney, E. 1995, Instrument Science Report WFPC2 95-02
(Baltimore: STScI)
Heyer, I. 2001, Instrument Science Report WFPC2 01-09 (Baltimore: STScI)
Heyer, I., Richardson, M., Whitmore, B. C. & Lubin, L. 2002, this volume, 333
Holtzman, J., et al. 1995a PASP, 107, 156
Holtzman, J., et al. 1995b PASP, 107, 1065
Karkoschka, E. 2003, this volume, 315
Karkoschka, E. & Biretta, J. 2001, Instrument Science Report WFPC2 01-07 (Baltimore:
STScI)
Koekemoer, A. M., Biretta, J., & Mack, J. 2002, Instrument Science Report WFPC2 02-02
(Baltimore: STScI)
Koekemoer, A. M., et al. 2002a, Instrument Science Report WFPC2 02-06 (Baltimore:
STScI)
Koekemoer, A. M., et al. 2002b, HST Dither Handbook (Baltimore: STScI)
Kozhurina-Platais, V., et al. 2003, this volume, 354
Landolt, A. 1992, AJ, 104, 340
Mackenty, J. C. 1995, in The 1995 HST Calibration Workshop, eds. A. Koratkar & C. Leitherer (Baltimore: STScI)
McMaster, M. & Whitmore, B. 2002, Instrument Science Report WFPC2 02-07 (Baltimore:
STScI)
Riess, A. 1999, Instrument Science Report WFPC2 99-04 (Baltimore: STScI)
Riess, A. 2000, Instrument Science Report WFPC2 00-04 (Baltimore: STScI)
Saha, A. 2002, private communication
Stetson, P. 2002, http://cadcwww.hia.nrc.ca/standards/
Stiavelli, M., et al. 2001, Instrument Science Report WFC3 01-05 (Baltimore: STScI)
Suchkov, A. & Casertano, S. 1997, Instrument Science Report WFPC2 97-01 (Baltimore:
STScI)
Trauger, J. T., et al. 1995, in The 1995 HST Calibration Workshop, eds. A. Koratkar &
C. Leitherer (Baltimore: STScI)
Whitmore, B. C. 1997, in The 1997 HST Calibration Workshop, eds. S. Casertano, R. Jedrzejewski, Tony Keyes, & Mark Stevens (Baltimore: STScI)
Whitmore, B. C. & Heyer I. 2002a, Technical Instrument Report WFPC2 02-04 (Baltimore:
STScI)
Whitmore, B. C. & Heyer I. 2002b, Instrument Science Report WFPC2 02-03 (Baltimore:
STScI)
Whitmore, B. C., Heyer I., & Casertano, S. 1999, PASP, 111, 1559
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
WFPC2 Calibration and Close-Out
A. M. Koekemoer
Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218
Abstract. This work summarizes the overall calibration strategy for WFPC2, covering the design of the observational programs as well as analysis of the results and
their incorporation into the calibration pipeline. This strategy comprises routine
long-term calibration monitoring programs, including darks, biases, flat fields, photometric and astrometric monitoring, as well as special calibration programs such
as CTE and PSF characterization, and photometric cross-calibration with ACS and
ground-based systems. In addition, we discuss special close-out calibration programs
planned for the remaining cycles of WFPC2 operation, and describe ways in which
community input can play a key role in further defining these plans.
1.
Introduction
The Wide Field and Planetary Camera 2 (WFPC2) has been the principal imaging camera
on board the Hubble Space Telescope (HST ) for the past nine years, after being installed
during the first Servicing Mission in December 1993. Its comprehensive suite of 48 filters,
spanning wavelengths from the far ultraviolet to one micron, and including wide, medium
and narrow-band as well as polarimetric and linear ramp filters, have facilitated an exceptionally wide range of scientific projects, with over 125,000 science exposures obtained to
date.
Cycle 12 is currently planned to be the last full cycle for WFPC2 operation, since it
will be removed in 2005 during Servicing Mission 4 and replaced with the Wide Field Camera 3 (WFC3). Therefore we are currently planning the final cycle of special “Close-Out”
calibration programs and related activities, aimed at maximizing the scientific value of the
wealth of archival WFPC2 data. In addition to our normal calibration plan for WFPC2
that is performed during each cycle, we are soliciting general input from the community as
to whether there are any additional calibration programs that should be carried out with
WFPC2 during this final cycle, in order to improve or augment our current calibration
accuracies or explore new types of calibration. Here we describe the normal WFPC2 calibration plan along with the special calibration programs that are currently underway, and
some ideas for other possible programs that may help to maximize the archival legacy of
WFPC2.
2.
Overview of WFPC2 Calibration Strategies
The general philosophy of WFPC2 calibration is divided broadly into three areas:
• Basic activities aimed at maintaining the general health and safety of the instrument.
Examples of this include the regular “decontamination” procedures (DECONs), along
with associated photometric observations and internal measurements to ensure that
the instrument continues to function as expected.
291
292
Koekemoer, et al.
• Routine calibration monitoring programs, which are carried out with sufficient frequency to allow the calibration accuracy to be maintained for instrument characteristics that are time-dependent. Examples of these include flatfields, bias and dark
frames, and additional photometric monitors, supplementing those obtained during
the regular DECONs.
• Special calibration programs, aimed at characterizing anomalous behavior or improving our knowledge of some aspect of the instrument. Examples of these have included
programs aimed at characterizing the Charge Transfer Efficiency (CTE) problem, or
astrometric measurements of the camera distortion, or photometric cross-calibration
with other instruments and filter systems.
At the start of each HST observing cycle, the WFPC2 group has assembled a calibration
plan outlining the various programs to be carried out for that cycle, along with a budget of
how many orbits would be required. The orbits are divided into “external” orbits (observations of real astronomical targets) along with “internal” exposures (such as darks, biases
and flats using the internal lamps) that may be obtained during occultation or in parallel
with another instrument observing as the prime instrument. In addition, after the completion of each cycle the WFPC2 group has issued a Calibration Close-Out report, describing
the principal results from the programs during that cycle. In Table 1 we present a summary
of all the WFPC2 calibration plans and close-out reports that have been published to date
by the WFPC2 group. During the first few on-orbit cycles for WFPC2, the total number of external orbits allocated to these programs generally ranged between about 70 and
90 for each cycle, while the number of internal exposures typically ranged between about
2000 and 3000 per cycle. The decreased demand for WFPC2 during recent cycles has led
to a decrease in the allocation of external orbits (61 and 40 orbits for Cycles 10 and 11
respectively), while the number of internal exposures remains at ∼2000 per cycle.
Table 1.
WFPC2 Calibration Plans and Close-Out Reports, as of October 2002
ISR Number
ISR WFPC2-96-08
ISR WFPC2-97-06
ISR WFPC2-99-02
ISR WFPC2-00-01
ISR WFPC2-01-03
ISR WFPC2-02-05
Date
Jul 23, 1996
Aug 18, 1997
May 19, 1999
May 25, 2000
May 15, 2001
Aug 15, 2002
WFPC2
WFPC2
WFPC2
WFPC2
WFPC2
WFPC2
Cycle
Cycle
Cycle
Cycle
Cycle
Cycle
Title
6 Calibration Plan
7 Calibration Plan
8 Calibration Plan
9 Calibration Plan
10 Calibration Plan
11 Calibration Plan
Authors
Casertano et al.
Casertano et al.
Baggett et al.
Baggett et al.
Baggett et al.
Gonzaga et al.
ISR
ISR
ISR
ISR
ISR
Dec 22, 1995
Feb 3, 1997
Apr 21, 1998
Dec 23, 1999
Jun 11, 2001
WFPC2
WFPC2
WFPC2
WFPC2
WFPC2
Cycle
Cycle
Cycle
Cycle
Cycle
4
5
6
7
8
Baggett et al.
Casertano et al.
Baggett et al.
Baggett et al.
Baggett et al.
WFPC2-95-07
WFPC2-97-02
WFPC2-98-01
WFPC2-99-05
WFPC2-01-06
Calibration Summary
Calibration Closure Report
Calibration Closure Report
Closure Report
Closure Report
For Cycle 12, we plan to continue the routine monitoring programs for WFPC2, with a
similar orbit allocation to Cycle 11. The plan for Cycle 12 will be finalized in Spring 2003,
therefore we are now soliciting from the community ideas for “special” calibration programs
that should be included, which can be either external (most likely limited to a few orbits),
or internal.
3.
Routine WFPC2 Calibration Monitoring Programs
The routine calibration programs for WFPC2 can be divided into DECONs (and the observations directly associated with them), and longer-term monitoring programs aimed at
extending or supplementing the observations obtained during the DECONs. Here we describe both classes of programs, currently executing for Cycle 11 and planned for Cycle 12.
WFPC2 Calibration and Close-Out
3.1.
293
Decontaminations and Related Observations
The WFPC2 is continually subject to the deposition of contaminants on the cold CCD
windows (−88◦ C) inside the camera, and the absorption from these contaminants significantly reduces the throughput of the instrument at Far-UV and Near-UV wavelengths. The
contaminants deposit gradually, typically producing ∼ 0.5–1% loss in throughput per day
in the F170W filter. Throughput losses of up ∼ 30% can be tolerated, thus throughout the
on-orbit life of WFPC2 we have scheduled “decontamination” visits (DECONs) approximately once per month, where the camera heads are heated to +22◦ C, usually for a 6 hour
period. This has been shown to completely evaporate the contaminants, after which they
start to deposit once more as soon as the instrument is cooled down to −88◦ C. Thus far no
permanent contamination has ever been observed in the instrument. In Cycle 11 the time
interval between DECONs was increased to 49 days, since the contamination rate has been
shown to have decreased considerably over recent years (McMaster & Whitmore 2003, this
volume).
In addition, the temperature increase during the DECONs serves to “anneal” most of
the hot pixels that form on a daily basis as a result of radiation damage (usually several
tens of pixels/day for each CCD). Therefore, the calibration programs associated with the
DECON visits include not only external photometric monitoring observations, but also
internal darks, biases and INTFLATs, to verify basic instrument performance. Finally,
these visits contain observations of the Kelsall spots (KSPOTS) in the WFPC2 pyramid,
that can be used to obtain valuable information on long-term movements of the four cameras
with respect to one another. In Table 2 we show all the observations that are associated
with each DECON visit, as described more fully in Cycle 11 calibration programs 9589 and
9590.
Table 2.
WFPC2 DECON Visits and Observations, executed once every 49 days
Type of Exposure
Filter
(Pre-and Post-DECON Observations)
GRW+70D5824 (WD star) F170W
GRW+70D5824
F160BW, F218W, F225W,
F300W, F336W, F439W,
F555W, F814W
DARK
—
BIAS
—
INTFLAT
F555W
KSPOTS
F555W
(Additional Post-DECON Observations)
INTFLAT
F336,F439,F555W,
F675W, F814W
3.2.
Notes
all 4 CCDs
Rotates among the
4 CCDs with each
DECON visit
GAIN
GAIN
GAIN
GAIN
=
=
=
=
7
7, 15
7, 15
15
GAIN = 7
Other Routine Monitoring Programs
Although the observations associated with the DECON visits provide sufficient information
to verify the basic operational functionality of the instrument, they are not performed with
sufficient frequency to allow us to base our calibration on these data alone, nor do they cover
the entire range of capabilities of the instrument. Instead, routine monitoring of the full
suite of WFPC2 capabilities is provided by two classes of additional programs: (1) relatively
294
Koekemoer, et al.
frequent internal observations, and (2) full sweeps of the entire WFPC2 filter set, performed
once during each cycle. Here we summarize these programs.
Daily and Weekly Internal Monitors. The internal monitors consist primarily of weekly
programs that provide 30 minute DARK frames (five with CLOCKS=NO and one with
CLOCKS=YES), along with INTFLATS and BIAS frames taken at both gain settings
(Cycle 11 programs 9592 and 9596). The darks with CLOCKS=NO are used to create the
weekly darks, while those with CLOCKS=YES are simply obtained as a service to those
observers who wish to use the less-supported mode of leaving the serial clocks on during the
exposure. The weekly darks are in turn used to create the superdark that is the basis for the
dark reference file for each DECON cycle that is used in the calibration pipeline, while the
biases are used to create the superbias on an approximately annual basis. The INTFLATs
are used to monitor the gain stability of the instrument. In addition, we obtain up to 3
shorter exposure darks (1000s each) on a daily basis, as a service to GOs who may wish to
create their own darks from data that may be closer to the time of their observations than
the standard weekly darks. During Cycle 11 these supplemental daily darks are obtained
in programs 9593, 9594 and 9595. In Table 3 we summarize all these exposures.
Table 3.
WFPC2 Daily and Weekly Internal Monitors
Type of Exposure
DARK
DARK
BIAS
NTFLAT
Frequency
3 times/day
5 times/week
4 times/week
2 times/week
Notes
GAIN
GAIN
GAIN
GAIN
=
=
=
=
7;
7;
7,
7,
exptime=1000s
exptime=1800s
15
15
In addition to the internal monitors, we regularly obtain exposures of the bright earth
(EARTHFLATs) when WFPC2 is not observing as prime instrument. These are taken
continually throughout the year in a range of narrow-band optical and UV filters (Cycle 11
programs 9598 and 9599 respectively), and are aimed at monitoring long-term changes in
the flatfields. The optical EARTHFLATs consist of 200 exposures in each of four narrowband filters (F375N, F502N, F656N, F953N), and 50 exposures in each of 10 other
filters (F160BW, F336W, F343N, F390N, F437N, F469N, F487N, F631N, F658N, F673N).
The UV program contains 100 exposures in each of 6 filters (F170W, F185W, F218W,
F255W, F300W, and F336W) along with 20 exposures in each of 4 crossed filter sets
(F170WxF606W, F218WxF450W, F300WxF814W, F336xF814W) in order to assess and
remove the redleak contribution. These observations have been used to update the pipeline
flatfield reference files, and also allow the possibility of substantial improvements in the
pixel-to-pixel flatfields (Koekemoer, Biretta & Mack 2002).
Annual Routine Monitoring Programs. A number of monitoring calibration programs are
carried out on a less regular basis, either because they use limited resources (external
orbits, or usage of the VISFLAT lamps which are decaying with time), or because they
track changes that are relatively slow.
The annual photometric filter sweep (Cycle 11 program 9590) contains exposures of
the WFPC2 standard star GRW+70D5824 at GAIN=15 in the filters F675W, F450W,
F467M, F606W, F791W, F850LP, and F1042M, in all four CCDs, with exposure times
ranging between 2s and 40s (except for F1042M which has 2x300s). Other filter sweeps
include the UVFLAT sweeps in the filters F160BW, F185W, F122M, F170W, and F336W
(generally with exposure times ranging from 400 to 1000s, except for F336W which is 30s),
and VISFLAT sweeps in the filters F439W, F555W, F675W, F814W, and FR533N. The
WFPC2 Calibration and Close-Out
295
UVFLAT sweeps are predominantly aimed at characterizing the long-term evolution of the
filters, for example the F160BW which is known to have been developing pinholes over time.
The other annual filter-related program involves a complete sweep of all the standard
optical and UV filters through the INTFLAT lamps (Cycle 11 Program 9597). These
provide long-term monitoring of the pixel-to-pixel response of the flatfields, and also provide
a backup database in the event that the VISFLAT lamp can no longer be used. This program
also contains a linearity check of the CCDs, consisting of a series of exposures in F555W
covering a range of exposure times (6–18 s at GAIN=7, and 8–36 s at GAIN=15), and using
both Blade A and Blade B for the flat. In addition to the linearity check, these allow for
long-term monitoring of the shutter behavior.
This program also contains a set of EARTHFLAT observations through the linear ramp
filters FR418NxF437N, FR533P33xF520N, FR680NxF631N, and FR868NxF953N, as well
as VISFLAT observations of the same set of ramps uncrossed with any other filters, along
with VISFLAT exposures through additional filters (FQUVN and FQCH4N). In addition
to providing long-term monitoring of the calibration of these filters, these exposures allow
checks of the repeatability of the filter wheel positioning mechanism.
4.
Examples of Some Special WFPC2 Calibration Programs
In addition to the routine monitoring programs related to instrument health and safety, we
also have a variety of “special” calibration programs during each cycle, which are generally
aimed at better characterizing some specific aspect of the instrument. These may either be
carried over from one cycle to the next, or otherwise need to be done in only one or two
cycles. Examples of these programs have included characterization of CTE, improved photometric zeropoints, and astrometric characterization. In Table 4 we summarize the special
programs from recent cycles along with the current cycle, after which we briefly describe
some of the current programs. Further details of all the programs and their products are
available from the WFPC2 web site: http://www.stsci.edu/instruments/wfpc2/.
Table 4.
WFPC2 Special Calibration Programs since Cycle 7
Program Title
Photometric Characterization
PSF Characterization
CTE Characterization
Astrometric Monitor
Polarization
Noiseless Preflash
Photometry of Very Red Stars
CTE for Extended Sources
Plate Scale Verification
Wavelength Stability of LRFs and Narrow-band filters
Redleak Check
Astrometric Effects of CTE
Clocks ON Verification
Methane Quad Filter Check
WFPC2-ACS Photometric Cross-Calibration
Cycle
7,8,9,10,11
7,8,9,10,11
7,8,9,10,11
7,8,9,10,11
8
8
8
8
8
8, 9
9
10
10
10
11
Program IDs
7628, 8451, 8818, 9251, 9601
7629, 8452, 8819, 9257, 9600
7630, 8447, 8821, 9254, 9591
7627, 8446, 8813, 9253, 9600
8453
8450
8455
8456
8458
8454, 8820
8814
9255
9252
9256
9601
We next describe a selection of three of the more recent special calibration programs,
which we present here primarily to serve as examples of the types of studies that these types
of programs typically consist of. Space limitations do not permit us to include detailed
descriptions of all the programs, but all this information, along with the results derived
from these programs, are available at the aforementioned WFPC2 web site.
296
4.1.
Koekemoer, et al.
CTE Characterization
The principal aims of the CTE proposals for WFPC2 have been to characterize the effects
of CTE on photometry and astrometry of point sources, along with some more recent
proposals aimed at investigating CTE effects for extended sources. Not only do the data
allow the effect to be described as a function of flux and location on the chip, but they also
track its evolution over time. The first results from Cycles 5 and 6 (programs 6192 and
6937 respectively) also led to the identification of the “long-vs.-short” problem, an apparent
non-linearity of the photometric calibration.
The observational approach has generally consisted of observing a relatively rich starfield
(Omega Centauri) in a number of broad-band photometric filters, for a range of exposure
times and also using a range of preflash levels. The details have varied from one cycle to the
next, depending upon available orbit allocation as well as the specific tests to be performed
during each cycle. An example is the Cycle 7 program 7630, which contained observations
in F814W with exposure times of 10s, 40s, 100s, 300s, and 1000s, and preflash levels of 0,
5, 10, 100, and 1000 electrons for each exposure time. Smaller subsets of exposure/preflash
combinations were obtained in the filters F555W and F300W, in order to provide photometric checks for the other filters. In subsequent programs the preflash tests were reduced,
while serving to continue the essential monitoring observations. These programs have thus
far led to a number of detailed discussions on how the photometric effects of CTE may
be quantified (e.g., Whitmore, Heyer & Casertano 1999; Dolphin 2000, 2002; Whitmore &
Heyer 2002 and references therein).
Although the photometric effects of CTE have now been well characterized, less is
known about its effects on astrometry. A study by Riess (2000) showed that extended
sources suffer some degree of distortion due to CTE, indicating that the astrometry of
sources must also be affected. For example, the relative separation of a faint source from
a bright source may depend on all the factors that influence CTE (position on detector,
observing epoch, brightness in electrons, and image background). Therefore, some of the
more recent CTE proposals have attempted to quantify the astrometric effects of CTE
by measuring: (1) the relative separation of a bright source vs. a faint target at different
positions on the PC1 CCD, and (2) the relative motion of a source on the CCD compared
to very precise slews performed with the FGSs. These tests are conducted for point and
extended targets at several different intensity levels.
For these astrometric tests, the target is observed with the PC chip in a 2 × 2 grid,
20 on a side, repeated over two orbits (with the second orbit being slightly offset by
half a PC pixel). The 2-orbit sequence is done at different background light levels, using
exposures from 100s in F450W to 1200s in F622W, repeated three times for two targets
(total 12 orbits). The targets include the dense star field in Omega Centauri, and an
extragalactic field of faint galaxies. Both fields are chosen to have a bright star surrounded
by fainter objects. While most of the test is performed on the PC, the WFC CCDs are
also important, as they can provide a sanity check on the motions made with the FGSs.
The motions on the WFC CCDs is a smaller number of pixels, hence less subject to CTE
variations.
The most recent CTE proposal (Cycle 11, program 9591) aims to do an additional check
of CTE characterization, by carrying out the observations in the 2×2 on-chip binning mode.
This mode has been seldom used on WFPC2 due to the relatively large size of the WFC
pixels (0.1). However, since the level of CTE depends upon the number of pixels that are
being read out, it is possible that the severity of the effect can be reduced by using on-chip
binning. This test is primarily intended with newer instruments in mind (ACS, WFC3)
since their smaller native pixel size (0.05) may make on-chip binning a more appealing
option if it is indeed found to substantially mitigate CTE.
WFPC2 Calibration and Close-Out
4.2.
297
WFPC2 Astrometric Characterization
This program has been executing twice per year until Cycle 10, after which it has been
executing once per year. It consists of observing a rich star field (Omega Centauri) with a
pattern of large shifts designed to move the same set of stars onto each of the four detectors,
as well as small shifts aimed at providing sub-pixel dithering to improve the sampling of the
PSF. The principal part of the program has involved observations at F555W, which allows
both the determination of a geometric solution as well as long-term monitoring of possible
changes. This has been supplemented in some cycles by observations in two other filters,
F300W and F814W, which are aimed at allowing the measurement of the color dependence
of the solution. Observations are also obtained at a range of position angles, which can be
used to constrain additional geometric parameters such as skew. Results from this program
have been published in the form of an ISR by Casertano & Wiggs (2001), and by King &
Anderson (2003, this volume), as well as Kozhurina-Platais et al. (2003, this volume).
4.3.
WFPC2-ACS Photometric Cross-Calibration
This proposal is aimed at providing photometric zeropoint cross-calibration between the
commonly used WFPC2 photometric filter sets and those that will be used for ACS programs. The proposal includes observations of globular clusters covering two extremes of
metallicity: the metal-rich cluster 47 Tucanae and the metal-poor cluster NGC 2419. In
addition, the proposal obtains WFPC2 observations of the primary ACS standard star
BD+17D4708. This program will produce a valuable tie-in between the WFPC2, ACS and
Sloan filter photometric systems.
The observations of 47 Tucanae are all at the same position and orientation as the
observations in an earlier related proposal (8267), with the following filters: F300W, F336W,
F410M, F439W, F450W, F467M, F547M, F569W, F606W, F675W, F702W, F850LP (note
that extensive observations already exist for this object in the F555W and F814W filters).
Similarly, the observations for NGC 2419 are at the same position and orientation as those
in earlier related proposals 5481 and 7628, with these filters: F300W, F336W, F410M,
F439W, F450W, F467M, F547M, F569W, F606W, F675W, F702W, F850LP (again, this
object also has extensive F555W and F814W observations, therefore they do not need to
be repeated here).
Finally, the observations of the Sloan primary standard star BD+17D4708 are obtained
using the PC and WF3 chips, in all of the following filters: F300W, F336W, F410M, F439W,
F450W, F467M, F547M, F555W, F569W, F606W, F675W, F702W, F814W, F850LP.
It is expected that this program will provide ∼ 1–2% zeropoint accuracy for baseline
observations using most commonly-used filters (e.g., F439W, F555W, F675W, F814W), and
will enable a direct photometric tie-in not only to the ACS but also to the ground-based
SDSS system.
5.
Possible Ideas for WFPC2 Close-Out Calibration Programs in Cycle 12
As was mentioned earlier, Cycle 12 is currently planned to be the final full cycle for WFPC2
operations before it is de-orbited. Therefore, it is imperative that the final “close-out” round
of calibration programs address critical issues that may help to improve the quality of the
science or otherwise aid in improving our understanding of the behavior of the instrument.
Community input to this process is essential , and we actively solicit any ideas that observers
in the general community may have.
5.1.
Our “Top 10” Current Ideas for Close-Out Programs
In this section we summarize some of our current ideas for special close-out calibration
programs during Cycle 12. If we receive sufficient community input then we may modify
298
Koekemoer, et al.
these or supplement them with additional programs. Our general philosophy is to improve
the calibration accuracy of some filter modes that may provide enhanced scientific results,
as well as carrying out observations aimed at improving our knowledge of specific aspects
of the instrument (e.g, PSF behavior). We may also make use of this opportunity to carry
out programs that may have been difficult to do earlier in the life of the instrument (for
example, characterizing the nature of all the “permanent” hot pixels that are not easily
removed by the regular DECONs), if such programs may be helpful for some of the newer
instruments.
It is reasonably likely that we will carry out the first three or four programs in this
list, depending somewhat upon the reactions that we receive from the community. We will
attempt to carry out as many of the other programs further down the list as our orbital
allocation will allow, although we may move some of these up in priority if there is sufficient
interest from the community in a particular program. Finally, this list is not intended to
be complete and we would be happy to consider adding additional programs that might be
suggested by the community, if there is sufficient interest.
1. Fully characterizing CTE behavior for extended sources. While the effects of CTE on
point sources are now relatively well understood for a wide range of count-rates and
background levels, there are as yet no well-developed methods for calculating the effect
of CTE on extended targets. This work would primarily consist of two parts, analytical
and observational. The analysis effort would involve developing algorithms to perform
iterative calculation of CTE to determine the true underlying flux distribution for an
arbitrary source in an image. The observational component may consist of additional
exposures to our Cycle 12 CTE calibration program to obtain data that may be used
in testing the algorithms, although it is possible that such tests may be carried out
using data already in the archives.
2. Creating broad-band skyflats using data from external science exposures. While observations of the bright earth are useful in creating flatfields for narrow-band filters, the
optical broad-band filters generally saturate too quickly to enable useful data to be obtained. Therefore, on-orbit flats for the broad-band filters have to date been obtained
using the internal lamps, but unfortunately these contain their own field-dependent
variations that are not present in external exposures. This proposal would aim to
analyze large numbers of archival external WFPC2 exposures in order to construct
sky flats directly from the background light levels in the exposures.
3. Improving the geometric distortion characterization. Currently the WFPC2 distortion has been explicitly obtained for only three filters, namely F300W, F555W and
F814W. Since the distortion is strongly color-dependent, it may be desirable to obtain
observations at a wider range of filters, perhaps at even shorter wavelengths, in order
to improve the constraints. Further work also needs to be done on determining the
degree of skew in the solutions, which requires analysis of observations at a range of
different orientation angles.
4. Reducing the errors in photometric zero points between the WFPC2 filters. Currently
the photometric zero points between filters are known to levels of ∼ 2%. These may
be reduced to less than 1% by examining measurements of larger numbers of stars
in selected regions, as well as by improving the color terms and comparison with
ground-based photometric systems.
5. Improving the calibration of narrow-band and linear ramp filters. Many of the narrowband filters are calibrated to accuracies of only ∼ 5%, while the linear ramp filters
can be off by as much as 10%. These levels of calibration may be improved by a
program of observations aimed at bright emission-line photometric standards, such as
WFPC2 Calibration and Close-Out
299
planetary nebulae or extragalactic objects (QSOs, AGN) that have been well-studied
spectroscopically with other instruments, for example STIS. The flatfields for these
filters could also be improved using wealth of on-orbit data currently in the archives.
6. Improved characterization of the efficiency of some filters, for example the z-band,
F785LP and 1042M. The red broad-band filters suffer from unique limitations in
that the objects of greatest interest for these filters are often extremely red, thus
the color terms involved are much larger than for many of the other filters and the
potential photometric errors are thus much higher. The calibration of these filters can
potentially be significantly improved by programs aimed at observations of samples
of red spectrophotometric standards.
7. Improved measurements of filter red leaks, including characterization of their spatial
dependence. A number of the broad-band filters (F336W and blueward) are known
to have significant red leaks, which are characterized only by their ground-based filter
throughput traces. It is possible that these effects may vary spatially across the
filters, which can be measured using observations of photometric standards at a range
of locations across the chips. Similarly, the time evolution of the red leaks could
be better characterized using the database of photometric calibration observations
currently in the archive.
8. Measuring possible changes in the central wavelengths of some of the narrow-band
filters. Ground-based tests reveal the possibility of changes in the photometric properties of the narrow-band filters due to the large numbers of coatings on these filters.
Although they were extensively baked out prior to launch, some observations have
suggested the possibility of subsequent changes on-orbit. This program would involve
observations designed at measuring such changes if they have occurred.
9. Measuring the extended wings of the point-spread function on large scales across the
chips. Some programs involve observations of faint targets in the vicinity of bright
objects, for example objects near bright stars, or planetary moons in our own solar
system. It is often desirable to model the PSF of the bright sources on large scales, to
better constrain the properties of the faint targets of interest. This calibration program
would aim at complementing the current suite of PSF library data by extending PSF
information to scales of several hundred pixels. Care would need to be taken to
properly account for the effects of scattered light and telescope focus variations.
10. Characterizing the photometric effects of intra-pixel variations resulting from focus
changes due to telescope breathing. During an orbit the thermal “breathing” of the
telescope can amount to a focus variation of several microns. This changes the amount
by which a star’s light is distributed among the pixels, and can potentially introduce
time-dependent photometric variations if several exposures are obtained during the
orbit. This program would likely involve analysis of a large number of suitable exposures from the archive to determine the magnitude of this effect, although new
additional observations may be proposed if the current data are not adequate.
5.2.
How the Observing Community Can Help
If anyone in the community would like to give us input for possible close-out programs to
consider, please simply send email to “help@stsci.edu” with the subject of “WFPC2 CloseOut Calibration.” We will gladly examine all the suggestions that we receive, and would
be happy to discuss possible programs with interested observers.
In general, any proposals aimed at obtaining new observations will be performed by
the STScI WFPC2 group as part of our calibration program for Cycle 12. Thus, feedback
that we receive on observational aspects of close-out will be included in these programs.
300
Koekemoer, et al.
If observers in the community would like to carry out calibration-related analysis on
any archival WFPC2 data, then funding for such work can be obtained by submitting a
“Calibration Outsourcing Proposal” as part of the general Phase I HST call for proposals.
In this case, the WFPC2 group plays a consulting role but the actual analysis is done
directly by the proposer, who then deliver the products back to STScI for inclusion in our
calibration database (for example, improved zeropoints, or a refined geometric distortion
solution, or software that deals with CTE).
Acknowledgments. We are pleased to acknowledge all those in the community who
have so far given us very valuable feedback, including Abi Saha, Ivan King, Jay Anderson,
Andrew Dolphin, Erich Karkoschka, Dave Zurek. We also thank the many people at STScI,
especially Stefano Casertano, Sylvia Baggett, Shireen Gonzaga, Ron Gilliland, and Brad
Whitmore, who have all contributed to the success of the WFPC2 calibration programs.
References
Baggett, et al., 1999, Instrument Science Report WFPC2-99-05 (Baltimore: STScI)
Baggett, S., Casertano, S., & Biretta, J., 1995, Instrument Science Report WFPC2-95-07
(Baltimore: STScI)
Baggett, S., Casertano, S., Biretta, J., Gonzaga, S., & the WFPC2 Group, 1999, Instrument
Science Report WFPC2-99-02 (Baltimore: STScI)
Baggett, S., Casertano, S., & the WFPC2 group, 1998, Instrument Science Report WFPC298-01 (Baltimore: STScI)
Baggett, et al., 2001, Instrument Science Report WFPC2-01-06 (Baltimore: STScI)
Baggett, S., Gonzaga, S., Biretta, J., Casertano, S., Heyer, I., Koekemoer, A., McMaster,
M., O’Dea, C., Riess, A., Schultz, A., Whitmore, B., & Wiggs, M. S. 2000, Instrument
Science Report WFPC2-00-01 (Baltimore: STScI)
Baggett, S., Biretta, J., Heyer, I., Koekemoer, A., Mack, J., McMaster, M., & Schultz, A.
2001, Instrument Science Report WFPC2-01-03 (Baltimore: STScI)
Casertano, S., & Baggett, S., 1997, Instrument Science Report WFPC2-97-02 (Baltimore:
STScI)
Casertano, et al., 1996, Instrument Science Report WFPC2-96-08 (Baltimore: STScI)
Casertano, S. & the WFPC2 Group, 1997, Instrument Science Report WFPC2-97-06 (Baltimore: STScI)
Casertano, S., & Wiggs, M., 2001, Instrument Science Report WFPC2-01-10 (Baltimore:
STScI)
Dolphin, A. E. 2000, PASP 112, 1397
Dolphin, A. E. 2002, astro-ph/0212117
Gonzaga, et al., 2002, Instrument Science Report WFPC2-02-05 (Baltimore: STScI)
King, I. & Anderson, J. 2003, this volume
Koekemoer, A. M., Biretta, J., & Mack, J. 2002, Instrument Science Report WFPC2-02-02
(Baltimore: STScI)
Kozhurina-Platais, V., Casertano, S. & Koekemoer, A. M. 2003, this volume
McMaster, M. & Whitmore, B., 2003, this volume
Riess, A., 2000, Instrument Science Report WFPC2-00-04 (Baltimore: STScI)
Whitmore, B., Heyer, I., & Casertano, S. 1999, PASP 111, 1559
Whitmore, B. & Heyer, I. 2002, Instrument Science Report WFPC2-02-03 (Baltimore:
STScI)
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
WFPC2 CTE Characterization
Andrew E. Dolphin
National Optical Astronomy Observatories, Tucson, AZ 85726; dolphin@noao.edu
Abstract. The limiting factor of the accuracy of WFPC2 photometry is the CTE
loss, which has increased to the level of 50% or more for faint stars at the top of the
chips. I describe recent work on characterizing this effect, and provide improved equations for CTE correction. I also examine issues affecting background measurement,
which if not done correctly can introduce artificial nonlinearities into photometry.
1.
Introduction
Several obstacles inhibit the obtaining of accurate photometry from WFPC2 images. The
most severe of these is charge loss during readout, commonly known as CTE loss. While no
CCD is likely to be perfect, the effect is pronounced on WFPC2, where initial measurements
showed that a star at the top of the chip (y = 800) would lose approximately 10% of its
charge while being read out. This loss was reduced by cooling the camera from −76◦ C to
−88◦ C, and Holtzman et al. (1995) found that the CTE loss could be corrected to acceptable
levels with a correction of 0.04 y/800 magnitudes to photometry. Unfortunately, this has
increased over time, and faint stars at the top of the chip now can lose well over half their
light to CTE loss.
Improved characterizations of the charge loss were produced by Stetson (1998), Whitmore, Heyer, & Casertano (1999), and Saha, Labhardt, & Prosser (2000) by analysis of
larger data sets using DAOPHOT, IRAF apphot, and DoPHOT, respectively. While there
were several issues of agreement between these studies, there were also significant discrepancies. For example, Stetson (1998) found no significant time dependence, while Whitmore
et al. (1999) did. Likewise, Saha et al. (2000) found no X-CTE loss and no nonlinearity,
while both were observed in the other studies. The fact that a different photometry package was used in each study raised the possibility that the observed CTE loss was partly a
function of the package used.
Another possible source of photometric error was reported from measurements of the
same fields in long and short exposures, in which objects appeared fainter in the short
exposure than in the long exposure. This was independent of the position on the chip,
and the phenomenon has become known as the “long vs. short anomaly.” Casertano &
Mutchler (1998) characterized this effect using short and long observations of NGC 2419,
and found it to be a function of counts rather than exposure time. The effect they found
was a very large one; rather than the 5% effect reported previously, their correction is
0.18 magnitudes at 280 electrons (50 ADU at gain of 7). However, Stetson (1998) solved
for a position-independent effect as part of his CTE study and found none. Given that his
data set included the NGC 2419 observations, this only added to the questions about how
well these effects were really understood.
Dolphin (2000a; hereafter D00) presented a study of the WFPC2 CTE based on reductions of 843 WFPC2 images of ω Cen and NGC 2419 using his HSTphot photometry
package (Dolphin 2000b). In addition to providing yet another CTE solution based on yet
another photometry package, this work shed light on the discrepancies noted above. The
lack of time dependence seen by Stetson was observed to be primarily the result of an in301
302
Dolphin
sufficient time baseline and the fact that all of his high-background data were obtained at
the end of the time baseline. The lack of a nonlinearity in the study of Saha et al. was
found to result from their background measurements, which contained significant amounts
of starlight. This meant that their CTE correction (explicitly a function of background
only) was implicitly a function of brightness as well. Finally, it was demonstrated that the
long-short anomaly is primarily a result of poor background measurements, as my reductions of the NGC 2419 data showed no significant discrepancies after the CTE loss was
corrected.
In this paper, I describe ongoing efforts to improve upon D00 and present the latest
results.
2.
Observations
The basis of my CTE study (both D00 and the present work) has been the comparison of
instrumental magnitudes of WFPC2 images with ground-based photometry. This allows
the use of the ground-based data as the standard stars, and any discrepancy is understood
as resulting from a combination of CTE loss and calibration. This is not the only way to
do such a work; Whitmore et al. (1999) based their study on relative photometry of stars
as they were moved around the chips. Both techniques have their drawbacks; the main
drawback of the route I chose is that errors in the functional form can be harder to find.
Most notably, a position-independent nonlinearity such as the long-short anomaly could be
fit out as CTE loss. I address this concern in the next section.
In D00, I used images of the ω Cen standard field and of NGC 2419. The mixture of two
fields was necessary because, at the time, no images of the ω Cen standard field had high
background levels; however this introduced the possibility of errors in the CTE correction
caused by inconsistent calibrations of the two ground-based data sets. This compromise is
no longer necessary, as high-background images of ω Cen have been taken, and thus in this
work I use only those data. The ground-based photometry is that of Walker (1994), which I
have transformed to the expected WFPC2 flight system magnitudes by using the Holtzman
et al. (1995) transformations.
A total of 1216 images were photometered in this project, covering all observations of
the ω Cen field in B, V, R, I filters through August 2002. The majority of the images (800)
were taken in the F555W and F814W filters. All images were photometered with HSTphot.
Because of the huge number of stars observed, I eliminated all data I considered suspect
because the aperture corrections were unusual or were based on too few stars.
After matching the stars on the WFPC2 images to the list of ground-based standards,
many of the standards were eliminated because they were resolved into multiple stars or
small extended objects by WFPC2. A few other standards were deemed to be poorly
photometered because no CTE correction was capable of producing photometry that agreed
at the 10% level; these were also eliminated from the sample.
The end result was a list of 36983 stars on the WFPC2 images that had been matched
to any of 202 of Walker’s standard stars. From this list, the magnitude differences between
the WFPC2 magnitudes and ground-based magnitudes were fit as functions of CTE loss
and zero point differences. Since this contribution describes CTE results, the zero points
will not be discussed further. They are, however, available from the author’s web site.
3.
Characterization
It is important to bear in mind that the correction formulae below are based on the functional forms that best fit the data, and are not based on a physical understanding of the
charge transfer process. It is likely that a “perfect” correction would be much more complex;
what is presented below is the best fit using a minimum number of parameters.
WFPC2 CTE Characterization
303
Figure 1.
Ratio of counts in short (14 second) image to counts in long (100 second) image, separated by star brightness in the short image. Small dots are measurements of individual stars; the square and diagonal line are the fit to the trend.
The triangles show the data after CTE correction using old correction formulae,
while the horizontal line shows the expected ratio of 0.14. Note that the bottom panel shows a significant overcorrection, indicated by the triangles falling well
above the horizontal line.
A feature of the CTE correction procedure is that it includes the effect of CTE loss on
the CTE loss. That is, as a star reads out and becomes fainter, CTE loss (in magnitudes
per pixel) increases. This leads to a more complex functional form, but gives a correction
that is accurate down to about 60 electrons (< 10 ADU at gain 7) instead of only to about
100 electrons. At much fainter levels (20–30 electrons), it is clear that the CTE loss is less
than what is predicted, since otherwise noise peaks in the background level itself (which
can be thought of as faint stars) would be truncated. The improvement can be seen in
Figures 1 and 2, which show count ratios between observations in short and long images as
a function of y position and counts. It is clear that the corrections above 100 electrons work
well either way, but that the newer correction (Figure 2) is much better for stars between
60 and 100 electrons.
It should also be noted that the ground-based standard magnitudes were not used in
creating Figures 1 and 2; rather they are based entirely on instrumental WFPC2 magnitudes. It is clear that the corrections performed well, dispelling any concern that the use
of absolute photometry comparisons rather than relative photometry comparisons impaired
the solution. Most notably, had a long-short anomaly been present, the y intercept of the
uncorrected photometry (the diagonal line in each panel) would have been lower than 0.14,
while the CTE correction would have overcorrected stars with high y values and undercorrected those with low y values since the long-short anomaly would have been fit by CTE
loss. We can clearly see from the figure that this is not the case. If long-short anomaly
were at the level seen by Casertano & Mutchler (1998), the y intercept would have been at
0.11 in the 100–400 e− panel and at 0.13 in the 400–1000 e− panel. (A comparison with the
bottom panel is unwarranted, as their correction was only valid to 200 electrons.) Thus it
is clear that any long-short error is at least an order of magnitude smaller than what they
reported.
304
Dolphin
Figure 2. Same as Figure 1, but using the new CTE corrections. Note that the
bright stars are still fit well, while the overcorrection for faint stars is eliminated.
4.
CTE Correction Recipe
The procedure for correcting for CTE loss is outlined below. The XCTE correction depends
only on x and the background (in electrons). Note that the background value used should
be the true background at the position of the star, rather than the background measured
nearby a star (which contains some starlight). In images with variable background, this
requires a knowledge of the amount of starlight contained in the background measurement.
bg ≡
1 + background2 − 10
(1)
XCTE = 0.0194e−0.00085bgx/800
(2)
The YCTE loss depends on y, background, brightness (also in electrons), and the date
of the observation.
lbg ≡ 0.5 ln 1 + background2 − 1
(3)
lct ≡ ln(brightness) + 0.921XCTE − 7
(4)
yr ≡ (MJD − 50193)/365.25
c1 ≡ 0.0143 0.729e−0.397lbg + 0.271e−0.0144bg
1 + 0.267 yr − 0.0004 yr2 y/800
(5)
(6)
c2 ≡ 2.99e−0.479lct
(7)
YCTE = ln[(1 + c2 )e − c2 ]/0.441
(8)
c1
Both XCTE and YCTE corrections are in magnitudes, and should be subtracted from
instrumental magnitudes to make the correction.
Figures 3, 4, 5, and 6 show the magnitude differences between the WFPC2 magnitudes
and the ground-based standard magnitudes before and after correction. In all figures, the
top panel shows the uncorrected WFPC2 minus ground-based magnitude and the bottom
panel shows the corrected difference.
WFPC2 CTE Characterization
Figure 3. WFPC2 magnitude errors (observed minus ground-based standard
magnitudes) vs. y, before (top) and after (bottom) corrections were applied. Note
that stars with large y values have fainter raw magnitudes, but no discernible trend
remains in the CTE-corrected magnitudes.
Figure 4. Like Figure 3, plotted vs. x. Note that stars with large x values have
slightly fainter raw magnitudes, but no discernible trend remains in the CTEcorrected magnitudes.
305
306
Dolphin
Figure 5. Like Figure 3, plotted vs. background. Note that stars on low background have fainter raw magnitudes, but no discernible trend remains in the CTEcorrected magnitudes.
Figure 6. Like Figure 3, plotted vs. brightness. Note that faint stars have fainter
raw magnitudes, but no discernible trend remains in the CTE-corrected magnitudes.
WFPC2 CTE Characterization
Figure 7.
5.
307
Histogram of PC data values from a short (14 second) image.
Background Measurement
Although this work focuses on CTE corrections, another critical issue in obtaining accurate
photometry is background measurement. If the background is mismeasured, one will introduce a nonlinearity into the photometry. In fact, it is likely that the Casertano & Mutchler
(1998) study of the long vs. short anomaly was influenced by background determination.
Two pieces of evidence point to this. First, their residuals are better fit by such a function
than by their reported correction formula. Second, their discrepancies were larger when using larger photometry apertures. In addition, Hill et al. (1998) noted that the short vs. long
error they measured could be explained by a 2 e− per pixel loss only in pixels containing
stars, which is the same as a 2 e− per pixel overestimation of the sky.
The reason special care must be taken in calculating background levels is the very
low background levels of WFPC2 observations. The commonly-used IRAF packages were
designed for reducing ground-based data, for which the histogram of background pixel levels
can be approximated as a Gaussian or other smooth function. However, when the noise in
the background is less than 1 ADU, the histogram is dominated by digitization. A sample
histogram of sky values from a 14 second exposure is shown in Figure 7.
Naturally any sky-measuring algorithm that assumes a smooth distribution is prone to
failure when handling these data. Figure 8 shows measurements made by various IRAF sky
algorithms on this image, with most of the default parameters left in place. This analysis
is similar to the examination done by Ferguson (1996) on simulated data. In all panels,
the x values of the sky are taken from HSTphot measurements, which use a sigma-clipped
mean algorithm that has been verified to work in cases like these.
From a comparison of the plots, it is clear that not all of the algorithms function as
intended when facing digitization-dominated sky histograms. The Gaussian and median
routines appear to be affected by the histogram peak at ∼ 0.6 ADU, the cross-correlation
algorithm is extremely unstable, and the optimal filtering algorithm is biased. The “mode”
calculation is not a true mode, but rather a calculation including the median routine and is
thus not suitable. Additional tests have been made involving simulated images with known
background levels, and these routines handle the simulated data no better than the real
data.
In more detailed tests, it is observed that the centroid algorithm is also prone to small
biases, while a straight mean will be overly affected by deviant pixels (uncorrected bad
pixels or faint stars). The errors in the centroid algorithm affect measurements at only the
0.03 ADU level, however, which is acceptable for most purposes. Likewise, the stability of
the mean calculation can be increased by use of sigma clipping. Thus the recommendation
is that either routine can be used successfully, but that a sigma-clipped mean is preferable.
308
Dolphin
Figure 8.
6.
Sky determinations from various IRAF algorithms.
Ongoing Work
While the current CTE corrections work very well down to extremely faint count levels,
there are two key problems yet to be addressed in a comprehensive manner. First is the
issue of extremely faint stars. An examination of the CTE correction equations shows that
the expected CTE loss becomes very large for faint stars. Specifically, the noise peaks in the
background itself should be destroyed by CTE loss, as those peaks should be treated as 1 or
2 ADU stars. The histogram in Figure 7 indicates that CTE does affect the background
but not at the level predicted by the CTE correction equations. The dilemma is that
it is prohibitively difficult to obtain accurate photometry of stars of brightness 4 ADU
(28 electrons at gain of 7). However, recent calibration data have been obtained that
should allow this issue to be addressed with repeated short exposures of NGC 2419 that
can be coadded to improve the photometry.
A second problem affecting photometry is that the star’s profile is not dimmed uniformly. In a fractional sense, the highest charge loss comes from the bottom edge and sides
of the star, while the top edge can actually have charge added. Furthermore, the ratio of
dimming of the wings to the dimming of the peak will be a function of the star brightness.
An example of this is shown in Figure 9, which shows the difference between a short and
long image, scaled so that they should cancel out. The trail above the star in the image is
the light that was lost by the star and released from the trap several readout steps later.
The result of this is that PSF-fitting photometry will have some trouble dealing with
faint stars, which will have different PSFs from the bright stars. Furthermore, the possibility
exists for small systematic differences between PSF-fitting and aperture photometry. While
both errors are dwarfed by the random error in measuring faint stars, there are some
applications for which they are significant. More significantly, the effects of CTE loss on
the profiles of extended objects can be quite large. Riess (2000) created a simple model for
trapping and recreate the CTE effects on the profiles of extended sources. The challenge will
WFPC2 CTE Characterization
309
Figure 9.
Smoothed difference image (short − scaled long image), showing
charge loss (dark) at star’s position and trail (bright) above the star. The brightness of the trail is ∼ 0.1 counts.
be to improve such a model so that it can quantitatively reproduce the CTE loss measured
in stellar photometry. Given the complexity of the correction equations, it is clear that this
is not a trivial task. However, once this is achieved, one can invert the charge loss model to
obtain the true image prior to readout, effectively correcting the image for CTE loss rather
than the photometry.
7.
Summary
I have developed and described a CTE correction procedure to supersede that of D00. There
are several significant improvements over the previous corrections. First, a single uniform
set of ground-based photometry is used to provide the comparisons. Second, HSTphot
has been improved several times over the intervening time. Third, more stringent cuts
were applied to the data to eliminate bad points. Finally, an improved functional form of
the CTE correction accounts for the changing CTE loss as a star becomes fainter during
readout.
I have also examined a commonly-overlooked aspect of stellar photometry, background
measurement. In short exposure images, the background value is sufficiently low that the
digitization effects dominate the histogram. If not handled correctly, this can result in
significant artificial nonlinearities added to the data. After exploring the options offered by
IRAF, the recommendation is that one use the “mean” algorithm with sigma clipping.
Finally, I conclude by describing ongoing efforts to improve the CTE corrections for
faint sources and to obtain an image-based CTE correction. My WFPC2 calibration web
site (http://www.noao.edu/staff/dolphin/wfpc2 calib/) will be kept updated with improvements to the CTE corrections and photometric zero points as they become available.
310
Dolphin
References
Casertano, S. & Mutchler, M. 1998, Instrument Science Report WFPC2 98-02 (Baltimore:
STScI)
Dolphin, A. E. 2000a, PASP, 112, 1397
Dolphin, A. E. 2000b, PASP, 112, 1383
Ferguson, H. C. 1996, Instrument Science Report WFPC2 96-03 (Baltimore: STScI)
Hill, R. J., et al. 1998, ApJ, 496, 648
Holtzman, J. A., et al. 1995, PASP, 107, 156
Riess, A. 2000, Instrument Science Report WFPC2 00-04 (Baltimore: STScI)
Saha, A., Labhardt, L., & Prosser, C. 2000, PASP, 112, 163
Stetson, P. B. 1998, PASP, 110, 1448
Walker, A. R. 1994, PASP, 106, 828
Whitmore, B., Heyer, I., & Casertano, S. 1999, PASP, 111, 1559
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
An Improved Distortion Solution for WFPC2
Ivan R. King
Astronomy Department, Box 351580, University of Washington, Seattle, WA
98195-1580
Jay Anderson1
Astronomy Department, University of California, Berkeley, CA 94720-3411
Abstract. This is a brief account of work that is published in detail elsewhere. We
have derived a greatly improved set of distortion corrections for the individual chips
of WFPC2. We also track the relative positions of the chips with time. We end with
a description of interactions between distortion and scale that we do not understand.
1.
Introduction
Most of this discussion will describe our recent redetermination of the geometric distortion
corrections needed for WFPC2 images. We begin, however, with the motivation for this
study.
Astrometry has two parts. One is the measurement of good positions that are free of
systematic measuring errors; the other is the combination of positions measured in different
images. The first, the measurement of positions, we discussed two years ago (Anderson
& King 2000). The essence of the methods described there is to use as many stars as
possible to derive an extremely accurate PSF. We iterate between improving the individual
positions from which the PSF is created, so as to fit them together correctly, and improving
the PSF, so as to get a better set of positions next time round. The demon to be exorcised is
pixel-phase error, i.e., a systematic position error that depends on how each star is centered
with respect to pixel boundaries. That is the basic purpose that our accurate PSF-building
accomplishes.
The other part, the combination of positions measured on different images, usually in
different dither positions and sometimes in different orientations, is much more complicated.
It always requires a transformation from the coordinate system of one image to that of
another image, and here is where distortion gets in the way.
The problem is that in order to derive the transformation from one image frame to
another, one has to use the positions of a number of stars in each image, to derive a linear
transformation between them. But if the distortion has not been totally removed, the
true relationship will not be linear, because when the same star falls in different places in
two images, these positions suffer different distortions. The non-linearities of course grow
with separation in the image, so that what we are forced to do is to derive a separate
transformation for each individual star, from the positions of other stars in its immediate
neighborhood. But the larger the distortions that remain, the smaller is the set of neighbors
that we can use, and the accuracy of the transformation suffers. Ideally, we would like
to remove all the distortion, so as to be able to use a single global transformation over
1
Present address, Department of Physics & Astronomy, MS 108, Rice University, 6100 Main Street, Houston,
TX 77005
311
312
King & Anderson
the whole image. Unfortunately that still is not possible, but minimizing the remaining
distortion allows us to use more surrounding reference stars and therefore make better
transformations.
That is our own interest in improving distortion corrections. But we should also note
that it is only in the rich globular-cluster fields of our projects that one can use the localtransformation work-around; in sparse fields, astrometry is completely at the mercy of the
distortion correction. Thus the work that we describe here not only serves our own needs
but also contributes to the public welfare.
The summary that we present here will be quite brief, however, because by the time
this account appears, our work will have been published in detail (Anderson & King 2003,
AK03).
2.
The Data Set
For the basic distortion solution we had an excellent data set available. In the so-called
inner calibration field of ω Centauri there are 80 exposures with F555W, at all sorts of
orientations. The variety of orientations turns out to be crucial, because when overlapping
images are all at the same orientation there is no way of solving for the part of the distortion
that consists of skewing.
3.
The True Nature of Distortion
One tends to think of distortion as a problem that consists only of non-linearities, but that
is not so. As we will see, a considerable part of the improvement that we make in the
distortion correction for WFPC2 is the discovery of a hitherto unrecognized skewing in the
PC chip.
This recognition of skewing as a distortion that is mathematically linear leads in turn to
a consideration of what kinds of linear transformations should be used in various situations.
The distinction that we make is that on the one hand we combine positions by using full
6-parameter linear transformations, in order to get the coordinate systems of the two images
to match as well as possible, under conditions where some distortion may remain.
If, on the other hand, we have two coordinate systems each of whose star positions are
completely distortion-free, then these systems are related by a 4-parameter transformation.
The parameters represent a translation, a rotation, and a scale change. In AK03 we refer to
this as a conformal transformation, because within the whole class of linear transformations,
it is the sub-class that can be characterized as angle-preserving. In that paper, in fact, we
make this an operational definition of undistorted images—that positions of stars in any
pair of undistorted images can be related by a conformal transformation—and we apply this
definition quite literally in our search for a set of corrections that will render every image
undistorted.
4.
The Method of Solution
For the actual solution, the conventional approach would be to take a set of overlapping
images and do a least-squares solution for the positions of all the images relative to each
other, and at the same time for the distortion coefficients that get the images best to conform
with each other. But we disliked the black-box aspect of this, so we chose instead to start by
applying the best distortion corrections that we had, and then examining all the position
residuals of individual star images from the mean position of that star, as a function of
location in the chip, in order to see empirically what further correction was needed. For
all the thousands of stars, in all the overlaps between 80 sets of 4 chips each, we had more
An Improved Distortion Solution for WFPC2
313
than a million residuals, and there were enough of them in each small region of the chip
that we could see directly what the mean distortion was there. After a few iterations, we
had the best distortion correction that we could get.
Another aspect of our distortion solution is important to point out: there is a separate
solution for each chip. Although all four have to be solved for simultaneously, because so
many of the overlaps are from one chip to another, the distortion solution for each chip is
independent of the others. To put this in another way, we did not look for a meta-chip
solution, in which all four chips are forced into the same coordinate system. The chips have
moved with respect to each other, over the years, so that there is no way in which we could
have accommodated our multi-year data set in a single fixed system. Thus we treated the
chips separately, and looked at their relative positions later, as a separate problem.
5.
The Results
Like previous corrections, ours is a third-order polynomial. We believe that it is now good
to 0.01–0.02 pixel within each chip. We should note that our solution excluded, and should
not be applied to, pixels in the first 100 rows and columns of each chip. We found that
these are badly behaved, presumably because the OTA spherical aberration spills over the
edges of the reflecting pyramid. At first glance the errors look smaller for the WFs than for
the PC, but in both cases they actually run around 1 mas.
The big surprise was that there is a quite appreciable skewing in the PC, amounting
to about half a pixel over the length of its edge. Previous solutions, which did not have
overlaps that were rotated, had been unable to detect this; and the fact that the otherwise
excellent solution by Casertano & Wiggs (2001) was a meta-chip solution caused the skew
in the PC somewhat to infect the corrections in the other chips.
For the relative positions of the chips we used the outer calibration field in ω Centauri,
which has more than 1200 observations spread from 1994 to 2002. We used successive
overlaps in a rich set of observations at a single epoch to find the positions of all the stars
in a meta-chip coordinate system, and then simply matched each chip in each exposure to
this system, to find where it was in relation to a fiducial point in WF3. The result is a
graph of positions against time for each of the other chips. These are good only to about
a tenth of a pixel, so relative positions within a chip can be measured with much better
accuracy than positions from one chip to another.
All of this is described in detail in our paper (AK03). For the convenience of WFPC2
users there is a link to that paper in the WFPC2 web pages. On a more practical level, the
web pages also have a copy of a Fortran subroutine which, when given the name of a chip
and column and row values (x, y), returns the distortion-corrected position.
6.
Desiderata
Unfortunately our distortion solution is not as good as we would wish. An accuracy of
0.01 pixel is better than ever before, but in many of our measurements we need accuracies
of 0.001 pixel or better, so we are still unable to make transformations that extend over
more than a small fraction of a chip. The problem is that we have been unable to find
distortion changes that are independent of, or even that allow for, the scale changes that
are constantly taking place from one HST image to the next.
The scale changes have three contributors. Two of them come from changes in the
OTA focal length. Occasional adjustments are made in the position of the secondary, but
our main problem comes from orbital breathing, which produces a range of scale change
of about 10−4 around each orbit, in the WFs, and about half that much in the PC. (The
difference is a simple consequence of the Gaussian optics of the transfer systems.)
314
King & Anderson
The third contributor to scale change is velocity aberration. An appendix to our paper
(AK03) shows that velocity aberration produces a scale change that is v/c times the cosine
of the angle of the target from the plane of motion. Since v/c is 10−4 for the motion of the
Earth around the Sun, and the HST orbital velocity is also about 1/4 as large as that, it is
clear that velocity aberration can also be an important contributor to scale changes.
For the study of scale changes due to breathing we had available another valuable data
set, what we like to refer to as the “Gilliland stare,” 8 days of nearly continuous CVZ
imaging of 47 Tucanae (GO-8267).
There are interesting correlations of scale with orbital phase and with time of day. We
simply do not understand them, and have not studied them further.
By far the most intriguing correlation we found was between scale and distortion. What
we did was to look at the positional residuals in each image, and fit them with Legendre
polynomials (just for convenience, because they are quicker to fit than powers). We find
that each of the coefficients correlates closely with scale, so that there is a clean, systematic
change in distortion as the scale changes. (By the way, this is not just a question of whether
we apply the distortion correction before or after the scale change; we tried it both ways, and
it makes hardly any difference.) The frustrating thing about this scale-dependent distortion
is that whereas we could in principle correct for this within the Gilliland data set where
we know all the relative scales, for a randomly chosen image of another field we do not
know where we are in the range of scales, so that we don’t know how to correct for this last
bit of distortion. This is only a small effect—about a hundredth of a pixel from center to
edge—but we wish we could fix it. Again, a hundredth of a pixel does not sound like much,
but in some of our current work it is just not good enough accuracy.
We do not propose to study these scale effects any further, but hope that some one
else will. We will happily make available the scale factors of the more than 1200 Gilliland
images.
Acknowledgments. This work was supported by STScI grant AR-8738.
References
Anderson, J. & King, I. R. 2000, PASP, 112, 1360
Anderson, J. & King, I. R. 2003, PASP, 115, in press (AK03)
Casertano, S. & Wiggs, M. S. 2001, Instrument Science Report WFPC2 2001-10 (Baltimore:
STScI)
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
WFPC2 Flatfields with Reduced Noise and an Anomaly of Filter
FQCH4N-D
E. Karkoschka
Lunar and Planetary Lab, University of Arizona, Tucson, AZ 85721
A. Koekemoer
Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218
Abstract. The transmission of the filter FQCH4N-D varies by 20 percent across
the filter while the mean wavelength shifts by 3 nm. For objects with a flat spectrum across the main bandpass (889–897 nm), the flatfielding removes the spatial
variations except for the outer corner, for which we give the necessary photometric
correction. On the other hand, for objects with steep spectral features within the
bandpass, such as Jupiter and Saturn, the spectral shift causes photometric variations
of some 30 percent across the filter which are not taken out by flatfielding. We give
the magnitude and direction of the shift to account for these variations. Flatfields
with reduced noise are described in the Instrument Science Report WFPC2 200107 http://www.stsci.edu/instruments/wfpc2/Wfpc2 isr/wfpc2 isr0107.html and not
repeated here, except for the abstract: We examine the noise contributed by the
WFPC2 flatfields during normal calibration, and provide new low-noise flats for
41 filters. Highly exposed science images (> 20, 000 electrons per pixel) will show
significant noise reduction if these new flats are used; this is especially true for images on the PC1 chip. For some ultraviolet filters a significant improvement occurs
even for much lower exposure levels. Potential photometric issues are also discussed.
The new flats are available in the HST data archive as calibrated science data (i.e.,
data which have already calibrated with the normal flatfields) to obtain the noise
reduction. These corrections may be incorporated in the normal pipeline flatfields
at some future date for selected filters.
1.
Introduction
The filter FQCH4N of WFPC2 is a quad filter which selects narrow bandwidths in four
methane absorption bands. This gives unique vertical probing of planetary atmospheres
and reduces stray light from planets when imaging nearby rings or satellites. Therefore,
this filter has been the filter most often used for planetary imaging. Eighty percent of the
observations with the filter FQCH4N use the quad with the deepest methane absorption,
the filter FQCH4N-D, which is the focus of this study.
In 1994, we found that images of Jupiter in the filter FQCH4N-D could not be modeled,
unlike many other observations. The images showed an unexplained discrepancy of about
30 percent intensity between the east and west limb of Jupiter. On the other hand, Galilean
satellites had consistent counts across the field of view. We concluded that the filter has a
spatial change of the spectral response which affects the photometry of objects depending
on their spectrum. A warning of a possible spatial variation was posted on the WFPC2
web site at http://www.stsci.edu/instruments/wfpc2/Wfpc2 phot/wfpc2 ss phot.html.
Another indication that this filter is unusual is documented in its flatfield displaying
an anomalous brightness variation of 30 percent with respect to other filters at similar
315
316
Karkoschka & Koekemoer
Figure 1. Transmission curves for the four methane quad filters with the wavelengths scaled to the same mean. For FQCH4N-D, the original and adjusted
transmission curves are shown.
wavelengths. Flatfield variations between different filters of similar wavelengths are typically
on the order of one or a few percent; a variation of 30 percent is only present with one
other WFPC2 filter, which is FQCH4N-C. While this filter could have similar problems as
FQCH4N-D, an investigations of its properties would benefit very few programs because of
the low usage of FQCH4N-C.
In September, 2001, one orbit was devoted to characterize the spatial variation of
filter FQCH4N-D. Nine images of Saturn were taken with identical exposure times but
with different locations covering the whole unvignetted field of view. The rings of Saturn
yielded consistent photometry across most of the field of view as expected since they have
similar spectra as the Galilean satellites. Counts on Saturn’s globe showed a photometric
discrepancy of about 25 percent across the field of view as expected since Saturn’s globe
has a similar spectrum as Jupiter. We use these observations to characterize the filter
FQCH4N-D.
Before we describe the observations further in Section 4, we look at the basic calibration data of the filter, its spectral transmission measurements (Section 2) and its flatfield
(Section 3). Section 5 describes the usable field of view. Section 6 gives a recommended
adjustment to the flatfield. Section 7 explains the observed photometric discrepancy. The
last section concludes with summarizing suggestions for users of the filter FQCH4N-D.
2.
Spectral Transmission Curve
The measured spectral response curve of the filter FQCH4N-D is available at the WFPC2
web site: http://www.stsci.edu/instruments/wfpc2/Wfpc2 thru/fqch4nd.txt. It is plotted
in Figure 1 along with the transmission curves of the other three methane quad filters,
which have been scaled in wavelength to allow the comparison. For transmission values
above 1 percent, all four curves have a similar shape. However, for the scan between 880
and 900 nm, the low transmission numbers of the filter FQCH4N-D seem to level off near
0.65 percent without decreasing any further. This is unlike the other three filters which
plunge steeply below 0.01 percent transmission. We assume that the leveling off at 0.0065
WFPC2 Filter FQCH4N-D
317
transmission for the FQCH4N-D filter is not real but an artifact of the measurement, such
as a constant contribution from background light. We adjust the transmission values of the
filter FQCH4N-D by subtracting 0.0065 of each value whenever the original value is higher
and setting it to zero otherwise. This is the adopted transmission, shown by the open circles
in Figure 1. It follows the shape of the other three curves quite well.
The original transmission curve has a higher integrated throughput than the adjusted
one. For objects with a flat spectrum, the increase is 8 percent. For objects with methane
absorption, the increase is estimated at 38, 34, 14, 21, and 17 percent for Jupiter, Saturn,
Titan, Uranus, and Neptune, respectively, based on published spectra (Karkoschka 1998).
Thus, this adjustment yields a very significant photometric correction.
3.
The FQCH4N-D Flatfield
Figure 2 (top left) displays the flatfield of FQCH4N on WF3 divided by the flatfield of
F850LP which has a similar mean wavelength as FQCH4N-D. Note that the flatfield is displayed here in brightness units while STScI flatfields are usually given in inverse brightness
so that they can be multiplied into the raw image. Obvious are the curved edges near the
center and the bottom right where light through the FQCH4N-D quad is vignetted and
light through the FQCH4N-C and FQCH4N-A filters, respectively, starts to contribute.
Inside the unvignetted field of view, the brightness increases from the bottom to the
upper right. A least-square fit to the data gives a gradient direction of 37 degrees counterclockwise from horizontal. In the perpendicular direction, the brightness scatters by only
one percent or less (Figure 3). Thus, the observed spatial variation is a function of only one
variable, plotted on the x-axis of Figure 3. We chose the center of the WFPC2 pyramid as
the origin of this axis.
Figure 3 also displays another ratio of two flatfields of similar mean wavelengths,
F785LP and F850LP. In this case, the ratio remains close to unity throughout the field
of view. Other ratios behave similarly. The strange slope of FQCH4N-D cannot be due to
variations in spectral sensitivity of pixels. It is due to a spatial variation in the transmission
properties of the filter.
4.
Image Processing
The nine images of Saturn taken for this program were processed with the standard WFPC2
calibration pipeline. Then, a total of 777 pixels were identified which had elevated counts,
mostly due to cosmic ray strikes. The flatfielded counts of those pixels were replaced by
counts interpolated from pixels outside the contaminated areas. The filter FQCH4N-D was
used in its rotation FQCH4N, where it extends mostly across the WF3 chip. Its small
section on the WF2 chip did not produce data suitable for photometry. Figure 2 (top right)
shows the nine calibrated images laid on top of each other, with the maximum data number
at each pixel displayed.
The next image processing step was image navigation, which was performed for each
of the nine images to an accuracy of about 0.05 pixels, taking into account the distortion
for the WF3 chip. After the relative offsets of Saturn in the nine images were determined,
the nine coordinate pairs for a location on Saturn can be calculated. This calculation was
performed for some 100,000 locations. The interpolation of data numbers to fractional
pixels used the 64 pixels of the 8 × 8-pixel box centered on the fractional pixel with cubic
interpolation in both axes.
A mean image of Saturn was created by averaging the nine images accounting for the
appropriate offsets. At each location, the weighting for the averaging was largest for pixels
in the center of the field of view and zero outside the unvignetted field of view (described
318
Karkoschka & Koekemoer
Figure 2. Top left: Ratio of the flatfield brightness of FQCH4N-D/F850LP. Top
right: Nine overlaid flatfielded images of Saturn in FQCH4N-D on WF3. Bottom:
The same nine images after division by the mean image of Saturn. To avoid
cluttering, images are shown either on the left or on the right side. The white
lines mark the adopted edge of the unvignetted field of view.
WFPC2 Filter FQCH4N-D
319
Figure 3. Flatfield brightness ratios on WF3 for FQCH4N-D (left) and F785LP
(right) relative to F850LP. The spatial variable is the x-coordinate in a coordinate
system rotated 37 degrees counterclockwise from the WF3 system with the pyramid
center as the origin (column 30, row 47 of WF3).
in Section 5). The weighting function was the product of the distances to the edges of the
unvignetted field of view.
Each of the nine images was divided by the shifted mean image. In a perfect world,
all divided images should have data numbers of unity. Deviations from unity show various
imperfections (Figure 2, bottom). First, we note that vignetted regions do not divide out
well at all. We need to restrict ourselves to the unvignetted part of the field of view,
described in Section 5. Second, the rings seem to perform close to perfect, except for the
upper right-hand corner where data numbers drop, which is investigated in Section 6. Third,
Saturn’s globe is brightest in the bottom images and faintest in the upper right-hand images,
which is investigated in Section 7. Fourth, the division is not perfect at sharp edges of the
planet. This is probably due to imperfections in the knowledge of the distortion. Therefore,
we discard data within three pixels of those edges, which leaves more than 90 percent of
the data. Finally, we note a bright latitudinal feature just south of Saturn’s equator. Since
it rotates during the time period of the nine exposures, it does not divide out. Therefore,
we exclude this latitude zone from our analysis. No other longitudinal features are obvious
on Saturn. However, a close investigation of divided images yields more features of low
contrast, typically 1–2 percent. Thus, results on Saturn’s globe will be limited to that
accuracy.
5.
Vignetting
For the determination of the edges of the usable field of view, we use Saturn’s rings and exclude locations within three pixels of Saturn’s globe. We distinguish five edges (cf. Figure 2,
top left): the rounded edge near the center towards the quad FQCH4N-C, the horizontal
edge at the bottom towards WF2, the rounded edge at the bottom right towards the quad
FQCH4N-A, the right edge of WF3, and its top edge.
320
Karkoschka & Koekemoer
Figure 4. Deviation of divided ring counts from unity for FQCH4N on WF3,
displayed as function of distance from the adopted edges of the unvignetted field
of view (left), and displayed as function of radius from the pyramid center for the
original and adjusted flatfield (right). The arrow corresponds to the dashed line
in Figure 5.
We define the usable field of view as the area where the divided ring counts are within
about one percent of unity, which means that photometry is consistent to one percent across
this area. Figure 4 (left) displays the mean deviation of counts from unity in the divided
ring images as a function of distance from the adopted locations of the edges. According
to Figure 4, photometric errors rise quickly above several percent outside the adopted
edges, where an image might be still useful for feature recognition, but not for photometry.
Furthermore, a spot of 10 pixel radius around pixel (543,767) does not flatfield out. It is
visible on the left side of Figure 2, near the very top.
For the rotation FQCH4P15 of the filter FQCH4N-D, we estimated the shift and used
the rotation of 15 degrees to derive its unvignetted field on PC1. The adopted unvignetted
fields are displayed in Figure 5. Its definition for WF3 is: |y − 677| > 397 − x and
(x − 135)2 + (y − 415)2 > 370 and |y − 153| > 397 − x and y > 72 and |x − 523| > 162 − y
and (x − 785)2 + (y + 100)2 > 370 and x < 800 (or x < 790 for higher precision) and
y < 800, and (x − 543)2 + (y − 767)2 > 10. The definition for PC1 is: y > 100 and
y > 1.73x − 623 and y < 665 − 0.58x and y < 1.73x − 258 (x is the column number and y
is the row number).
6.
Flatfield Imperfection
Figure 4 (right) displays divided ring counts near the upper right-hand corner of WF3.
Starting near a radius of r = 88.5 arc-seconds from the center of the pyramid (x = 30 and
y = 47), the ring counts deviate by up to nine percent from unity. However, by multiplying
ring counts by 1 + 2 × 10−6 (r − 84)2 , all counts are brought back to unity within the typical
scatter. We think that the upper right-hand corner of WF3 can be used photometrically
after applying the given correction. The decrease of ring counts near that corner seems to
derive from excess light in the flatfield (Figure 3). Thus, the best way to correct this defect
WFPC2 Filter FQCH4N-D
321
Figure 5.
Outlines of the unvignetted fields of view for the filter rotation
FQCH4N on WF3 and FQCH4P15 on PC1. The area outside the dashed line
requires a flatfield correction.
is a change in the flatfield of FQCH4N on WF3. With this correction, the average deviation
of divided ring counts from unity inside the unvignetted field of view is 0.4 percent. Thus,
the filter is suitable for excellent photometry.
7.
Spectral Shift
We divide Saturn into three sections: the rings and both latitude regions south and north of
the latitude zone with the obvious longitudinal feature. Again, locations within three pixels
of a boundary are excluded from all sections. Counts of divided images vary systematically
by 2, 30, and 20 percent across the field of view for the rings, southern latitudes, and the
Equatorial Zone, respectively. Least-square fits to both variations on Saturn’s globe yield
gradient directions near 37 degrees counterclockwise from the x-axis of WF3, the same as
for the flatfield variation discussed in Section 3. Thus, in Figure 6 we plot the divided
counts as functions of the same variable used for Figure 3.
Most likely, the filter FQCH4N-D has a spatial variation such as a variation in the
thickness of a layer. This causes the variation in the flatfield (Section 3) as well as the
different variations seen in Figure 6. While the flatfield variation can be explained by
transmission values changing in the same way for all wavelengths, the variation seen in
Figure 6 requires a spectral variation. The easiest explanation is a spectral shift of the
whole transmission curve.
A small spectral shift can cause the observed variations since Saturn displays steep
spectral features within the bandpass of FQCH4N-D (Figure 7). Notably, from the short
wavelength end of the filter’s passband to its opposite end, Saturn’s flux increases by a
factor of two or more.
We assume that the transmission curve of the filter FQCH4N-D was measured at the
aperture FQCH4NW3, that it is shifted by 2 nm longward at the aperture FQCH4P15, and
that the shift is linear and in the direction 37 degrees counterclockwise from the horizontal
axis of WF3. This assumption can explain the approximate size of the variations seen for
322
Karkoschka & Koekemoer
Figure 6. Divided counts on three parts of Saturn as function of location (dots).
The curves are the expected variations for the adopted wavelength shift of the
filter FQCH4N-D.
WFPC2 Filter FQCH4N-D
323
Figure 7.
Adopted filter transmission curves for two apertures and the spectra of two regions on Saturn, taken from the same observations as published by
Karkoschka (1998).
both sections of Saturn’s globe (curves in Figure 6). It also explains the approximate shape
of the variations. The fact that the curves do not match the dots perfectly may result
from observational limitations such as rotating longitudinal features on Saturn or spectral
variations within each of the two selected sections. In view of these limitations, better fits
possible with a more sophisticated dependency of the spectral shift as function of location
seem unwarranted.
Based on the relative flatfield brightness of the four methane quads, we estimate that
the flatfield source had a color temperature near 3000 K. The spectrum of the ring is close
to a solar spectrum with a color temperature near 6000 K. This difference causes a slight
variation of flatfielded ring counts across the field of view which is even hinted at by the
data, the slightly sloping data points for the rings in Figure 6.
The adopted shift at each pixel can be calculated by (602 − x)0.0032 nm + (608 −
y)0.0024 nm for the rotation FQCH4N on WF3 and 2 nm + (x − 400)0.0011 nm + (y −
312)0.0014 nm for the rotation FQCH4P15 on PC1. The total wavelength shift across
the whole unvignetted field of view of the filter FQCH4N-D is 3.3 nm. This shift causes
brightness variations across the filter FQCH4N-D of factors of 1.29, 1.24, 1.05, 1.01, and
1.02 for Jupiter, Saturn, Titan, Uranus, and Neptune, respectively based on spectra by
Karkoschka (1998). However, spectral variations across each planetary disk are significant
so that these averages are only a rough guide for actual variations of specific features on
each disk.
8.
Recommendations
1. An object should be placed inside the unvignetted field of view, shown in Figure 5.
For the WF3, a placement towards the upper right yields higher signals due to the
higher transmission of the filter.
324
Karkoschka & Koekemoer
2. A placement outside the dashed line of Figure 5 is only recommended if necessary
because of some uncertainty of the flatfield adjustment.
3. The flatfield of FQCH4N-D in rotation FQCH4N on WF3 should be adjusted according to Section 6.
4. The transmission values of filter FQCH4N-D should be reduced by 0.0065 according
to Section 2.
5. The transmission curve of FQCH4N-D should be spectrally shifted as explained in
Section 7. This shift is most important for photometry on Jupiter and Saturn.
6. The observations of Saturn for this work gave a guide to understand the characteristics
of filter FQCH4N-D, but they cannot replace measurements of the filter transmission
curve at various locations of the filter. While this was not done before the installation
of WFPC2 on HST, it may be possible to do when WFPC2 is brought back to the
ground.
7. Other WFPC2 filters such as F953N also have anomalous brightness variations of the
flatfield and measured transmission values which do not seem to reach zero. While
they are smaller than those of FQCH4N-D, they still may be significant for some
applications.
8. Measurements of filter transmission curves are preferentially performed at several
locations since some filters at not spatially homogeneous.
References
Karkoschka, E. 1998, Icarus, 133, 134
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
Using MultiDrizzle to combine Dithered WFPC2 Images
Gabriel Brammer, Anton Koekemoer, and Bulent Kiziltan
Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218
Abstract. This poster presents a guide to using the new MultiDrizzle Pyraf script
to combine sets of dithered WFPC2 images. The MultiDrizzle script has condensed
the steps of drizzling multiple images, as shown in the HST Dither Handbook (Koekemoer, et al. 2002), into a single Pyraf command with a number of parameters governing its behavior. This is aimed at greatly improving the ease with which images can
be registered, cleaned of cosmic rays, and combined together using drizzle and related tasks. Images that have been produced using MultiDrizzle to combine WFPC2
datasets from the Dither Handbook examples are presented, and the results from a
variety of parameter settings are explained and compared.
1.
Introduction
Previously, a set of tools were provided in the IRAF/STSDAS dither package to analyze
dithered data obtained with HST. These tools address a variety of issues, such as registration, cosmic ray cleaning, and combination. The dither tasks are extremely flexible, but
are also extremely complex with a large number of parameters.
A new technique is now available—MultiDrizzle (Koekemoer et al. 2003, this volume,
p. 337)—which automatically calls the dither package scripts along with the drizzle program
(Fruchter & Hook 2002) and the PyDrizzle script(Hack & Jedrzejewsky 2002), using default
parameters designed to work for a wide range of images (while still allowing the parameters
to be changed as necessary). MultiDrizzle is run as a single command in Pyraf, following
standard IRAF syntax.
2.
Some MultiDrizzle Parameters
The initial steps carried out by the MultiDrizzle script consist of the identification of bad
pixels, subtraction of the sky background, and running drizzle to transform each of the
individual input images onto a set of output images that are registered on a common frame.
The script then combines these registered images to create a clean median image, which
is subsequently transformed back to the original frame of each input image using the blot
task.
The next step involves the creation of the cosmic ray mask file. This is done by
comparing each original input image with its counterpart “blotted” clean image, together
with a third image that represents the spatial derivative of the “blotted” image. This
comparison is carried out by the task driz cr, and it uses the following algorithm:
|original image − blotted image| > scale × derivative + snr × rms .
(1)
The two important parameters in the step are:
• driz cr scale—this takes into account the possibility that slight offsets in the shifts may
be present, which could cause inappropriate rejection of valid pixels, such as bright
325
326
Brammer, et al.
stellar cores. Increasing the value of scale will help ensure that such misidentifications
occur less often; good values to use for scale typically range from 1–2.
• driz cr snr —this is simply a multiplicative scaling of the rms, which has been calculated by taking into account the sky background (stored in the header), together with
the readnoise and gain, as well as the observed flux in the pixel. Typical values for
snr that will yield good results are in the range 3–5; higher values will lead to fewer
cosmic rays being rejected, while lower values will cause more frequent rejection of
pixels that are not necessarily cosmic rays.
After creating the cosmic ray masks, the final step is to drizzle all the input images onto
a single output image, using the information from the masks to exclude pixels in each input
image that have been affected by cosmic rays. The task drizzle has several parameters, but
two of the most import are:
• final scale—the size of output pixels relative to the input pixels. If the sub-pixel
space is reasonably well sampled by more than a few dithers, then it is acceptable
to consider setting the value of scale small enough to provide critical sampling of the
PSF. Typically, this means that values down to ∼ 0.4–0.5 can be considered for scale.
• final pixfrac—the size by which input pixels are “shrunk” before being mapped onto
the output grid. The values for this range between 0 and 1: if pixfrac = 0, then this is
equivalent to interlacing (each input pixel will only ever contribute to a single output
pixel), while at the other extreme, pixfrac = 1 corresponds to “shift-and-add.” In this
case the output image is convolved by the full size of the input pixels. For a typical
case of ∼4 sub-pixel dithers and scale = 0.5, reasonable values for pixfrac would be
in the range 0.6–0.8. This allows some sharpening of the PSF relative to pixfrac = 1,
while at the same time still retaining reasonably uniform coverage of the output pixel
grid.
For more information on the many MultiDrizzle parameters, as well as a detailed description of the script’s intermediate actions and products, please refer to the paper by
Koekemoer et al. (2003, this volume) and to the MultiDrizzle web page, located at:
http://www.stsci.edu/∼koekemoe/multidrizzle/
3.
Executing MultiDrizzle on a Set of WFPC2 Images
This section goes through a step-by-step explanation of how to run MultiDrizzle on a set of
3 dithered exposures of the edge-on spiral galaxy NGC 4565 (ID 6092,PI: Keith Ashman)
used in Example 2 of the HST Dither Handbook (Koekemoer et al. 2002). Note that
MultiDrizzle products require substantial free disk space. The GEIS-formatted images for
this example can be downloaded from:
http://www.stsci.edu/instruments/wfpc2/dither/examples.html
MultiDrizzle is run in the Pyraf command environment, loaded with pyraf at the unix
command prompt. More information about Pyraf can be obtained at the STScI Pyraf home
page:
http://pyraf.stsci.edu
Move to the directory containing the images for this demonstration, create an input
image list, and set up MultiDrizzle:
$ cd /data/mydir/
$ ls -1 *.c0h > files.list
MultiDrizzle and WFPC2
327
$ pyraf
pyexecute(’/data/wallaby1/anton/multidrizzle/multidrizzle iraf.py’,
--> import multidrizzle
--> unlearn multidrizzle
files.list should contain:
u31s0101t.c0h
u31s0102t.c0h
u31s0103t.c0h
Run MultiDrizzle for the WFPC2 images with shifts calculated from the image headers:
--> multidrizzle output=’final’ filelist=’files.list’ inst=’WFPC2’
The drizzled output images are 4-chip mosaics with the pixel scale of the PC (0.05/pixel).
Since the WF chips have twice the pixel scale of the PC (0.1/pixel), the output images are
approximately
2 chips × 800 pixels × 2 pixel scale
= 3200
(2)
f inal.scale = 1
pixels on a side. Display the drizzled science and weight images (see Figure 1), and compare
to an individual input frame:
--> display final sci.fits 1 zr- zs- z1=-.1 z2=.5
--> display final wht.fits 2 zr- zs- z1=0 z2=500
--> display u31s0101t.c0h[3] 3 zr- zs- z1=0 z2=100
4.
MultiDrizzle Products
The output science images have units of counts s−1 , and the weight image is a map of the
effective exposure time per pixel. Note how the final science image stitches together the
four WFPC2 chips and how the cosmic rays seen in the input image are eliminated. The
slight level differences between the chips, (e.g., between WF2 and WF3 at upper and lower
left, respectively) arise from difficulties in determining the sky background level for images
of an extended source that entirely fills the camera’s field of view.
The weight map shows how pixels affected by cosmic ray events on one or more input
exposures have lower effective exposure times than those unaffected by cosmic rays. In this
example with three exposures of 160 s each, a pixel affected by cosmic rays in all three
images would have an effective exposure time (or weight) of zero, while a clean pixel in
all three exposures would have an effective exposure time of 480 s. The PC and WF chips
scale differently in the weight image because of the difference in pixel scales discussed above.
MultiDrizzle corrects for the geometric distortions of the WFPC2 chips, and this correction
is also visible in the weight images. From the distortion, the area of a single pixel projected
onto the sky decreases towards the corners of the chips, and this decrease is incorporated
in the flatfield file. Thus, the weight of the pixel needs to be increased correspondingly in
order to preserve total flux.
5.
Conclusions
MultiDrizzle produces clean, registered drizzled images of dithered exposures through an
interface that greatly simplifies the dither process. For the example described above, the
process of producing properly drizzled products that required more than one hundred commands using the IRAF dither package has been reduced to a single Pyraf command.
328
Brammer, et al.
Figure 1.
(left) MultiDrizzle output science image, final sci.fits, with inverted color scale. (right) MultiDrizzle output weight image, final wht.fits.
The display of the weight image has been stretched to highlight the geometric
distortion corrections in the WF chips. The gradient from the centers to the edges
of the chips is about 3% of the modal WF pixel value. The PC shows a similar
effect, but at a level that is outside the image stretch as a result of the difference
in the pixel scales of the PC and WF chips.
References
Fruchter, A. & Hook, R. 2002, Drizzle: a method for the linear reconstruction of undersampled images, PASP, 114, 144.
Hack, W. & Jedrzejewsky, R., 2002, Pydrizzle User’s Manual,
http://stsdas.stsci.edu/pydrizzle
Koekemoer, A. M., et al. 2002, HST Dither Handbook, Version 2.0 (Baltimore: STScI)
Koekemoer, A. M., Fruchter, A. S., Hook, R., & Hack, W. 2003, this volume, 337
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
WFPC2 Pointing Uncertainties
Gabriel Brammer, Brad Whitmore, and Anton Koekemoer
Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218
Abstract. This paper examines the absolute pointing of WFPC2 using repeat
observations of GRW+70D5824, our photometric monitoring star. An offset of 1–2is
found between the intended pointing position—pixel (420,424.5) on the PC (PC1)—
and the observed pointing positions. This offset manifests itself as a circular annulus
of data points centered around PC1 due to changing roll angles of HST during the
year, which hence increases the observed pointing scatter from the expected ∼1 (due
to uncertainties in the guide stars) to 2–3 . This issue is not critical for most WFPC2
users since the camera has a large field of view. Users requiring better absolute
astrometric precision should measure positions relative to astrometric standards on
the same image frame, if they exist.
1.
Introduction
The absolute pointing of WFPC2, with its wide field of view, is not as critically important as
it is for some of the other HST science instruments that require placing a target on apertures
as small as a 5 square, (e.g., STIS). However, occasionally such accuracy is desired for
specific target placement on the WFPC2 chips. The absolute pointing of WFPC2 has not
been extensively studied in the past, and a more careful characterization of the absolute
pointing could be useful for estimating the reliability of both past and future WFPC2
pointings.
Following Servicing Mission 3B, a jump was noticed in the relative alignments between
the three Fine Guidance Sensors of the FGS. The team monitoring the FGS alignment
wanted to see if this jump was visible in WFPC2 observations, motivating the accumulation
of the dataset described here.
2.
Dataset
The observations used here to monitor the pointing of WFPC2 come from observations of the
WFPC2 standard star GRW+70D5824 taken for the monthly photometric monitor program
(Gonzaga et al. 2003). The star has been consistently observed since 1994 in 9 filters on
both the PC and WF3 chips, with less frequent observations on the other two WF chips.
The sample analyzed here contains 92 observations beginning Jan 2, 1999, through Aug 11,
2002, using the PC and filter F555W.
These data provide an opportunity to monitor the absolute pointing of WFPC2, as the
star would ideally be consistently placed at the same pixel location on the chip with some
associated scatter due to uncertainties in the guide star positions (∼1 ). The expected
pixel position here corresponds to the PC1 aperture located at (420,424.5) on the PC. The
observed positions of GRW+70D5824 are shown in Figure 1.
329
330
Brammer, et al.
Figure 1.
(a) GRW+70D5824 positions on the PC from Jan 1999 to Aug 2002.
The different point styles indicate which Fine Guidance Sensor was dominant for
the particular observation. The offsets are relative to the PC1 aperture position, marked by “+”. The plate scale of the PC is ∼0.05/pixel. (b) x and y
GRW+70D5824 positions vs. time. The sinusoidal oscillations in each direction
have a period of ∼365 days.
3.
Analysis
The standard star pointings plotted in Figure 1a do not scatter randomly around PC1.
Rather, they are distributed in an approximately circular annulus centered around the PC1
aperture location. If the points of Figure 1a are plotted consecutively in time, then the
star’s location on the chip appears to rotate in the counterclockwise direction about PC1.
This rotation is manifested in the out-of-phase sinusoidal variations of x and y position with
time, shown in Figure 1b.
Such a rotation is caused by a pointing offset in RA and/or declination whose projection on the camera x and y axes changes as the orientation of the telescope rotates. The
orientation angle rotates through 360◦ over the course of the year. To estimate the WFPC2
pointing uncertainty without the effect of the pointing offset, we modeled the rotation of
the data points by “de-rotating” them by an angle φ, relative to an arbitrary reference
observation. The angle φ (in radians) used is then:
φ=
(M JD − M JDref )
× 2π.
365
The “de-rotated” result is shown in Figure 2a, with M JDref = 51371 (Jul 12, 1999).
Another model fitting a sinusoidal function to the points of Figure 1b for each guidance
sensor individually and plotting the residuals was used to compare with the rotation model.
The amplitude and vertical offset were fit for each FGS in both x and y, while the phase and
period were set to the values that produced the best fit in FGS3. For each FGS, the fitted
vertical offset was added to the residuals to predict the offset caused by the misalignment.
The fit residuals are shown in Figure 2c for comparison with the results of the “de-rotation”
model. A comparison of the rms scatter of the points for each FGS before and after applying
both models is shown in Table 1.
WFPC2 Pointing Uncertainties
Table 1.
331
Rms before and after De-rotating (in arcseconds)
Dominant FGS
FGS1r
FGS2r
FGS3
initial
(x)
0.65
0.91
0.62
“de-rotated”
(x)
0.61
0.41
0.21
“sine-fit”
(x)
0.28
0.47
0.18
initial
(y)
0.66
0.63
0.51
“de-rotated”
(y)
0.56
0.90
0.29
“sine-fit”
(y)
0.25
0.44
0.21
Figure 2.
(a) “De-rotated” pointing positions. The positions fall into distinct
groups according to which FGS was dominant for a particular pointing. The “θ”
position of the groups about the origin depends on the choice of the arbitrary
reference observation, so this model does not indicate the physical direction of
the misalignment offsets for each FGS. (b) Predicted distribution of pointings
after the aperture update implemented Oct 20, 2002, to correct the FGS-FGS
misalignments. This plot centers the mean position of the pointings for each FGS
on (0,0). (c,d) Same as a,b for the residuals of the sine-fit model. Note how the
points are more tightly distributed, and how they continue to fall into discrete
groups according to dominant FGS.
4.
Conclusions
The most obvious feature of the plots before and after modeling the rotation of the data
points is that the pointing locations clearly fall into discrete regions according to which Fine
Guidance Sensor was dominant for a given pointing. This indicates a relative misalignment
between the guidance sensors, which has been previously seen in FGS monitors using STIS.
The pointings using FGS3 as dominant show the smallest rms after removing the
rotation effect, shown in Figures 2a,c and lie closest to the expected pointing of the PC1
aperture location. The rms scatter of these pointings decreases by a factor of about 2.5
in both x and y after compensating for the rotation. Where FGS 1r was dominant, the
pointings show a smaller decrease in scatter after removing the rotation, and there is minimal
improvement seen for pointings using FGS 2r. Both the scatter in the pointings, and the
amount by which the pointings miss the aperture position, are largest for pointings in which
FGS 2r was dominant.
These results confirm STIS monitors that indicate that pointings using FGS 3 are the
most accurate and precise of the three, while FGS 2r pointings show significant scatter and
experience more frequent pointing failures. These results can be used as a “before” picture
of the WFPC2 pointings to compare with results of pointings made after the observatory
332
Brammer, et al.
aperture file was updated on Oct 20, 2002 to correct for the misaligned guidance sensors.
After the aperture correction, we would expect all of the pointings in all three guidance
sensors to be distributed randomly about PC1, or the desired pointing position, with a
small rms scatter caused by the astrometric accuracy limits of the guide star catalogs. To
estimate the improvement in pointing presision after the update, we shifted the pointing
position for each FGS to a common mean and plot the results in Figure 2b,d. Combining
this shift and the rotation from above, the rms scatter of all of the pointings decreases
by 70–80% in x and 10–40% in y. Since the orientations of x and y are arbitrary in the
model, the characteristic rms after the aperture update should show a decrease of 50–70%
after Oct 20, 2002. We will determine the actual updated WFPC2 pointing statistics from
subsequent GRW+70D5824 observations taken over the coming months.
Table 2.
Rms before and after De-rotating (in arcseconds)
Dominant FGS
FGS1r
FGS2r
FGS3
initial
(x)
0.65
0.91
0.62
“de-rotated”
(x)
0.61
0.41
0.21
“sine-fit”
(x)
0.28
0.47
0.18
initial
(y)
0.66
0.63
0.51
“de-rotated”
(y)
0.56
0.90
0.29
“sine-fit”
(y)
0.25
0.44
0.21
Acknowledgments. We would like to thank Olivia Lupie and Colin Cox for their FGS
expertise.
References
Gonzaga, S., Ritchie, C., Baggett, S., Whitmore, B., Casertano, S. 2003, Technical Instrument Report WFPC2 (Baltimore: STScI), to be released in early 2003
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
The Accuracy of WFPC2 Photometric Zeropoints
Inge Heyer, Marin Richardson, Brad Whitmore, Lori Lubin
Space Telescope Science Institute, Baltimore, MD 21218
Abstract. The accuracy of WFPC2 photometric zeropoints is examined using two
methods. The first approach compares the zeropoints from five sources: Holtzman
(1995), the HST Data Handbook (1995 and 2002 versions), and Dolphin (both 2000
and 2002 versions). We find the mean scatter between the different studies to be:
0.043 mag for F336W, 0.034 mag for F439W, 0.016 mag for F555W, and 0.018 mag
for F814W.
The second approach is a comparison of WFPC2 observations of NGC2419 with
ground-based photometry from Stetson (from his website) and Saha et al. (private
communication). The tentative agreement between these comparisons is similar to
the historical zeropoint comparisons. Hence we conclude that the true uncertainty of
WFPC2 zeropoints is currently about 0.02–0.03 magnitudes. Since Poisson statistics
would predict that 1% absolute accuracy should be attainable, we conclude that
there are still systematic error sources which have not yet been identified.
1.
Goals and Approach
The ultimate goal of this project is to determine if 1% absolute photometry is possible using
WFPC2. In principle this should be attainable, as evidenced by the fact that the shortterm rms in our photometric monitoring observations for the primary broadband filters are
< 1%. The challenge is to: 1) understand the various systematic errors well enough (e.g.,
CTE loss, variable focus, geometric distortion, etc.) and 2) match the zeropoints to existing
standards with enough precision to make this possible. In this poster we address the second
issue by examining the accuracy of WFPC2 photometric zeropoints using two methods.
The first approach compares the zeropoints from five sources: Holtzman (1995), HST
Data Handbook (1995), HST Data Handbook (2002), Dolphin (2000), and Dolphin (2002).
See Whitmore (2003) for a discussion and a figure. These five studies use largely independent
methods to determine zeropoints (e.g., the Data Handbook uses a single photometric monitoring star and SYNPHOT while Dolphin uses ground-based photometry of Omega Cen
and NGC 2419). Hence the resulting scatter provides an empirical estimate of the true
uncertainty.
The second approach is a comparison of WFPC2 observations of NGC 2419 with
ground-based photometry from Stetson (2002; from his website) and Saha et al. (2002; private communication). The resulting scatter between these two determinations, along with
the historical scatter outlined above, provides our best estimate of the true uncertainty in
the WFPC2 zeropoints. A weighted combination of all determinations will eventually be
used to determine new WFPC2 zeropoints for the F336W, F439W, F555W, F675W, and
F814W filters. At present, we have results for F555W and F814W.
Caveat: The current results should be considered tentative, pending some additional
checks. Please refer to the Instrument Science Report (when completed) and the WFPC2
WWW site for the final values.
333
334
Heyer, et al.
Table 1.
Filter
F555W
F814W
a
Chip
PC1
WF2
WF3
WF4
PC1
WF2
WF3
WF4
Averaged Means and Mean Residuals of the Zeropoint Deltas
Sample
Sizea
22
29
32
16
46
64
68
48
Stetson’s
Averaged Mean
−0.0087
0.0106
−0.0083
−0.0048
−0.0197
−0.0255
−0.0310
−0.0143
Stetson’s
Mean Residual
0.0303
0.0412
0.0403
0.0270
0.0346
0.0374
0.0306
0.0313
Sample
Sizea
10
13
13
12
8
16
14
11
Saha’s
Averaged Mean
−0.0229
0.0010
−0.0025
−0.0264
−0.0708
−0.0573
−0.0654
−0.0746
Saha’s
Mean Residual
0.0412
0.0293
0.0321
0.0531
0.0394
0.0304
0.0269
0.0414
‘Sample’ refers to different datasets observed between 1995 and 2002. Each sample typically consists of 5–20
stars.
2.
Data Reduction
The images were first multiplied by a geometric distortion correction image, since we are
doing point-source rather than surface photometry. Aperture photometry was performed
on each dataset using a 0.5 radius, and the values were corrected to infinity by subtracting
0.1 magnitudes (Holtzman 1995). Very bright stars and very faint stars were trimmed from
the sample, due to suspected saturation and excessive noise, respectively. Searches were
then performed to identify stars that matched stars from Stetson’s (2002; WWW site) data
files. The Dolphin (2002) CTE correction and the Holtzman color transformations were
applied. The sample was further trimmed by applying graduated isolation criteria with a
limit approximating a 4-magnitude difference at 5 distance. Finally, plots were produced
for each dataset showing the magnitude and V − I versus the delta between the observed
magnitude (using the Data Handbook 2002 zeropoint) and the comparison study.
3.
Results
We present the results of our examination for the target NGC 2419 in the filters F555W
and F814W. Table 1 shows the averaged means and mean residuals of the deltas of the
zeropoints for each filter and chip.
Figures 1–4 show the mean as a function of exposure time and observation date (in
MJD) for F555W and F814W. The circles show the results from the comparison with
Stetson’s stars, and the triangles show the results from the comparison with Saha’s stars.
4.
Conclusions
1. The true uncertainty in the current WFPC2 zeropoints, as judged by either the historical zeropoints (see Figure 3 in Whitmore 2003) or comparisons of HST observations of
NGC 2419 with ground-based photometry is about 0.02 mag for F555W and F814W,
and about 0.03–0.04 mag for F439W and F336W. The F814W comparison with Saha
(2002) appears to be slightly worse.
2. The short-term rms scatter would predict that an accuracy of 1% should be attainable.
The fact that the true uncertainty is currently about 0.02–0.03 magnitudes indicates
that there are as yet unidentified error sources.
The Accuracy of WFPC2 Photometric Zeropoints
Figure 1.
Delta vs. Exposure Time for F555W.
Figure 2.
Delta vs. Observation Date for F555W.
Figure 3.
Delta vs. Exposure Time for F814W.
Figure 4.
Delta vs. Observation Date for F814W.
335
336
Heyer, et al.
3. While there appear to be some possible trends in the zeropoint deltas versus exposure
time and time of observation, the lack of agreement in these trends for the different
filters suggests that the underlying source of the error is still unknown.
4. Results of the various methods used here will be averaged together to produce new
values for the zeropoints. These will be included in a WFPC2 Instrument Science
Report and on the WFPC2 WWW site at a future date.
References
Baggett, S. ed., HST Data Handbook (WFPC2), Version 4.0, October 2002 (Baltimore:
STScI)
Dolphin, A. E. 2000, PASP, 112, 1397
Dolphin, A. E. 2002, http://www.noao.edu/staff/dolphin/wfpc2 calib/
Holtzman, J., et al. 1995, PASP, 107, 1065
Leitherer, C. ed.,HST Data Handbook, Version 2.0, December 1995 (Baltimore: STScI)
Saha, A. 2002, (private communication)
Stetson, P. 2002, http://cadcwww.dao.nrc.ca/cadcbin/wdb/astrocat/stetson/query/
Whitmore, B. 2003, this volume, 281
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
MultiDrizzle: An Integrated Pyraf Script for Registering, Cleaning
and Combining Images
Anton M. Koekemoer, Andrew S. Fruchter, Richard Hook,1 Warren Hack
Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218
Abstract. We present the new PyRAF-based ‘MultiDrizzle’ script, which is aimed
at providing a one-step approach to combining dithered HST images. The purpose
of this script is to allow easy interaction with the complex suite of tasks in the
IRAF/STSDAS ‘dither’ package, as well as the new ‘PyDrizzle’ task, while at the
same time retaining the flexibility of these tasks through a number of parameters.
These parameters control the various individual steps, such as sky subtraction, image registration, ‘drizzling’ onto separate output images, creation of a clean median
image, transformation of the median with ‘blot’ and creation of cosmic ray masks,
as well as the final image combination step using ‘drizzle’ . The default parameters of all the steps are set so that the task will work automatically for a wide
variety of different types of images, while at the same time allowing adjustment of
individual parameters for special cases. The script currently works for both ACS
and WFPC2 data, and is now being tested on STIS and NICMOS images. We
describe the operation of the script and the effect of various parameters, particularly in the context of combining images from dithered observations using ACS and
WFPC2. Additional information is also available at the ‘MultiDrizzle’ home page:
http://www.stsci.edu/∼koekemoe/multidrizzle/
Introduction
The MultiDrizzle task is designed to provide a seamless, integrated approach to using all the
various tasks in the IRAF/STSDAS dither package to register, clean, and combine dithered
images. It has quite a few parameters but in principle can be run very simply from the
PyRAF command line, specifying only the output filename and an input file list, e.g.:
multidrizzle output=‘outputfilename’ filelist=‘files.lis’
The other parameters can be specified on the PyRAF command line or alternatively can be
edited using the standard IRAF ‘epar’ mechanism before running the task. It is designed
to carry out the following steps, either in a single pass or alternatively by selecting various
steps individually:
1. StaticMask
- Identify negative bad pixels, based on examining all the images,
and include them in the dq file
2. SkySub
- Sky-subtract each frame
3. Driz Separate - Drizzle the input images onto separate, registered outputs
(using shifts computed from the headers)
4. Median
- Create a median image from the separate drizzled images
5. Blot
- Blot the median image back to the original input frames
- Use each blotted image to create a derivative image,
6. Driz cr
and compute CR masks
7. Driz Combine - Do the final drizzle combination
Here we describe the details of the parameters involved in running each of these steps.
1
Space Telescope European Coordinating Facility, Karl-Schwarzschild-Str. 2, Garching, D-85748, Germany
337
338
Koekemoer, et al.
1. Create the Static Mask
Parameters:
staticfile
static goodval
=
=
1.0
Name of (optional) input static bad-pixel mask
Value of good pixels in the input static mask
Output files:
modified dq array in the original input files
This step goes through each of the input images, calculates the r.m.s value for each chip,
and identifies pixels that are below the median value by more than 5 times the r.m.s.
This is aimed at identifying pixels that may have high values in the dark frame that is
subtracted during calibration, but may not necessarily have high values in the images, thus
the subtraction gives them strongly negative values. Such pixels are not always flagged in
the dq file, hence this step allows them to be identified. Sometimes such pixels fall on bright
objects so they would not be negative, but instead would be positive although lower than
surrounding pixels. However, if the images are dithered then they should land on blank sky
at least some of the time, in which case they will appear negative and will be flagged.
2. Perform Sky Subtraction
Parameters:
skytype
skyname
skywidth
skystat
skylower
skyupper
=
=
=
=
=
=
‘single|quadrants’
‘SKYSUM’
50.0
‘mode|mean|median’
−50.0
200.0
Type of subtraction (e.g., amplifier quadrants)
Header keyword containing sky value
Interval width for sky statistics
Sky correction statistics parameter
Lower limit of usable data for sky (in DN)
Upper limit of usable data for sky (in DN)
Output files:
modified science array in the original input files
This can subtract either the entire chip (skytype = ‘single’) or specific regions corresponding
to each of the four individual amplifiers on the ACS/WFC chips (skytype = ‘quadrants’).
The other parameters correspond directly to those in the sky task in the dither package,
and are passed to it exactly as they are specified here.
3. Create Separate Drizzled Images
Parameters:
driz sep outnx
driz sep outny
driz sep kernel
driz sep scale
driz sep pixfrac
driz sep rot
driz sep fillval
=
=
=
=
=
=
=
‘square|point|gaussian|turbo|tophat’
1.0
1.0
0.
INDEF
Output image x-size
Output image y-size
Drizzle kernel
Size of output pixels
Size of ‘drop’
Rotation (anticlockwise)
Value for undefined pixels
Output files:
* single sci.fits (drizzled output image for each input image)
* single wht.fits (weight image corresponding to each drizzled image)
This task drizzles the input images onto separate output images. By default it uses the
drizzle ‘turbo’ kernel, and drizzle parameters of pixfrac = 1 and scale = 1. These can
be changed; for example masks can be substantially improved by specifying a smaller value
of scale (e.g., 0.5 or 0.66), with the trade-off being larger images (their size increases as
the inverse square of the value of scale), and increased computation time. The shifts used
here are calculated from the image headers by PyDrizzle.
MultiDrizzle: Automatic Image Combination
339
4. Create the Median Image
Parameters:
median newmasks
combine type
combine reject
combine nsigma
combine nlow
combine nhigh
combine grow
=
=
=
=
=
=
=
yes
‘minmed|average|median’
‘minmax|ccdclip|crreject|avsigclip’
63
0
1
1.0
Create new masks?
Type of combine operation
Type of rejection
Significance for min. vs median
Number of low pixels to reject
Number of high pixels to reject
Radius for neighbor rejection
Output files:
output med *.fits
* single wht maskhead.pl
This creates a median image from the separate drizzled input images, allowing a variety of
combination and rejection schemes. If combine type is set to ‘median’ or ‘average’, then the
routine calls the IRAF task imcombine, passing to it the values of combine reject (usually
expected to be ‘minmax’), along with combine nlow and combine nhigh (the number of low
and high pixels to reject) and combine grow, the amount by which flagged pixels can grow.
If median newmasks = ‘yes’, then pixels are flagged using the static bad pixel masks. If this
parameter is ‘no’ then the task will simply use whatever masks are specified in the ‘BPM’
header keyword of each image (which you could create yourself). In general, however, it is
recommended to use the static bad pixel masks that are generated by default.
If combine type is set to ‘minmed’, then this task will use a slightly more sophisticated
algorithm than the ones in imcombine, to create a cleaner combined image. The basic
concept in this case is that each pixel in the output combined image will be either the
median or the minimum of the input pixel values, depending on whether the median is
above the minimum by more than a certain number of sigma. An estimate of the ‘true’
counts is obtained from the median image (after rejecting the highest-valued pixel), while
the minimum is actually the minimum unmasked (‘good’) pixel. This algorithm is designed
to perform optimally in the case of combining only a few images (3 or 4), where tripleincidence cosmic rays often pose a serious problem for more simplified median combination
strategies. It performs the following steps:
1. Create median image, rejecting the highest pixel and applying masks
2. Use this median to estimate the true counts, and thus derive an r.m.s.
3. If the median is above the lowest pixel value by less than the first value in combine nsigma,
then use the median value, otherwise use the lowest value.
If combine grow > 0, repeat the above 3 steps for all pixels around those that have already
been chosen as the minimum, this time using a lower significance threshold specified as the
second value in combine nsigma. This is very successful at flagging the lower-S/N ‘halos’
around bright cosmic rays that were flagged in the first pass.
5. Blot Back the Median to the Frame of the Original Images
Parameters:
[none]
Output files:
* blt.fits
This takes the median image and uses the dither package task blot to apply the geometric
distortion and transform it back to the reference frame of each of the original individual
input images, in preparation for the subsequent step of cosmic-ray rejection.
340
Koekemoer, et al.
6. Create Cosmic Ray Masks
Parameters:
driz cr snr
driz cr scale
=
=
‘3.0 2.5’
‘1.2 0.7’
driz cr.SNR parameter
driz cr.scale parameter
Output files:
* blt deriv.fits
* cor.fits
* crderiv.pl
This uses the original input images, the blotted images, and the derivative of the blotted images (created using the dither.deriv task) to create cosmic ray masks (using the dither.driz cr
task), stored as separate files, which can later be combined with other masks. This step also
creates a ‘ cor’ image, where bad pixels are replaced with pixels from the blotted median
image. These relatively clean ‘ cor’ images can be used to determine shifts, if desired.
7. Perform Final Drizzle Combination
Parameters:
final outnx
final outny
final kernel
final scale
final pixfrac
final rot
final fillval
=
=
=
=
=
=
=
‘square|point|gaussian|turbo|tophat’
1.0
1.0
0.
INDEF
Output image x-size
Output image y-size
Drizzle kernel
Size of output pixels
Size of ‘drop’
Rotation (anticlockwise)
Value for undefined pixels
Output files:
output sci.fits
output wht.fits
This takes the original input images, together with the final cosmic ray masks, and drizzles
them all onto a single output image. The standard drizzle parameters of kernel, scale,
pixfrac and rot can be specified for this task. By default the pixel scale of the output
image is 1, but feel free to experiment with other options (e.g., when combining at least 4
sub-pixel dithered images, scale = 0.5 and pixfrac = 0.7 yields a sharper output PSF).
Future Enhancements
The next major enhancement to this script will consist of the ability to iteratively refine
the shifts between images. This involves the use of image-based catalogs, or alternatively
cross-correlation techniques, to directly determine the shifts from the data in the images.
Prototype versions of these techniques are being successfully implemented, and a robust
version will be included in a subsequent public release of MultiDrizzle.
Acknowledgments. We are pleased to acknowledge very valuable contributions from
a large number of people who have contributed ideas or feedback, including Ed Smith,
Max Mutchler, Gabe Brammer, Bill Sparks, Benne Holwerda, Stefano Casertano, Harry
Ferguson, Shireen Gonzaga, Megan Sosey, Linda Dressel, as well as many others.
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
WFPC2 Re-Commissioning After Servicing Mission 3B
Anton M. Koekemoer, Shireen Gonzaga, Inge Heyer, Lori M. Lubin, Vera
Kozhurina-Platais, and Brad Whitmore
Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218
Abstract. We describe here the results from an extensive series of tests and observations that we carried out with WFPC2 as part of the Observatory Verification
program during March to April 2002, after SM3B. These tests included UV monitoring of possible contamination, performance checks of the biases, darks and other
internal calibrations, as well as PSF and flatfield verification. The results from these
tests show that there are no significant changes in the characteristics of the camera
with respect to its pre-SM3B performance.
1.
Introduction
In March 2002 Servicing Mission 3B (SM3B) was carried out, which included the addition
of the NICMOS Cryo-cooler System (NCS) and the Advanced Camera for Surveys (ACS).
While these will facilitate a wide range of science programs, WFPC2 retains a number
of unique scientific capabilities. Thus, the WFPC2 SM3B plan involved protecting the
health and safety of WFPC2 during and immediately after SM3B, and evaluating possible
changes in its performance. Throughout SM3B and the subsequent 12 days of BrightEarth Avoidance (BEA), WFPC2 was maintained in an inactive Protect Decon mode with
the camera heads warm (+22◦ C), the shutter closed and the F785LP filter in place, to
minimize the risk of potential contaminants entering the instrument and depositing on the
optical surfaces. On March 23 2002, WFPC2 was cooled down to its nominal operating
temperature of −88◦ C, and an extensive four-week program of Servicing Mission Orbital
Verification (SMOV3B) calibration observations were commenced, to verify that the camera
performance and characteristics remained essentially unchanged. Here we describe the
analysis and results from these programs.
2.
UV Contamination Monitoring
A critical component of WFPC2 cool-down involved intensive monitoring of the UV throughput, to ensure no permanently degradation by contamination deposited on the cold (−88◦ C)
CCD windows. We began monitoring the standard star GRW+70d5824 immediately after
cool-down, using the F170W filter in all four chips, repeated at 3, 6, 12, 18, 24, 36 hours,
and 2, 3, 4, 5, 6 days after cooldown, after which a Decon was scheduled. UV observations
were also obtained before and after each subequent Decon during SMOV.
The results are presented in Figure 1, and can be summarized as follows: (1) at no
point did any of the cameras exceed a 10% drop in throughput, well removed from the
30% limit; (2) the daily contamination rate was slightly higher than normal, but still below
those during previous servicing missions; (3) the SMOV3B Decons successfully restored the
F170W UV throughput to its nominal value; (4) the daily contamination rates now appear
to have returned to their nominal values. Thus we conclude that our program of delayed
cooldown, pro-active UV monitoring, and frequent decontaminations during SMOV3B were
successful in fully retaining the F170W UV throughput capabilities of WFPC2.
341
342
Koekemoer, et al.
WFPC2 F170W GRW+70d5824 Measured count-rates
PC
WF4
counts/s
200
180
160
140
counts/s
200
WF2
WF3
180
160
140
0
0
5
10
15
20
25
Time since cooldown (days)
1.00
0.95
0.90
WF2
1.05
Normalized counts/s
10
15
20
25
Time since cooldown (days)
WFPC2 F170W GRW+70d5824 Normalized count-rates
PC
WF4
1.05
Normalized counts/s
5
WF3
1.00
0.95
0.90
0
5
10
15
20
25
Time since cooldown (days)
0
5
10
15
20
25
Time since cooldown (days)
Figure 1.
Measured decrease in the observed count-rate of GRW+70d5824 during the first month after cooldown, plotted for each of the cameras separately. The
top panels show the measured count-rates, while the bottom panels show countrates normalized to the pre-SMOV3B values. It can be seen that after each decon,
the throughput is effectively returned to its nominal value.
3.
Lyman-α Monitoring
During a servicing mission, contaminants could potentially settle on the WFPC2 pick-off
mirror which is exposed to the HST hub area. We used F122M F160BW, by themselves
and crossed with F130LP, to monitor the far-UV Lyman-α throughput for any decrease
due to such contaminants. We used the historical GRW+70d5824 data to compare the
contamination rate, and calculated the red-leak-corrected F122M data by subtracting the
count rate measured in F122M+F130LP from that measured in the single F122M filter.
We find that all the data taken in F160BW and F160BW+F130LP are within 2 sigma of
the mean value. For the F122M filter, the SMOV3B count rates are lower than the average
value by 2−3 sigma, but this is in agreement with the historical long-term CTE degradation.
The observed deviations in the red-leak-corrected F122M data are also consistent with those
found after the previous two Servicing Missions. Thus, we conclude that the Lyman-α
throughput shows no significant contamination as a direct result of SM3B.
4.
Photometric Verification
This check was aimed at verifying the photometric accuracy to levels of 1 − 2%. The
standard star GRW+70d5824 was observed in F160BW, F170W, F185W, F218W, F255W,
F300W, F336W, F439W, F555W, F675W, and F814W, in all 4 CCDs. For each filter and
CCD, we first fitted the long-term evolution of the photometric measurements, to account
for CTE. We then expressed the post-SM3B measurements for each filter as a difference
from the mean fitted trend, in units of the standard deviation of the historical data.
Figure 2 shows the statistical distribution of the post-SM3B measurements, normalized
by the 1-sigma error for each filter. The distribution has a mean of 0.34 ± 0.26 sigma, thus
WFPC2 Re-Commissioning After Servicing Mission 3B
343
Figure 2. Statistical distribution of the post-SM3B photometry data points
around the predicted values based on the fits obtained for each filter independently, and normalized by the 1-sigma error bars for each filter.
there is no obvious change in the throughput across SM3B. Since none of the filters deviate
strongly from the mean, we conclude that the response of WFPC2 is essentially unchanged
by SM3B, and that the long-term throughput decline is entirely consistent with the expected
CTE loss.
5.
Flat Field Verification
We examined observations of the bright Earth (“Earthflats”) to test the flat field stability
and to verify that there is no unexpected OTA obscuration. We began with 124 pre-SM3B
Earthflats in F502N, selected those obtained within 7 days after a Decon, and discarded
images with mean PC1 counts < 500 DN and mean WF counts > 3200 DN, and with
bad streaks. The remaining images were combined with the task streakflat to produce a
pre-SM3B flat. A post-SM3B flat was created similarly, and divided by the pre-SM3B flat.
The only changes are on large scales at levels below 0.1−0.2%, well characterized based
on long-term evolution of the camera vignetting (Koekemoer et al. 2002). Other evidence
of small on-going geometric changes is seen in KSPOT images (Casertano and Wiggs 2001).
The pixel-to-pixel fluctuations (over the central 400 x 400 pixels) in the ratio image are
∼ 0.4% r.m.s. for the WFC CCDs and 0.8% r.m.s. for PC1, entirely consistent with photon
noise. No change in chip-to-chip sensitivity is seen in on any levels above ∼ 0.3%, and there
is no evidence of obscuration or other changes in the OTA. Thus we conclude that there
are no significant changes in the flat fields due to SM3B.
6.
PSF Monitoring
Following the procedures used after the previous two servicing missions (see Biretta et al.
1997; Casertano et al. 2000), images of Omega Cen were obtained to characterize the PSF.
344
Koekemoer, et al.
Figure 3.
Radial profile of the PSF in the central region of the WFPC2 PC
image of Omega Cen taken after SM3B. A composite stellar image was created
using about 30 isolated, unsaturated stars near the chip center with the IRAF
task psf , and the radial profile measured using the task radprof . The solid line
shows the best-fit Gaussian model with FWHM=0.066, comparing well with the
pre-SM3B value of 0. 064.
Four dithered images were obtained in F555W and F814W, sub-stepped by one-third of
a pixel to provide a critically sampled PSF. The images were combined using the dither
package. We used the task psf to construct a two-dimensional composite PSF from about
30 bright, unsaturated, isolated stars across the PC chip.
We then used the IRAF task radprof to measure the radial profile of the composite
PSF (see Figure 3). The best-fit Gaussian has a FWHM of 0. 066 ± 0.002, which compares
well with the pre-SM3B value of 0.064 (Biretta et al. 1997; Casertano et al. 2000). We
also measured the composite PSF across the rest of the chip. As previously noted (Krist &
Burrows 1995), the off-center PSFs are more asymmetric due to coma and astigmatism. We
find that this behavior is unchanged in our current measurements. Therefore we conclude
that there are no significant changes in the WFPC2 PSF after SM3B.
7.
Internal Monitoring
The internal monitoring observations for program 8950 were commenced immediately after cooldown, repeated regularly throughout SMOV3B, and included biases, dark frames,
INTFLATs, VISFLATS, and Kellsall-spot (KSPOT) images.
The bias frames were used to determine the read-out noise, using the 16 bias frames
obtained after SM3B, and comparing with a similar number from before SM3B (Figure 4).
The dark current was measured using 40 darks obtained after SM3B and the same number
obtained before. No changes are evident to levels below 0.1 − 0.2 DN. We used the 16
post-SM3B INTFLATs to compare with a similar pre-SM3B dataset, and found no changes
above ∼0.5%, all consistent with well-document small long-term changes in the INTFLAT
lamp. We also compared VISFLATS from before and after SM3B and found that the gain
ratios of all the chips remained constant to within 1%. Finally, we obtained 16 KSPOT
images during SMOV3B, and compared these with a similar pre-SM3B dataset. We found
that the spot locations for each chip agree to within a few mas, fully consistent with the
slow long-term evolution previously discussed by Casertano and Wiggs (2001).
8.
Summary
Overall, WFPC2 appears to be very stable, exhibiting only the minor changes expected due
to known long-term evolution, and there are no significant changes attributable to SM3B.
WFPC2 Re-Commissioning After Servicing Mission 3B
345
Figure 4.
Comparison of the r.m.s. read-noise measured from bias frames taken
before and after SM3B, for the four WFPC2 chips (PC, WF2, WF3, WF4), at
gain=7. There is no significant change at all in the read-noise for any of the chips,
at either gain=7 or 15.
References
Biretta, J., et al. 1997, Instrument Science Report WFPC2-97-09 (Baltimore: STScI)
Casertano, S., et al. 2000, Instrument Science Report WFPC2-00-02 (Baltimore: STScI)
Casertano, S. & Wiggs, M. 2001, Instrument Science Report WFPC2-01-10 (Baltimore:
STScI)
Koekemoer, A. M., Biretta, J., & Mack, J., 2002, Instrument Science Report WFPC2-02-02
(Baltimore: STScI)
Koekemoer, A. M., Gonzaga, S., Lubin, L., Whitmore, B., & Heyer, I., 2002, Technical
Instrument Report WFPC2-02-03 (Baltimore: STScI)
Koekemoer, A. M., et al. 2002, HST Dither Handbook, Version 2.0 (Baltimore: STScI)
Krist, J. & Burrows, C. 1995, Applied Optics 34, 4952
Lubin, L. M., Whitmore, B., Koekemoer, A. M., & Heyer, I. 2002, Technical Instrument
Report WFPC2-02-05 (Baltimore: STScI)
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
Photometry of Saturated Stars in CCD Images
J. Maı́z-Apellániz1
Space Telescope Science Institute, Baltimore, MD 21218, USA
Abstract. We describe here a simple general method to correct for the effects of
saturation in CCD observations of point sources where the A/D saturation level
is significantly lower than the pixel full-well capacity. The method is intended to
complement the results from a PSF-fitting or aperture photometry package and is
somehow different from the one developed by Gilliland (1994) for WFPC2 data. The
current implementation has been tested with WFPC2 and STIS CCD data, yields
uncertainties between 0.02 and 0.10 magnitudes, and can be easily adapted to other
HST or non-HST imaging CCDs.
1.
Introduction
The accuracy of bright-star photometry derived from CCD observations is limited by the
A/D saturation level, the electron capacity of the pixel well, and the possible onset of
non-linearity at high count levels (see, e.g. Howell 2000). The four WFPC2 chips and the
STIS CCD are known to be highly linear below the A/D saturation level (Dolphin 2000b,
Gilliland et al. 1999). Furthermore, the similarity between the full-well capacity and the
A/D saturation level for the GAIN = 4 setting of the STIS CCD allows the detector to
behave in a linear fashion even beyond saturation (Gilliland et al. 1999), since charge is
simply transferred to neighboring pixels (a phenomenon known as bleeding). The behavior
for the GAIN = 1 setting of the STIS CCD or for the WFPC2 chips is rather different, given
that the full-well capacity is between one-and-a-half and four times the A/D saturation level,
leading to a possible substantial loss of counts in the data.
2.
Description of the method
We can distinguish two different types of saturated point sources: weak, where no pixel
reaches its full-well capacity, and strong, where at least one does and bleeding occurs.
Weakly saturated stars retain an approximate circular symmetry while strongly saturated
ones are elongated since bleeding takes place preferentially in the vertical direction (Fig. 1).
Given their pixel sizes and PSFs, the transition between the two regimes is expected to
take place for the GAIN = 7 setting of the WFPC2 and the GAIN = 1 of the STIS CCD
somewhere in between 4 and 7 total saturated pixels (for a single star).
Our method consists of selecting ten or more unsaturated stars in the same chip where
the saturated stars are observed and using them as a reference to correct for the information lost due to saturation. Aperture photometry is performed on the unsaturated stars
blocking the central pixels in a manner which simulates what would happen if the stars
where brighter and saturated. Thus, for each star we obtain its real magnitude and the
magnitudes we would obtain if any number of pixels between 1 and Nmax were saturated
and the central pixels were not used for the calculation. Three choices can be used for the
1
ESA Space Telescope Division
346
Photometry of Saturated Stars in CCD Images
347
Figure 1.
Examples of weakly (left) and strongly (right) saturated stars representative of WFPC2 (GAIN = 7) or STIS CCD (GAIN = 1) observations. The
darkest shade is used to indicate pixels that have reached the full-well capacity,
the intermediate one for those that are saturated but not completely filled, and
the lightest one for not-saturated pixels that have suffered bleeding from pixels
immediately below or above (only vertical bleeding is assumed here). The pixels
possibly affected by vertical bleeding are marked with a cross and those possibly
affected by horizontal (but not vertical) bleeding are marked with a diagonal line.
blocked central region: no bleeding affects adjacent non-saturated pixels, so only saturated
ones are excluded (appropriate for the weak case); bleeding possibly affects vertically adjacent pixels, so those are also excluded; and bleeding possibly affects both vertically and
horizontally adjacent pixels, extending the exclusion to them (see Fig. 1). The magnitude
differences for the reference stars are grouped by number of blocked pixels and within each
group also by analogous geometrical kernels in order to account for the differences in pixel
centering. For each kernel a mean magnitude correction and its dispersion is then calculated. The chip is then scanned for saturated stars, the pixels to be blocked are identified in
each case, aperture photometry without the blocked pixels is performed, and the correction
corresponding to that geometrical kernel is applied. Further corrections (CTE, geometrical
distortions, aperture corrections) can easily be included.
This method has been implemented in an IDL code which currently handles up to
Nmax = 6 and the first two choices for blocked pixels. The other cases will be included in
future versions. The advantages of this method are that no previous knowledge of the PSF
or accurate calibration of the detector (in the form of e.g. some filter-dependent parameterization) are needed. Its main disadvantage is that the results will not be accurate if used
for non-point sources.
3.
Results
In order to test the accuracy of the method, we selected archival observations for which both
short (unsaturated) and long (saturated) exposures of the same field existed. For WFPC2
(GAIN = 7), we used the F555W and F814W 30 Doradus data from program 5114 (P.I.:
Westphal) and for STIS (GAIN = 1) the 50CCD NGC 6752 data from program 8415 (P.I.:
Gilliland). The reference unsaturated photometry was obtained using HSTphot (Dolphin
2000a) for WFPC2 and a custom-made aperture photometry IDL code for STIS. We applied
the method to the stars that had between 1 and 6 saturated pixels in the long exposure
data. For WFPC2, only the PC and WF2 were tested and the stars in R136 were excluded
348
Jesús Maı́z-Apellániz
m
0.50
13.0
13.5
14.0
14.5
m
15.0
15.5
16.0
13.0
14.5
15.0
15.5
16.0
0.50
F814W WF2
0.25
0.25
0.00
0.00
−0.25
−0.25
−0.50
0.50
−0.50
0.50
F555W PC
F555W WF2
0.25
0.25
0.00
0.00
−0.25
−0.25
−0.50
13.0
13.5
14.0
14.5
15.0
15.5
16.0
13.0
13.5
14.0
14.5
m
15.0
15.5
16.0
∆m (sat.−unsat.)
∆m (sat.−unsat.)
14.0
∆m (sat.−unsat.)
∆m (sat.−unsat.)
F814W PC
13.5
−0.50
m
Figure 2. Measured magnitude difference between the saturated and unsaturated exposures as a function of magnitude for the weakly saturated stars in the
WFPC2 observations of 30 Doradus. The saturated magnitudes were obtained using the method described in this poster while the unsaturated ones were obtained
with HSTphot. Stars shown in gray are those with large values of χ2 in HSTphot.
m
16.75
17.00
17.25
17.50
17.75
18.00
18.25
18.50
0.2
0.2
0.1
0.1
0.0
0.0
−0.1
−0.1
−0.2
−0.2
16.75
17.00
17.25
17.50
17.75
18.00
18.25
∆m (sat.−unsat.)
∆m (sat.−unsat.)
STIS CCD
18.50
m
Figure 3. Same as Fig. 2 for the STIS CCD observations of NGC 6752. Here
the unsaturated magnitudes were obtained by aperture photometry.
Photometry of Saturated Stars in CCD Images
∆m / (σ +σ
2
sat
18
−5.0
−2.5
∆m / (σ +σ
0.5
2
unsat
2
sat
)
0.0
2.5
5.0
17
16
WFPC2
mean = 0.06
13
sigma = 1.30
)
0.0
2.5
17
21
21
20
20
19
19
18
18
14
17
13
16
22
17
16
STIS CCD
mean = 0.26
sigma = 1.10
15
15
12
12
11
11
10
10
12
12
9
9
11
11
10
10
9
9
8
8
7
7
14
8
N
14
13
7
7
6
6
5
5
6
6
4
4
5
5
4
4
3
3
2
2
3
3
2
2
1
1
1
0
0
0
−5.0
−2.5
0.0
∆m / (σ2sat+σ2unsat)0.5
2.5
5.0
N
8
13
N
N
5.0
22
15
14
−2.5
18
16
15
−5.0
349
0.5
2
unsat
1
−5.0
−2.5
0.0
2.5
5.0
0
∆m / (σ2sat+σ2unsat)0.5
Figure 4. Measured magnitude difference normalized by its uncertainty for the
WFPC2 (left) and STIS (right) results. The continuous curve shows the expected
distribution. Gray blocks identify the gray data points in Fig. 2.
due to the heavy crowding there. The two blocking options (weak and strong saturation)
were tested. Results for the first one are shown in Figs. 2, 3, and 4 in terms of ∆m, the
difference between the magnitudes derived from the unsaturated and the saturated data.
• The weak-saturation blocking option yields photometric uncertainties of 0.02–0.06
(PC), 0.05–0.10 (WF2), and 0.02–0.04 (STIS) magnitudes.
• The systematic bias generated by the algorithm is very small, of the order of 0.01
magnitudes for STIS and even smaller for WFPC2. In both cases the bias is also
small in comparison with the measured uncertainty.
• No significant differences were found between the weak- and strong-saturation blocking
options. Furthermore, no tendency in ∆m as a function of m is observed in Fig. 2
and only a slight one is observed in Fig. 3. Therefore, the transition between the two
modes of saturation appears to take place around N = 5–6 pixels for the STIS CCD
and possibly at higher values for the WFPC2 chips.
2 + σ2
• The distribution of ∆m/(σsat
unsat ) closely resembles a normal distribution (the
limit N → ∞ of a Student’s t distribution) for both the WFPC2 and STIS results.
The only small divergences are the existence of a small bias in the STIS case and
the presence of an extended left wing in the distribution for both cases. The latter
phenomenon could be caused by the presence of stars with strong saturation or nearresolved binaries. This last interpretation is favored by the location of the stars with
high values of χ2 in the HSTphot output in Figs. 2 and 4.
References
Dolphin, A. E. 2000a, PASP, 112, 1383
Dolphin, A. E. 2000b, PASP, 112, 1397
Gilliland, R. L. 1994, ApJ, 435, L63
Gilliland, R. L., Goudfrooij, P., & Kimble, R. A. 1999, PASP, 111, 1009
Howell, S. B. 2000, Handbook of CCD Astronomy, CUP
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
Updated Contamination Rates for WFPC2 UV Filters
Matt McMaster and Brad Whitmore
Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218
Abstract. Photometric monitoring observations of a white dwarf standard have
been used to update the contamination rates for WFPC2 UV filters. Observations
from April 1994 through May 2002 were used in the analysis. In general, the contamination rates have declined by roughly 50 fits have been made to the data to allow
observers to remove the effects of contamination in their WFPC2 observations.
1.
Introduction
Contaminants within the WFPC2 instrument gradually build up on the cold CCD faceplate
which results in a decrease of the UV throughput. Approximately once per month, these
contaminants are melted off of the faceplate by means of a decontamination procedure
(decon) which restores the UV throughput to its nominal value.
This study extends earlier analyses and provides least-squares fits to the yearly contamination rates. The resulting formulae provide both more accurate and more easily used
corrections for the effects of contamination on WFPC2 UV data.
For a more detailed version of this paper, please see McMaster and Whitmore (2002).
2.
Data
The data used in this study were taken from the F160BW, F170W, F218W, F255W, F336W,
and F439W photometric monitoring observations of the DA3 white dwarf, GRW+70D5824
and can be found at:
http://www.stsci.edu/instruments/wfpc2/Wfpc2 memos/wfpc2 stdstar phot3.html. The
data cover the time period from April 1994 (after the cool down from −76◦ C to −88◦ C) to
May 2002.
3.
Analysis
The first step in the analysis was determining the contamination rate as a function of the
number of days since decontamination (DSD) for each year. The resulting least-square fits
for the period 4/97–4/98 and 4/01–4/02 are shown below for the F170W and F336W filters.
As a matter of clarity, only the 97–98 data are compared with the 01–02; for similar plots
of earlier data, please consult earlier work done on this subject.
As seen on the left hand side of Figure 1, the count rates (DN/sec) for F170W in the
PC at or near 0 DSD for 2001–2002 are generally higher than those for 1997–1998. The
opposite is true for the F336W data (right hand side of Figure 1) where the difference in
count rates for the PC near 0 DSD is lower by 3–7 percent. The F160BW data behave
similarly to the F170W data, while the data for the F218W, F255W, and F439W filters
(not shown) follow the trend of F336W. Since the NUV filters studied (F218W, F255W,
F336W, and F439W) are less affected by contamination than the FUV filters (F160BW
and F170W), this may suggest that some of the long-lived contaminants (i.e., those that
350
UV Contamination Rates
351
Figure 1.
Contamination rates for 1997–1998 (open circles) and 2001–2002 (filled
circles) for F170W (left) and F336W (right).
Figure 2.
Fits to the contamination rates for F170W (left) and F439W (right).
See McMaster & Whitmore (2002).
are not removed by the monthly decons) have outgassed from the PC over time. Also, it
is now known that throughput loss due to CTE has been increasing linearly with time and
it is probably this increase which has caused the count rates to decline in the NUV data.
Since a decrease in count rates is not seen for F160BW and F170W, it can be concluded
that these filters are more affected by the loss of long-lived contaminants than they are by
the increase in CTE loss.
4.
Phase II
The second part of the analysis consisted of plotting the yearly contamination rates as a
function of time and making a least-squares fit to the data. These fits can be used as correction formulae, and are made possible due to the longer temporal baseline. The subsequent
smoothing and averaging of the data provide an improvement of our past technique of listing the yearly rates, where occasional increases were inferred due to observational scatter.
The plots below are a representative sample of the fits.
Note the step-like structure for F170W (Figure 2, left); this is probably an indication
of an effect after Servicing Missions 2 and 3A, MJD = 50490.03 and 51531.16 respectively
(dashed lines in the plots). The asterisks to the left of the dashed lines are the contamination
352
McMaster & Whitmore
rates before the servicing missions (from several months to a few days before), and those
to the right are the contamination rates after the servicing missions (from a few days to
a few months). Note that all of the asterisks to the right of the dashed lines are higher
than those to the left, indicating an increase in the contamination rate just after a servicing
mission. This would explain the step-like structure seen in the plots. Note also that while
the contamination rates immediately after a servicing mission can be quite high (almost
twice as high after SM3A), they return to the general trend soon afterward. Only the solid
points were used in determining the fits. Despite an increase in the contamination rates
shortly after a servicing mission, a simple linear fit appears to represent the data quite well.
An exception to this is the F439W data (Figure 2, right), where due to observational
scatter, it appears that the rates for WF2 and WF3 are increasing. Since this is not thought
to be true, a mean rate (the dashed lines in the plot and the numbers in parentheses in
Tables 1 and 2) is also provided for each of the chips; we recommend using this mean rate.
Table 1.
Fits to Yearly Contamination Rates for PC and WF2
Filter
F160BW
F170W
F218Wb
F255W
F336W
F439Wc,d
Table 2.
PC
Slope
−1.348E-4
−7.440E-5
−5.185E-5
−4.240E-5
−1.084E-5
−3.275E-6
(0.025)
Error
3.042E-5
1.219E-5
6.497E-6
6.803E-6
1.569E-5
7.541E-5
(0.017)
Const
1.499
1.004
0.829
0.462
0.276
0.128
WF2
Slope
−2.985E-4
−1.659E-4
−1.519E-4
−8.050E-5
−6.245E-5
+9.022E-6
(0.140)
Error
4.821E-5
2.144E-5
1.926E-5
2.212E-5
3.912E-5
(0.050)
Fits to Yearly Contamination Rates for WF3 and WF4
Filter
F160BW
F170W
F218W
F255W
F336W
F439Wc,d
a
Consta
0.857
0.562
0.478
0.246
0.081
0.029
Consta
1.344
1.038
0.863
0.459
0.234
0.062
WF3
Slope
−2.401E-4
−1.755E-4
−1.387E-4
−6.093E-5
−5.005E-5
+2.874E-5
(0.092)
Error
2.210E-5
1.934E-5
3.091E-5
3.958E-5
1.862E-5
1.281E-5
(0.028)
Const
1.285
0.835
0.795
0.387
0.214
0.069
WF4
Slope
−2.443E-4
−1.457E-4
−1.651E-4
−7.822E-5
−6.210E-5
−4.115E-6
(0.063)
Error
4.474E-5
2.106E-5
1.673E-5
1.750E-5
2.597E-5
2.647E-5
(0.044)
The value of the percentage throughput loss at MJD = 49500 (May 28, 1994)
Due to a lack of sufficient data, the values for WF2 are an average of those for WF3 and WF4 and are
recommended when determining the contamination correction for this filter/chip combination.
c
A plus (+) sign indicates where the rate appears to be increasing, though this is probably due to observational
scatter rather than a larger number of contaminants falling on the chips.
d
The numbers in parentheses are the mean values of the yearly contamination rates (in percent throughput
loss per days since decontamination) and should be used when correcting for contamination effects in this
filter (i.e., with Equation 2).
b
UV Contamination Rates
5.
353
Correction Formulae
The following formula can be used in correcting data taken in the F160BW, F170W, F218W,
F255W, and F336W filters
COUN T Scorr =
COUN T Sobs
1.0 − (Slope(M JD − 49500) + Constant) (DSD/100)
(1)
where COUNTS obs is the count rate, in DN/second, measured from the image; Slope
is the percentage throughput loss per day per MJD; MJD is the Modified Julian Date
which can be determined from the header parameter EXPSTART; Constant is the value
of the throughput loss per day at MJD = 49500; and DSD is the number of Days Since
Decontamination which can be found on the WFPC2 Decontamination Date web page:
http://www.stsci.edu/instruments/wfpc2/Wfpc2 memos/wfpc2 decon dates. html.
For an object observed in the F439 filter, the correction formula would be:
COUN T Scorr =
COUN T Sobs
1.0 − (M ean (DSD/100))
(2)
where COUNTS obs is the count rate, in DN/second, measured from the image; Mean is the
average value of the yearly contamination rate in percentage throughput loss per days since
decontamination (the number in parentheses in Tables 1 and 2); and DSD is the number
of days since decontamination.
6.
Red Leaks
Some of these UV filters have substantial red leaks which have not been accounted for in
the photometric monitoring data. In fact, red leaks can account for a significant percentage
in the overall count rate for an observation in the FUV filters and strictly speaking, the
corrections presented here are valid only for an object with the same spectral distribution as
GRW+70D5824 (a white dwarf). The red leak is minimal for this case since the star is a very
blue DA3 star. Table 3.13 in the WFPC2 Instrument Handbook indicates that roughly 6%
of the light comes from the red leak. We caution users about using the contamination rates
presented here for very red sources. SYNPHOT (the Space Telescope Science Institute’s
SYNthetic PHOTometry package) can be used to determine more realistic corrections for
these objects.
7.
Conclusions and Recommendations
Photometric monitoring observations from April 1994 to May 2002 have shown a decrease in
the rate of contamination in the UV filters, with F160BW and F170W showing the steepest
drop off. While it appears that the data show a correlation with the Servicing Missions,
especially in the F170W filter, a least squares fit seems to be adequate. These fits, listed in
Tables 1 and 2, along with the equations given above, can be used to correct for the effects
of contamination in WFPC2 UV photometric data.
The corrections presented here supersede those given in previous work and when using
SYNPHOT to account for contamination in the obsmode parameter
(e.g., wfpc2,2,a2d7,f170w,cont#MJD).
References
McMaster, M. & Whitmore, B. C. 2002, Instrument Science Report WFPC2 02-07 (Baltimore: STScI)
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
Toward a Multi-Wavelength Geometric Distortion Solution for
WFPC2
V. Kozhurina-Platais, S. Casertano, A. Koekemoer
Space Telescope Science Institute, Baltimore, MD 212128
Abstract. The inner calibration field of omega Cen has been used to examine the
geometric distortions of WFPC2 as a function of wavelength. We used multiple
observations of this field shifted in the range of 0.25” to 35” and exposed through
F300W, F555W, and F814W filters. All observations have been reduced using the
IRAF/PSF/ALLSTARS package which yields the standard error of a single position
0.08, 0.05 to 0.06 pix, depending on the filter. The master catalog of all positions
was used to obtain new distortion coefficients by differential method. Although
the chosen set of observations do not allow us to find the non-perpendicularity of
coordinate axes (skew), they provide clues on the scale change from filter to filter
and within uncertainties confirm the values of previously found distortion coefficients.
Future improvements will include more observations with rotated fields of stars and
at selected epochs.
1.
Introduction
The geometric distortion of HST WFPC2 has received repeated attention in terms of astrometric calibrations (e.g., Holtzman et al. 1995, Anderson 2001, Casertano & Wiggs 2001,
Anderson & King 2003). The goal of astrometric calibration of HST WFPC2 is not only
to obtain a world coordinate system (WCS) free of distortion down to a precision level of
1 mas, but also to obtain a contiguous and seamless image over the field of view of the
entire CCD mosaic. To achieve this goal, coordinates of individual CCD chips must be
translated into the WCS and fluxes stacked up, employing point spread function fitting. If
the knowledge of the PSF across the whole CCD mosaic frame, precise CCD mosaic metrology, and a distortion-free WCS are neglected, image resampling and stacking will produce
an unwanted blurring of the objects of scientific interest.
In this study, the inner calibration field of ω Cen exposed through filters F300W, F555W
and F814W has been used to examine the geometric distortion of WFPC2 as a function
of wavelength. Although the chosen set of observations do not allow us to find the nonperpendicularity of the coordinate axes (skew), it provides clues on the scale change from
filter to filter, and, within uncertainties, confirms the values of previously found distortion
coefficients. Future improvements will include more observations with rotated fields of stars
at different selected epochs.
2.
Data Set and Reductions
The WFPC2’s wavelength-dependent geometric distortions have been computed from a series of overlapping images of the globular cluster ω Cen, taken in three WFPC2 bandpasses—
F300W, F555W and F814W—over a single five hour period in June 13, 1997. Each set of
the F330W, and F814W images consists of two central pointings with a 0. 25 shift and four
outer pointing pairs with 35 shifts (with an offset of 0.25 in each pair). The set of F555W
images has offsets at 35, 15 , and 0. 25. The IRAF/PSF/ALLSTARS tasks were used to
354
Multi-Wavelength Distortion Solution for WFPC2
355
Figure 1. Contour plot of the composite PSF as a function of position on the
WF2 chip (filter F555W). The panels, from left to right, represent the observed
PSF in the center, lower left, lower right, upper right and upper left of the chip,
respectively.
obtain the star positions on 136 CCD chips for all 34 images. The PSF fitting implements
the empirical PSF fitting as a sum of an analytical function and look-up tables of residuals
between the actual PSF and the fitting function (Stetson 1987, Stetson, Davis, & Crabtree
1990). An analytical Moffat function and six lookup tables were chosen to provide the best
representation of the undersampled image cores and extended wings of the PSF in WFPC2’s
data in order to calculate the image centers. An IRAF script was written to automate the
detection of objects, selection of PSF stars, fit the analytical PSF to detected images and
provide an output with X, Y and instrumental magnitudes for all four WFPC2 CCD chips.
About 50–100 bright, unsaturated and isolated stars, well distributed over the chip, were
selected to generate a template PSF. If the normalized standard scatter in the fit of an
analytical PSF to the observed PSF stellar profile exceeded 0.020, the stars selected for the
PSF template were examined interactively, poor images were deleted and the PSF fit was
repeated until a normal scatter of 0.02 was achieved.
Figure 1 presents the contour plots of the composite PSF as a function of position on
the WF2 chip. It is well known that the observed PSF varies as a function of position on
WFPC2 chips, with the off-center PSFs being noticeably more asymmetric due to coma and
astigmatism. The quality of the PSF fit for each CCD chip was monitored by examining χ2
as a function of X, Y and magnitude. The stars with poor images (high χ2 , large magnitude
errors) were rejected from the subsequent astrometric reductions, and only stars with good
measurements (Figure 2) were used to calculate the geometric distortions. On average,
there are over 30,000 such stars per filter.
3.
The Model of Geometric Distortion
The geometric distortion model for WFPC2 has been described by Holtzman et al. (1995),
which, is essence, is a polynomial transformation between observed and distortion-free coordinates. Recently, Casertano & Wiggs (2001) attempted to improve the geometric solution
for WFPC2 using data which were specifically designed to provide good sampling in all
regions of the field of view. Here we use the Holtzman formalism with Casertano’s modification to calculate the geometric distortions of WFPC2 at three different wavelengths.
The geometric correction is based on a bicubic polynomial which transforms the pixel
coordinates (x, y) of each WFPC2 chip to geometrically corrected coordinates (Xg ,Yg ),
matching the scale and orientation of the PC1 chip. The transformation operates on a
single 800×800 CCD image in a coordinate system with its origin at the center of a CCD,
i.e., X = x − 400 and Y = y − 400:
356
Kozhurina-Platais, et al.
Figure 2. The distribution of errors in magnitude as a function of magnitude
for the WF2 chip. Only stars denoted by the bold dots are kept in the astrometric
reductions.
Figure 3.
The distortion correction maps for the F300W filter and all four chips.
The grid is depicted using raw pixel coordinates with the zeropoint at 400,400. The
size of the longest arrow corresponds to ∼ 6 pixels for PC1 (upper left panel), and
twice as much for the remaining panels (WF2, WF3, WF4). The distortion maps
for the other two filters are nearly identical.
Multi-Wavelength Distortion Solution for WFPC2
357
Figure 4. Differences in the distortion correction in the sense “F555W–F814W.”
The small amount of these differences along with a fairly random pattern change
from chip to chip indicate that the differences are negligible.
Xg
=
C1 + C2 X + C3 Y + C4 X 2 + C5 XY + C6 Y 2 + C7 X 3 + C8 X 2 Y + C9 XY 2 + C10 Y 3
Yg
=
D1 + D2 X + D3 Y + D4 X 2 + D5 XY + D6 Y 2 + D7 X 3 + D8 X 2 Y + D9 XY 2 + D10 Y 3
where the different sets of coefficients C and D are defined for each CCD/detector. Details
on actual calculations can be found in the papers listed above.
4.
WFPC2 Geometric Distortion
Holtzman et al. (1995) point out that there should be a small but perceptible difference in
the amount of expected distortion as a function of wavelength. The required corrections to
account for cubic distortion (see coefficients C4 –C10 and D4 –D10) are displayed in Figure 3.
Nominally they all look the same, however, the differences can be quantified by examining
the maximum distortion at the location X = ±375 and Y = ±375 pixels. Cubic distortion
in the F300W filter is definitely larger than in the F555W filter. The average increase in
distortion for the F300W filter is 3%, or 0.37 PC1 pixels. The cubic distortion is essentially
identical in the F555W and F814W filters. This is illustrated by Figure 4 which shows the
differences in cubic distortion in the sense “F555W–F814W” for the WFPC2 chips. The
longest vector is 0.12 PC1 pixels or 5.4 mas, which is comparable with the image centroid
precision. It appears that the existing sets of geometric distortion coefficients cannot fully
account for the cubic distortion in the F300W filter. More CCD frames with dense stellar
fields like ω Cen are required to obtain more accurate distortion coefficients for this filter.
5.
Conclusions
The traditional PSF fitting techniques have been applied to obtain working sets of x, y
coordinates for an assorted set of HST WFPC2 frames with the globular cluster ω Cen
358
Kozhurina-Platais, et al.
in three bandpasses. We used a bicubic polynomial model to derive geometric distortions
in F300W, F555W, and F814W filters. The main conclusion of this study is that existing
sets of geometric distortions do not fully represent distortions in the F300W filter. The
differences in such distortions between the F555W and F814W filters are only at the level
of a few mas, which is comparable with the attained precision of image centering. Since the
chosen set of observations have not been rotated with respect to each other we could not
find the non-perpendicularity of coordinate axes—a skew parameter.
Acknowledgments. We are grateful to B. Whitmore for the support and keen interest
in this study. V.K.-P. thanks Ronald Gilliland for helpful comments and suggestions at
various stages of this project.
References
Anderson, J. 2001, American Astronomical Society, DDA meeting, No. 32, poster 04.11
Anderson, J. & King, I. 2003, PASP(in print)
Casertano, S. & Wiggs, M. 2001, in WFPC2 Instrument Handbook V.6.0 (Baltimore:
STScI)
Holtzman, J., Hester, J. J., Casertano, S., Trauger, et al. 1995, PASP, 107, 156
Stetson, P. B. 1987, PASP, 99, 191
Stetson, P. B., Davis, L. E., & Crabtree, D. R. 1990, in ASP Conf. Ser. 8, CCDs in
Astronomy, ed. G. H. Jacoby, 289
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
Charge Transfer Efficiency for Very Faint Objects and a
Reexamination of the Long-vs.-Short Problem for the WFPC2
Brad Whitmore and Inge Heyer
Space Telescope Science Institute, Baltimore, MD 21218
Abstract. An analysis of WFPC2 observations of Omega Cen and NGC 2419 leads
to the following results.
1. The correction formula developed by Whitmore, Heyer, and Casertano (1999; hereafter WHC99) does a reasonable job of correcting for CTE loss down to extremely low
count levels. There is no sharp cutoff to the detection threshold for very faint stars.
2. A comparison of the WHC99 formula with the Dolphin (2000; hereafter D00) formula
shows reasonable agreement for bright and moderately bright stars, with the D00
formula giving better results. However, at very faint levels, the D00 formula overestimates, and the WHC99 formula underestimates, the correction by tens of percent.
Note: Our current recommendation is to use the new Dolphin 2002 (hereafter D02)
formula for CTE loss correction, which is an improvement on the D00 formula.
3. A reexamination of the long-vs-short nonlinearity shows that the effect is very small (a
few percent) or nonexistent for uncrowded fields, with less than ∼1000 stars per chip.
However, for crowded fields, with ∼10,000 stars per chip, apparent nonlinearities of
tens of percent are possible. We believe this is due to difficulties in measuring the sky
values for the short exposures.
4. Preflashing may be a useful method of reducing the effects of CTE loss for certain
observations (moderately bright objects on very faint backgrounds), but the effects of
added noise and longer overheads limit its effectiveness.
5. The detection thresholds for typical broad band observations have been reduced by
∼0.1–0.2 mag in the ∼9 years since WFPC2 was launched. For worst-case observations
(F336W) the effect is currently ∼0.5 magnitudes.
1.
A Comparison of WHC99 and D00 CTE Correction Formulae
The Charge Transfer Efficiency (CTE) of the WFPC2 is declining with time, as first shown
by Whitmore (1998). By February 1999, for worst case observations (very faint objects
on very faint backgrounds) the effect of CTE loss from the top of the chip had reached
levels of ∼40% (WHC99). For more typical observations, CTE loss was 5–6% by this date.
Formulae have been derived by various groups to attempt to correct for the effects of CTE
loss (WHC99; Saha, Lambert & Prosser 2000; D00). Various techniques are currently being
studied to minimize the effect of CTE loss on new HST instruments (ACS and WFC3
include preflash capabilities).
D00 examines CTE loss by comparing WFPC2 observations with ground based observations of Omega Centauri and NGC 2419, using a baseline through March 2000, roughly a
year longer than available for an earlier study by WHC99. In general, D00 finds good agreement with the WHC99 results, and the longer baseline and more extensive data set used
359
360
Whitmore, et al.
Figure 1.
This shows the ratio of counts between a 14 sec and 100 sec exposure
for stars in Omega Cen vs. the Y position for stars on all three WF chips. The
raw values (filled circles) fall below a ratio of 0.14 due to CTE loss. The different
panels are for different target brightness, as selected on the 100 sec exposure
and described by the labels. The filled squares show the values corrected using
the Whitmore, Heyer & Casertano (1998) formula while the filled triangles show
the values corrected using the Dolphin (2000) formula. Note that neither of the
two correction formulae is very good for the faintest stars (∼5 DN on the short
exposure). Also note that the extrapolation of the raw data to Y = 0 (the sloped
line) is consistent with the predicted value of the throughput ratio based on the
exposure times, hence the long-vs.-short anomaly is not a problem for this data
set (from Dolphin 2002). Dolphin has recently updated his formulae to improve
the agreement for the faint stars.
by D00 result in less scatter in the residuals. In particular, D00 finds similar corrections to
within a few hundredths of a magnitude in almost all cases.
The details of this study can be found in Whitmore and Heyer (2002), available at
http://www.stsci.edu/instruments/wfpc2/Wfpc2 isr/wfpc2 isr0203.html
2.
A Reexamination of the Long-vs.-Short Anomaly
Suggestions of a long-vs-short photometric nonlinearity between short and long exposures
was first discussed by Stetson (1995), and then examined in more detail by Kelson et al.
(1996) and Casertano and Mutchler (1998). More recent studies, however, have found less
evidence for the existence of the “long-vs.-short” problem (Dolphin 2000). Dolphin (2000)
suggest that the apparent long-vs.-short anomaly may be caused by overestimating the
value of the sky by a few electrons in the shorter exposure.
It has been suggested that the long-vs-short anomaly may be caused by difficulties in
measuring the sky on very crowded images. We can address this question by separating the
measurement of the local sky and the measurement of the object. The top left in Figure 2
shows a measured slope consistent with the normal CTE effect. The intercept at Y = 0 is
0.00985 ± 0.00012, very near the theoretical value of 0.01. Hence, there appears to be little
or no long-vs-short problem for the ratio of the object observations. However, the ratio of
CTE for Very Faint Objects and the Long-vs.-Short Problem for the WFPC2
361
Figure 2.
The ratios of the counts in the 10 sec exposure to the counts in the
1000 sec exposure for the crowded NGC 2419 field (left side). The top panel shows
the ratios in the object apertures using a constant sky value of 0.30 DN for the
short exposure and 30 DN for the long exposure. The bottom panel shows the
ratio for the local sky measurements. The sky ratio is well above the predicted
value of 0.01, demonstrating that the long-vs.-short effect is caused by the sky
measurements rather than the object measurement. The dashed line shows a least
squares fit for the object ratio data.
the local sky values (as measured in an annulus between 5 and 10 pixels) shows a obvious
tendency to be above the theoretical value of 0.01, with a median value of 0.0135. This
appears to be the cause of the long-vs.-short anomaly in this data set; the sky values in the
10 sec exposure appears to be overestimated by about 35%, relative to the predicted value
based on the sky measurement of the 1000 sec exposure. See Whitmore et al. (2002) for
more details and a possible explanation for this discrepancy.
3.
Detection Threshold as a Function of Time
The CTE correction formulae can be used to estimate the evolution of the degradation
in S/N for stars due to CTE loss as a function of time, and hence allow us to determine
how the detection threshold (defined at S/N = 10) evolves. We show an estimate of S/N
for a 1000s exposure using the F555W filter for a typical observation with a background
of 50 electrons. We assume aperture photometry with object/inner-sky/outer-sky values
of 2/5/10 pixels. Noise components are read noise (5 electrons/pixels), Poisson noise, and
the uncertainty in the CTE formula (25%, based on a comparison of the WHC99 and D00
formulae).
362
Whitmore, et al.
Table 1 shows the Signal-to-Noise at various magnitudes for a 1000s exposure using
the F555W filter at the top of the chip (i.e., y = 800 pixels). This represents the worst
case. The CTE effects are essentially half as large for a random distribution of objects (with
mean Y = 400 instead of Y = 800). Table 2 shows the limiting magnitudes for the V and
U bands for a 1000s exposure.
Table 1.
Signal-to-Noise at Various Magnitudes for a 1000s Exposure
V mag
22.0
23.0
24.0
25.0
25.5
26.0
27.0
Table 2.
04/24/1994
103.1
67.4
39.9
21.2
14.7
10.0
4.4
2005
101.5
64.4
36.5
18.3
12.4
8.2
3.4
2010
100.8
63.0
35.1
17.2
11.5
7.5
3.0
Limiting Magnitudes for Two Different Filters in 1000 Seconds
Date
04/24/1994
2000
2005
2010
4.
2000
102.2
65.7
38.0
19.5
13.4
9.0
3.8
V-Mag
26.00
25.87
25.76
25.67
delta
0.13
0.24
0.33
U-mag
23.16
22.81
22.59
22.40
delta
0.35
0.57
0.76
To Preflash Or Not To Preflash...
CTE loss can be reduced by increasing the background, hence filling some of the traps
before the target reaches them. One can artificially enhance the background by adding a
preflash. This removes the dependence on CTE correction formulae, which introduce their
own uncertainties. The problem with this approach is that it also adds noise. Figure 3 shows
a calculation based on the WHC99 correction formula, assuming a very low background for
the raw image (0.1 electron, appropriate for a very short exposure, a narrow-band exposure,
or an exposure in the UV) versus an exposure which has been preflashed with 25 electrons.
The ratio of the S/N for the preflashed image versus the raw image is plotted vs. the Log
of the target brightness. The three curves show the effects for a star near the bottom of
the chip (X = 400, Y = 100, where the preflash is never an advantage since CTE loss is
low and the preflash adds noise), near the center of the chip, and near the top of the chip
(where the preflash is an advantage for the brighter targets). Therefore, for fainter targets
there will be nothing gained by preflashing, while for brighter targets, the amount of gain
will depend on the location on the chip.
An additional factor to consider is that the preflash exposure requires 2–5 minutes
of overhead per exposure (depending on the necessity of filter changes and read-out, and
length of the preflash exposure). In some cases these internal flat preflashes can be taken
during occultations by the Earth, hence not affecting the effective integration time. The
effect of the overhead time has not been included in the calculation since the exposure time
for a given target will vary. However, for the typical case of two 1000 sec exposures per
CTE for Very Faint Objects and the Long-vs.-Short Problem for the WFPC2
363
Figure 3. Calculation based on the WHC99 correction formula, assuming a very
low background for the raw image, versus an exposure which has been preflashed
with 25 electrons.
orbit, with the first preflash being taken during the occultation, the result would generally
be that only two 900 sec exposures would fit into one orbit, resulting in a decrease of ∼5% in
the S/N, and ∼0.1 mag in detection threshold (assuming the noise is dominated by Poisson
statistics; the effective change would be smaller if other sources of noise dominate). For
shorter exposures the effect can be much larger, especially since a smaller percentage of
the preflash exposures can be taken during the occultation. For example, for 3 minute
exposures the S/N would be diminished by ∼40%, offsetting any advantage of a preflash
over nearly the entire chip, even for best-case-scenarios.
References
Casertano, S. & Mutchler, M. 1998, “The Long vs. Short Anomaly in WFPC2 Images,”
Instrument Science Report WFPC2 98-02 (Baltimore: STScI)
Dolphin, A. E. 2000, PASP, 112, 1397 (D00)
Dolphin, A. E. 2002, private communication,
http://www.noao.edu/staff/dolphin/wfpc2 calib/
Kelson, D. D., et al. 1996, ApJ, 463, 26
Saha, A., Labhardt, L., & Prosser, C. 2000, PASP, 112, 163
Stetson, P. 1995 (unpublished, reported in Kelson et al. 1996)
Whitmore, B. C. 1998, “Time Dependence of the Charge Transfer Efficiency on the WFPC2,”
WFPC2 TIR 98-01 (Baltimore: STScI)
Whitmore, B. & Heyer, I. 2002, “Charge Transfer Efficiency for Very Faint Objects and a
Reexamination of the Long-vs.-Short Problem for the WFPC2,” Instrument Science
Report WFPC2 02-03 (Baltimore: STScI)
Whitmore, B., Heyer, I., & Casertano, S. 1999, PASP, 111, 1559 (WHC99)
Part 5. Other Instruments
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
Optical Interferometry with HST /FGS at V > 15
E. Nelan and R. Makidon
Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD, 21218
Abstract. The Hubble Space Telescope’s Fine Guidance Sensor FGS 1r has been
used to observe cool white dwarf stars with apparent magnitudes that are near the
FGS’s faint limit. We had expected to discover that about 10% of these stars are binary white dwarf systems. We also expected the binaries to have angular separations
much larger than the size of the FGS white light fringes, making them easy to resolve. Although we did find about 10% percent of the stars to be binaries, most have
angular separations less than 25 mas, well below the HST diffraction limit. Instead
of two widely separated fringes, we observed fringes that displayed subtle differences,
in amplitude and morphology, from those of point sources. A major complication for
our program was the need to address and remove the effects of the detector’s dark
current, which for the faintest targets contributed up to 40 percent of the counts.
This paper outlines the process we employed to retrieve the science from the data.
1.
Introduction
In Cycle 10 we used HST to observe cool white dwarf (WD) stars in an effort to discover
binary systems composed solely of white dwarfs, hereafter referred to as double degenerate
(DD) systems. We hoped to identify systems with separations suggesting orbital periods
less than 25 years. Such binaries would be ideal candidates for follow up studies for deriving
orbital elements, and ultimately the mass of each component. This would facilitate a more
comprehensive calibration of the WD mass-radius relation and cooling curve for a variety
of WD core and envelope compositions which are currently calibrated by only 4 WDs with
dynamically measured masses. Based upon the incidence of binarity and the distribution of
periods among G dwarf stars in the solar neighborhood (Duquennoy & Mayor 1991), and
allowing for the expectation that systems with initial separations less than about 2 A.U.
would evolve into unresolvable short period systems due to the orbital shrinkage expected to
result from common envelope evolution (Iben & Livio 1993), we anticipated that about 10%
of the WDs in our sample would be resolved as DDs with separations larger than 100 mas
(all of the stars in our sample are within 50 pc). To optimize our prospects for resolving a
DD, we restricted our target list to include only WDs cooler than about 9000 K since any
companion could not be much cooler, and hence not much fainter than the primary.
Although we expected to discover DDs with separations wide enough to be resolved
by WFPC2, the superior angular resolution achievable (8 mas) with the Fine Guidance
Sensor 1r (FGS 1r) made it the instrument of choice in the event that binaries with small
separations, or unfavorable projection angles might be encountered. However, the FGS 1r
faint limit at V = 17 is set by the instrument’s dark current. Many of the targets in our
observing program would be fainter than V = 16, implying that the contribution from
the dark counts would be comparable to that from the source. However, this presents an
analysis problem only for binary systems with projected angular separations less than the
size of the FGS white light fringe packet (50 mas). Few such systems were anticipated to
be encountered.
367
368
2.
Nelan and Makidon
Observations
We used FGS 1r in its high angular resolution observing mode (Transfer mode). After
acquiring the star, the instrument’s instantaneous field of view (IFOV) is scanned across
the star’s photocenter along a 45◦ diagonal path (in FGS X, Y detector space). The length
of the scan is specified by the observer. We used scan lengths of 1 asec and step sizes of
0.8 mas. The total number of scans (typically 50) was set by the length of the observing
window (we specified the maximum number possible). Each step in the scan is a 25 msec
integration. The data contains the 40 Hz measurement of the location of the FGS 1r IFOV
and the photon counts from the instrument’s four photomultiplier tubes (PMTs). These
data are used to reconstruct the target’s observed fringes along the FGS 1r X, Y axis.
For a detailed description and discussion of the FGS, please consult the FGS Instrument
Handbook and the HST Data Handbook for the FGS, both of which can be obtained from
the FGS web site (follow the links from http://www.stsci.edu/hst/fgs.)
3.
Analysis
Data from the individual scans are auto correlated, binned, and co-added to produce a pair
of high signal to noise ratio (SNR) interferograms for the observation, one for each of the
instrument’s two orthogonal baselines. The photon noise in the individual scans for stars
fainter than V = 15.5 makes the auto correlation unreliable as it injects false jitter. In
such cases the scans are correlated using the HST guide star data (from the guiding FGSs)
prior to being binned and co-added. After co-adding, the interferograms are smoothed by
application of a cubic spline.
If an object is resolved by FGS 1r to be a binary system, the observed fringes will
depart from those obtained from a point source. To analyze binary star observations, one
finds the best fitting linear superposition of point source fringes which have been scaled and
shifted to represent the relative brightness and separation of the two sources comprising
the binary system. The data along the FGS X-axis are analyzed independently from those
along the Y -axis, however the relative brightness of the components must agree.
The point source fringes are available from observations of standard stars made as part
of the HST /FGS calibration program. For the best SNR, only bright standards are observed
(V < 10, generally). However, when observing a star as faint as V = 16.4, the instrument
dark current contributes about 40 percent of the counts accumulated by the detector. Since
this is incoherent with the light from the star, the fringe visibility is reduced, as can be seen
in Figure 1 which compares the observed fringes of two point sources, one at V = 11.5 and
the other at V = 16.4. Clearly, before any headway can be made in analyzing the data from
scans of faint stars, the effect of the dark counts must be addressed. Rather than trying to
remove the dark counts from the faint star data, it is better to use the data from a bright
calibration source to model the fringes of a faint point source. In other words, we scale the
observed photometric counts of the bright standard down to the level they would be had the
star been as faint as the science target, we add in the dark counts, and then regenerate the
fringes. Figure 2 shows the bright star fringe after being adjusted to simulate a V = 16.4
point source, and compares it to the observed fringes of the faint source.
4.
Results
In this section we show how our analysis of the observed fringes of the faint, V = 16.2 star
WD1818+126 has allowed us to resolve it as a binary system. Bergeron et al. (1997) note
that the object’s spectra is best modeled as a composite from two stars, a DA and a cooler
DC, to explain the shallow depth of the hydrogen lines. They also note that the inferred
Optical Interferometry with HST/FGS at V > 15
Figure 1.
Comparison of fringes from two point sources, one at V = 11.5, the
other at V = 16.4. The dark counts have reduced the amplitude of the faint star’s
fringes.
Figure 2. The bright star fringe, after being adjusted as if it can from a faint
source, is compared to the observed fringes of a faint source.
369
370
Nelan and Makidon
Figure 3.
Comparison of an adjusted V = 11.5 fringe to that of WD1818+126.
This shows that the white dwarf is not a point source.
surface gravity (derived from its luminosity and temperature) is too low, implying either an
unusually low mass (large radius) or that it is a binary system composed of white dwarfs.
Figure 3 compares the observed fringes of WD1818+126 to those from a bright point
source which has been modeled as a faint star. Clearly WD1818+126 is not a point source.
Figure 4 compares the fit of a model which uses the adjusted calibration fringe to determine
the separation and relative brightness of the components along the FGS X-axis. Figure 5
shows the model’s fit along the FGS 1r Y -axis. The object is a wide binary with separation
of 176 mas, but its position angle was nearly aligned with the FGS 1r Y -axis, thereby projecting a small separation along the X-axis and a large, easily resolved separation along the
Y -axis. The best fitting model yields an angular separation along (X, Y )=(15.7, 173.7) mas,
with the secondary being 0.94 magnitudes fainter than the primary. Note that the faint
companion is nearly V = 17, very close to the FGS 1r’s limiting sensitivity.
5.
Conclusions
We had expected to detect only wide DDs, such as WD1818+126. However, the wide
systems turned out to be fewer in number than the close (sep<30 mas) systems. The
chance alignment of WD1818+126 as projected onto the FGS 1r yields a wide, easily resolved
separation along the Y -axis and a small, sub-diffraction-limited separation along the X-axis.
The Y -axis result clearly established the binary nature of the object. This in turn firmly
establishes the reliability of the analysis process which resulted in the X-axis detection.
Therefore, we conclude that performance of FGS 1r is not significantly impaired near the
instrument’s faint limiting magnitude, and that sub-diffraction-limited angular resolution
remains viable, provided the dark counts are properly modeled.
The prevalence of binary white dwarf systems with projected angular separations suggesting physical separations less than 1 A.U. indicates that common envelope evolution does
not necessarily result in unresolvable short period binary systems with small physical separations. This has implications for the range of values that can be taken on by the common
envelope efficiency parameter αCE . Finally, our program to identify DD systems suitable for
follow up studies for dynamical mass determinations has yielded several good candidates.
Optical Interferometry with HST/FGS at V > 15
Figure 4.
The FGS 1r X-axis fringe of WD1818+126 is best modeled as a composite of two point sources with a projected angular separation of 15.7 mas and a
magnitude difference of 0.94.
Figure 5.
The wide projected angular separation of WD1818+126’s components
along the FGS 1r Y -axis leave no doubt that this is a binary system. The measured
magnitude difference agrees well with the result from the analysis of the X-axis
data.
371
372
Nelan and Makidon
It is important for HST to continue observing the longer period systems (P > 15 years) in
order to establish an accurate baseline for a hand off to SIM in the event that SIM’s mission
lifetime is otherwise too short for deriving the orbital elements of these systems.
Acknowledgments. This work is based upon observations made with the NASA/ESA
Hubble Space Telescope, which is operated by the Space Telescope Science Institute, of the
Association of Universities for Research in Astronomy, Inc. under NASA contract NAS526555. This work is supported through grant GO-09169 (E. Nelan, P.I.)
References
Bergeron, P., Ruiz, M. T., & Leggett, S. K. 1997, ApJS, 108
Duquennoy, A. & Mayor, M. 1991, A&A, 248,485
Iben, I. & Livio, M. 1993, PASP, 105, 13731
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
The Optical Field Angle Distortion Calibration of HST Fine
Guidance Sensors 1R and 3
B. McArthur, G. F. Benedict1 and W. H. Jefferys
Astronomy Department, University of Texas, Austin, TX 78712
E. Nelan
Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218
Abstract. To date five OFAD (Optical Field Angle Distortion) calibrations have
been performed with a star field in M35, four on FGS 3 and one on FGS 1, all analyzed by the Astrometry Science Team. We have recently completed the FGS 1R
OFAD calibration. The ongoing Long Term Stability Tests have also been analyzed
and incorporated into these calibrations, which are time-dependent due to on-orbit
changes in the FGS. Descriptions of these tests and the results of our OFAD modeling are given. Because all OFAD calibrations use the same star field, we calibrate
FGS 1 and FGS 3 simultaneously. This increases the precision of our input catalog, particularly in regards to proper motion, resulting in an improvement in both
the FGS 1 and FGS 3 calibrations. Residuals to our OFAD modeling indicate that
FGS 1 will provide astrometry superior to FGS 3 by ∼ 20%. Past and future FGS
astrometric science supported by these calibrations is briefly reviewed.
1.
Introduction
The largest source of error in reducing star positions from observations with the Hubble
Space Telescope (HST ) Fine Guidance Sensors (FGSs) is the Optical Field Angle Distortion
(OFAD). Description of previous analyses can be found in McArthur et al. (1997), Jefferys
et al. (1994), and Whipple et al. (1994,1996). The precise calibration of the distortion
can only be determined with analysis of on-orbit observations. The Long Term STABility
tests (LTSTAB), initiated in fall 1992, are an essential component of the OFAD calibration,
and provide information on temporal changes within an FGS. They also provide indicators
that a new OFAD calibration is necessary. This paper reports the results of the continuing
OFAD calibration of FGS 3 and a new OFAD calibration for FGS 1, including the LTSTAB
tests. Past astrometry produced by FGS 3 and future astrometric results anticipated from
FGS 1 are briefly reviewed.
2.
Motivation and Observations
A nineteen-orbit OFAD (Optical Field Angle Distortion) was performed in the spring of 1993
for the initial on-orbit calibration of the OFAD in FGS 3. The first servicing mission made
no changes to the internal optics of the three Fine Guidance Sensors (FGS) that are used
for guiding and astrometry on HST. However, the subsequent movement of the secondary
mirror of the telescope to the so-called “zero coma” position did change the morphology
1
G. F. Benedict presented the paper at the 2002 Calibration Workshop.
373
374
McArthur, et al.
of the FGS transfer functions (Ftaclas et al. 1993). Therefore, a five-orbit post servicing
mission delta-OFAD calibration plan was designed and executed. After detection by the
LTSTAB of increasing incompatibility with the spring 1994 delta-OFAD calibration, an
11 orbit OFAD was performed in the fall of 1995 to recover the error budget for astrometry,
after In the spring of 1997 a five-orbit OFAD was performed on FGS 3 after the second
servicing mission. In December of 2000, a 14 orbit OFAD was performed on FGS 1R, which
replaced FGS 3 as the prime astrometer for scientific observations. FGS 1R, an enhanced
FGS with an adjustable fold-flat mirror that can be commanded from the ground, had
replaced the original FGS 1 instrument in February of 1997 in SM2 (Servicing Mission 2).
Seventy LTSTABS (Long Term Stability Tests) have been performed in both FGS 1R and
FGS 3 to assess time-dependent changes. A current list of the OFAD and LTSTAB tests is
shown in Table 1.
3.
Optical Field Angle Distortion Calibration and Long Term Stability Test
The Optical Telescope Assembly (OTA) of the Hubble Space Telescope (HST ) is a Aplanatic
Cassegrain telescope of Ritchey-Chrètien design. The aberration of the OTA, along with
the optics of the FGS comprise the OFAD. The largest component of the design distortion,
which consists of several arcseconds, is an effect that mimics a change in plate scale. The
magnitude of non-linear, low frequency distortions is on the order of 0.5 seconds of arc
over the FGS field of view. The OFAD is the most significant source of systematic error
in position mode astrometry done with the FGS. We have adopted a pre-launch functional
form originally developed by Perkin-Elmer (Dente, 1984). It can be described (and modeled
to the level of one millisecond of arc) by the two dimensional fifth order polynomial:
x = a00 + a10 x + a01 y + a20 x2 + a02 y 2 + a11 xy + a30 x(x2 + y 2 ) + a21 x(x2 − y 2 )
+a12 y(y 2 − x2 ) + a03 y(y 2 + x2 ) + a50 x(x2 + y 2 )2 + a41 y(y 2 + x2 )2
+a32 x(x4 − y 4 ) + a23 y(y 4 − x4 ) + a14 x(x2 − y 2 )2 + a05 y(y 2 − x2 )2
y = b00 + b10 x + b01y + b20 x2 + b02 y 2 + b11 xy + b30 x(x2 + y 2 ) + b21 x(x2 − y 2 )
+b12 y((y 2 − x2 ) + b03 y(y 2 + x2 ) + b50 x(x2 + y 2 )2 + b41 y(y 2 + x2 )2
+b32 x((x4 − y 4 ) + b23 y(y 4 − x4 ) + b14 x(x2 − y 2 )2 + b05 y(y 2 − x2 )2
(1)
where x, y are the observed position within the FGS field of view, x , y are the corrected
position, and the numerical values of the coefficients aij and bij are determined by calibration. Although ray-traces were used for the initial estimation of the OFAD, gravity release,
outgassing of the graphite-epoxy structures, and post-launch adjustment of the HST secondary mirror required that the final determination of the OFAD coefficients aij and bij be
made by an on-orbit calibration.
M35 was chosen as the calibration field. Since the ground-based positions of our target
calibration stars were known only to 23 milliseconds of arc, the positions of the stars were
estimated simultaneously with the distortion parameters. This was accomplished during
a nineteen-orbit calibration, executed on 10 January 1993 in FGS number 3. GaussFit
(Jefferys 1988), a least squares and robust estimation package, was used to simultaneously
estimate the relative star positions, the pointing and roll of the telescope during each orbit
(by quaternions), the magnification of the telescope, the OFAD polynomial coefficients,
and these parameters that describe the star selector optics inside the FGS: ρA and ρB (the
arm lengths of the star selectors A and B), and κA and κB (the offset angles of the star
selectors). Because of the linear relationship between ρA , ρA , κA and κB , the value of κB
The OFAD Calibration of HST Fine Guidance Sensors 1R and 3
Table 1.
Orbit
1
2
3-21
22
23
24
25
26
27
28
29
30
31
32
33-37
38
39
40
41
42
43
44
45
46
47
48
49
50–60
61
62
63
64
65
66
67
68
69
70
71
72–76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
LTSTAB and OFAD Observations
Julian Date
2448959.340822
2448971.061435
2448997.782164
2449082.954086
2449095.742836
2449096.613044
2449226.341817
2449255.529236
2449283.771053
2449309.341898
2449379.838241
2449408.794850
2449437.560417
2449468.662153
2449469.602118
2449593.554884
2449624.182975
2449652.274942
2449683.371435
2449711.665382
2449749.996910
2449780.160903
2449811.662894
2449838.070301
2449990.553542
2450018.625255
2450042.360197
2450052.674838
2450112.122350
2450133.837824
2450158.835440
2450174.716192
2450199.778704
2450321.550822
2450353.777465
2450377.443275
2450416.366701
2450480.031933
2450518.768090
2450560.517523
2450717.416169
2450743.225891
2450783.224190
2450822.077315
2450847.955266
2450904.886979
2450924.644942
2451054.361725
2451113.296366
2451121.224560
2451153.943299
2451163.019213
2451184.786771
2451189.556088
Year
1992
1992
1993
1993
1993
1993
1993
1993
1993
1993
1994
1994
1994
1994
1994
1994
1994
1994
1994
1994
1995
1995
1995
1995
1995
1995
1995
1995
1996
1996
1996
1996
1996
1996
1996
1996
1996
1997
1997
1997
1997
1997
1997
1998
1998
1998
1998
1998
1998
1998
1998
1998
1999
1999
Day
337
349
10
95
108
109
238
268
296
321
27
56
85
116
117
241
271
299
330
359
32
62
94
120
273
301
324
335
29
51
76
92
117
239
271
294
333
31
70
112
268
294
334
8
34
91
111
240
299
307
340
349
6
11
FGS
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
1
3
3
3
1
3
1
1
3
Observation
LTSTAB
LTSTAB
OFAD
LTSTAB
LTSTAB
LTSTAB
LTSTAB
LTSTAB
LTSTAB
LTSTAB
LTSTAB
LTSTAB
LTSTAB
LTSTAB
Spring Delta-OFAD
LTSTAB
LTSTAB
LTSTAB
LTSTAB
LTSTAB
LTSTAB
LTSTAB
LTSTAB
LTSTAB
LTSTAB
LTSTAB
LTSTAB
Fall Delta-OFAD
LTSTAB
LTSTAB
LTSTAB
LTSTAB
LTSTAB
LTSTAB
LTSTAB
LTSTAB
LTSTAB
LTSTAB
LTSTAB
Spring Delta-OFAD
LTSTAB
LTSTAB
LTSTAB
LTSTAB
LTSTAB
LTSTAB
LTSTAB
LTSTAB
LTSTAB
LTSTAB
LTSTAB
LTSTAB
LTSTAB
LTSTAB
Coefficient Set
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
4
3
3
3
4
3
4
4
3
375
376
McArthur, et al.
Table 1.
Orbit
91
92
93
94
95
96
97
98
99
100
101–114
115
116
117
118
119
120
121
122
123
124
Continued
Julian Date
2451300.596829
2451300.664236
2451416.507917
2451430.269572
2451555.127963
2451555.199688
2451649.638229
2451653.660590
2451783.159410
2451830.321088
2451899.105289
2451968.923102
2452021.654896
2452137.970671
2452201.355764
2452263.961701
2452274.313264
2452295.219942
2452370.867882
2452384.694618
2452520.528970
Year
1999
1999
1999
1999
2000
2000
2000
2000
2000
2000
2000
2001
2001
2001
2001
2001
2001
2002
2002
2002
2002
Day
122
121.2
238
251
11
11
106
110
239
286
355
59
112
228
291
354
364
20
96
110
246
FGS
3
1
3
1
1
3
1
3
1
1
1
1
1
1
1
1
1
3
1
1
1
Observation
LTSTAB
LTSTAB
LTSTAB
LTSTAB
LTSTAB
LTSTAB
LTSTAB
LTSTAB
LTSTAB
LTSTAB
OFAD
LTSTAB
LTSTAB
LTSTAB
LTSTAB
LTSTAB
LTSTAB
LTSTAB
LTSTAB
LTSTAB
LTSTAB
Coefficient Set
3
4
3
4
4
3
4
3
4
4
4
4
4
4
4
4
4
3
4
4
4
is constrained to be zero. A complete description of that calibration, the analysis of the
data, and the results are given in Jefferys et al. (1994).
In late fall 1992, just prior to the 1993 OFAD calibration, a series of one-orbit longterm stability tests (LTSTAB) was initiated. These tests had two seasonal orientations, a
spring orientation taken from an orbit of the OFAD, and a fall orientation, which was a 180
degree flip of the spring orientation. LTSTABs have been performed several times in each
of the orientations, spring and fall, every year.
The LTSTAB is sensitive to scale and low order distortion changes. It is an indicator of
the validity of the current OFAD coefficients and the need for recalibration. The LTSTAB
series immediately showed that the scale measured by the FGS was changing with time. The
indication of this change was seen in the large increase with time in the post-fit residuals
from a solution that solved for a constant sets of star positions, star selector encoder (SSE)
parameters, and OFAD parameters. The amount of scale change is too large to be due to
true magnification changes in the HST optical telescope assembly. These changes could be
due to water desorption in the graphite-epoxy components within the FGS. Initially the
scale-like change was modeled by allowing a variation in the star-selector-A effective lever
arm(ρA). Since 1995, the change has been modeled by allowing a change in both ρA and
κA (the offset angle of the star selector).
A five-orbit delta-OFAD was performed on 27 April 1994 after the first servicing mission
to assess the distortion changes caused by the secondary mirror movement to the zero coma
position. Significant effects in the OFAD (in addition to the scale-like changes) at the level
of 10 mas were found. The LTSTAB tests have revealed continued permutations in the FGS.
In addition to the scale changes, in mid-1995 we began to recognize higher order distortion
changes. These changes manifested themselves as something that looks like a radial scale
variation and is fairly well modeled by alterations in the third order terms in Eq. (1). We
had also noted that the residuals from the fall orientation LTSTABS are consistently higher
than for the spring in FGS 3.
The OFAD Calibration of HST Fine Guidance Sensors 1R and 3
377
An eleven-orbit delta-OFAD was performed in the late fall of 1995, to analyze temporal
changes, and upgrade the y-axis coverage.The star catalog was redetermined with input from
the three OFAD experiments of 1993, 1994 and 1995 to minimize the OFAD distortion that
could have been absorbed by the catalog positions. A more complete analyses of this deltaOFAD can be found in McArthur 1997.
In the spring of 1997 a second servicing mission replaced FGS 1. A five orbit deltaOFAD was performed in FGS 3, repeating the orientation of spring 1994. The coefficients
produced by this five-orbit delta-OFAD did not provide a better calibration than the 11 orbit Fall of 1995 delta-OFAD calibration, so these orbits were used instead as LTSTABS.
Two LTSTABS were performed in Spring 1997, one before and one after the second servicing mission. With scale and offset removed, a comparison yielded an rms of 0.965 mas,
indicating stability of FGS 3 across the servicing mission.
At the end of 2000, a 14 orbit OFAD was executed in FGS 1R, for a total of approximately observations. Figure 1 shows the rotations and offsets of FGS 1R in this
OFAD calibration. Because we now have a ten-year time span of M35 star positions, the
McNamara (1986) proper motion values were entered as observations with error in a quasiBayesian fashion, instead of being applied as constants. They then combine with the HST
observations to determine the proper motions. For this calibration, we ran a model which
performed a simultaneous solution of OFAD polynomials, star selector encoder (sse) parameters, proper motions, drift parameters, and catalog positions. This model had over 12,000
equations of conditions using all 124 OFAD and LTSTAB plates. Only the OFAD plates
determined the OFAD polynomials and complete sse parameters, while the LTSTAB combined with the OFAD plates contributed to a time-varying ρA and κA , proper motions, and
catalog positions. Each plate formed its own drift and rotation parameters. A systematic
signature in the X residuals from the four OFAD analysis remains. This signature differs
between FGS 3 and FGS 1. It appears as a very distinctive curve in the x component
residuals as a function of position angle in the FGS field of view (Figure 2). The curve
cannot be modeled by the fifth order polynomial. We have used a four frequency Fourier
series to remove this effect. The size of this effect, in an RMS sense over the entire field of
view of the FGS, is about one millisecond of arc. However, the peak-to-peak values near the
center of the field of view can be as large as 7 mas in FGS 3. The FGS 1 systematic is much
smaller with a peak to peak of about 2.5 mas. The source of this unexpected distortion is
not yet known but it may be due to the way the FGS responds to the spherically aberrated
HST beam.
On the basis of almost ten years of monitoring the distortions in FGS 3 we have
concluded that at the level of a few milliseconds of arc, the optical field angle distortion in
HST FGS 3 changes with time. These changes can be monitored and modeled by continuing
the LTSTAB tests, which also alerts us to the need for a new OFAD calibration. There
remains some dichotomy between the OFAD calibration data taken in the spring and that
taken in the fall.
Five sets of OFAD coefficients (Eq. 1) and star selector parameters (M , ρA , ρB , κA
and κB ) have been derived for reductions of astrometry observations. The average plate
residuals for these determinations are listed in Table 2. Comparisons of grids created with
each set of FGS 3 OFAD coefficients and distortion parameters indicate that the OFAD has
changed around 10 milliseconds of arc in non-scalar distortion between calibrations (which
have spanned 12–18 months)in FGS 3.
Each LTSTAB is associated with a specific set of coefficients Table 1. In the boundary
area between two OFAD experiments, the LTSTAB observations were reduced with both
sets of OFAD separately to determine which coefficients produce the best ρA κA fit of the
LTSTAB.
The values of ρA and κA determined by the LTSTABS and OFADS in FGSs 1 and 2
are illustrated in Figure 6, 5, 4 and 6. The error bars for these determination are smaller
378
McArthur, et al.
Figure 1.
Rotation and Offsets of FGS 1R Winter 2000 OFAD.
6
FGS3
FGS3
FGS3
FGS1
5
4
1993
1994
1995
2000
correction in arcseconds
3
2
1
0
-1
-2
-3
-4
-5
-6x10
-3
500
400
300
200
100
0
-100
X position in arcseconds
-200
-300
-400
-500
Figure 2.
Four frequency Fourier series correction of systematic signature in
X Residuals.
The OFAD Calibration of HST Fine Guidance Sensors 1R and 3
Table 2.
379
OFAD Residuals in milliseconds of arc
OFAD
Spring 1993
Spring 1994
Fall 1995
Spring 1997
Winter 2000
FGS
3
3
3
3
1
Xrms
1.90
1.96
2.09
1.85
1.87
Yrms
2.48
2.47
2.49
2.62
1.95
RSS
2.77
2.71
2.78
2.78
2.32
Number of Residuals
490
90
312
101
420
Orbits
19
5
11
5
14
-0.60
-0.61
κ A in FGS3
-0.62
-0.63
-0.64
-0.65
-0.66
-0.67
-0.68
2.4490
Figure 3.
2.4495
2.4500
2.4505
Julian Date
2.4510
2.4515
2.4520x106
κA fit of the LTSTABS in FGS 3.
than the symbols. For reduction of science astrometry data, the ρA κA parameters are
determined by interpolation of the two nearest LTSTABS in time.
4.
Past and Ongoing Astrometric Science with HST FGS
FGS 3 has been used to determine the first astrometrically determined mass of an extrasolar
planet, which is around the star GL 876 (ApJL, in press). It has been used to obtain many
trigonometric parallaxes. Targets included distance scale calibrators (δ Cep—Benedict et al.
2002b; RR Lyr—Benedict et al. 2002a), interacting binaries (Feige 24—Benedict et al. 2000),
and cataclysmic variables (RW Tri—McArthur et al. 1999; TV Col—McArthur et al. 2001;
SS Cyg, U Gem and SS Aur—Harrison et al. 1999). It was also involved in an intensive
effort to obtain masses and mass ratios for a number of very low-mass M stars (for example,
GJ 22, GJ 791.2, GJ 623, and GJ 748—Benedict et al. 2001). The average parallax precision
resulting from FGS 3 was σπ = 0.26 mas.
FGS 1 is being used to determine the parallaxes of several cataclysmic variables (EX Hya,
EF Eri, V1223 Sgr), parallaxes of a representative set of AM CVn stars, an independent
parallax of the Pleiades, and the masses of extrasolar planets around 2 Eridani and υ Andromeda. FGS 1 is also involved in an ongoing effort to obtain masses and mass ratios for
additional sets of low-mass M stars.
A continued program of LTSTAB monitoring and OFAD updates is essential to the
success of these long-term investigations with FGS 1.
380
McArthur, et al.
6.9090
ρ A in FGS3
6.9080
6.9070
6.9060
6.9050
6.9040
6.9030
2.4490
Figure 4.
2.4495
2.4500
2.4505
Julian Date
2.4510
2.4515
2.4520x106
ρA fit of the LTSTABS in FGS 3.
0.839
0.838
0.837
0.836
κ A in FGS1
0.835
0.834
0.833
0.832
0.831
0.830
0.829
0.828
0.827
2.4510
Figure 5.
5.
2.4512
2.4514
2.4516
2.4518
Julian Date
2.4520
2.4522
2.4524x106
κA fit of the LTSTABS in FGS 1.
Conclusions
We have shown that continued OFAD calibration of the Fine Guidance Sensors can reduce
this source of systematic error in positions measured by the FGSs to the level of 2 mas.
However, changes in the FGS units continue to occur, even twelve years after launch. These
changes require periodic updates to the OFAD to maintain this critical calibration.
Acknowledgments. The Astrometry Science Team is supported by NASA NAG51603. We are grateful to Q. Wang for the initial modeling of the OFAD and D. Story and
A. L. Whipple for their earlier contributions to this calibration. We thank L. Reed for
her long-term contribution to our knowledge of FGS 3. We thank Gary Welter and Keith
Kalinowski for their interest and assistance at Goddard Space Flight Center. We thank all
the members of the STAT, past and present for their support and useful discussions.
The OFAD Calibration of HST Fine Guidance Sensors 1R and 3
381
6.8148
ρ A in FGS1
6.8144
6.8140
6.8136
6.8132
6.8128
6.8124
6.8120
2.4510
Figure 6.
2.4512
2.4514
2.4516
2.4518
Julian Date
2.4520
2.4522
2.4524x106
ρA fit of the LTSTABS in FGS 1.
References
Benedict, G. F., McArthur, B., Chappell, D. W., Nelan, E., Jefferys, W. H., van Altena,
W., Lee, J., Cornell, D., Shelus, P. J., Hemenway, P. D., Franz, O. G., Wasserman,
L. H., Duncombe, R. L., Story, D., Whipple, A., & Fredrick, L. W. 1999, AJ, 118,
1086
Benedict, G. F., McArthur, B. E., Franz, O. G., Wasserman, L. H., Henry, T. J., Takato, T.,
Strateva, I. V., Crawford, J. L., Ianna, P. A., McCarthy, D. W., Nelan, E., Jefferys,
W. H., van Altena, W., Shelus, P. J., Hemenway, P. D., Duncombe, R. L., Story, D.,
Whipple, A. L., Bradley, A. J., & Fredrick, L. W. 2001, AJ, 121, 1607
Benedict, G. F., McArthur, B. E., Franz, O. G., Wasserman, L. H., Nelan, E., Lee, J.,
Fredrick, L. W., Jefferys, W. H., van Altena, W., Robinson, E. L., Spiesman, W. J.,
Shelus, P. J., Hemenway, P. D., Duncombe, R. L., Story, D., Whipple, A. L., &
Bradley, A. 2000, AJ, 119, 2382
Benedict, G. F., McArthur, B. E., Fredrick, L. W., Harrison, T. E., Lee, J., Slesnick, C. L.,
Rhee, J., Patterson, R. J., Nelan, E., Jefferys, W. H., van Altena, W., Shelus, P. J.,
Franz, O. G., Wasserman, L. H., Hemenway, P. D., Duncombe, R. L., Story, D.,
Whipple, A. L., & Bradley, A. J. 2002a, AJ, 123, 473
Benedict, G. F., McArthur, B. E., Fredrick, L. W., Harrison, T. E., Slesnick, C. L., Rhee,
J., Patterson, R. J., Skrutskie, M. F., Franz, O. G., & Wasserman, L. H. 2002b, AJ,
124, 1695
Benedict, G. F., Nelan, E., McArthur, B., Story, D., van Altena, W., Ting-Gao, Y., Hemenway, P. D., Shelus, P. J., Whipple, A. L., Franz, O. G., Fredrick, L. W., & Duncombe,
R. L. 1993, PASP, 105, 487
Bradley, A., Abramowicz-Reed, L., Story, D., Benedict, G., & Jefferys, W. 1991, PASP,
103, 317
Dente, G. & Abramowicz-Reed, L. 1984, Space Telescope Optical Field Angle Distortion
Calibration Algorithm, Perkin-Elmer ST-SE-3336
Ftaclas, C., Nonnenmacher, A., Weindling, F., Story, D., & Nelan, E. 1993, Ap. Opt., 32,
1696
Harrison, T. E., McNamara, B., Szkody, P., McArthur, B. E., Benedict, G. F., Klemola,
A., & Gilliland, R. L. 1999, ApJ, 515, 93
Jefferys, W., Fitzpatrick, J., & McArthur, B. 1988, Celest. Mech., 41, 39
382
McArthur, et al.
Jefferys, W. J., Whipple, A., Wang, Q., McArthur, B., Benedict, G. F., Nelan, E., Story,
D., & Abramowicz-Reed, L. 1994, in Calibrating Hubble Space Telescope, eds. J. C.
Blades & S. J. Osmer (Baltimore: STScI), 353
McArthur, B. E., Benedict, G. F., Jefferys, W. H., & Nelan, E. 1997, in The 1997 HST
Calibration Workshop with a New Generation of Instruments, eds. S. Casertano,
R. Jedrzejewski, C. D. Keyes, & M. Stevens (Baltimore: STScI), 472
McArthur, B. E., Benedict, G. F., Lee, J., Lu, C.-L., van Altena, W. F., Deliyannis, C. P.,
Girard, T., Fredrick, L. W., Nelan, E., Duncombe, R. L., Hemenway, P. D., Jefferys,
W. H., Shelus, P. J., Franz, O. G. & Wasserman, L. H. 1999, ApJ, 520, 59
McArthur, B. E., Benedict, G. F., Lee, J., van Altena, W. F., Slesnick, C. L., Rhee, J.,
Patterson, R. J., Fredrick, L. W., Harrison, T. E., Spiesman, W. J., Nelan, E.,
Duncombe, R. L., Hemenway, P. D., Jefferys, W. H., Shelus, P. J., Franz, O. G., &
Wasserman, L. H. 2001, ApJ, 560, 907
McNamara, B. & Sekiguchi, K. 1986, AJ91, 557
van Altena, W. F., Lu, C.-L., Lee, J. T., Girard, T. M., Guo, X., Deliyannis, C., Platais, I.,
Kozhurina-Platais, V., McArthur, B., Benedict, G. F., Duncombe, R. L., Hemenway,
P. D., Jefferys, W. H., King, J. R., Nelan, E., Shelus, P. S., Story, D., Whipple, A.,
Franz, O. G., Wasserman, L., Fredrick, L. W., Hanson, R. B., Klemola, A. R., Jones,
B. F., Mendez, R., Tsay, W.-S., & Bradley, A. 1997, ApJL, 486, 123
Wang, Q. & Jefferys, W. 1990, in Statistical Analysis of the Optical Field Angle Distortion
of the Hubble Space Telescope Using the Bootstrap Method (Austin: University of
Texas, Department of Astronomy Institute), 1
Whipple, A., Jefferys, W., Wang, Q., McArthur, B., Benedict, G. F., Nelan, E., Story, D., &
Abramowicz-Reed, L. 1994, in Calibrating Hubble Space Telescope, eds. J. C. Blades
& S. J. Osmer (Baltimore: STScI), 375
Whipple, A., McArthur, B., Wang, Q., Jefferys, W., Benedict, G. F., Lalich, A. M., Hemenway, P. D., Nelan, E., Shelus, P. J., and Story, D. 1995, in Calibrating Hubble Space
Telescope: Post Servicing Mission, eds. A. Koratkar & C. Leitherer (Baltimore:
STScI), 119
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
Wide Field Camera 3: Design, Status, and Calibration Plans
John W. MacKenty
Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218
Abstract. The mission of Wide Field Camera 3 (WFC3) is to assure the continuance of HST s superb imaging capability until 2010 while adhering to a cost
capped development approach. It will provide HST with a UVIS channel from 200
to 1000 nm and an infrared channel from 850 to 1700 nm with a rich set of filters.
WFC3 is based on the heritage of the existing HST instruments and follows a philosophy of extensive re-use of designs, components, and procedures. Its calibrations
and data products are based on the approaches used by the ACS and NICMOS
instruments.
1.
Introduction
WFC3 will be the first “panchromatic” camera on HST with two channels covering from
the near-ultraviolet into the near-infrared. The WFC3 UVIS channel uses a CCD detector.
This channel backs up ACS WFC capability while providing in addition a vastly improved
near-ultraviolet wide field science capability for HST. The WFC3 IR channel extends the
infrared capabilities on HST beyond the NICMOS instrument with a seven times larger field
of view, improved sensitivity where HST is most advantageous compared to ground-based
observatories, and a design compatible with operation until the end of the HST mission.
WFC3 is a facility instrument being developed on behalf of the HST user community.
It will replace the Wide Field Planetary Camera 2 (WFPC2) in Hubble’s radial science
instrument slot during Servicing Mission 4. The primary purpose of WFC3 is to assure
continued world class HST imaging science to the end of mission (now expected around
2010). To this end, NASA decided to develop WFC3 as a facility instrument without the
GTO team associated with prior HST instruments. The scientific goals and oversight of
WFC3 are provided by a NASA appointed Scientific Oversight Committee (SOC) chaired
by Dr. Robert O’Connell of the University of Virginia. Day to day development of the
instrument is conducted by an Integrated Product Team (IPT) formed by teams experienced
in the development of prior HST instruments. Led by NASA’s Goddard Space Flight Center
(GSFC), the IPT includes the Space Telescope Science Institute (STScI), Ball Aerospace
Corporation, Swales Aerospace Corporation, the Jet Propulsion Laboratory (JPL), and
many industrial suppliers. The IPT is led by Dr. Randy Kimble (GSFC) as Instrument
Scientist [who replaced Dr. Ed Cheng (GSFC) in September 2002], Dr. John MacKenty
(STScI) as Deputy Instrument Scientist, and Thai Pham (GSFC) as Instrument Manager.
2.
2.1.
UVIS Channel
CCD Detector
The UVIS channel has a Focal Plane Array consisting of two 2048 × 4096 pixel backside
illuminated CCDs. These were manufactured by E2V (then Marconi) Corporation in the
United Kingdom. They will provide a field of view of 162 × 160 arcsecond with 0.039 arcsecond projected pixel size. This is comparable to the existing WFPC2 Planetary Camera
383
384
MacKenty
Figure 1.
CCD Focal Plane Array.
channel and somewhat better than the ACS/WFC (0.050 arcsecond) sampling. These CCDs
have blue/near-UV optimized anti-reflection coatings that extend their sensitivity down to
200 nm. These coatings extend the wide field imaging into the near-UV at the expense
of sensitivity in the green-red region (where ACS is optimized). Further, WFC3 uses Aluminum reflective optics rather than protected silver (as employed by ACS/WFC) resulting
in a further red light performance advantage for ACS.
At this time, E2V/Marconi has completed all the WFC3 program CCD detector deliveries. These are exceptional devices, with extremely uniform behavior from device to
device, ultraviolet (250 nm) quantum efficiencies of 50 to 60 percent, readout noise of 2 to
2.5 e− rms (approximately 3 e− rms including the flight electronics), and excellent CTE.
These CCD detectors were extensively characterized at the GSFC Detector Characterization Laboratory by the WFC3 team. At the present time, Ball Aerospace has bonded two
pairs of these CCDs into 4K ×4K focal plane substrates and is nearing completion of their
assembly into flight units.
2.2.
UVIS Filter Elements
The complement of filters was selected by the WFC3 SOC after extensive input from the
astronomical community. It represents a carefully considered balance between continuing
the presence of heavily used WFPC2 and ACS filters and offering new capabilities. WFC3
also benefits from recent improvements in filter technology that reduce pinholes, improve
out of band rejection, and in-band throughput and bandpass shape.
The UVIS channel has a 48 element selectable, optical filter assembly (SOFA) that is
the actual unit flown in the WF/PC-1 instrument from 1990–1993. This refurbished unit has
been populated with new filters designed and manufactured for WFC3 by Barr Associates
and Omega Optical (two filters were obtained from the stock of WFPC2 spares). There
are 42 full field of view filters and 5 quad-filters which provide different passbands in each
quadrant of the image. There is also an ultraviolet grism to provide slitless spectroscopy
that was originally developed for the WF/PC-1 instrument. These filters represent the
state of the art in astronomical filters with especially excellent broad-band near-UV filters.
Combined with the enhanced UV detector sensitivity, WFC3 is several magnitudes more
sensitive than WFPC2 in the UV.
WFC3: Design, Status and Calibration Plans
Figure 2.
2.3.
385
SOFA Mechanism.
UVIS Calibration Considerations
The UVIS channel is nearly identical to the ACS Wide Field Channel (WFC). It uses the
same detector format, electronics, flight and ground software. The data sent to observers
has the same format and is processed by nearly identical pipeline calibration software.
There are two significant new features of its operation: (1) we have added support for 2 × 2
and 3 × 3 pixel on-chip binning, and (2) we will replace post-flash with charge-injection for
mitigation of the decline in charge transfer efficiency with on-orbit radiation damage.
We have obtained a full characterization of the detectors including monochrometer flat
fields in the red to provide the basis of a fringing correction with narrow emission line
sources. We are placing a high priority in obtaining extensive calibration in the ultraviolet
during the system level ground testing. WFC3 will be equipped with internal deuterium
and tungsten lamps for differential calibration.
An important consideration for calibration of WFC3 (also present to nearly as large
an extent in the ACS) is the significant geometric distortion in the field of view. We
anticipate that WFC3 will fully re-use the CALACS (drizzle) pipeline and that the majority
of observations will be reconstructed using dithered observations.
3.
3.1.
IR Channel
HgCdTe Detector
The Infrared Channel has a focal plan array consisting of a single 1024 × 1024 pixel HgCdTe
detector array. This array is a Rockwell Scientific Hawaii-1R device with a custom WFC3
mounting. The array provides a 1014 × 1014 pixel imaging area with 5 non-light-sensitive
reference pixels along each edge. This array provides a 139 × 123 arcsecond field of view
386
MacKenty
Figure 3.
Infrared Focal Plane Array mounted on 6 stage TEC.
(0.130 × 0.120 arcsecond projected pixel size). While not fully Nyquist sampled, the IR
sample represents a balance between maximizing the field of view and sampling the point
spread function. With dithered observations, it is expected that the full diffraction limited
resolution of HST will be preserved at wavelengths longwards of 1 micron.
The HgCdTe detector has a short wavelength cutoff at 0.82 to 0.85 microns determined
by its CdZn substrate and a long wavelength cutoff turned to 1700 nm. The long wavelength
cutoff was selected to provide acceptable dark current for operation at 150 K. This temperature is the minimum practical with the use of a solid state thermal electric cooler(TEC).
Compared to NICMOS’s original limited lifetime stored cryogen or current power-hungry
mechanical power cooler, the TEC cooling permits low power and long lifetime operation
and has strong design inheritance from the TECs that have cooled CCD detectors in STIS,
ACS, and the WFC3 UVIS channel.
The IR detector program was less mature at its inception than the CCD detectors and
required the production of multiple lots of devices. At this time, recent detectors approach
the desired specification and are in evaluation. The WFC3 IR channel is expected to have
somewhat better point source sensitivity than the NICMOS. In the broad J and H filters,
the detector dark current and noise, plus the instrument and telescope background, is
comparable to the zodiacal dust emission. Combined with its larger field of view (and
improved sampling over the large field), it should greatly increase HST ’s infrared survey
capability. The NICMOS instrument will remain the only HST instrument with infrared
coronographic and polarization capability, and with response beyond 1.7 microns.
3.2.
IR Filter Elements
The IR channel has a single 18 element filter wheel located in a cold enclosure near the cold
stop (and HST ’s pupil). This provides 15 bandpass filters and two grisms that offer coverage
from 850 to 1700 nm at broad, medium, and narrow bands (mapped to astronomically
interesting features).
3.3.
IR Calibration Considerations
The IR Channel is closely patterned on the NICMOS instrument. While NICMOS supported five detector operation modes, WFC3 only supports MULTI-ACCUM since this
was used for essentially all NICMOS observations. Also known as sample-up-the-ramp,
WFC3: Design, Status and Calibration Plans
Figure 4.
387
Infrared Filter Wheel.
MULTI-ACCUM obtains a sequence of readouts while signal accumulates in the detector.
The observer selects from a menu of stored exposure Sample Sequences (for which dark calibrations will be maintained) and obtains a selected number of readouts from that sequence.
WFC3 is capable of obtaining up to 16 readouts during a single exposure; this is limited by
onboard data storage.
The data format follows the NICMOS model with provision for cosmic ray removal
from individual datasets and the combination of datasets. The addition of the reference
pixels may help track drifts in the baseline (pedestal) signal. As with the UVIS channel,
we expect these datasets will be compatible with the ACS second stage (drizzle) imaging
combination pipeline software. This has several benefits: combination of multiple samples
to reduce noise and artifacts, improved spatial resolution, and correction for the geometric
distortion of the field of view.
We have equipped the IR Channel with the capability to read subarrays of 256 × 256,
128 × 128, and 64 × 64 pixels. This capability helps with data volume but, perhaps more
importantly, enables shorter exposures. Since the IR Channel does not have a mechanical
shutter (the detector reset serves to start an exposure), the minimum exposure time is the
3 to 4 second detector readout time when the entire detector is read out. Subarrays reduce
this time approximately proportionally to their area.
4.
Optical and Mechanical Design
The WFC3 instrument started from the foundation of the returned hardware of the original
WF/PC-1 instrument. Early on it was decided to design and fabricate a new optical bench
rather than attempt to re-use the WF/PC-1 bench. However, the external enclosure and
radiator were retained and reworked for WFC3. The physical layout captures the center of
the HST field of view with a pickoff mirror (essentially identical to WF/PC-1 and WFPC2),
passes the light past a Channel Select Mechanism (CSM) that either reflects the beam into
the IR Channel or lets it pass unhindered into the UVIS channel. The UVIS channel has
388
MacKenty
Figure 5.
WFC3 Optical Assembly showing UVIS and IR light paths.
the light fall onto an adjustable mirror (in tip, tilt, and piston) that steers the beam onto
a mirror containing the correction for the HST spherical aberration. This design, and the
actual corrector mechanism, are a close copy of the ACS WFC. The beam then transits the
SOFA, shutter mechanism (copied from the ACS WFC shutter), and then enters the CCD
detector enclosure (also copied from the ACS WFC design).
With the CSM in the IR Channel position the beam is directed onto a fold mirror, then
re-imaged using a pair of optics (one positioned on an identical tip-tilt-piston correction
mechanism as used in the IR channel). The beam then enters a cold enclosure (−35C) that
reduces both the cooling requirements of the IR detector and the internal background at
infrared wavelengths. Within this enclosure it passes through a refractive corrector element
(to remove the HST spherical aberration), a cold mask (for the HST pupil), and the infrared
filter element. The detector is housed in a enclosure with heritage from the STIS and ACS
detector enclosures. The use of a transmission correction for the spherical aberration has
the decided advantage of making a clean pupil available for the cold mask. This design is
both physically compact and minimizes oversizing of the cold mask (and thereby minimizes
the resulting throughput loss).
WFC3 makes considerable use of its attached dedicated thermal radiator. This is
divided into two zones. The hot zone dumps heat from the electronics and reduces WFC3’s
thermal load into the aft shroud of HST where the other science instruments are located.
The cold zone provides the first stage of cooling for both the UVIS and IR detectors plus
the IR cold enclosure. This is accomplished via a bank of 18 single-stage TEC units.
5.
Instrument Status and Plans
WFC3 is presently in the Integration and Test phase. The optical bench assembly and
testing has been completed by Ball Aerospace and delivered to GSFC in early December
2002. The team is presently concentrating on getting the electronics and bench integrated
into the enclosure at GSFC. Final selection of IR detectors, detector packaging and installation, system level thermal vacuum testing, and science calibration will be accomplished
WFC3: Design, Status and Calibration Plans
Figure 6.
389
Solid Model of WFC3 showing re-use of components.
during the coming year. While at the time of the Calibration Workshop the SM4 mission
was expected in April 2004, it is now rescheduled for early 2005.
Acknowledgments. The WFC3 instrument is an ongoing effort of a large and talented
team of people. Additional information may be obtained from the World Wide Web sites at
GSFC (http://wfc3.gsfc.nasa.gov) and at STScI (http://www.stsci.edu/instruments/wfc3/).
A more detailed discussion of WFC3 (and its anticipated performance as of August 2002)
is available in the WFC3 Mini-Handbook (http://www.stsci.edu/instruments/wfc3/wfc3docs.html).
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
Calibration Status of the Cosmic Origins Spectrograph Detectors
Steven V. Penton, Stéphane Béland, and Erik Wilkinson
Center for Astrophysics and Space Astronomy, University of Colorado, Boulder,
CO 80309
Abstract. COS has two distinct ultraviolet channels covering the spectral range
from 1150Å to 3200Å. The NUV channel covers the range from 1700Å to 3200Å
and uses the Hubble Space Telescopes STIS spare MAMA. The FUV channel uses
a micro channel plate detector with a cross-delay line readout system to cover the
range from 1150Å to 1900Å. Due to the analog nature of the readout electronics of
the FUV detector, this system is sensitive to temperature variations and has nonuniform pixel size across its sensitive area. We present a step-by-step description
of the calibration process required to transform raw data from the COS into fully
corrected and calibrated spectra ready for scientific analysis. Initial simulated raw
COS data is used to demonstrate the calibration process.
1.
Introduction
During the HST servicing mission currently scheduled for Spring 2005 (SM4), the Cosmic Origins Spectrograph (COS, Sembach 2002) is scheduled to be installed in the science
bay currently occupied by COSTAR. COS contains two ultra-violet (uv) channels, which
share two common 2.5 diameter apertures (the primary science aperture, PSA, and the 1%
transmission bright object aperture, BOA). The far uv (FUV, 1150–1900 Å) and the near
uv (NUV, 1700–3200 Å) channels employ independent detectors but cannot be operated
simultaneously. The one-bounce COS FUV channel uses holographic gratings to simultaneously disperse and correct the aberrated HST beam onto a two segment cross-delay line
microchannel plate (MCP) similar to that flown on FUSE. The NUV channel uses a fourbounce optical path to disperse and correct the HST beam into three non-contiguous strips
on a spare STIS MAMA detector. In this brief update on the calibration status of COS,
we discuss the progress of the COS detector calibrations.
2.
The FUV channel
The two FUV segments, ‘A’ and ‘B’, employ time-delay anodes in both the dispersion (‘X ’)
and cross-dispersion (‘Y ’) directions. The anodes are used in conjunction with 85 × 10 mm
MCP stacks (McPhate et al. 2000). The detectors do not have physical pixels, instead the
time-delay detector represents the event location as an analog value. Each detector segment
is represented ∼ 14, 000 × 400 digital elements (DEs). The physical ‘size’ represented by
each DE is variable across the detector. The background count rate is low, ∼ 2 counts
DE−1 month−1 . The FUV detector deadtimes are well characterized, and are < 10% at
10,000 counts s−1 . In this section, we will discuss the known distortions and the ground
calibrations employed to correct them (Vallerga et al. 2001). Unless otherwise stated, the
values discussed here are those for the ‘A’ segment of the ‘FUV01’ detector.
390
Calibration Status of the Cosmic Origins Spectrograph Detectors
2.1.
391
Thermal Distortions
The mapping function from reported photon location to DE value is temperature dependent. The thermal distortions, introduced prior to digitization, are well characterized as
a combination of a shift and stretch of the position to pixel value mapping. Electronic
stim pulses, representing events at fixed locations, are injected into the detector electronics
and digitized as if actual photon events. The electronic stim pulses appear in the photon
list at positions to the lower left and upper right of the MCP active area. To correct the
photon list for thermal distortions, the lower left stim pulse position, and with it the rest
of the photon list, is adjusted to a predetermined baseline position. The photon list is then
stretched or compressed to force the upper right stim pulse position to fall at its baseline
position. Results from the testing with the available ground flat field data (§2.3) suggest
that this algorithm is more than sufficient for correcting the expected thermal distortions.
2.2.
Geometric Distortion
As described in detail in Wilkinson, et al. 2001 and Béland et al. 2002, the mapping function
from physical photon location on the FUV detector to analog DE value is not a straightforward linear mapping. Distortions in the FUV readout electronics and MCPs result in
DEs of variable size. A typical row of the segment ‘A’ shows uncorrected DE sizes of
5.96 ± 0.01µm × 24.2 ± 0.1µm. To determine the geometric distortion correction (GDC),
also referred to as the integral non-linearity (INL), an opaque mask with a regularly spaced
grid of pinholes was imaged by the detectors. By comparing the known physical centers
of the pinholes to their digital values, the GDC is determined for each segment. Since the
DEs well sample the resolution element (RE, ≈ 6 × 12 DE), the physical size of the DEs can
be forced to be a constant size of 6.0 × 24.0µm without affecting the scientific value (wavelength, resolution, etc.) of the detected events. The GDC is determined with thermally
corrected data, and the correction is always applied after thermal correction. An opaque
mask of slits was also imaged during ground testing, providing an independent method of
determining the accuracy of the GDC. The GDC was applied to the measured position of
the slits, and the known physical X position of the slits was then compared to the GDC
corrected positions. These residuals show a Gaussian distribution with a residual error (1σ)
of < 0.5 DE in the dispersion direction, corresponding to ∼ 1/12 of a RE.
2.3.
Flat Fields
During ground testing, 114 flat field images were obtained for each segment. These flat
fields were thermally and geometrically corrected, then combined into deep flat fields (DFF)
images. Each DFF contains ∼ 109 photons, with a mean number of counts RE−1 of ∼ 10000.
These combined flat fields contain information on both the illumination of the detectors,
the L-flat, and the DE-to-DE variations, the D-flat (or more traditionally, the P -flat). To
separate the L and D-flats from the DFF, for each column (Y ), the DFF was smoothed
in the X direction, then a low-order polynomial function approximating the illumination
pattern was fit along each column. The L-flat was constructed from the least-squares fits
(assuming Poisson statistics), then the D-flat was derived by dividing the DFF by the Lflat. The signal-to-noise ratio, S/N , of the DFFs are ∼ 100 RE−1 . A portion of the ‘A’
segment D-flat is shown in Figure 1. In this figure, the intensity scale has been modified to
show the variations, the actual variations are Gaussian with a width of 5%. This implies
that the flat-field variations must be removed to achieve COS FUV data with S/N > 20.
To test the quality of the flats, the original flat field images were randomly selected to
create two independent DFFs for each segment. Each DFF was divided by the appropriate
L-flat and D-flat to create two independent flatfielded test images. A 12 DE (Y ) strip, shown
in Figure 1, was extracted from each test image. The strips were collapsed per RE to form
a spectrum following the same algorithm used for a spectral extraction. The two S/N ∼ 70
RE−1 spectra were then divided to test the quality of the combined thermal, geometric, and
392
Penton, Béland, & Wilkinson
Figure 1. D-flat for a small section of the FUV segment ‘A.’ The intensity scale
of the image has been modified to show variations. The actual variations are
Gaussian with a width of ≈ 5% (S/N ≈ 20). The dashed lines indicate the 12 DE
(Y) strip used to extract the simulated ’spectrum.’
Figure 2. The DFF data set was divided into two independent data sets, each
corrected for thermal, geometric, and flat field distortions. A strip of each test
image was extracted as if a science spectrum. The ‘spectra’ were divided, with
the above distribution indicating that COS FUV observations of S/N = 50 are
achievable (or S/N = 100 if using FP-SPLIT = 4).
flat field extraction. The results of this test are shown in Figure 2. The resulting distribution
is Gaussian with a width implying a final S/N of≈ 50. This is indeed the expectation for
two data sets limited only by Poisson statistics, (1/70)2 + (1/70)2 ≈ 1/50. The data set
was then dithered
to approximate a FPSPLIT = 4 algorithm, with the expected result of
√
a S/N = 4 × 50 = 100 extraction. COS contains an on-board flatfield lamp, which will
be used on-orbit to achieve S/N ≥ 70 RE−1 flatfields for the regions of the COS detector
employed for obtaining scientific data.
2.4.
Wavelength Accuracy
The limiting factor in wavelength accuracy is the ability to center targets in the aperture
during target acquisition (TA). The COS TA algorithms have been tested with ray-trace
simulations (Penton, 2001), and will be able to center the target in the aperture to within
0.1. This corresponds to a 3σ wavelength velocity error of less than 10 km s−1 for the
FUV medium resolution gratings (64 km s−1 for the low resolution).
3.
The NUV channel
The NUV detector is a STIS spare 1k ×1k NUV MAMA, which has 25×25µm pixels defined
by physical structures in the anode. The thermal and geometric distortions of NUV detector
are small, and will not require monitoring, calibration, or correction. Based upon ground
measurements, and in-flight performance of the STIS MAMAs, the background count rate
is expected to be ∼ 34 counts s−1 cm−2 or 1.3 counts hour−1 pixel−1 . Following a flat-field
procedure similar to that of the FUV channel, the measured NUV P-flat pixel distribution
is Gaussian with a width of 0.044 (S/N ∼ 20 pixel−1 ). Extracting a spectral region gives a
Calibration Status of the Cosmic Origins Spectrograph Detectors
393
Gaussian distribution with a S/N ∼ 50 RE−1 , or S/N ∼ 100 RE−1 for an FPSPLIT = 4
observation. Like the FUV channel, on-orbit NUV flat fields will be used for calibrating
HST +COS NUV data. NUV wavelength accuracy is also limited by the TA centering
accuracy. Simulations (Penton, 2001) indicate that the TA introduced 3σ wavelength error
for medium resolution NUV observations should be < 19 km s−1 (200 km s−1 for low
resolution).
4.
Conclusions
Ground-based calibration data, combined with extensive modelling using our current understanding of the COS detectors and optical system, indicate that S/N of 100 observations
should be possible with both COS channels. On-going calibrations are on schedule to provide the highest quality data with HST +COS. Some calibrations, such as measuring the
sensitivity function, can only be performed on-orbit. More information on COS can be
found at http://cos.colorado.edu and http://www.stsci.edu/instruments/cos. This work
was supported by the HST COS project (NAS5-98043).
References
Béland, S., Penton, S. V., & Wilkinson, E., 2002, Proc. SPIE, in press, 2002
Sembach, K. R., et al. 2002, in COS Instrument Mini-Handbook, version 1.0, (Baltimore:
STScI), available at http://www.stsci.edu/instruments/cos/cos docs.html.
McPhate, J. B., Siegmund, O. H., Gaines, G., Vallerga, J. H., & Hull, J., 2000, The COS
FUV Detector, Proc. SPIE 4139, 25
Morse, J. A., et al. 1998, Performance overview and science goals of COS for HST, Proc.
SPIE 3356, 361, (also see COS-01-0001, available at
http://cos-arl.colorado.edu/OP01/)
Penton, S. V. 2000, TAACOS: Phase I FUV Report, COS Internal Document, COS-0110016, available at http://cos-arl.colorado.edu/TAACOS/
Penton, S. V. 2001, TAACOS: Phase I NUV Report, COS Internal Document, COS-0110024, available at http://cos-arl.colorado.edu/TAACOS/
Vallerga, J. V., McPhate, J. B., Martin, A. P., Gaines, G. A., Siegmund, O. H., Wilkinson,
E., Penton, S. V., & Béland, Stéphane, 2001, Proc. SPIE 4498, 141
Wilkinson, E. 2002, COS Calibration Requirements and Procedures, COS Internal Document, COS-01-0003, available at http://cos-arl.colorado.edu/AV03/
Wilkinson, E., Penton, S. V., Béland, S., Vallerga, J. V., McPhate, J., & Sahnow, D. J.
2001, Proc. SPIE, 4498, 267
2002 HST Calibration Workshop
Space Telescope Science Institute, 2002
S. Arribas, A. Koekemoer, and B. Whitmore, eds.
Coronagraphic Imaging: Keck II AO and HST ACS Compared
Paul Kalas1
Astronomy Department, University of California, Berkeley, CA 94720
David Le Mignant
W. M. Keck Observatory, Kamuela, HI 96743
Franck Marchis1 and James R. Graham1
Astronomy Department, University of California, Berkeley, CA 94720
Abstract. Coronagraphic imaging reduces PSF wings by 0.6 mag using Keck adaptive optics in the NIR, and 1.5 mag using ACS in the visible. The PSF suppression
attained is roughly comparable for the two instruments. Future work should test the
relative contrast gains from PSF subtractions.
1.
Introduction
High contrast imaging is necessary to search for and study sub-stellar objects and debris
disks surrounding bright stars. Here we compare the performance of coronagraphs used with
ground-based adaptive optics (AO) systems to those implemented in HST science cameras.
The following data are used in the present investigation:
1. With Keck II (10 m) AO, we used the coronagraphic mode of the NIRC2 science
camera (1024 × 1024 InSb array). NIRC2 has 10 focal plane occulting spots between
100 mas and 2000 mas diameter, five selectable pupil plane stops, and three selectable
plate scales. The V = 8.8 star HD 162052 was observed at K , with a 300 mas focal
plane mask, a pupil plane mask matched to the telescope pupil (large hexagonal),
120 sec cumulative integration time, and with 0.01/pix plate scale. Measured Strehl
ratios are S∼50%.
2. For HST ACS, we adopted the performance specifications for the F606W filter given
in the ACS Instrument Handbook for Cycle 12 Version 3.0. John Krist also supplied
us with the F606W coronagraphic images of Arcturus shown in Chapter 5 of the ACS
Instrument Handbook (offspot and onspot, with the High Resolution Channel Lyot
mechanism in place).
2.
PSF Differences
In general, PSFs have static and temporal characteristics, and are composed of both scattered light (the seeing halo) and diffracted light (the Airy disk). The encircled energies
of the Keck AO and the HST ACS data are shown in Figure 1. Keck AO produces near
1
Center for Adaptive Optics, University of California, Santa Cruz, CA 95064
394
Coronagraphic Imaging: Keck II AO and HST ACS Compared
395
Encircled Energy (normalized to 1.0 at 1.2 arcsec)
1
0.9
0.8
0.7
0.6
0.5
0.4
HST ACS offspot
Keck NIRC2 offspot
0.3
0.2
0
0.2
0.4
0.6
Radius (arcsec)
0.8
1
1.2
Figure 1. Encircled energies for the stars used to test coronagraph performance,
before occultation by a focal plane occulting mask.
diffraction-limited central cores (FWHM = 54 mas), but the uncorrected atmosphere contributes a scattered seeing halo beyond 0.1 radius (2.5λ/D) that contains ∼40% of the
light.
A coronagraph should have the following effect on the PSF:
• Light scattered into the seeing halo by the atmosphere or by the Optical Telescope
Assembly (OTA) is not suppressed.
• Scattered light within the science camera is suppressed because the PSF core is absorbed by the focal plane occulting mask.
• Diffracted light due to the OTA is suppressed, depending on the sizes of the Lyot stop
and the occulting spot. The larger the occulting spot, the more light is pushed to the
edges of the pupil plane image, which is then masked by the Lyot stop.
Based on these principles, the a priori expectation is that the ACS coronagraph will
further diminish the intensity of the PSF wings shown in Figure 1 because the PSF is
dominated by diffraction. Ground-based, AO coronagraphic data should demonstrate only
modest improvement over a non-coronagraphic observations because the seeing halo is dominated by atmospheric scattered light.
3.
Coronagraph Suppression
Figure 2 is made by taking the median azimuthal value for each occulted PSF as a function
of radius, and converting it to a relative magnitude based on the peak intensity of the
star in an unocculted image. We find that the coronagraph suppresses the AO PSF wings
by a median value of 0.6 mag in the range 0.2 − −1.2 radius. Thus the suppression of
instrumental and diffracted light by the coronagraph gives a somewhat unexpected contrast
396
Kalas, et al.
0
Kalliope B
(Marchis et al. 2002)
4
Relative Magnitude
HST ACS F606W, offspot
HST ACS F606W, 1800 mas spot
Keck AO K', 300 mas spot
Keck AO K', offspot
Patroclus B
(Merline et al. 2001)
2
6
Sylvia B (Brown & Margot 2001)
8
HR 7672 B (Liu et al. 2002)
HR 4796A (Schneider et al. 1999)
10
Gl 86 B (Els et al 2001)
12
14
16
0
0.5
1
1.5
2
2.5
3
Radius (arcsec)
Figure 2. Radial PSF intensity normalized relative to the peak intensity and
converted to relative magnitude. Solid circles are sub-stellar objects, solid ellipses
are circumstellar disks, and solid squares are asteroidal companions discovered by
direct observations with AO systems at Lick, Gemini, and Keck.
gain for AO images. The ACS coronagraph suppresses the PSF by a median value of 1.5 mag
in the range 1.0 − −3.0 radius.
Overall, the differences between HST and Keck when comparing the occulted (spot)
or unocculted (offspot) PSFs are not significant. Clearly the region within 1 radius of
the star is accessible with Keck AO, whereas the large size (radius 0.9) of the smallest
ACS occulting spot prevents coronagraphic imaging of the sub-arcsecond region. Recently
detected objects follow an envelope just above the PSF curves, indicating that, to first
order, a simple radial plot of the PSF corresponds to detection limits.
4.
Detectability Simulations
Ultimately the sensitivity of HST and Keck should be tested by simulating the science
objects and working through a variety of observing modes and data reductions (e.g., Kalas
& Jewitt 1996, Schneider et al. 2001). A crucial step is subtraction of the PSFs shown in
Figure 2 by either self-subtraction after a field roll, or by observing a nearby reference PSF
contemporaneously. Because the Keck telescope has an alt-az mount that produces field
rotation, we are presently testing the efficacy of roll deconvolution from the ground.
For the present investigation, we merely insert a model dust disk into the coronagraphic
PSFs without any further data reduction. The model disk is described in Kalas & Jewitt
(1996). We fix the disk inclination (73◦ ), central hole radius (1 ) and peak surface brightness
(12 mag arcsec−2 ), to correspond to the HST NICMOS image of HR 4796A (Schneider et al.
1999). The main difference is that the model has no fixed outer extent, whereas the real
HR 4796A disk is a confined ring. Figure 3 shows that the ACS coronagraph would detect
the dust scattered light if the disk were in fact extended rather than confined. In addition
to detecting the outer disk, Keck AO reveals the central hole in the dust distribution.
Coronagraphic Imaging: Keck II AO and HST ACS Compared
397
Figure 3. Model HR 4796-like disk inserted into the ACS occulted PSF at
PA=135◦ (top) and into the Keck AO PSF at PA = 90◦ (bottom) . The normal
PSF shows an azimuthally symmetric halo, whereas the distortion evident here is
due to the circumstellar disk. Arrows designate the brightest portion of the disk
within which a central disk hole is present.
5.
Conclusions and Future Directions
Coronagraphic imaging reduces the PSF wings by 0.6 mag using Keck AO in the NIR, and
1.5 mag using the ACS coronagraph in the visible. In circumstellar regions where these
tests overlap, the occulted and unocculted PSF wings show roughly comparable sensitivity
for Keck AO and HST.
Ground-based AO has the advantage over HST ACS in imaging circumstellar regions
within 1 radius. The HST NICMOS coronagraph has a smaller occulting spot than ACS,
but imaging is still limited to ∼ 0.6 radius (Schneider et al. 1999). High-order AO systems
will further improve the ground-based capability of imaging the sub-arcsecond circumstellar
region (Sivaramakrishnan et al. 2001, Lloyd et al. 2001).
The results shown here have a limited scope. Future work must test the sensitivity gains
attained with PSF subtractions and longer integration times. Comparisons of ground-based
AO to HST NICMOS and STIS coronagraphic data are also necessary.
Acknowledgments. This work has been supported by the NSF Center for Adaptive
Optics (managed by the University of California, Santa Cruz, under cooperative agreement
AST-9876783), and AURA (award HST-GO-09475.01-A).
References
Brown, M. E. & Margot, J. L. 2001, IAUC 7588
Els, S. G., Sterzik, M. F., Marchis, F., Pantin, E., Endl, M. & Kuerster, 2001, A&A, 370,
L1
Kalas, P. & Jewitt, D. 1996, AJ, 111, 1347
Liu, M. C., Fischer, D. A., Graham, J. R., Lloyd, J. P., Marcy, G. W., & Butler, R. P. 2002,
ApJ, 571, L519
Lloyd, J. P., Graham, J. R., Kalas, P., et al. 2001, SPIE, 4490, 290
Marchis, F., Descamps, P. , Hestroffer, D., et al. 2002, submitted to Icarus
Merline, E. J., Close, L. M., Siegler, N., et al. 2001, IAUC 7741
398
Kalas, et al.
Pavlovsky, C., et al. 2002, ACS Instrument Handbook for Cycle 12 Version 3.0 (Baltimore:
STScI)
Schneider G., Becklin, E. E., Smith, B. A., Weinberger, A. J., Silverstone, M., & Heines,
D. C. 2001, AJ, 121, 525
Schneider, G., Smith, B. A., Becklin, E. E., et al. 1999, ApJ, 513, L127
Sivaramakrishnan, A., Koresko, C. D., Makidon, R. B., Berkefeld, T., & Kuchner, M. J.
2001, ApJ, 552, 397
399
Author Index
Alexov, A., 162, 176
Anderson, J., 13, 311
Argabright, V., 86
Arribas, S., 215, 263
Ayres, T. R., 171
Hill, R. J., 197
Hill, R. S., 148, 197
Hines, D. C., 259
Hodge, P., 97
Hook, R. N., 337
Barrett, P., 97
Béland, S., 390
Benedict, G. F., 373
Bergeron, L. E., 222, 263
Blakeslee, J. P., 3, 23, 65
Böker, T., 215, 222, 263
Bohlin, R. C., 3, 23, 31, 86, 97, 115, 189,
232
Bouwens, R. J., 65
Bowers, C. W., 127, 137, 148, 184, 197,
209
Brammer, G., 325, 329
Bristow, P., 162, 176
Brown, T. M., 97, 180, 189, 201
Bushouse, H., 271
Busko, I., 97, 205, 209
Illingworth, G. D., 3
Calzetti, D., 215, 232, 263
Casertano, S., 354
Cheng, E. S., 197
Clampin, M., 3, 65, 86
Collins, N. R., 184, 193
Cottingham, D. A., 197
Cox, C., 58, 65, 86
Davies, J. E., 97, 180, 201
de Marchi, G., 23, 31, 65, 86
Diaz-Miller, R. I., 97, 189
Dickinson, M., 215, 232, 263
Dolphin, A. E., 301
Dressel, L., 97
Ford, H. C., 3, 65, 86
Freudling, W., 241
Fruchter, A. S., 337
Gilliland, R. L., 3, 23, 31, 58, 61
Gonzaga, S., 341
Goudfrooij, P., 97, 105, 205
Grady, C. A., 137, 148
Graham, J. R., 394
Gull, T. R., 137, 148, 184, 197
Hack, W., 70, 337
Hartig, G. F., 3, 65, 86
Heap, S. R., 137
Heyer, I., 333, 341, 359
Jansen, R. A., 193
Jee, M. J., 82
Jefferys, W. H., 373
Johnson, S. D., 197
Kalas, P., 394
Karkoschka, E., 315
Kerber, F., 162, 176
Kim Quijano, J., 97, 189, 209
Kimble, R. A., 86, 105, 137, 197
King, I. R., 311
Kiziltan, B., 325
Koekemoer, A. M., 70, 291, 315, 325, 329,
337, 341, 354
Kozhurina-Platais, V., 341, 354
Krist, J., 3
Le Mignant, D., 394
Lindler, D., 65, 127, 137, 148, 184, 189,
197, 209
Lubin, L. M., 333, 341
Mack, J., 23, 31
MacKenty, J. W., 383
Maı́z-Apellániz, J., 97, 346
Makidon, R., 367
Malhotra, S., 215, 263
Malumuth, E. M., 137, 148, 197
Marchis, F., 394
Martel, A. R., 3, 65, 82, 86
Mazzuca, L., 215, 222, 263
McArthur, B., 373
McMaster, M., 350
Meurer, G. R., 3, 65, 86
Mobasher, B., 97, 201, 215, 263
Mutchler, M., 70
Nelan, E., 367, 373
Noll, K., 215, 263
Pasquali, A., 38, 74, 90
Pavlovsky, C., 31
Penton, S. V., 390
Pirzkal, N., 38, 74, 90
Plait, P., 137
400
Author Index
Potter, M., 97
Proffitt, C. R., 97, 137, 189, 201, 205
Ratnatunga, K. U., 78
Richardson, M., 333
Rieke, M., 232
Riess, A., 47, 61
Rosa, M. R., 162, 176
Roye, E. W., 215, 263, 267
Sahu, K. C., 97, 189, 205
Schneider, G., 249
Schultz, A. B., 215, 263, 267, 271
Sirianni, M., 3, 31, 65, 82, 86
Sosey, M., 215, 222, 232, 263, 275
Sparks, W. B., 53, 82
Stys, D. J., 97, 205
Tennant, D., 148
Thompson, R. I., 241
Tran, H. D., 65, 86
Valenti, J., 97, 189, 209
van der Marel, R., 23
Van Orsow, D., 82
Walborn, N. R., 97, 205
Walsh, J. R., 38, 74, 90
Wen, Y., 197
Whitmore, B. C., 281, 329, 333, 341, 350,
359
Wiklind, T., 215, 263
Wilkinson, E., 390
Windhorst, R. A., 193
Woodgate, B. E., 137, 197
Xu, C., 215, 222, 263
401
Subject Index
ACS papers begin on page 3, STIS papers
begin on page 97, NICMOS papers begin
on page 215, WFPC2 papers begin on page
281, and papers for other instruments begin on page 367.
ACS slitless extraction software (aXe), 38,
74, 90
ACS/HRC, 3, 13, 23, 31, 38, 47, 53, 65,
90, 337
ACS/SBC, 3, 31, 38, 53, 65, 86, 337
ACS/WFC, 3, 13, 23, 31, 38, 47, 53, 58,
61, 65, 70, 74, 78, 82, 90, 337
ampglow, 97, 222
anneal, 3, 47
anomalies
image, 3, 180, 267
instrument, 97, 193
aperture corrections, 61, 201, 232
aperture location, 97, 137, 329, 373
aperture photometry
ACS, 23, 31, 47, 61
STIS, 105, 201
WFPC2, 281, 301, 333, 346, 359
astrometry
ACS, 13, 53, 58, 65
FGS, 367, 373
WFPC2, 281, 291, 311, 329, 354
background
astronomical, 249, 301, 325, 359
instrument, 3, 97, 180, 193, 271, 367
bad pixels, 189, 222
bias, 82, 222
breathing, 184, 281
calacs, 23, 31, 53, 58, 65, 70, 82
calfos, 162
calibration plan, 263, 291
calnica, 275
calnicb, 271, 275
calstis, 97, 127, 180, 184, 209
charge transfer efficiency (CTE)
ACS, 3, 47
STIS, 97, 105, 115, 176, 205
WFPC2, 281, 291, 301, 350, 359
charge transfer traps, 47, 97, 176, 301
chip-to-chip normalization, 13, 23, 82
contamination, 86, 127, 281, 341, 350
coronagraphy, 3, 137, 249, 394
COS, 390
cosmic rays, 47, 58, 78, 325
cross-dispersion profile, 184
dark, 86, 97, 162, 180, 189, 222, 367
dark generator, 275
decontamination, 281, 291, 350
detector quantum efficiency, 3, 31, 61, 127,
222
dither, 47, 53, 70, 281, 311, 325
drizzle, 53, 70, 281, 325, 337
echelle, 97, 127
Exposure Time Calculator (ETC), 275
FGS, 281, 329, 367, 373
field distortion, 13, 58, 65, 78
flat generator, 275
flatfield, 341
ACS, 23, 53, 74
COS, 390
STIS, 162, 189
WFPC2, 281, 315
focus, 184, 267, 281, 373
FOS, 162
fringing, 90, 148, 197
geometric distortion
ACS, 13, 53, 58, 65, 78
COS, 390
STIS, 162
WFPC2, 281, 311, 325, 354
grating efficiency, 127
hot pixels, 3, 47, 222
HST-phot, 301, 333, 346
image quality, 3, 86, 267
image registration, 58, 78, 311, 325, 329,
354
imaging
ACS, 3, 58, 70, 78, 86
NICMOS, 249, 267, 271
STIS, 97, 180, 193, 201
WFC3, 383
WFPC2, 291, 311, 337, 341, 354
infrared (IR), 215, 249, 259, 263, 267, 383
instrument design, 383
Interactive Data Language (IDL), 193
intrapixel sensitivity, 13
line spread function (LSF), 148, 184
long vs. short anomaly, 281, 301, 359
mosaicing, 271, 325
402
Subject Index
MultiDrizzle, 53, 70, 325, 337
narrow band imaging, 90, 315
narrow band photometry, 90
NCS, 215, 222
NICMOS/NIC1, 215, 222, 232, 259, 263,
267, 271, 337
NICMOS/NIC2, 215, 222, 232, 249, 259,
263, 267, 271, 337
NICMOS/NIC3, 215, 222, 232, 241, 263,
267, 271, 337
NICMOSlook, 215, 241
optical field angle distortion (OFAD), 373
parallel observing, 193, 271
persistence, 222
photometric monitor, 291, 341
photometric transformations, 281, 301, 333
photometric zeropoint
ACS, 31, 61
NICMOS, 232
STIS, 162
WFPC2, 281, 291, 333
photometry
ACS, 23, 31, 47, 61
FGS, 367
NICMOS, 232
STIS, 105, 184
WFPC2, 281, 301, 315, 329, 333, 346,
350, 359
plate scale, 13, 58, 311, 373
point spread function (PSF)
ACS, 3, 13, 31, 58, 61, 86, 394
KECK, 394
NICMOS, 249, 267
STIS, 137, 148
WFPC2, 281, 301, 311, 341, 354
polarimetry, 259
preflash, 359
pydrizzle, 53, 70, 325, 337
PYRAF, 53, 70, 325, 337
read noise, 3, 97, 180, 193, 222
red leaks, 350
repeller wire, 86, 189
saturated data, 346
scattered light, 3, 61, 97, 209
sensitivity
ACS, 3, 31, 47, 61, 86
STIS, 97, 105, 115, 127, 201, 205
WFPC2, 333, 359
servicing mission, 281, 329, 341, 350, 373
shading, 222
sky subtraction, 301, 325, 359
slitless spectroscopy
ACS, 3, 38, 74, 90
NICMOS, 241
STIS, 97, 197
WFC3, 383
spectroscopy
COS, 390
STIS, 97, 115, 148, 162, 171, 180, 189,
193, 197, 201, 209
STIS/CCD, 97, 105, 115, 137, 148, 162,
176, 180, 193, 197, 201, 205, 346
STIS/FUV-MAMA, 97, 115, 127, 162, 171,
189, 201, 205, 209
STIS/NUV-MAMA, 97, 115, 127, 162, 189,
201, 205, 209, 390
STSDAS, 53, 70, 97, 105
surface photometry, 82
SYNPHOT, 97, 281, 333, 350
throughput
ACS, 3, 31, 47, 61
NICMOS, 241
STIS, 115, 201
WFPC2, 315, 341
uv throughput, 86, 201, 281, 341, 350
velocity aberration, 58
vignetting, 3, 315
wavelength dispersion solution
ACS, 38
COS, 390
NICMOS, 241
STIS, 127, 162, 171, 184
WFC3/IR, 383
WFC3/UVIS, 197, 383
WFPC2, 281, 291, 301, 311, 315, 325, 329,
333, 337, 341, 346, 350, 354, 359