WFPC2 Instrument Science Report 95-061
A Field Guide to WFPC2
Image Anomalies
J. Biretta, C. Ritchie, and K. Rudloff
November 2, 1995
ABSTRACT
We describe many of the less-common artifacts and anomalies which appear in WFPC2
images. Examples are given with brief descriptions of their cause, correction, and prevention.
1. Introduction
Maximizing the science return of WFPC2 images requires familiarity with various artifacts and anomalies which can occur. Common artifacts, such as cosmic rays and hot pixels, are familiar to most observers and there are effective observing strategies and analysis
tools for removing them. Other anomalies and artifacts are less well-known and less common, and yet impact many hundreds of images per year. These lesser-known anomalies
are the subject of this report. Fortunately, with careful planning and data analysis, most
can be prevented or removed from the data.
Several anomalies are associated with observations of bright stars and other objects attaining high exposure levels.These include ghost images, OTA diffraction effects, and horizontal smearing. These effects all make it difficult to detect faint companions of bright
targets, but can be eliminated with proper observation planning and careful data analysis.
Residual image effects are also associated with high exposure levels, though they primarily impact subsequent exposures of faint targets.
There are also artifacts associated with bright objects which are merely near the WFPC2
field of view. These include PC stray light and the “dragons breath” patterns. Here the
simplest cure is prevention by placing the offending object farther from the field of view.
Contamination patterns, or “worms,” are readily visible only in planetary and flat field
1. Copies of this report may be obtained from the Science Support Division, Space Telescope Science Institute, 3700 San Martin Drive, Baltimore MD 21218, by e-mail to help@stsci.edu, by anonymous FTP to stsci.edu directory instrument_news/WFPC2, and by WWW at http://www.stsci.edu/
instruments.html
1
exposures, but yet affect the calibration of all images at the 1% to 2% level. These can be
reduced by planning observations early in decontamination cycles, or by correcting
images during analysis.
Some artifacts occur completely at random, and no reasonable degree of planning will
eliminate them. While careful analysis techniques can reduce their effects, the only complete cure is taking multiple data frames of the same field and being prepared to discard
affected image regions. These anomalies include moving targets, missing data, jumps in
bias level, and bias ripples. Similarly, scattered Earth light, which impacts long exposures
in broad filters, will be difficult to avoid during planning, but can be eliminated when combining multiple frames of a given field.
Herein we describe these image anomalies, show examples, and discuss steps which can
be taken to prevent or correct them. The anomalies are ordered starting with those internal
to the CCD, and proceeding backwards along the light path, through the OTA, and finally
concluding with those caused by data transmission to the ground. The final section (Section 14) discusses correction of some anomalies by rejecting bad image regions during
image combination. Common artifacts, including cosmic rays and hot pixels, are covered
in the WFPC2 Handbook (v. 3.0) section 4.2.
Observers who feel their data have been severely compromised by some anomaly should
be aware that they can file a HST Observation Problem Report (HOPR) within three
months of the data distribution and request a repeat of the observation. The procedure for
this is described at http://presto.stsci.edu/public/hopr_instruct.html. Once a repeat observation is granted the current policy is for the corrupted image(s) to become public. This
prospect should be weighed when deciding whether to request a repeat observation.
2
Table 1: Contents of This Report
Section
Topic
Page
1
Introduction
1
2
Jumps in Bias Level
4
3
Bias Ripples
14
4
Horizontal Smearing of Saturated Images
16
5
Residual Images
19
6
Contamination Worms
36
7
PC1 Stray Light Patterns
38
8
“Dragon’s Breath” Reflection
55
9
Ghost Images
58
10
OTA Diffraction Patterns
64
11
OTA Earth Reflection Pattern
69
12
Moving Targets
76
13
Missing Data
79
14
Combining Dithered Images with Artifact Rejection
81
15
References
85
3
2. Jumps in Bias Level
Description: One or more horizontal bands appear when an image is displayed at high
contrast near the background level. Band(s) are typically about 0.5 DN higher or lower
than rest of image. Effect is usually limited to one (or rarely two) CCDs.
Cause: The CCD bias level is unstable and changes during readout. A small bias voltage
is applied to the CCD signal during readout in order to prevent zero-volts crossings. During pipeline calibration a separate bias level is estimated for each CCD from values in the
CCD over-scan columns (.x0h files), and then a single bias level is subtracted from each
group (CCD). Since the pipeline processing system assumes the voltage is constant, a
jump or anomaly is seen in the image if the bias voltage changes during readout. The voltage changes are thought to be caused by other spacecraft activities occurring during CCD
readout, but the exact cause is not known. This is perhaps the most common single anomaly and impacts roughly one to two percent of the images taken.
Examples: Figure 2.1 shows a bias jump in PC1. Figure 2.2 shows different jumps in
WF3 and WF4. Figure 2.3 shows a pair of weak jumps, both in WF4. In each case these
appear as a discontinuity in the background level of the affected CCD(s). The “edge” of
the discontinuity runs along the CCD X direction (normal to the readout direction).
Impact: These jumps tend to have little impact since most of the effect is correctable and
occurs near the CCD noise level. A few CCD rows right at the jump may be difficult to
correct and this might prevent detection of faint targets in those rows.
Correction: Bias jumps are perhaps the most easily corrected anomaly. The normal pipeline calibrated data (.c0h file) are first displayed and examined to get a rough location of
the discontinuity (Figures 2.4 and 2.5). Using the image display cursor it appears that the
jump occurs near row Y=285 in group [3]. Next we use the .x0h overscan data and
IMSTAT (in IRAF area cl.image) to locate the jump more precisely. The IMSTAT parameters are set by the command “epar imstat.” Only columns 5 through 14 of the .x0h file are
used since the other columns usually contain artifacts.
PACKAGE = images
TASK = imstatistics
images
(fields
(lower
(upper
(binwid
(format
=
=
=
=
=
=
u2650z08t.x0h[3][5:14,283]
image,npix,min,max,mean
INDEF
INDEF
0.1
yes
4
With result:
#
IMAGE
NPIX
u2650z08t.x0h[3][5:14,283] 10
MIN
MAX
305.
MEAN
307.
305.7
We then repeat this for a few more rows with results:
#
IMAGE
NPIX
u2650z08t.x0h[3][5:14,284] 10
MIN
#
IMAGE
NPIX
u2650z08t.x0h[3][5:14,285] 10
MIN
#
IMAGE
NPIX
u2650z08t.x0h[3][5:14,286] 10
MIN
MAX
305.
MEAN
307.
MAX
305.9
306.
MEAN
308.
MAX
306.7
306.
MEAN
307.
306.3
From the “mean” values in each row of the .x0h file, it is apparent that the jump is between
rows 284 and 285. Next we use IMSTAT to compute the mean of the .x0h file before and
after the jump:
imstat u2650z08t.x0h[3][5:14,1:284]
with result:
#
IMAGE
NPIX
MIN
MAX
MEAN
u2650z08t.x0h[3][5:14,1:284] 2840 304. 308. 305.666
And after the jump:
imstat u2650z08t.x0h[3][5:14,285:800]
with result:
#
IMAGE
NPIX
MIN
MAX
MEAN
u2650z08t.x0h[3][5:14,285:800] 5160 304. 309. 306.666
From these results it appears that the bias value is 305.666 DN before the jump, and
306.666 DN after the jump. We also need the mean value of the bias level in all the .x0h
rows, which can be computed from the above statistics as
2840 × 305.666 + 5160 × 306.666
---------------------------------------------------------------------------------- = 306.311
2840 + 5160
Next we generate a one-group correction image and apply it to group [3] of the data. We
start by making an image which contains the difference from the mean value 306.311
above and below the jump using IMCALC in stsdas.toolbox.imgtools. Above the jump the
image value will be 306.311-305.666=0.645, and below the jump it will be 306.311306.666=-0.355:
5
PACKAGE = imgtools
TASK = imcalc
input
output
equals
(pixtype
(nullval
(verbose
=
=
=
=
=
=
u2650z08t.c0h[3]
corr1.hhh
if y.lt.285 then (0.645) else (-0.355)
old
0.
yes
We must apply the flat field correction to this image. The same flat field reference file used
in pipeline calibration should be used, and it can be obtained through the usual STARVIEW interface to the HST archive. Here the flat field reference file is e391434fu.r4h.
imarith corr1.hhh * e391434fu.r4h[3] corr2.hhh
If the exposure time is shorter than 10 sec. and a shutter shading correction was applied in
the pipeline (SHADCORR=COMPLETE in .c0h header) this correction should also be
applied here. Here e371355eu.r5h is the shutter shading file used in the pipeline (from
SHADFILE in .c0h header) and the image has an exposure time of 6 sec.
PACKAGE = imgtools
TASK = imcalc
input
output
equals
(pixtype
(nullval
(verbose
=
=
=
=
=
=
corr2.hhh,e371355eu.r5h[3]
corr3.hhh
im1 * (1. + im2 / 6.)
old
0.
yes
Finally we make a duplicate copy of the pipeline calibrated file while applying the bias
jump correction to group [3]:
imcopy u2650z08t.c0h[1] output.c0h[1/4]
imcopy u2650z08t.c0h[2] output.c0h[2]
imarith u2650z08t.c0h[3] + corr3.hhh output.c0h[3]
imcopy u2650z08t.c0h[4] output.c0h[4]
And “output.c0h” is the final 4-group image (Figure 2.6). If multiple frames of the same
field are to be combined later on, it may be useful to mask-out or discard a few image rows
near the jump where the correction is sometimes imperfect. See Section 14.
Prevention: At present bias jumps cannot be prevented. They may be preventable in the
future by changing the way spacecraft activities are scheduled.
6
Figure 2.1. Mosaicked image showing bias jump on PC1. In mosaicked images the CCDs
(PC1, WF2, WF3, and WF4) are oriented as they appear on the sky, with the readout direction rotated 90 degrees between CCDs. Unless otherwise indicated, the intensity scales
throughout this report are inverted, so that high count levels appear black, and low count
levels appear white.
PC1 bias jump
WF4
PC1
WF2
WF3
7
Figure 2.2. Image with bias jumps on both WF3 and WF4.
8
9
Figure 2.3. Multiple bias jumps on WF4.
10
11
Figure 2.4. Bias jump in WF3 of image u2650z08t.c0h before correction. Jump is near
row X=285.
Bias jump in WF3 near row 285
12
Counts
Figure 2.5 Plots illustrating bias jump and its correction. These plots are made using the
PCOLS task in cl.plot; e.g. (A) is generated by the command “pcols u2650z08t.x0h[3] 5
14”. Figure (A) shows the bias jump in Figure 2.4 as seen in the .x0h overscan columns;
(B) shows the same jump as it appears in the data file before correction; and (C) shows the
data after correction (output.c0h). See text. After correction a small glitch remains near the
location of the jump.
13
Figure 2.6. Image u2650z08t.c0h after correction of bias jump in WF3 using method
described in text (image output.c0h).
Bias jump after correction
14
(Page intentionally left blank.)
15
3. Bias Ripples
Description: Faint horizontal bands appearing in one or more CCDs. Bands are regularly
spaced along CCD y direction with typically period of 5 or 6 pixels. Peak-to-peak amplitude of ripples is typically <7 electrons. Sometimes effect appears / disappears within
given image as it is read out.
Cause: Exact cause is unknown. May be caused by extraneous signals in spacecraft leaking into CCD readout electronics.
Example: Figure 3.1 shows a 400 x 400 pixel section of WF2 illustrating the ripple pattern.
Impact: Will make detection of faint objects difficult.
Correction: To the extent that the pattern is periodic and uniform, it may be possible to
model and remove it with Fourier analysis techniques. See King, Bartels, Nowak, and
Hack (1993) for treatment of similar problems in FOC data.
Prevention: Cannot be prevented.
14
Figure 3.1. Example of bias ripples; center 400x400 pixels of WF2.
15
4. Horizontal Smearing of Saturated Images
Description: Region to the right of saturated columns has elevated count levels. The effect
can also wrap around the CCD edge and give slightly elevated counts to the left of saturated columns.
Cause: Highly saturated pixels temporarily affect the readout amplifier causing elevated
output. The effect increases when more pixels per row are saturated and decays exponentially away from the saturated columns. If there are many saturated pixels per row, the
effect can last long enough to impact readout of the following CCD row. Hence the elevated background can wrap around to the left side of the CCD and back on to the right
side. See WFPC2 Handbook (v.3) section 4.2.4 and WFPC2 Instrument Science Report
94-01 for more discussion.
Examples: Figures 4.1 and 4.2 show examples of highly saturated stars which have
bloomed to the top and bottom of the CCD. The total counts are approximately 1.2 x 108
electrons in both images. Background regions immediately to the right of the bloom have
slightly higher count levels (about 28 electrons higher) than those to the left. In rows with
many saturated pixels, the effect wraps around the CCD edge and causes horizontal smearing on both the left and right sides of the target. Figure 4.2 shows a discontinuity in the
background level stretching across the image where the number of saturated columns per
row increases from 2 to 3.
Impact: The effect is largely correctable during analysis and hence there is little impact.
Correction: Images may be approximately corrected using algorithm given in WFPC2
Handbook (v.3) section 4.2.4 and WFPC2 Instrument Science Report 94-01.
Prevention: Reduce exposure time so as to avoid saturation. Can reduce effect by using
A-to-D gain = 15, as effect appears smaller in the gain 15 readout electronics.
16
Figure 4.1. Highly saturated star showing horizontal smearing of saturated columns on
WF4. Note increased background counts immediately to right of saturated columns, and
wrapping of smear off right CCD edge and back on to left edge. A-to-D gain 7.
Wrapping of horizontal
smear off right CCD edge
and back onto left edge
Bloom from
saturated star
17
Horizontal smear from
bright saturated star
Elevated background
to right of bloom /
saturated columns
Figure 4.2. Highly saturated star on WF3 showing background discontinuity where number of saturated columns per row changes. A-to-D gain 7.
Wrapped
discontinuity
Increase in
number of
saturated
columns
18
Background discontinuity in row
where number of
saturated columns
changes.
5. Residual Images
Description: Bright streaks along CCD Y-direction a few DN above background level.
Can also appear as small spots or large patches. More prevalent in long exposures. Pattern
can repeat in back-to-back images.
Cause: There are three distinct effects that produce residual images. All are aggravated by
high exposure levels. The first effect occurs in images exceeding full well capacity, and is
caused by charge being trapped at the Si-SiO2 interface in the CCD. This effect is similar
to that seen in WF/PC-1. The trapped charge is slowly released back into the imaging layers and appears in subsequent images. The residual image decays exponentially with a
time constant of ~15 minutes; residuals from strongly saturated images can require several
hours to completely clear.
A second effect, called “bulk residual image,” is peculiar to front-side illuminated devices
such as WFPC2. At long wavelengths, light can pass through the imaging layers of the
CCD and produce charge in the substrate. This charge slowly diffuses into the imaging
layers, and appears as a residual in subsequent images.
A third effect is also peculiar to WFPC2 and is caused by charge transfer efficiency (CTE)
problems. At all exposure levels some charge becomes temporarily bound to impurities
throughout the silicon. Over time the charge is slowly freed and appears in subsequent
images as a residual image. The effect is more prevalent after images with high exposure
levels (>10000 electrons per pixel), since the charge fills a larger volume of the silicon and
hence reaches more capture sites. The residual image appears both at the location of
objects during earlier exposures, and also in pixels below those objects (lower Y values)
where their charge passed during readout. The timescale for dissipation of CTE residuals
is uncertain, but is thought to be of order minutes.
Note that the CTE effect also impacts photometry, since the charge which we see as
“residual” in one image is, in fact, missing from some previous image. The photometry
effects and their correction are detailed by Holtzman, et al. 1995 (a and b).
Examples: Figures 5.1.a and 5.1.b show a saturated star on PC1 and the residual image
seen in an 1800 sec. dark calibration frame which was started six minutes later. Note that
the residual image has three distinct parts: (1) a narrow vertical line from the top to the
bottom of the CCD about 7 electrons above background where charge bled during the
exposure; (2) a small bright spot at the location of the star about 80 electrons above the
background; and (3) a broad vertical streak about 5 electrons above background from the
location of the target to the bottom of the CCD where charge passed during readout. This
last effect is caused by the CTE problem.
Figures 5.2.a and 5.2.b show a heavily exposed (but not saturated) image of Venus and the
subsequent CTE residual image. The peak counts in Venus are about 36,000 electrons per
pixel, which is significantly below the CCD saturation level of ~80,000 electrons. The
subsequent image is an 1800 second dark calibration frame which was started 12 minutes
after the Venus image. The CTE effect causes a 4 electron residual image and trail along
the readout direction.
19
Figure 5.3.a shows a flat field with saturation and blooming in the corners of PC1, WF2,
and WF4. Figure 5.3.b shows the corresponding residual in a 400 second exposure taken
25 minutes later; the residual image is about 15 electrons above background. Note that no
CTE trail is seen downstream from the saturated regions; this is in part because the traps
there were already filled by the flat field exposure. This residual pattern persists into later
exposures; an 1800 second dark frame taken 67 minutes after the flat still shows the same
pattern, 4 electrons above background.
Figure 5.4 illustrates the CTE residual images which occur in a dark calibration frame
taken immediately after an image of a star field. Both saturated and unsaturated stars leave
trails from their location to the bottom of each CCD along the readout direction. Here
images were taken in a sequence: star field, dark, star field, dark, etc., and the star fields
and darks have each been combined to remove cosmic rays.
Figure 5.5 shows CTE residuals which can arise in a sequence of back-to-back images of
the same field. Here a 1500 sec. F336W image (shown) follows 4 minutes after a 700 sec.
F814W image of the same field. Residual charge from the prior F814W exposure is clearly
visible as trails from the location of each star to the bottom of the CCD along the readout
direction.
As Figure 5.6 illustrates, residual images need not be associated with highly exposed data
frames. They can be generated anytime the shutter is opened (e.g. during cal-channel mirror moves, etc.).
Impact: Sky level will be difficult to determine in regions with residuals. Can reduce
effect by combining pointing-dithered images of the same field (see below).
Correction: Narrow streaks and small spots can be removed by masking (discarding)
these regions when combining multiple pointing-dithered frames of the same field. See
Section 14.
Prevention: Observers can take several steps to reduce residual image problems. It is relatively rare for observations in one science program to be impacted by residuals from a previous science program. This is because the nominal slew and setup times between
programs are usually enough to allow most residuals to dissipate. Observers who are concerned about the slight possibility of residual images from previous programs should
schedule their least sensitive observations first within the program; this will allow residuals to dissipate before more sensitive observations commence.
Within a multiple-orbit science program, residual image problems can be reduced by
scheduling the most sensitive exposures and those with the lowest count levels first on
each orbit (except possibly the first orbit, as noted above). Exposures with high count levels should be placed at the ends of orbits, so as to allow dissipation of the residual during
Earth occultation. Also, dithering the telescope pointing with POS TARGs between exposures will allow affected image regions to be discarded when combining multiple frames.
A common strategy is to take exposures at both POS TARG 0, 0 and POS TARG 0.501,
0.501, which is nearly an integral pixel shift on both the PC and WFC. See Section 14 for
an example of image combination with artifact rejection.
20
Observers who anticipate taking highly saturated images are requested to identify such
exposures in their Phase II proposal comments. This will help STScI staff in scheduling
subsequent science programs so as to eliminate residuals. Automatic bright target warnings are under study, and might eventually eliminate cross-contamination between science
programs.
Eventually it may be possible to eliminate the CTE problem and its residuals by preflashing the CCDs with a low exposure level or by filling charge capture sites with INTFLATSs
or VISFLATs prior to science exposures. These methods are currently being tested.
21
Figure 5.1.a. PC1 image of star with saturation and blooming.
22
23
Figure 5.1.b. Residual image in dark exposure resulting after saturated star in 5.1.a.
24
narrow line
bright spot
broad streak
from readout
(CTE residual)
Readout Direction
25
Figure 5.2.a. Non-saturated image of Venus on WF2 (mosaicked)
26
Figure 5.2.b. Residual image left by exposure and readout of Venus image.
27
Readout Directions
CTE
Residual
Image
28
Figure 5.3.a. Saturated / blooming flat field exposure (linear ramp filter).
29
30
Figure 5.3.b. Residual image left by saturated flat field in Figure 5.3.a.
31
Residual Image
Readout Directions
Readout direction
32
Figure 5.4.a. Image of star field. Four images have been combined to remove cosmic rays,
and then the 4 CCDs were mosaicked together.
WF4
PC1
WF2
WF3
33
Figure 5.4.b.Combined series of four dark calibration frames showing residual images left
by the star field in Figure 5.4.a. Each of the four 1800 sec. dark images followed one of the
star field exposures by about 8 minutes. Trails caused by CTE problem can be seen starting at the location of each star, and then continuing along the readout direction.
Readout Directions
34
Figure 5.4.c. PC1 image from Figure 5.4.a showing detail of star locations.
35
Figure 5.4.d. PC1 image showing detail of CTE residual trails caused by previous exposure (Figure 5.4.c). Brightest trails are about 4 electrons high. Compare against star locations in 5.4.c. Figures 5.4.a and 5.4.b show 4-CCD mosaic of these images.
36
Figure 5.5.a. Image of star field showing CTE residual trails from prior exposure of same
field. This 1500 sec F336W image follows 4 minutes after a 700 sec. F814W exposure,
which is responsible for the trails. Brightest trails are about 6 electrons high.
Readout Directions
37
Figure 5.5.b. Detail of PC1 image from Figure 5.5.a (central 500 x 500 pixels). Each moderately bright star has a trail extending downward along the readout direction. Trails are
caused by CTE problem in previous exposure of same field.
38
Figure 5.6. Faint image of Mars appearing in a 2200 sec. F160BW VISFLAT. A 200x200
pixel detail of WF4 is shown. This image is thought to be a “bulk residual” resulting not
from a prior science image, but instead from saturation occurring while the shutter was
open (2-3 sec.) to allow positioning of the cal-channel flip mirror for the VISFLAT. Mars
happened to be in the field of view with the FR868N filter in place during the mirror move.
Brightest Mars regions (white) are ~ 100 electrons high.
39
(Page intentionally left blank.)
40
6. Contamination Worms
Description: Weak, irregular lines in images of bright extended targets (e.g. planets) with
high count rates.
Cause: Contaminant particles slowly collect on cold CCD windows (field flatteners) in a
quasi-uniform layer. For reasons which are not entirely clear, the contaminants avoid linear regions on the windows, thus giving rise to reduced scattering and higher counts in
these areas at visible wavelengths. In the far UV these linear features have low transmission, suggesting the presence of some permanent contaminant which repels new material.
Approximately once per month the CCDs are warmed and the contaminants are boiled off;
this “decontamination” temporarily eliminates the worm pattern until the contaminant
layer reforms.
Example: Ratio of F439W internal VISFLATs taken before and after a decontamination
for CCD WF2. The ratio is roughly 0.988 over most of the image due to light scattering
from contaminant layer in pre-decon. VISFLAT. Ratio is 1.002 in the “worm” features,
indicating an absence of contaminants.
Impact: Contributes false structure to extended targets. Since it is a weak effect (1-2%), it
will only be visible at high count levels.
Correction: In principle, the internal VISFLATs which are taken weekly can be used to
estimate and remove the pattern. One must also consider that the flat field reference files
used for calibration will also contain the worm pattern to some degree, and will partially
correct the worm pattern in science images. Eventually “deltaflats” may be made available
to allow easy correction of contaminant structures.
Prevention: In critical situations, one might request that the observations be scheduled
just after one of the monthly decontaminations. This request can be made in the RPS2
Phase II Visit Level Comments, and by contacting the STScI Program Coordinator for
your program.
36
Figure 6.1. Ratio of F439W internal VISFLATs taken before and after a decontamination
for CCD WF2. The image taken before the decontamination has linear regions (worms)
where the counts are about 1% larger (darker in this display) than the surrounding regions.
37
7. PC1 Stray Light Patterns
Description: One or more broad arcs of light on PC1. In some cases there is a single
bright arc with sharp edges near the pyramid apex (type 1 or “PC direct stray light” pattern). In other cases there are multiple arcs with a softer appearance together with a diffraction spike at the outer edges of the PC (type 2 or “PC diffraction stray light” pattern).
In both cases the arcs continue into the pyramid shadows along the X~0 and Y~0 edges of
the CCD.
Cause: There are two similar but distinct patterns. Both are caused by a bright star which
is on the PC pyramid facet, but outside the region imaged by the PC CCD.
The single arc with sharp edges near the pyramid apex (type 1 or “PC direct stray light”
pattern) is caused by a moderately bright star (usually m<14) falling near the outer corner
of the PC1 pyramid facet. This stray light is caused by incomplete baffling of the PC Cassegrain relay secondary mirror, which allows light to take a non-standard path through the
PC1 camera relay optics -- light passes directly from the articulating fold mirror (AFM) to
the CCD, without the usual re-imaging. The pattern has been successfully reproduced by
ray-tracing the WFPC2 optics. The intensity of the stray light can be estimated from the
count rate of the offending star, the F/24 OTA focal ratio, the specification of the AFM
(concave with radius 587 cm and located 16.0 cm from OTA focal plane), and the distance
from the AFM to the CCD (110.0 cm). Stars producing this pattern have a blur diameter of
5.0 cm, or ~3500 PC pixels, at the CCD. Hence a star which would generate 1 electron
sec-1 when imaged in the normal manner, will thus produce a “direct” pattern with an average intensity of 1.1 x 10-7 electrons sec-1 pixel-1. This is why the stray light pattern is only
seen for bright stars -- the light is severely out of focus so patterns due to faint stars are lost
in the background and read noise. The pattern can reach two or three times this average
intensity along its edges, where diffraction effects are important.
The type 2 or “PC diffraction stray light” pattern is caused by a moderately bright star
(usually m<13) located anywhere in the L-shaped region which is reflected by the pyramid, but is not imaged on the PC (i.e. the “missing” region of the PC field). Light from
objects in this region cannot pass directly from the AFM to the CCD; however, some combination diffraction, scattering, and multiple reflection still allows light to reach the CCD.
The maximum intensity in the pattern is about 1.3 x 10-7 electrons sec-1 pixel-1 for a star +
filter producing 1 electron sec-1 (similar to mean intensity in the “direct” pattern).
Examples: Figure 7.1.a shows a typical example of a “PC direct stray light” pattern with a
sharp arc near the pyramid apex. The GASP (Sky Survey) chart in Figure 7.1.b shows the
presence of an 11.0 (+/- 0.2) magnitude star at the outer corner of the PC1 pyramid facet.
For this star, the F702W filter, and the 140 sec. exposure, we would predict a pattern with
5+/-1 electrons pixel-1, which is close to the 7+/-1 electrons pixel-1 in the observed pattern.
Figures 7.2.a and 7.2.b show another example -- a 1500 sec. exposure in F606W with an
m=13.8 star at the outer corned of the PC pyramid facet. Here we predict 6 +/-1 electrons
pixel-1 in the pattern, and 6 electrons pixel-1 are observed. Figures 7.3.a and 7.3.b show a
variation where the offending star (m=11.4) is along an outer edge of the PC pyramid
facet, but ~20 arcseconds from the outer corner.
38
Figure 7.4.a shows a type 2 or “PC diffraction stray light” pattern with two arcs which are
softer in appearance. The diffraction spike provides a hint of the bright star just off the
PC1 field and Figure 7.4.b clearly shows the situation. Figures 7.5.a and 7.5.b show a similar example where three arcs are seen.
Figure 7.6.a and 7.6.b show an intermediate case between type 1 and type 2. The star is
located along the outer edge of the PC1 pyramid facet, but far from the facet corner.
This anomaly also impacts flat fields for PC1. For both internal and external flats, light
striking the outer corner of the PC pyramid facet is imaged into a broad arc on PC1 near
the pyramid apex. Hence the flats are too high in this region, and the flat fielded data will
be too low. The affected region is illustrated by Figure 7.7 which shows a POLQN18 +
F555W VISFLAT divided by an F555W VISFLAT. The negative 18 degree rotation of the
polarizer filter wheel has caused the clear (or open) filter hole on that wheel to move in
front of the outer corner of the PC pyramid facet, while light striking the PC through the
normal optical path still passes through the polarizer. Hence the pattern is enhanced over
its normal intensity, and it becomes clearly visible in the ratio image as a broad arc near
the pyramid apex in Figure 7.7. The effect on flats can be deduced from the polarizer
transmission and the 8% height of the observed stray light pattern -- flats will be in error
by 4% to 5% over this region of PC1. We note that the sky background will be properly
flattened, since the illumination is identical to that of the flats, but objects will suffer this
4% to 5% error.
Impact: Faint targets can be lost over large regions of PC1. Also, objects will suffer 4% to
5% flat fielding errors on PC1 in the “direct” pattern region near the pyramid apex (see
Figure 7.7).
Correction: Can attempt to derive and subtract local background inside the pattern.
Prevention: STScI routinely provides observers with GSSS (or GASP) finding charts of
all targets prior to final proposal scheduling. Observers can use these charts and the overlay in Figure 7.8 to check for bright stars in problem regions of the WFPC2 field of view.
The standard target positions (aperture locations PC1, etc.) are indicated on the overlay.
The overlay also contains images of stars which are taken from the GSSS images, which
can be used to estimate stellar magnitudes (+/- 0.5 mag.). These star images are given for
sky survey plate exposures of 5, 20, and 70 minutes (listed on GSSS charts as “Exposure
(min)”). The overlay should be rotated counter-clockwise about the selected aperture
(PC1, etc.) by the angle specified in the ORIENT special requirement in the phase II proposal. The avoidance magnitudes indicated on the overlay assume a 600 sec. exposure in
filter F555W, and would result in stray light intensities reaching about 1 electron pixel-1.
39
Figure 7.1.a. Mosaicked image showing type 1 “direct” stray light pattern on PC near pyramid apex.
40
Figure 7.1.b. GASP image for type 1 pattern in Figure 7.1.a, showing bright star (11.0
magnitude) near the outer corner of the PC pyramid facet.
41
Figure 7.2.a. Mosaicked image showing type 1 “direct”pattern on PC near pyramid apex.
42
43
Figure 7.2.b. GASP image for type 1 pattern in Figure 7.2.a, showing a moderately bright
star (13.8 magnitude) near outer corner of the PC pyramid facet.
44
Figure 7.3.a.Variation of PC type 1 “direct” stray light pattern. Note that pattern is shifted
relative to those in Figure 7.1 and 7.2. The shift occurs when the offending star is along
outer edge of PC1 pyramid facet, instead of at outer corner.
45
Figure 7.3.b. GASP chart for Figure 7.3.a showing bright star (m=11.4) near outer edge of
PC1 pyramid facet.
46
Figure 7.4.a. PC type 2 or “diffraction” pattern with two arcs and spike.
46
47
Figure 7.4.b. GASP image for PC type 2 or “diffraction” pattern in Figure 7.4.a, showing
bright (10.8 magnitude) star on the PC pyramid facet just outside PC field of view.
48
Figure 7.5.a. PC type 2 or “diffraction” pattern showing multiple arcs and spike.
49
50
Figure 7.5.b. GASP image for Figure 7.5.a showing bright (9.4 magnitude) star on the PC
pyramid facet just outside the PC field of view.
51
Figure 7.6.a. PC1 stray light pattern which is intermediate between type 1 and type 2.
“Direct” ghost pattern 8%
above normal level
52
[Vignetted region due
to polarizer rotation]
Figure 7.6.b. GASP image for Figure 7.6.a. A bright star is present along outer edge of PC
pyramid facet, but far from the outer corner.
53
Figure 7.7 Ratio of a POLQN18 + F555W visflat divided by a F555W visflat. The “direct”
stray light pattern is visible as a bright arc in the lower left (near pyramid apex). The
POLQ rotation causes light forming the pattern to pass through the open position on the
filter wheel, and hence the pattern is intensified and becomes apparent in this ratio image.
54
Figure 7.8. GASP chart overlay showing bright star avoidance regions near WFPC2 field
of view. Stated avoidance magnitudes assume 600 sec. exposure in F555W producing 1
electron pixel-1 stray light pattern. Sample GASP star images with approx. magnitudes are
shown at bottom.
U3
Rotate about aperture
(PC1, etc.) to
Phase II
“Orient” angle
from North
PC direct ghost region avoid < 15th mag stars
PC diffraction ghost
region - avoid < 14th
mag stars
PC1
WF2
WF4
WFALL
WF3
“Dragon’s breath”
region - avoid < 15th mag
stars
120”
Magnitude
GASP Exp.
10.5
11
11.5
12
12.5
5 min.
20 min.
70 min.
55
13
13.5
14
14.5
15
15.5
(Page intentionally left blank.)
56
8. “Dragon’s Breath”
Description: Fan-shaped light pattern entering from outer edge of WFPC2 field of view.
Cause: Scattered light from bright star just outside WFPC2 field of view. Details are
unclear, but the effect almost certainly involves scattering of light in the region along the
outer edge of the pyramid mirror which is not normally imaged by the CCDs. This “unimaged” border on the pyramid is approx. 6 arcseconds wide. See figure 7.8.
Example: Figure 8.1.a shows pattern on WF4 caused by bright star (m=11.4) which is 6
arcseconds outside field of view (Figure 8.1.b). The OTA diffraction spikes from the star
are clearly visible. Local maxima in the pattern reach several thousand electrons pixel-1, or
about 1.3 x 10-5 of the total counts in the offending star.
Impact: Data lost / severely corrupted over small portion of field.
Correction: None.
Prevention: Use GSSS (GASP) charts provided by STScI during proposal scheduling,
along with overlay in Figure 7.8, to check for stars near WFPC2 field of view. It is probably sufficient to avoid m<15 stars, though in principle, stars as faint as m~19 could produce noticeable artifacts (intensity 1 electron pixel-1 assuming 600 sec. exposure in
F555W). The exact limits of the problem region are unknown, however, it should be sufficient to avoid bright stars within 6 arcsec. of the WFPC2 field (region reflected by pyramid
mirror, but not imaged onto CCDs). If one includes guide star position uncertainties, an 8
arcsecond avoidance band is perhaps better. Use POS TARGs in Phase II proposal to move
bright stars farther from WFPC2 field.
55
Figure 8.1.a. Mosaicked image showing dragon’s breath pattern on WF4.
“dragon’s breath”
56
Figure 8.1.b. GASP chart for Figure 8.1.a. showing bright star just off WF4 which causes
“dragon’s breath” anomaly.
57
9. Ghost Images
Description: Weak doughnut-shaped or fan-shaped artifacts near bright stars.
Cause: Internal reflections in CCD field-flattener lenses or in filters.
Each CCD has a plano-concave field-flattener lens made of MgF2. A ghost arises when
light passing though the lens is internally reflected from the surface near the CCD, to the
surface opposite the CCD, and then back through the lens and onto the CCD. The fieldflattener ghosts are always on the side of the star opposite the CCD center. The distance
between the star and ghost increases as the distance from the CCD center increases; if D is
the distance from the CCD center to the star, the ghost is between ~0.36D to ~0.41D from
the star on the WF CCDs, and about ~0.25D to ~0.30D from the star on the PC. The total
light in the ghost is determined by the refractive index of MgF2; the ghost contains about
0.15% of the light in the main image. In the WFC the ghost is doughnut-shaped with
diameter ~10 pixels, which is determined by the thickness of the lens which adds about 2
mm of defocus to the ghost optical path, and the camera f-ratio. The doughnut-shape is
simply the shape of the defocussed OTA + camera relay pupil. On the PC the ghosts are
only about 3 pixels in diameter due to the slower F-ratio.
The filter ghosts are fan-shaped (comatic), and are caused by reflections between different
surfaces within a filter. The coma results from misalignment of the OTA pupil on the corrective camera relay secondaries for the ghost optical path. Their location varies from filter
to filter, but for a given filter the offset from the star to the ghost (in sky coordinates) will
be constant over the field of view. They tend to be stronger for narrow band filters, which
are interference filters with little internal absorption. The strength of the ghost varies from
filter to filter, but is generally less than 1%. Multiple ghosts can be seen, which are due to
repeated internal reflections.
Examples: Figures 9.1 through 9.3 show a variety of field-flattener ghosts. In all cases the
star is located on a line from the CCD center and through the ghost image.
Figure 9.4 shows a filter ghost for the F502N filter. The first order ghost contains about 2%
of the total starlight, while the second order ghost contains about 0.04% (0.022).
Impact: Will hamper searches for faint objects near bright ones.
Correction: The field-flattener ghosts can be removed when combining multiple images
of the same field which are observed with large (>100 pixel) position dithers. See Section
14. (Note, however, that optical distortion will complicate analysis of images with large
dithers.) The filter ghosts can be removed when combining images of the same field which
are observed at different spacecraft roll angles (ORIENTs).
Prevention: Observe with position dithers, or different spacecraft roll angles, to allow
removal of artifacts during analysis.
58
Figure 9.1. Example of field-flattener ghost in WF4. Image shows entire CCD.
Bright Star
Ghost
59
Figure 9.2. Example of field-flattener ghost in WF2. Image shows entire CCD.
Bright star
60
Ghost
Figure 9.3.a. Another field-flattener ghost on WF4. Here the star contains about 7 x 106
electrons. Note that the ghost image always lies on a line which passes through the CCD
center, through the star, and finally through the ghost. Image shows entire CCD.
Bright star
Ghost
61
Figure 9.3.b. Detail of field-flattener ghost on WF4 in Figure 9.3.a.
Bright star
Ghost
62
Figure 9.4. Example of filter ghosts in F502N filter. Intensity scale is logarithmic. The star
contains about 4 x 108 electrons. Of this about 2% are in the first-order ghost, and 0.04%
are in the second-order ghost.
1st order ghost
2nd order ghost
63
10. OTA Diffraction Patterns
Description: Faint rings and spikes surrounding bright stars.
Cause: Inner rings are caused by normal Airy disk. Roughness in OTA contributes to
outer rings. Spikes are caused by secondary mirror support spiders.
Examples: Figure 10.1 shows a typical example of the inner Airy rings; rings have peak
intensities near 0.5% that of star. Figures 10.2 and 10.3 show PSF structure at larger radii.
Figure 10.4 shows a plot of azimuthally averaged brightness of PSF.
Impact: Will be difficult to detect faint targets near bright stars.
Correction: Subtract model of diffraction pattern derived from other stars, or from TinyTIM or TIM models. Take separate exposures of bright stars during science program to
use as PSF model. Observe at different spacecraft roll angles (ORIENTs) to move spikes
and features in PSF relative to any faint targets on sky.
Prevention: Can minimize impact of diffraction spikes and other features by observing at
different spacecraft roll angles (ORIENTs).
64
Figure 10.1. Section of PC image showing two stars with OTA diffraction rings. Peak
counts in each star are about 14,000 electrons per pixel; peak counts in diffraction rings
are about 80 electrons per pixel. Radius of ring is about 7 pixels. Filter is F814W.
65
Figure 10.2. Region of PC image (250 x 250 pixels) showing heavily exposed star with
OTA diffraction rings and spikes. Filter F502N. Star image has total of 3.6 x 106 electrons
with blooming and approximately 75 saturated pixels. Peak intensities for visible diffraction rings range from ~2000 (inner rings) to ~10 electrons per pixel (outer rings). Intensities in diffraction spikes range from 3500 electrons per pixel near star, to a few electrons at
image corners. Filter ghost is also visible below and right of star.
66
Figure 10.3. Region of PC image (500 x 500 pixels) showing heavily exposed star with
OTA diffraction rings and spikes. Filter F502N. Star image has total of 3.6 x 108 electrons
with blooming and approximately 4200 saturated pixels. Peak intensities for visible diffraction rings range from ~1400 (at radius 60 pixels) to ~140 electrons per pixel (at radius
120 pixels). Peak intensities in diffraction spikes range from ~7000 electrons (at radius 60
pixels) to about 400 electrons (at image corners). Filter ghost is also visible below and
right of star.
67
Figure 10.4. Azimuthally averaged PSF brightness as function of radial distance. Star is
located near the center of PC1. The discontinuity between PC and WF4 results from scattering within the PC camera. From Krist 1995.
68
11. OTA Earth Reflections
Description: Elevated background level throughout field of view. Dark diagonal X’s and /
or bars in some CCDs which are a few DN below background.
Cause: Light from bright sun-lit Earth is reflected off the OTA baffles and spider (secondary mirror supports) and into WFPC2. These reflections tend to occur when the bright
Earth is < 25° degrees from the OTA axis, such as at the beginning and end of exposures,
and for CVZ targets. The default Earth avoidance angle is 20° , and science observations
are not scheduled at smaller angles. At low Earth angles light is scattered throughout the
WFPC2 field raising the overall background level. Dark diagonal bars can appear in some
CCDs if bright OTA spider vane(s) pass into shadows of the WFPC2 camera relay spiders,
blocking the stray light along diagonal regions of the CCDs. Figure 11.1 illustrates these
effects and the formation of a dark bar in WF4. Earth light scattering tends to be significant only in broad visual and infrared filters (F555W, F606W, F814W, etc.).
Examples: Figure 11.2.a shows a typical example where a dark X pattern appears in WF2,
and diagonal bars appear in other CCDs. The darkest portion of the X is about 40 electrons
below the background level, which is ~100 electrons. The circumstances of the reflection
pattern are illustrated in Figure 11.2.b. Along diagonal regions of each CCD, the bright
OTA spider vanes fall into the shadows of the relay optics spiders. In both WF2 and WF4
the bright OTA spider vane at 10:30 o’clock is hidden by the relay optics spider at 10:30,
hence forming a dark bar from the top left to bottom right of these CCDs. In PC1, WF2,
and WF3 the dimmer OTA spider vane at 7:30 is similarly blocked, forming dark bands
from the top right to bottom left of these CCDs. The angle between the OTA axis and
bright Earth limb was roughly 21 degrees throughout this 600 sec. exposure in F814W; the
target was in the continuous viewing zone (CVZ).
Figure 11.3.a shows another example where diagonal bars are seen and Figure 11.3.b illustrates the circumstances of the reflection. A pattern of strong bars is seen on PC1, WF3,
and WF4; a second pattern of very weak bars is seen on WF2 and WF4. This 1100 sec.
exposure in F814W begins with the bright sun-lit Earth limb 21 degrees from the OTA
axis; the Earth angle steadily increases and reaches 34 degrees by the end of the exposure.
Comparison with later images in this exposure sequence (same filter and exposure) indicate an excess background light of ~13 electrons in this image, of which ~5 electrons are
blocked by the camera spiders (i.e. depth of diagonal bar below the background level).
Impact: Difficult to get accurate sky subtraction over small regions of image (Xs and
bars); unfortunately these regions include the default aperture positions. May be able to
eliminate impact when combining multiple frames of same field, if pattern is present in
only one or two images. Note that CVZ targets are highly susceptible to this anomaly,
where observers are usually trying to reach very faint targets.
Correction: Measure scattered light intensity away from Xs and bars, and then subtract
this level from images (after flat fielding). Combine multiple frames of the same field
while discarding X and bar regions of affected image(s). See Section 14.
69
Prevention: Avoid small angles (< 25° degrees) between bright Earth and OTA axis. Consider requesting “LOW SKY” in Phase II proposal, where HST is in Earth’s shadow, but
note that this will eliminate much of the advantage of CVZ observations.
70
Figure 11.1. Interaction of OTA spider Earth reflections with WF4 relay optics spider at
different points on WF4 CCD. Approximate alignments of OTA and camera pupils and
spiders are shown at nine points in the field of view. This example is taken from WF4 in
Figure 11.2. (Pupil diagrams not to scale.)
OTA pupil
Camera pupil
Earth reflections
on OTA spider
Dark band across WF4 where
one reflection passes behind
camera spider
Pyramid apex
71
Figure 11.2.a. Typical Earth reflection pattern showing dark Xs and diagonal bars where
reflection is blocked by camera spiders.
72
Figure 11.2.b. Circumstances of reflection pattern in 11.2.a.
SPIDER
EARTH
REFLECTIONS
OTA
PC 1
CAMERA
OTA
WF 4
CAMERA
DARK BANDS
(REFLECTIONS HIDDEN
BY CAMERA SPIDER)
WF 3
CAMERA
OTA
WF 2
CAMERA
OTA
OTA
To
Sun
OTA
Door
73
To Earth
Figure 11.3.a. Earth reflection pattern with dark diagonal bars in PC1, WF3 and WF4.
74
Figure 11.3.b. Circumstances of reflection pattern in 11.3.a.
SPIDER
EARTH
REFLECTIONS
OTA
OTA
WF 4
CAMERA
PC 1
CAMERA
DARK BANDS
(REFLECTIONS HIDDEN
BY CAMERA SPIDER)
WF 2
CAMERA
WF 3
CAMERA
OTA
OTA
.
75
12. Moving Targets
Description: Bright streak(s) across WFPC2 field of view at random orientation. Streaks
are usually only a few pixels across although some up to ~50 pixels wide have been seen.
May appear in one or more CCDs. A mosaicked image will show a continuous streak
across full field of view when multiple CCDs are involved.
Cause: Bright spacecraft / space junk passing through the WFPC field of view during
exposure.
Examples: Figures 12.1 and 12.2 show mosaicked images with typical moving targets.
Impact: Data lost over small region of field in single images. Reduce impact by combining multiple images of same field.
Correction: Mask out (discard) affected pixels when combining multiple frames of the
same field. See Section 14.
Prevention: Occurrence cannot be prevented. Can reduce impact by taking multiple
images of field.
76
Figure 12.1 Mosaicked image showing moving target across WF2, PC1, and WF4.
moving target
77
Figure 12.2. Another moving target across WF2, PC1, and WF4.
moving target
78
13. Missing Data
Description: Very dark regions of image usually in the form of horizontal bars (in Xdirection). Closer examination will show these regions have pixel values of -100 (the
default fill value) and are set to 16 in the .q0h data quality files.
Cause: Data lost during transmission from HST to ground station.
Examples: Figure 13.1 shows dark bar encompassing about 45% of WF4.
Impact: Data lost over part of field. Can reduce / eliminate impact by combining multiple
frames of same field.
Correction: Mask out (discard) affected region when combining multiple frames of the
same field. See Section 14 herein. Image regions with missing data are automatically set to
the “fill value” -100 DN during pipeline processing, and so should not require manual
masking.
Prevention: None.
79
Figure 13.1. Mosaicked image showing broad bar of missing data on WF4. Here white
indicates bright regions of the image; black indicates low counts. Missing data regions are
set to -100 during pipeline processing.
missing data
80
14. Combining Dithered Images with Artifact Rejection
Image artifacts which are either transient in time, or fixed in position relative to the pixel
grid, can be removed when combining multiple position-dithered images of the same field.
There are many ways to accomplish this; here we describe one possible method.
The cosmic ray rejection task CRREJ in IRAF stsdas.hst_calib.wfpc can be used to discard bad image regions during image combination. In this example, we first mask bad
image regions by setting them to value -1000. The images are then aligned, scaled for
exposure time and sky brightness differences, and finally combined with CRREJ. We specifically consider a series of four images with properties given in Table 2. The second two
images are observed with a position offset POS TARG 0.501, 0.501 relative to the first two
images.
Table 2: Properties of images used in example
Image Name
Exposure
Time (sec.)
POS TARG
(arcsec.)
Background Level Measured with
IMSTAT (DN)
PC1
WF2
WF3
WF4
u2fh0101t.c0h
2600
0, 0
25.5
119.2
118.8
119.1
u2fh0102t.c0h
2700
0, 0
26.3
124.2
124.1
124.3
u2fh0103t.c0h
2700
0.501, 0.501
25.5
122.1
122.3
122.4
u2fh0104t.c0h
2700
0.501, 0.501
24.6
119.1
119.2
118.5
Once the images are displayed, and the residual regions are noted, masking is achieved by
setting these image regions to -1000 DN. First create a copy of the original data to work
with. Use GCOPY in the IRAF stsdas.toolbox.imgtools package.
gcopy u2fh0101t.c0h data1_masked.c0h groups=ALL
Then mask affected regions in a science image using the IMREPLACE task in the IRAF
cl.proto package. Here we mask the region [300:305,1:531] in group [3] of the first science image:
imrep data1_masked.c0h[3][300:305,1:531] -1000.
This step is repeated for each image region that is to be discarded. The science images are
then aligned to remove position dithers applied during observation. The case of integral
pixel dithers is simple; the alignment can be done merely by copying appropriate image
sections to an output image. In this example we consider a dither by POS TARG 0.501,
0.501, which is equivalent to ( ∆ x, ∆ y) = (11,11) PC pixels or (5,5) WFC pixels. To align
the third science image (4-group image) which was dithered by POS TARG 0.501, 0.501
relative to the first image (POS TARG 0, 0), one would:
81
gcopy data3_masked.c0h data3_aligned.c0h groups=ALL
imrep data3_aligned.c0h[1] -1000.
imrep data3_aligned.c0h[2] -1000.
imrep data3_aligned.c0h[3] -1000.
imrep data3_aligned.c0h[4] -1000.
imcopy data3_masked.c0h[12:800,12:800] data3_aligned.c0h[1][1:789,1:789]
imcopy data3_masked.c0h[2][6:800,1:795] data3_aligned.c0h[2][1:795,6:800]
imcopy data3_masked.c0h[3][1:795,1:795] data3_aligned.c0h[3][6:800,6:800]
imcopy data3_masked.c0h[4][1:795,6:800] data3_aligned.c0h[4][6:800,1:795]
where missing border regions are set to the mask value -1000. The fourth science image
would be shifted in the same manner. Alignment of images dithered by non-integral pixel
amounts is more complex; consult the REGISTER, XREGISTER, and IMLINTRAN
tasks in the IRAF cl.image area.
The CRREJ task does not (at present) automatically handle images of different exposure
time. Provided the exposure times are at least similar (within ~20%) this can be handled
by scaling the input images. Here we scale the first 2600 sec. image to the 2700 sec. exposure time of the other images. The resulting CRREJ image will thus have an effective
exposure time of 2700 sec. The weighting of the first image in the CRREJ output will be
the same as the others, which is slightly non-optimal, but the effect is small. The first
image is thus multiplied by 2700 / 2600 = 1.03846. One can use IMCALC in the stsdas.toolbox.imgtools package.
imcalc data1_masked.c0h data1_scaled.c0h ″ im1 * 1.03846 ″ pixtype=”real”
It is possible for the background level in images to vary from one exposure to another,
especially in long exposures in broad band filters. This is due to changes in the spacecraft
position relative to the Sun and Earth. Variations more than a few tenths of a DN should be
removed before combining images. The background level can be estimated using the
IMSTAT task in IRAF cl.images. Once a rough background level is determined by displaying each group of an image, the median background is computed as shown below.
Here we compute the median in a small range (15 to 35 DN) about the approximate background noted when displaying the image. Only the central 400 x 400 pixels are used:
PACKAGE = images
TASK = imstatistics
images
(fields
(lower
(upper
(binwid
(format
(mode
=
=
=
=
=
=
=
data1_scaled.c0h[1][200:600,200:600]
image,npix,midpt
15
35
0.1
yes
ql
82
With result:
#
IMAGE
NPIX
MIDPT
data1_scaled.c0h[1][200:600,200:600] 155070
25.5
This is then repeated for each group ([2], [3], and [4]) of each image, with results as shown
in Table 1. The PC will usually have a background which is about 4.78 times lower than
the other CCDs. Differences in background level are then removed. Here we adjust each
image to have the same background level as the first image (25.5 DN in group [1], 119.2
DN in group [2], etc.). For example, the second image is processed as follows:
imarith data2_scaled.c0h[1] - 0.8 data2_bg.c0h[1/4]
imarith data2_scaled.c0h[2] - 5.0 data2_bg.c0h[2]
imarith data2_scaled.c0h[3] - 5.3 data2_bg.c0h[3]
imarith data2_scaled.c0h[4] - 5.2 data2_bg.c0h[4]
Finally the aligned, scaled, and background corrected images are combined with CRREJ
in IRAF stsdas.hst_calib.wfpc. The value of the “minval” parameter (minval=-50) will
cause pixels below -50 (masked pixels) to be discarded when CRREJ computes the combined image.
PACKAGE = WFPC
TASK = CRREJ
infile
=
outfile =
outfile2 =
sigmas
=
radius
=
pfactor =
hotthresh=
minval
=
initial =
(noisepa =
(fillval =
(verbose =
(mode
=
@images.list
output.c0h
mask_sum.c0h
4,4,3,2
1
0
4096
-50
min
indef
yes
al
The file “images.list” is merely a text file containing the names of the input images. For
example, this file might look like:
data1_bg.c0h
data2_bg.c0h
data3_bg.c0h
data4_bg.c0h
83
Before running CRREJ, one should also check that the NOISEPAR parameter set contains
reasonable values. These can be checked and set using the command “epar noisepar”. A
reasonable set of values for WFPC2 are the following:
PACKAGE = wfpc
TASK = noisepar
(readnoise=
(gain
=
(scaleno =
(mode
=
0.8
7.
3.
al
If significant pointing errors occur during observations, bright stars may have too many
pixels rejected in the region where the PSF slope is largest. This can be partially corrected
by increasing the “scaleno” (scale noise) parameter, which adds the specified percentage
of the pixel counts to the noise model.
The output from CRREJ, output.c0h, is effectively an averaged 2700 sec. exposure image.
The second output file, mask_sum.c0h, shows the number of input images used at each
output pixel, and is useful for studying the behavior of CRREJ.
We thank H. Bond, S. Casertano, D. Elkin, R. Griffiths, J. Krist, T. Lauer, M. Malkan, L.
Petro, K. Ratnatunga, H. Richer, J. Trauger, and J. Westphal for assistance in preparing
this report.
84
15. References.
Burrows, et al., 1994, “Wide Field and Planetary Camera Handbook,” Version 3.0. Available from STScI Science Support Division.
Holtzman, et al., 1995, “The Performance and Calibration of WFPC2 on the Hubble Space
Telescope,” PASP 107, 156.
Holtzman, et al., 1995, “The Photometric Performance and Calibration of WFPC2,” PASP
in press.
King, Bartels, Nowak, and Hack, 1993, “Removal of Diagonal Stripes from FOC Images,”
in Calibrating Hubble Space Telescope, eds. Blades and Osmer, p. 128.
Krist and Burrows, 1994, “Large Angle Scattering in WFPC2 and Horizontal ‘Smearing’
Correction,” WFPC2 Instrument Science Report 94-01.
Krist, 1995, “WFPC2 Ghosts, Scatter, and PSF Field Dependence,” in Calibrating Hubble
Space Telescope: Post Servicing Mission, eds. Koratkar and Leitherer, p. 311.
Rockey and Simmons, 1993, “Instrument Description and User Handbook for Wide-Field
Planetary Camera II,” JPL publication 751-34, Rev. A, (SE-01), JPL D-11212.
Trauger, et al., 1993, “WFPC2 Science Calibration Report.”
85