SPACE
TELESCOPE
SCIENCE
INSTITUTE
Operated for NASA by AURA
Instrument Science Report STIS 2011-01(v1)
Wavelength Calibration Accuracy for
the STIS CCD and MAMA Modes
Ilaria Pascucci1 , Phil Hodge1 , Charles R. Proffitt1 & T. Ayres2
1
Space Telescope Science Institute, Baltimore, MD 21218
2
University of Colorado, Boulder, CO 80309-0389
March 29, 2011
ABSTRACT
Two calibration programs were carried out after Servicing Mission 4 to determine the
accuracy of the wavelength solutions for the most used STIS CCD and MAMA modes.
We report here on the analysis of this dataset and show that the STIS wavelength solution
has not changed after SM4. We also show that a typical accuracy for the absolute wavelength zero-points is 0.1 pixels while the relative wavelength accuracy is 0.2 pixels.
Contents
• Introduction (page 2)
• Data (page 2)
• Analysis (page 3)
• Conclusions (page 6)
• Recommendations (page 7)
• Change History (page 7)
• References (page 7)
• Appendix (page 21)
Operated by the Association of Universities for Research in Astronomy, Inc., for the National Aeronautics
and Space Administration.
Introduction
The Space Telescope Imaging Spectrograph (STIS, Kimble et al. 1998) operated onorbit from February 1997 until a malfunction in August 2004. During Hubble Servicing
Mission 4 (SM4), STIS was successfully repaired and resumed operations in May 2009.
To allow the wavelength scale for individual science observations to be corrected for
non-repeatability and drifts in the STIS optical element alignment, most observations of
external targets are accompanied by one or more wavelength calibration spectra. There
are three Pt/Cr-Ne lamps on STIS for this purpose. These lamps are designated LINE,
HITM1, and HITM2. The LINE lamp is located on the Insertion Mechanism platform.
When the lamp is used, the Calibration Insert Mechanism (CIM) is placed into the light
path and all external light is blocked. The HITM1 and HITM2 lamps are located on
the Hole In The Mirror platform and can be used both for wavelength calibration and
target acquisition (though the HITM2 lamp is considered the spare lamp and has not
been used for target acquisition). The LINE lamp, being the brightest of the 3 Pt/Cr-Ne
lamps, is being used for the least sensitive modes on STIS, especially the FUV echelle
modes. The optics of the HITM1 and HITM2 lamps are such that their lower flux on the
detector is more suited for the wavelength calibration of more sensitive modes on STIS,
such as the optical low-resolution modes.
In order to test whether there has been any change in the wavelength solution postSM4, we have obtained internal wavelength exposures covering many CCD and MAMA
modes, focusing primarily on those used in GO programs as well as those utilized in
previous calibration programs. This study measures the wavelength difference between
lines in the STIS spectra and reference lines from a catalogue. It shows that the STIS
wavelength solution has not changed after SM4 and that the accuracy is typically within
0.2 pixels.
Data
Wavelength calibration exposures have been obtained for many STIS CCD and MAMA
modes frequently used both in GO programs as well as in previous calibration programs. The CCD exposures, which cover 39 different modes, were taken under program PID 11858 while the MAMA exposures, covering 26 different modes, were taken
under program PID 11859. CCD lamp spectra were acquired with the 52 × 0.1′′ slit and
exposure times ranged from 4 to over 200 s. MAMA lamp spectra were obtained with
varying small apertures ranging from 0.2 × 0.06′′ in the echelle modes to 52 × 0.1′′ in
the first-order spectroscopic modes. A summary of the observations can be found in
Tables 1 and 2.
To compare the line spectra with catalogued Pt/Cr-Ne lines we used the original
compilation from Don Lindler as well as new compilations of Pt/Cr-Ne lines from the
Operated by the Association of Universities for Research in Astronomy, Inc., for the National Aeronautics
and Space Administration.
NIST database1 , Sansonetti et al. (2004) for wavelengths <1830 Å and of Cr I lines
from Wallace & Hinkle (2009) in the NUV portion of the STIS band (λ >2200 Å).
Spectral
Element
G230LB
G230MB
G230MB
G230MB
G230MB
G230MB
G230MB
G230MB
G230MB
G230MB
G230MB
G230MB
G230MB
G430L
G430L
G430M
G430M
G430M
G430M
G430M
G430M
G430M
G430M
G430M
G430M
G430M
G430M
G750L
G750M
G750M
G750M
G750M
G750M
G750M
G750M
G750M
G750M
G750M
G750M
Central
Wavelength
[Å]
2375
1713
2135
2276
1854
2416
2557
1995
2697
2794
2836
2976
3115
4300
4300
3165
3423
3680
3843
3936
4194
4451
4706
4961
5093
5216
5471
7751
5734
6252
6581
6768
7283
7795
8311
8561
8825
9336
9851
Aperture
Lamp
Time
[′′ ]
52X0.1
52X0.1
52X0.1
52X0.1
52X0.1
52X0.1
52X0.1
52X0.1
52X0.1
52X0.1
52X0.1
52X0.1
52X0.1
52X0.1
52X0.1
52X0.1
52X0.1
52X0.1
52X0.1
52X0.1
52X0.1
52X0.1
52X0.1
52X0.1
52X0.1
52X0.1
52X0.1
52X0.1
52X0.1
52X0.1
52X0.1
52X0.1
52X0.1
52X0.1
52X0.1
52X0.1
52X0.1
52X0.1
52X0.1
HITM1
HITM1
HITM1
HITM1
HITM1
HITM1
HITM1
HITM1
HITM1
HITM1
HITM1
HITM1
HITM1
HITM1
LINE
HITM1
HITM1
HITM1
HITM1
HITM1
HITM1
HITM1
HITM1
HITM1
HITM1
HITM1
HITM1
HITM1
HITM1
HITM1
HITM1
HITM1
HITM1
HITM1
HITM1
HITM1
HITM1
HITM1
HITM1
[s]
220.0
368.0
30.0
84.0
600.0
41.0
63.0
190.0
14.0
30.0
24.0
10.0
24.0
10.0
10.0
10.0
10.0
10.0
10.0
13.0
17.0
18.0
21.0
24.0
10.0
10.0
26.0
6.2
5.9
4.1
4.1
3.9
3.9
29.0
10.0
10.0
10.0
10.0
220.0
Table 1. Log of the observations from program PID 11858 (CCD)
Analysis
We reduced the lamp exposures with CALSTIS as if they were science images. To do
this we created a companion file (named oxxxxxxxx wav.fits) and invoked this file as
the argument of the keyword WAVECAL in the FITS header of the original file (named
oxxxxxxxx raw.fits). We also modified the keyword ASN MTYP to SCIENCE so that
the software treats the image as a normal science exposure and the WAVECAL keyword from OMIT to PERFORM. We further switched the keywords HELCORR (for
the heliocentric correction) and FLUXCORR (for the flux calibration) to OMIT. For
first-order spectroscopic modes, five spectra were extracted by summing 32 rows (EXTRSIZE keyword in CALSTIS equal to 32) from different locations on the flt image.
The five spectra were centered on cross-dispersion locations 128, 320, 512, 704, 896.
We used spectra from the middle of the chip (row 512) for the line identifications (see
1
http://physics.nist.gov/PhysRefData/platinum/contents.html
Instrument Science Report STIS 2011-01(v1) Page 3
Spectral
Element
E140M
E140H
E140H
E140H
E140H
E140H
E230M
E230M
E230M
E230M
E230H
E230H
E230H
G140M
G140M
G140M
G140M
G140M
G140M
G140L
G230M
G230M
G230M
G230M
G230L
E230H
Central
Wavelength
[Å]
1425
1526
1598
1234
1271
1343
1978
2415
2561
2707
1763
2013
2713
1218
1222
1321
1540
1420
1640
1425
1687
2338
2818
3055
2376
1963
Aperture
Lamp
Time
[′′ ]
0.2X0.06
0.2X0.09
0.2X0.09
0.2X0.09
0.2X0.09
0.2X0.09
0.2X0.06
0.2X0.06
0.2X0.06
0.2X0.06
0.1X0.09
0.1X0.09
0.1X0.09
52X0.1
52X0.1
52X0.1
52X0.1
52X0.1
52X0.1
52X0.1
52X0.1
52X0.1
52X0.1
52X0.1
52X0.05
0.1X0.09
LINE
LINE
LINE
LINE
LINE
LINE
LINE
LINE
LINE
LINE
LINE
LINE
LINE
LINE
LINE
LINE
LINE
LINE
LINE
LINE
LINE
LINE
LINE
LINE
LINE
LINE
[s]
663.2
300
400.5
550
400
550
190.3
135
120
87.2
800
500
500
450
72.4
72.6
94.3
75
100
140
310.0
40.0
30.0
22.0
47.6
800
Table 2. Log of the observations from program 11859 (MAMA).
next paragraph) and the spectra taken at different locations to check for bad pixels.
The output of the CALSTIS pipeline is the x1d file, which contains wavelength, net
count rate, error (sorted by order in case of echelle spectra). For echelle spectra a onedimensional spectrum covering the full wavelength range of the mode was created by
interleaving the flux-wavelength points in ascending wavelength order.
Following this approach, we created 65 1-D spectra covering all the gratingcentral wavelength combinations in programs 11858 and 11859 . Using a semi-automatic
line identification routine written in IDL, we identified laboratory lines in the STIS spectra and computed the difference between the observed and catalogued lines for each
grating-central wavelength combination. In order to verify our approach we also produced figures showing the identified lines on top of the STIS lamp spectra. The centroid
pixel location for each line was measured in two ways: i) by fitting a gaussian to the
lines, and ii) by computing a weighted mean using the counts over ±2 pixels from the
central one as a weighting factor. The errors on the centroid are the formal error on the
peak of the gaussian for i) and the inverse of the square root of the counts over 5 pixels
for ii). We then computed the mean and standard deviations of the differences per mode
but also checked if there is any trend of the differences with wavelength within each
mode.
CCD Modes
Our results on the wavelength accuracy of the CCD modes are summarized in Figs. 1–
10. Fig. 4 shows that when the Wallace & Hinkle (2009) line compilation is not included
there are many fewer lines for which the wavelength accuracy can be measured.
Instrument Science Report STIS 2011-01(v1) Page 4
None of the CCD modes shows a wavelength dependence in the difference between observed and laboratory lines. The mean of the difference distribution is typically <0.1 pixel and the standard deviation is < 0.2 pixels as measured before SM4. In
the two cases where standard deviations are ∼0.4 pixels (G230MB-2794 and G430M3936), the large value is very likely due to the paucity of identified lines as shown for
the G230MB modes when the Wallace & Hinkle (2009) compilation is not included.
MAMA Modes
Our results on the wavelength accuracy of the MAMA modes are summarized in Figs. 11–
16. As noted for the CCD modes, including the Wallace & Hinkle (2009) compilation
for those settings around 2300 Å increases the number of detected lines and hence gives
more reliable estimates for the wavelength accuracy, see e.g. the E230H-2713 Å settings
and several E230M settings.
We find that the high-resolution echelle settings (E140H and E230H) perform as
pre-SM4: The mean of the difference distribution is typically <0.1 km/s (or 0.08 pixels) and the standard deviation is ∼0.3 km/s (or 0.2 pixels). These values are well in
agreement with those reported in Ayres (2008) who used an independent approach to
analyze deep (hundred seconds) high signal-to-noise wavecal exposures obtained before SM4. In a few instances, such as for the E230H-1963 Å setting, we see that higher
order effects are not corrected by the current coefficients in the STIS dispersion relations. In the next section we report on tests that have been made using new coefficients
and an expanded wavelength solution derived by Ayres using pre- as well as post-SM4
deep wavelength exposures. The medium resolution settings (E140M and E230M) have
larger differences from laboratory wavelengths than the H settings, as expected from
the lower spectral resolution. In pixels, the mean of the difference distribution is always
less than 0.1 pixels while the standard deviation is typically <0.2 pixels but sometimes
a bit larger up to 0.26 pixels. In these settings we see a few clear trends of offsets with
wavelengths. Such trends have been independently identified by Ayres. We attempt to
correct for them with the new and expanded wavelength solution (see next Section).
The MAMA G modes (medium and low resolution) perform as expected and as reported in the STIS Instrument Handbook. Specifically, the standard deviations between
observed and laboratory wavelengths are between 0.15 and 0.5 pixels, with the G140L
and G230L having the smallest deviations 0.2 pixels and 0.15 pixels respectively.
Further Improvements on the STIS Echelle Wavelength Accuracy
The current dispersion relations for the echelle modes produce some small but systematic offsets of the measured versus laboratory lines, see e.g. the E230H-1963 mode in
Fig. 12. These offsets differ from one central wavelength to another. As part of two
archival and one GO calibration programs (10203, 11743, and 12280) Tom Ayres has
independently verified the accuracy of the STIS echelle modes and recently proposed
Instrument Science Report STIS 2011-01(v1) Page 5
a modified version of the STIS dispersion polynomial which could further improve the
wavelength accuracy (see Ayres 2010, HST calibration workshop). This new version
includes two additional terms: one in [m λ]3 and a second in m2 , where m is the echelle
order number. In addition, he re-derived all dispersion coefficients for the CALSTIS
pipeline polynomial using a more extensive set of lamp exposures obtained through the
entire STIS lifetime. We tested this new dispersion relation with CALSTIS and found
that while some improvements could be seen overall (e.g. for E230H-2013, see Fig. 17),
a few modes presented larger deviations than with the previous dispersion solution (e.g.
E140H-1526, see Fig. 17). These deviations were confirmed by an independent analysis of Ayres. After a series of tests which included verifying the A4CORR, SHIFTA2,
and YREF terms in the new and previous CALSTIS (see Appendix), we realized that
the larger deviations seen only in some modes could be due to the new set of lamp
exposures utilized to compute the new coefficients for the dispersion solution. In the
past, monthly MSM (Mode Selection Mechanism) offsets (∼20 pixels) were applied
regularly for STIS observations in order to uniformly age the MAMA detectors, meaning that the center of the spectrogram would fall on a different detector location then
with the default zero offsets position which is used in recent observations. Indeed, we
found that for those settings which presented the largest deviations in the new CALSTIS several lamp exposures were taken at large MSM offset positions (for instance
for E140H-1526 the new coefficients were computed from 3 deep lamp exposures, the
deepest of which was taken at MSM offset2 = -23 pixels). We are in the process of
testing a new set of dispersion coefficients now derived using only deep lamp exposures
taken at the default zero MSM offset position. This finding also led to the realization
that there are not enough deep lamp exposures for all available modes to adequately
improve upon the current wavelength solution (see Recommendations). In addition, we
are testing a slightly different processing strategy in CALSTIS to better account for the
spectral tilts seen in the wavecals with large y offsets (see Appendix for more details).
This new procedure runs the wavecal processing twice for echelle data: the first crosscorrelation between the lamp spectrum and the template spectrum finds the offset in
cross-dispersion direction (SHIFTA2), then uses the computed SHIFTA2 in a second
pass to apply the proper A4CORR. This last term is a coefficient in the CALSTIS dispersion model that compensates for spectral tilts which occur when the echelle pattern
is offset in the y direction from the nominal position for that setting.
Conclusions
We analyzed high signal-to-noise wavelength exposures obtained after STIS was successfully repaired and resumed operations in May 2009 to test the wavelength accuracy
of the CCD and MAMA modes. The settings obtained under programs PID 11858 and
11859 and analyzed here do not present any anomalous behaviors. The mean of the
difference between observed and laboratory wavelengths, which is a measure of the accuracy of the wavelength zero point, is typically better than 0.1 pixels, as for pre-SM4.
Instrument Science Report STIS 2011-01(v1) Page 6
Similarly, the standard deviation of the differences, a measurement of the relative wavelength accuracy, is typically better than 0.2 pixels. We have identified a few settings in
which differences have higher order wavelength trends and might be improved using an
expanded dispersion solution. Further tests are in progress.
Recommendations
We recommend to continue the yearly monitoring of the STIS CCD and MAMA dispersion solutions as an integral part of the long-term monitoring program. In addition, for
the STIS MAMA echelle settings, additional deep engineering wavecal exposures could
be acquired to improve the wavelength accuracy. A list of settings that would benefit
from additional exposures is summarized in Table 3.
Change History for STIS ISR 2011-01
Version 1: 29 March 2011 - Original Document
References
Ayres, T. R. 2008, ApJS, 177, 626 10, 100
Kimble, R. A., Woodgate, B. E., Bowers, C. W. et al. 1998, ApJ, L492, 83 10, 100
Sansonetti, C. J., Kerber, F., Reader, J., Rosa, M. R. 2004, ApJS, 153, 555 10, 100
Wallace, L. & Hinkle, K. 2009, ApJ, 700, 720 10, 100
Instrument Science Report STIS 2011-01(v1) Page 7
G230LB-2375
1.0
0.044, 0.26
0.059, 0.24
Offset [pixel]
0.5
0.0
-0.5
-1.0
1500
2000
2500
3000
Wavelength [Angstrom]
3500
Figure 1. Difference in pixels between observed and laboratory lines for the G230LB
grating. Black diamonds are differences for centroids measured with gaussian fits to the
observed lines while red asterisks for weighted mean centroids. The mean and standard
deviation of the difference distributions are also given on the right upper corner in black
for gaussian fits and in red for weighted mean centroids. One pixel corresponds to
∼175 km/s.
Spectral
Element
Central
Recommended
Wavelength
exposure
[Å]
[#×s]
priority 1: prime settings
E140H
1416
3×1,000
E230M
2707
2×1,000
priority 2: secondary settings
E140H
1453
2×1,000
E140H
1307
1×1,000
E230M
2269
1×1,500
E230H
2363
1×1,500
E230H
2263
1×1,500
E230H
2862
1×1,500
E230H
2663
1×1,500
E230H
1913
1×1,500
E230H
2413
1×1,000
E230H
2762
1×1,000
E230M
2561
1×1,000
Table 3. Settings that would benefit from additional deep high signal-to-noise lamp
exposures
Instrument Science Report STIS 2011-01(v1) Page 8
G230MB-1713
G230MB-1854
Offset [pixel]
1.0
1.0
0.083, 0.23
0.5
0.0
0.0
-0.5
-0.5
-1.0
1650
1700
1750
0.053, 0.25
0.5
1800
-1.0
1750
1800
G230MB-1995
Offset [pixel]
1900
1950
G230MB-2135
1.0
1.0
-0.02, 0.23
0.5
0.0
-0.5
-0.5
1950
2000
2050
-0.12, 0.08
0.5
0.0
-1.0
1900
2100
-1.0
2080
2100
G230MB-2276
2120
2140
2160
2180
G230MB-2416
1.0
Offset [pixel]
1850
1.0
0.010, 0.22
0.5
0.0
0.0
-0.5
-0.5
-1.0
2200
2250
2300
Wavelength [Angstrom]
0.113, 0.12
0.5
2350
-1.0
2350
2400
2450
Wavelength [Angstrom]
2500
Figure 2. Difference in pixels between observed and laboratory lines for G230MB gratings. Colors are as in Fig. 1. One pixel corresponds to ∼20 km/s at 2300 Å.
G230MB-2557
G230MB-2794
Offset [pixel]
1.0
1.0
-0.01, 0.33
0.5
0.0
0.0
-0.5
-0.5
-1.0
2450
2500
2550
2600
2650
0.039, 0.24
0.5
-1.0
2700
2750
G230MB-2836
2850
2900
G230MB-2976
1.0
Offset [pixel]
2800
1.0
0.019, 0.25
0.5
0.0
0.0
-0.5
-0.5
-1.0
2750
2800
2850
2900
0.124, 0.16
0.5
2950
-1.0
2940
2960
2980
3000
3020
Wavelength [Angstrom]
3040
3060
G230MB-3115
Offset [pixel]
1.0
0.052, 0.20
0.5
0.0
-0.5
-1.0
3020
3040
3060
3080
3100
3120
Wavelength [Angstrom]
3140
3160
Figure 3. Difference in pixels between observed and laboratory lines for G230MB gratings. Colors are as in Fig. 1. One pixel corresponds to ∼20 km/s at 2300 Å.
Instrument Science Report STIS 2011-01(v1) Page 9
G230MB-2557
G230MB-2794
Offset [pixel]
1.0
1.0
-0.05, 0.32
0.5
0.0
0.0
-0.5
-0.5
-1.0
2480
2500
2520
2540
2560
2580
-0.04, 0.49
0.5
-1.0
2780
2800
G230MB-2836
Offset [pixel]
2840
2860
G230MB-2976
1.0
1.0
0.131, 0.40
0.5
0.0
0.0
-0.5
2800
2820
2840
2860
2880
0.182, 0.15
0.5
-0.5
-1.0
2780
2820
-1.0
2980
2990
3000
3010
3020
Wavelength [Angstrom]
3030
3040
G230MB-3115
Offset [pixel]
1.0
0.059, 0.29
0.5
0.0
-0.5
-1.0
3000
3050
3100
3150
Wavelength [Angstrom]
3200
Figure 4. Difference in pixels between observed and laboratory lines for G230MB gratings. This is the same as Fig. 3 but without the Wallace & Hinkle (2009) compilation of
Cr I lines. Fewer lines are identified in these spectra. Colors are as in Fig. 1.
Instrument Science Report STIS 2011-01(v1) Page 10
G430L-4300
1.0
-0.05, 0.28
-0.02, 0.28
Offset [pixel]
0.5
0.0
-0.5
-1.0
2500
3000
3500
4000
4500
5000
Wavelength [Angstrom]
5500
6000
Figure 5. Difference in pixels between observed and laboratory lines for the G430L
grating (HITM1 lamp). Black diamonds are differences for centroids measured with
gaussian fits to the observed lines while red asterisks for weighted mean centroids. The
mean and standard deviation of the difference distributions are also given on the right
upper corner in black for gaussian fits and in red for weighted mean centroids. One
pixel corresponds to ∼240 km/s.
Instrument Science Report STIS 2011-01(v1) Page 11
G430M-3165
G430M-3423
Offset [pixel]
1.0
1.0
0.156, 0.23
0.5
0.0
0.0
-0.5
-0.5
-1.0
3000
3050
3100
3150
3200
-0.00, 0.26
0.5
3250
-1.0
3300
3350
G430M-3680
Offset [pixel]
0.0
0.0
-0.5
3550
3600
3650
3700
3750
3800
3850
-1.0
3750
3800
G430M-4194
3850
3900
3950
4000
G430M-4451
1.0
Offset [pixel]
3550
0.128, 0.43
0.5
-0.5
1.0
0.015, 0.13
0.0
-0.5
-0.5
4200
4250
Wavelength [Angstrom]
0.055, 0.19
0.5
0.0
-1.0
4150
3500
1.0
0.115, 0.18
0.5
0.5
3450
G430M-3936
1.0
-1.0
3500
3400
4300
-1.0
4300
4350
4400
4450
4500
Wavelength [Angstrom]
4550
4600
Figure 6. Difference in pixels between observed and laboratory lines for G430M gratings. Colors are as in Fig. 5. One pixel corresponds to ∼25 km/s at 4300 Å.
Instrument Science Report STIS 2011-01(v1) Page 12
G430M-4706
G430M-4961
Offset [pixel]
1.0
1.0
0.154, 0.28
0.5
0.0
0.0
-0.5
-0.5
-1.0
4550
4600
4650
4700
4750
4800
-0.05, 0.07
0.5
4850
-1.0
4800
4850
G430M-5093
Offset [pixel]
5050
5100
0.0
0.0
-0.5
-0.5
5050
5100
5150
5200
-0.02, 0.07
0.5
5250
-1.0
5050
5100
G430M-5471
5150
5200
5250
5300
5350
G430L-4300
1.0
Offset [pixel]
5000
1.0
0.056, 0.11
0.5
1.0
0.064, 0.12
0.5
0.0
0.0
-0.5
5350
5400
5450
5500
Wavelength [Angstrom]
5550
-0.01, 0.26
0.5
-0.5
-1.0
5300
4950
G430M-5216
1.0
-1.0
5000
4900
5600
-1.0
2500
3000
3500
4000
4500
5000
Wavelength [Angstrom]
5500
6000
Figure 7. Difference in pixels between observed and laboratory lines for G430M gratings except last panel where we show the G430L setting for the LINE lamp. Note that
the difference distribution from this last panel is the same as that for the G430L HITM1
lamp setting (Fig. 5). Colors are as in Fig. 5. One pixel corresponds to ∼25 km/s at
4300 Å for the medium-resolution settings.
Instrument Science Report STIS 2011-01(v1) Page 13
G750L-7751
1.0
-0.08, 0.13
-0.07, 0.15
Offset [pixel]
0.5
0.0
-0.5
-1.0
5000
6000
7000
8000
Wavelength [Angstrom]
9000
10000
Figure 8. Difference in pixels between observed and laboratory lines for the G750L
grating. Black diamonds are differences for centroids measured with gaussian fits to the
observed lines while red asterisks for weighted mean centroids. The mean and standard
deviation of the difference distributions are also given on the right upper corner in black
for gaussian fits and in red for weighted mean centroids. One pixel corresponds to
∼197 km/s at 7500 Å.
Instrument Science Report STIS 2011-01(v1) Page 14
G750M-5734
G750M-6252
Offset [pixel]
1.0
1.0
0.013, 0.07
0.5
0.0
0.0
-0.5
-0.5
-1.0
5500
5600
5700
5800
5900
6000
0.094, 0.08
0.5
-1.0
5900
6000
G750M-6581
Offset [pixel]
6400
6500
6600
0.0
0.0
-0.5
-0.5
6400
6500
6600
6700
-0.01, 0.04
0.5
6800
-1.0
6500
6600
G750M-7283
6700
6800
6900
7000
7100
G750M-7795
1.0
Offset [pixel]
6300
1.0
0.090, 0.14
0.5
1.0
0.039, 0.06
0.5
0.0
0.0
-0.5
7100
7200
7300
7400
Wavelength [Angstrom]
7500
0.120, 0.12
0.5
-0.5
-1.0
7000
6200
G750M-6768
1.0
-1.0
6300
6100
7600
-1.0
7500
7600
7700
7800
Wavelength [Angstrom]
7900
8000
Figure 9. Difference in pixels between observed and laboratory lines for G750M gratings. Colors are as in Fig. 8. One pixel corresponds to ∼22 km/s at 7500 Å.
G750M-8311
G750M-8561
Offset [pixel]
1.0
1.0
0.059, 0.09
0.5
0.0
0.0
-0.5
-0.5
-1.0
8000
8100
8200
8300
8400
8500
8600
0.021, 0.07
0.5
-1.0
8300
8400
G750M-8825
Offset [pixel]
8600
8700
8800
G750M-9336
1.0
1.0
0.039, 0.06
0.5
0.0
0.0
-0.5
8600
8700
8800
8900
-0.01, 0.10
0.5
-0.5
-1.0
8500
8500
9000
-1.0
9100
9200
9300
9400
Wavelength [Angstrom]
9500
9600
G750M-9851
Offset [pixel]
1.0
0.142, 0.12
0.5
0.0
-0.5
-1.0
9560
9580
9600
9620
9640
Wavelength [Angstrom]
9660
9680
Figure 10. Difference in pixels between observed and laboratory lines for G750M gratings. Colors are as in Fig. 8. One pixel corresponds to ∼22 km/s at 7500 Å.
Instrument Science Report STIS 2011-01(v1) Page 15
E140H-1234
E140M-1425
0.207, 0.61
2
0
-2
-4
1100
1200
1300
1400
1500
Wavelength [Angstrom]
1600
4
Offset v [km/s]
Offset v [km/s]
4
1700
-0.07, 0.34
2
0
-2
-4
1150
1200
E140H-1271
2
0
-2
-4
1150
1200
1250
1300
Wavelength [Angstrom]
1350
4
1400
-0.17, 0.38
2
0
-2
-4
1200
1250
E140H-1526
2
0
-2
-4
1400
1450
1500
1550
Wavelength [Angstrom]
1300
1350
Wavelength [Angstrom]
1400
1450
E140H-1598
0.013, 0.39
1600
4
Offset v [km/s]
Offset v [km/s]
4
1350
E140H-1343
0.031, 0.34
Offset v [km/s]
Offset v [km/s]
4
1250
1300
Wavelength [Angstrom]
1650
-0.06, 0.27
2
0
-2
-4
1500
1550
1600
1650
Wavelength [Angstrom]
1700
Figure 11. Difference in velocity between observed and laboratory lines for the E140H
and M gratings acquired within program PID 11859. The velocity scale is used to better
compare these results with those reported by Ayres (2008). Black diamonds are differences for centroids measured with gaussian fits to the observed lines while red asterisks
for weighted mean centroids. The mean and standard deviation of the difference distributions are also given on the right upper corner in black for gaussian fits. One pixel
corresponds to ∼ λ/228, 000 Å or ∼1.3 km/s for the H settings and to ∼ λ/91, 700 Å or
3.3 km/s for the M settings.
Instrument Science Report STIS 2011-01(v1) Page 16
E230H-1963
E230H-1763
4
4
-0.03, 0.34
0.012, 0.32
2
Offset v [km/s]
Offset v [km/s]
2
0
-2
-4
1600
0
-2
1650
1700
1750
1800
Wavelength [Angstrom]
1850
1900
-4
1800
1850
1900
1950
2000
2050
Wavelength [Angstrom]
4
-0.20, 0.26
-0.11, 0.26
2
Offset v [km/s]
2
Offset v [km/s]
2150
E230H-2713
E230H-2013
4
0
-2
-4
1850
2100
0
-2
1900
1950
2000
2050
2100
Wavelength [Angstrom]
2150
2200
-4
2550
2600
2650
2700
2750
Wavelength [Angstrom]
2800
2850
Figure 12. Difference in velocity between observed and laboratory lines for the E230H
gratings. Colors are as in Fig. 11. One pixel corresponds to ∼ λ/228, 000 Å or
∼1.3 km/s.
E230M-2415
E230M-1978
6
6
-0.19, 0.78
2
0
-2
-4
-6
1600
0.122, 1.27
4
Offset v [km/s]
Offset v [km/s]
4
2
0
-2
-4
1800
2000
2200
Wavelength [Angstrom]
2400
-6
1800
2000
E230M-2561
3000
6
-0.47, 1.31
2
0
-2
0.428, 1.08
4
Offset v [km/s]
4
Offset v [km/s]
2800
E230M-2707
6
-4
-6
2000
2200
2400
2600
Wavelength [Angstrom]
2
0
-2
-4
2200
2400
2600
Wavelength [Angstrom]
2800
3000
-6
2200
2400
2600
2800
Wavelength [Angstrom]
3000
3200
Figure 13. Difference in velocity between observed and laboratory lines for the E230M
gratings. Colors are as in Fig. 11. One pixel corresponds to ∼ λ/60, 000 Å or ∼5 km/s.
Instrument Science Report STIS 2011-01(v1) Page 17
G140L-1425
100
-6.85, 25.7
Offset v [km/s]
50
0
-50
-100
1200
1300
1400
1500
Wavelength [Angstrom]
1600
1700
Figure 14. Difference in velocity between observed and laboratory lines for the G140L
grating. Colors are as in Fig. 11. One pixel corresponds to 0.6 Å or 126 km/s.
G140M-1222
G140M-1218
20
-1.19, 3.04
10
0
-10
-20
1190
1200
1210
1220
1230
Wavelength [Angstrom]
1240
Offset v [km/s]
Offset v [km/s]
20
1250
-0.86, 5.48
10
0
-10
-20
1190
1200
-10
1300
1310
1320
1330
Wavelength [Angstrom]
1340
Offset v [km/s]
Offset v [km/s]
0
1350
0.750, 3.01
10
0
-10
-20
1510
1520
1530
1540
1550
Wavelength [Angstrom]
1560
1570
G140M-1640
G140M-1420
20
-0.11, 3.17
10
0
-10
1400
1410
1420
Wavelength [Angstrom]
1430
Offset v [km/s]
20
Offset v [km/s]
1250
20
-0.35, 4.46
10
-20
1390
1240
G140M-1540
G140M-1321
20
-20
1290
1210
1220
1230
Wavelength [Angstrom]
1440
1.008, 2.71
10
0
-10
-20
1610
1620
1630
1640
Wavelength [Angstrom]
1650
1660
Figure 15. Difference in velocity between observed and laboratory lines for the G140M
gratings. Colors are as in Fig. 11. One pixel corresponds to 0.05 Å, about 10 times
better than for the G140L setting.
Instrument Science Report STIS 2011-01(v1) Page 18
G230M-1687
G230L-2376
20
1.029, 30.8
50
0
-50
-100
1500
2000
2500
3000
Wavelength [Angstrom]
Offset v [km/s]
Offset v [km/s]
100
3500
-0.44, 3.22
10
0
-10
-20
1640
1660
G230M-2338
1740
20
-1.62, 3.44
10
0
-10
2300
2320
2340
Wavelength [Angstrom]
2360
Offset v [km/s]
Offset v [km/s]
1720
G230M-2818
20
-20
2280
1680
1700
Wavelength [Angstrom]
2380
-1.68, 2.30
10
0
-10
-20
2780
2800
2820
2840
Wavelength [Angstrom]
2860
G230M-3055
Offset v [km/s]
20
-2.85, 1.92
10
0
-10
-20
3000
3020
3040
3060
Wavelength [Angstrom]
3080
3100
Figure 16. Difference in velocity between observed and laboratory lines for G230L and
M gratings. Colors are as in Fig. 11. One pixel corresponds to 1.58 Å (or 199 km/s) at
the L setting and 0.09 Å for the M settings.
Instrument Science Report STIS 2011-01(v1) Page 19
E230H--2013
4
Mean - stddev -0.044787725 0.032009982 (new)
Mean - stddev 0.20195109 0.035371159 (old)
Offset v [km/s]
2
0
-2
-4
1850
1900
1950
2000
2050
Wave [AA]
2100
2150
2200
E140H--1526
4
Mean - stddev 0.19030626 0.078527406 (new)
Mean - stddev -0.013427360 0.035686447 (old)
Offset v [km/s]
2
0
-2
-4
1400
1450
1500
1550
Wave [AA]
1600
1650
Figure 17. Difference in velocity between observed and laboratory lines for two settings
that have been processed with the new CALSTIS dispersion solution from T. Ayres.
Black diamonds are differences using the new CALSTIS coefficients while red asterisks
using the old coefficients. This wavelength solution resulted in an improvement between
measured and observed lines for some settings (see upper panel) but not for all settings
(see lower panel). The anomalous settings now have been recognized as those affected
by wavecals taken at large y offsets. An improved approach is in progress to compensate
self-consistently for those spectral tilts.
Instrument Science Report STIS 2011-01(v1) Page 20
Appendix
The image of the echelle spectra does not always fall at the same location on the detector. In early STIS data there was a deliberate offset to improve detector lifetime, and
even now there will be a random offset of a few pixels in each axis. The offset in the
cross-dispersion direction results in a small skew of the image, a shift in the dispersion
direction that is directly proportional to the location in the cross-dispersion direction.
This is called the A4 correction (A4CORR) because it is accounted for by modifying
dispersion coefficients 0 and 4 (a0 and a4); the latter term is the one that depends only
on wavelength. When the A4 correction was first added to CALSTIS there was a keyword that gave the approximate value of the offset in the cross-dispersion direction, and
the value of that keyword was used in order to apply the A4 correction during wavecal
processing. The correction is applied when constructing a template image; the shifts are
determined by cross correlating the observed wavecal spectrum with the template. That
keyword is no longer populated (it is set to zero), so calstis was recently modified to do
the wavecal processing step twice, once to determine the offset in the cross-dispersion
direction, then a second time with the A4 correction included, using the new y offset.
The A4CORR, SHIFTA2, and YREF are linked to the a4 and a0 coefficients in the STIS
wavelength dispersion solution as follows:
ydif = ypos − YREF = SHIFTA2 + a2center − YREF
a4 = a4 + A4CORR × ydif
a0 = a0 − cenwave ∗ A4CORR ∗ ydif
where a2center is taken from the trace table (SPTRCTAB) for spectral order
MREF. YREF is the location of order MREF in the WAVECAL used to determine the
dispersion coefficients. (ypos - YREF) is the global offset of the lamp image in the Y
direction with respect to the template lamp image.
Instrument Science Report STIS 2011-01(v1) Page 21