Pre-Launch NUV MAMA Flats

Transcription

1 Instrument Science Report STIS Pre-Launch NUV MAMA Flats R. C. Bohlin, D. J. Lindler, & M. E. Kaiser 1997 May ABSTRACT A full set of flat field calibration files for the NUV MAMA has been generated from preflight laboratory data. An iterative technique with 12 passes through each raw flat data image is used to separate the illuminating lamp signature from the high frequency structure of the MAMA flat itself. The set of flats consists of one wavelength independent P-flat that corrects for the high frequency pixel-to-pixel sensitivity variations and 63 L-flats that correct for the lower frequency, wavelength dependent variations. The S/N of the counting statistics of the P-flat is over 300 per MAMA resolution element (2x2 lores pixels.) Onorbit tests are required to confirm the validity of the ground based flats for the reduction of flight data. 1. INTRODUCTION Bohlin, Lindler, & Baum (1996, hereafter BLB) defined the algorithms for deriving the STIS MAMA flats. However, one major and a few minor modifications to the BLB formalism are required. Since the BLB requirement that...the small scale polish on the slit jaws must be constant to 1% of the slit width... is not met for the narrower slits employed with the internal STIS flat field lamps, a correction for slit defects is necessary. Fortunately, the narrowest slit used to obtain high S/N P-flat data is the 52x0.1 arcsec slit, where the narrow slit defects do not exceed 20% transmission loss. The introduction of a correction, W, for slit width variations can correct all but the largest defects to the required 1% precision. W is an average over wavelength and is a function of pixel position along the slit. The 1-D function W is the raw flat data image collapsed along the spectral direction and is analogous to the orthogonal correction R L (λ), which is the average lamp spectrum collapsed along the slit direction and is used to correct the flat field images for narrow emission lines. The product of the L- and P-flats is the original data image with the instrumental signature removed; so that the BLB Eq. 4 becomes LP = R L W R L ( λ) V (Eq. 1.) 1

2 where R L is the original calibration lamp image and V is just the correction for large scale cathode sensitivity changes and any OTA+STIS vignetting, which cannot be distinguished from illumination variation along the slit direction until a star is observed. V is now defined as just the denominator of the BLB Eq. 3: V( p i, λ) = R * ( λ) R * ( p i, λ) (Eq. 2.) where R * is the response to the star after application of the L- and P-flats with the initial guess of V=1. The smooth 2-D function V defined for every pixel is the smooth fit to V(p i,λ) at the discrete ~11 p i positions along the slit direction. The average of the 11 stellar spectra is R * (λ). Figures 1 and 2 show examples of average lamp spectra R L (λ), aka spectral S-flat, and of W, aka slit width W-flat, for a broad 2 wide slit and a narrow 0.1 slit, respectively. The BLB Eq. 5 becomes L = LP = R L ( W R L ( λ) ) V (Eq. 3.) Since the image WR L (λ) is normalized to the average of R L, L is just the low frequency variation of the sensitivity with respect to the average sensitivity. Eq. 6 is now P = LP L = LP LP = R L ( W R L ( λ) ) R L ( W R L ( λ) ) (Eq. 4.) where the < > indicates a 9x9 median filter followed by a rebinning to a 128x128 image followed by a 2x2 box smoothing, instead of the 21x21 median and 256x256 rebinning of BLB. A 9x9 (lores px) median is sufficient to remove the small MAMA dark spots, while the heavier binning improves S/N in the short exposure images obtained for the typical L- flat. A straightforward application of Eq. 3-4 to the lab exposures with V set to unity, consists of the following steps: 1. Geometrically correcting the original co-added image to make the dispersion and spatial axes parallel to the x and y axes of the rotated image. 2. Collapsing this corrected image along the separate axes to obtain the S-flat and the W-flat averages. 3. Applying the inverse geometric distortion correction to transform the WS-flat product image back to the original distorted space to get the denominator of Eq. 3 WR L (λ) for division into the original image R L. 2

3 4. Filling fiducials in the non-dithered flats and masking other problem regions. 5. Filtering as specified above by the < > operation to get the L-flat from Eq Applying Eq. 4 to get the P-flat, i.e., removing the lamp signature from the original co-added R L image and then dividing by the L-flat as rebinned to the original 2048x2048 hires size. 7. The L-flat and the P-flat are both normalized to unity in the central region. This procedure succeeds in separating the illuminating lamp signature from the detector signature, because the spectral and slit axes are not aligned with the MAMA detector, which has large sensitivity variations between adjacent rows of the hi-res image of ~4x. The geometric correction is mostly a rotation of ~0.02 radian for G230M. Thus, the collapsed averages in the geometrically correct space have a residual ripple with a period of ~50 hi-res px that is caused by dropping contributions from one original line, while including a new independent line that can differ by 4x in signal. The effect of this residual can be demonstrated by applying the derived LP-flat correction to the original data and displaying one of the corrected rows, as in Figure 3. Instead of removing the instrumental signature and producing a smooth spectrum of the continuum lamp, a ~1% ripple with a half-cycle of ~50px appears. To solve this problem, iterations are necessary to apply the improved P-flat from one iteration to the original image in order to reduce the error caused by residual even-odd variation and get improved flats after the next iteration. In other words, the P-flat from Eq. 4 is applied to R L before solving Eq. 3 again. Figure 3 demonstrates the effectiveness of the iterative technique on a complete row of an image corrected by its own flat field. Figure 4 quantifies the improvement for a piece of the same row after removing the slope with a spline fit. An adequate convergence is found after an initial pass plus 11 iterations for a total of 12 solutions of Eq Figure 5 shows the effectiveness of the iterations in reducing the lumpiness of an L-flat, where the periodicity of 50 hires px corresponds to ~3px in the 128x128 rebinned L-flat. Similarly, Figures 1-2 demonstrate the improvement in the W-flats and S-flats for the standard 11 iterations. 2. P-flats An extensive set of flat field exposures were obtained during ground testing at Ball Aerospace and at GSFC. The STIS IDT has stored these images in an on-line database and has assigned a unique Flight Software (FSW) number to each image. Individual exposures have up to 3000 counts per resolution element (2x2 lores pixel), so that co-addition is required to achieve the CEI spec of 10,000 counts or S/N=100 per resolution element. Masks must be constructed to locate fiducials, so those portions of each image can be given zero weight in the co-addition. Data of P-flat quality exist at six wavelength settings in the G230M mode, as summarized in Table 1. The FSW entries in Table 1 are ordered by 3

4 wavelength and clumped within wavelength for the same lamp+slit setup. Because the entrance slit is often shifted along its length in the focal plane to move the positions of the fiducials and fill in the flat field under the fiducials, only clumps of images with the same lamp illumination along the slit can be co-added. Table column headings are standard IDT database nomenclature, where eg. a mode_id of 2.2 is G230M, OSWABSP is the slit wheel step position, and MGLOBAL is the total counts/s in the entire image. The column external source indicates either deuterium ( D2 ) for the external lamp or INTER- NAL for the internal deuterium flat field lamp. As the first step in the production of flats, each of the 12 clumps of data are co-added to produce 12 raw flat data images with at least a total of 2500 counts/res-el or a S/N=50. Eq. 4 with 11 iterations is applied to each of the 12 sets of data in Table 1 to obtain 12 independent P-flats. The narrow slit defect near pixel 1155 in Figure 2 for the 52x0.1 slit and a similar defect for the two data sets that use the 0.5 wide slit require masks, because the cumulative effect of the iterative procedure broadens these narrow features in the W- flat that is derived from the geometrically rectified image. More smoothing is caused by the unrectification procedure; and the use of this broadened correction to remove the instrumental signature causes an undercorrection, leaving ~2% feature in the P-flat. Figures 6 and 7 compare P-flat data taken at Ball and a more recent P-flat obtained at GSFC to the same denominator image obtained at Ball, where both numerator images are at the same wavelength and utilize the same 52x2 entrance slit. Figure 6 shows only random noise, while a pattern appears in Figure 7, indicating a change between 96Aug-Sep and 96Nov. Tables 2-3 and Figure 8 characterize the P-flats and quantify their differences. Table 2 is analogous to Table 5 of BLB and compiles the statistics of the 12 independent P-flats for the central 20% of the x-range by 10% of the y-range of the images. The columns are in the same order as the groupings in Table 1 that enumerate the FSW images comprising each P-flat. The Poisson counting statistics, the actual one sigma rms scatter, and the max and min are tabulated for each image in three bin sizes: per resolution element of which there are 512x512 in the whole image, per lores pixel with 1024x1024 px in an image, and per hires pixel of the 2048x2048 image. The scatter and range is much larger in the hires case because of the odd-even effect in the MAMA detectors. Table 3 compares each of the first 11 P-flats to the extreme wavelength 2977Å flat with the best statistics. Table 3 is similar to Table 2, except the entries are the one sigma values for ratios of images. Image ratios test the similarity of the flat fields at different wavelengths. The first row for each of the three image sizes is the expected sigma from counting statistics, the second row is the actual scatter in the ratio images, and the third row measures the actual difference between the two ratioed images. In other words, these third rows of each set are the actual scatter with the Poisson uncertainty removed in quadrature. The results are consistent with no wavelength dependence and little MAMA 4

5 contribution to the scatter per lores pixel or per resolution element. There is a residual scatter of a few percent in the hires ratios, which demonstrates the nearly complete removal of the large ~60% pixel-to-pixel scatter of the hires flats. Since the tabulated Poisson statistics utilize the average counts, the Poisson entries for the hires case are underestimates of sigma because of the large change in sensitivity between adjacent pixels due to the odd-even effect in the MAMA electronics. The corresponding hires residuals are overestimates. The time change illustrated in Figure 7 for the 96Nov GSFC data appears most strongly in the hires residual of 9.81% in the last row of Table 3. The other P-flat data obtained in 96Nov is at 2419Å and also has a high residual of 9.22% per hires pixel at image center. To better illustrate the time change over the whole NUV MAMA, Figure 8 displays the residual one sigma values for the entire ratio image of Figure 7. This ratio image is divided into 8x8 blocks and the residual for each of the three types of scatter are listed in each of the 64 blocks. All blocks show less than 1% change per resolution element; and even most of the lores entries are <1%. However, all regions of the GSFC data differ from the Ball data by a statistically significant amount. Even though all 12 independent flats agree within the 1% specification per resolution element, the GSFC data demonstrate a change from the internally consistent set of Ball images. Thus, all of the Ball data can be averaged to make a pure NUV MAMA superflat relevant to the Ball thermal vac time period. Since the 1769 and 1933Å flats have illumination along only the central third of the slit, the superflat is the combination of the Ball data at the G230M central wavelengths of 2176, 2419, 2659, and 2977Å and is shown in Figure 9. The Poisson statistic of 0.30% per resolution element for the superflat appears in the final column of Table 2 and corresponds to a S/N=333 in regions without fiducial or slit defect masks. On-orbit flats are required to quantify changes, which degrade S/N achieved when this superflat is applied to flight data. 3. L-flats There are 63 L-flats required for the prime and intermediate supported modes for the NUV MAMA, as summarized in Clampin & Baum (1996). Of these 63, measurements for 27 modes exist: the four that also provide the P-flat data and 23 more collated in Table 4 in the same style as Table 1. The other 36 must be manufactured from measured L-flats, using assumptions about the continuity of change with wavelength. A variety of complications arise in several combinations for the different modes during the process of making an L-flat from a long slit spectrum of a continuum calibration lamp. Slit irregularities and the odd spectral emission line must be removed by geometrically rectifying the co-added image, extracting the average spectrum (S-flat) and the W-flat, creating the average image, distorting this average back to the uncorrected geometry, and dividing by the original coadded image by this average. Sometimes an emission line is too strong or the geometric 5

6 correction coefficients are preliminary, so that a special line mask is required. Emission lines at the edge of images also require a special mask as do regions of no signal, such as below the cutoff for the GSFC nitrogen purge data. The 29 long slit is not long enough to cover the cross-dispersed X-modes, so the L-flats are set to unity beyond the slit ends. For those L-flats with undithered slit positions, the fiducials must be filled with an average of the local image beyond the edges of the fiducial. Another problem is that the iterative procedure fails in regions of the lowest S/N, which is at the fiducial positions that are dithered and filled by only two of three slit positions. When the average number of counts per pixel in the fiducial positions falls below a critical value, the fiducial positions are too bright in the L-flat. For 7 hires counts, the fiducial positions are ~1% too high, while the problem is <0.1% but can still be discerned in the L-flat image at 13 counts. Thus, all future L-flat data should have at least 13 x 16 = 208, or maybe 300 to be safe, counts per resolution element, which is S/N=17. The lab data for mode G230M at 2739, 2818, 2898, and 3055Å all have less than nine hires counts at the fiducial positions and do not produce useful L-flats. The best L-flats are made from the high S/N>50 per resolution element P-flat quality data at the G230M central wavelengths of 2176, 2419, 2659, and 2977Å. Table 1 lists three repeat observations at 2419Å, four at 2659Å, and two at 2977Å. The L flats derived from these observations reproduce to <0.3% at the same wavelength for the external deuterium lamp illumination as illustrated in Figure 10 by the ratio of a 2 to a 0.5 wide slit observation at 2659Å. Unfortunately, the highest S/N L-flat at 2659Å with the 0.1 wide slit and the internal deuterium lamp (Figure 5b) does not agree with the L-flat derived from external lamp illumination, as shown in the ratio image of Figure 11. Differences at the left and right edges are overplotted in Figure 11 and amount to a total difference from top to bottom corners at either side of 1.4%. In other words, the lamp energy distributions at the top and bottom of the slit are different for at least one of the internal or external illuminations. A constant shape for the illumination along the STIS entrance slit is essential for defining L-flats. The repeat observations of the external lamp can be combined to produce a set of high quality L-flats at 2176, 2419, 2659, and 2977Å. At shorter wavelengths, the G230M L-flat data from Table 4 at 1769, 1851, 1933, and 2014Å agree within uncertainties, despite the fact that the first three utilize the internal lamp, while 2014Å has external illumination. These four L-flats are averaged with the counts in each as a weight to produce a high quality L-flat at the weighted average wavelength of 1915Å. The core set of five L-flats at 1915, 2176, 2419, 2659, and 2977Å show a consistent change with wavelength, as illustrated in Figures Changes with wavelength approach 1% only between 2176 and 2419Å, so that interpolation of flats at intermediate wavelengths may have errors of <1%. For example, the ratio of the G230M measured L- flat at 2257Å to the interpolated L-flat at 2257Å is illustrated in Figure 16. A comparison of Figure 16 with the ratio images in Figures demonstrates that most of the devia- 6

7 tions from unity and lumpy appearance of the ratio image and of the tracing in Figure 16 are caused by the low statistical significance of the numerator image. Similarly, ratio images in Figures for two cross-dispersed modes show no systematic deviations as large as 1% from the denominator image, which is extrapolated in the case of Figure 18. For G230M, only the internal lamp L-flat at 1687Å is discrepant with the extrapolated L- flat; and for the cross-dispersed modes with internal illumination, X230M at 1975Å and X230H at 2010Å show significant differences with the interpolation. Because the G230L spectral direction is nearly along the MAMA row direction, the images are first divided by the super P-flat to speed the convergence of the iterative procedure. The resulting G230L L-flats with internal and external lamps are discrepant by a few percent. Cuts along a column at nearly constant wavelength do not agree with the corresponding cut that could be interpolated from the set of 5 core L-flats, although the external lamp L-flat at G230L shows a level of variation more comparable to the core set. Perhaps, the requirement that the lamp illumination is a constant spectral shape along the entrance slit cannot be met for the broad wavelength coverage of G230L. 4. Summary A super P-flat has been constructed and is applicable to all NUV MAMA modes for data obtained on the ground. Studies of L-flats revealed these facts: Both external and internal L-flats are repeatable. The internal deuterium lamp produces different L-flats than derived for external illumination, although the internal G230M L-flats at 1769, 1851, and 1933Å are plausible extensions of the external series to shorter wavelength. External L-flats show less deviation from unity over the detector than L-flats made with the internal lamp. There is a smooth progression with wavelength for the five core L-flats from 1915 to 2977Å; and most of the measured L-flats are adequately represented to better than 1% accuracy by interpolation or extrapolation from this set. External cross-dispersed L-flats agree with external G230M L-flats, while internal cross-dispersed L-flats differ from internal G230M. For example, internal X230M- 1975Å and X230H-2010Å differ radically from internal G230M data at nearby wavelengths. None of the three possible choices for the G230L L-flat are consistent. A full set of 63 L-flats can be generated from the core set of five; however, serious discrepancies still exist. Internal and external lamps produce different L-flats, and four of the 27 measured modes have L-flats that do not fit a pattern and cannot be confirmed with existing data. Hopefully, the on-orbit vignetting measurements using stars at 11 positions 7

8 along the 52x2 slit will resolve the question of proper L-flats. Because of these problems and because the L-flats corrections are only of order 1%, we recommend populating the pipeline with unit L-flat files, initially. An effect of this choice of L=1 is to make the sensitivity for point sources a function of position along the slit direction. 5. References Bohlin, R. C., Lindler, D. J., & Baum, S. 1996, Instrument Science Report, STIS , (Baltimore:STScI). Clampin, M., & Baum, S. 1996, Instrument Science Report, STIS A, (Baltimore:STScI). 8

9 Table 1. STIS prelaunch FWS test data catalog of P flat data 9 FSW Entry mode _id cenwave (Å) OSWABSP INTEG (s) SMS Name EXPSTART SLITSIZE (arcsec) MGLOBAL external source COMMENTS SEP :35 52 X D2 Band 2 FF Methodology SEP :13 52 X Band 2 FF Methodology SEP :53 52 X Band 2 FF Methodology SEP :32 52 X Band 2 FF Methodology SEP :10 52 X Band 2 FF Methodology SEP : 9 52 X D2 Band 2 FF Methodology SEP :36 52 X Band 2 FF Methodology SEP :58 52 X Band 2 FF Methodology SEP :19 52 X Band 2 FF Methodology SEP :40 52 X Band 2 FF Methodology OFLT2P7B 29-AUG :49 52 X D2+ND0+DIFFU B2 Extern Deut FFs OFLT2P7B 29-AUG :42 52 X D2+ND0+DIFFU B2 Extern Deut FFs OFLT2P7B 29-AUG :49 52 X D2+ND0+DIFFU B2 Extern Deut FFs OFLT2P7B 29-AUG :42 52 X D2+ND0+DIFFU B2 Extern Deut FFs OFLT2P7B 30-AUG :56 52 X D2+ND0+DIFFU B2 Extern Deut FFs OFLT2P7B 30-AUG :49 52 X D2+ND0+DIFFU B2 Extern Deut FFs TST AUG :36 52 X D2 B2 Extern Deut FFs TST AUG :29 52 X B2 Extern Deut FFs OFT2P10D 1-SEP :28 52 X D2+ND0.5 B2 EXT D2 FF OFT2P10D 1-SEP :14 52 X D2+ND0.5 B2 EXT D2 FF OFT2P10D 1-SEP :59 52 X D2+ND0.5 B2 EXT D2 FF OFT2P10D 1-SEP :59 52 X D2+ND0.5 B2 EXT D2 FF OFT2P10D 1-SEP :44 52 X D2+ND0.5 B2 EXT D2 FF NOV : 6 52 X D2 Verif. of DMA Timeout Fix NOV : 4 52 X Verif. of DMA Timeout Fix OMIECHK6 12-NOV :40 52 X Verif. of DMA Timeout Fix OMIECHK6 12-NOV :32 52 X Verif. of DMA Timeout Fix OMIECHK6 12-NOV :40 52 X Verif. of DMA Timeout OFLT2STB 23-AUG :56 52 X INTERNAL Bnd2 Flat Field Stability OFLT2STB 23-AUG :55 52 X Bnd2 Flat Field Stability OFLT2STB 23-AUG :57 52 X Bnd2 Flat Field Stability

10 Table 1. STIS prelaunch FWS test data catalog of P flat data (Continued) 10 FSW Entry mode _id cenwave (Å) OSWABSP INTEG (s) SMS Name EXPSTART SLITSIZE (arcsec) MGLOBAL OFLT2STB 23-AUG :57 52 X INTERNAL Bnd2 Flat Field Stability OFLAT2DF 25-AUG :42 52 X Band 2 Flat Field OFLAT2DF 25-AUG :41 52 X Band 2 Flat Field OFLAT2DF 26-AUG :43 52 X Band 2 Flat Field OFLAT2DF 26-AUG :43 52 X Band 2 Flat Field OFLAT2DF 26-AUG :45 52 X Band 2 Flat Field OFLAT2DF 26-AUG :45 52 X Band 2 Flat Field OFLAT2DF 26-AUG :47 52 X Band 2 Flat Field OFLAT2DF 26-AUG :47 52 X Band 2 Flat Field OFLAT2DF 26-AUG :49 52 X Band 2 Flat Field OFLAT2DF 26-AUG :49 52 X Band 2 Flat Field OFLAT2DF 26-AUG :51 52 X Band 2 Flat Field OFLAT2DF 26-AUG :51 52 X Band 2 Flat Field OFLAT2DF 26-AUG :53 52 X Band 2 Flat Field OFLT2STB 16-SEP :58 52 X Flat Field Stability OFLT2STB 16-SEP :57 52 X Flat Field Stability OFLT2STB 16-SEP :59 52 X Flat Field Stability OFLT2STB 16-SEP :59 52 X Flat Field Stability OFLT2EX3 29-AUG :22 52 X D2 B2 Extern Deut FFs OFLT2EX3 29-AUG :10 52 X B2 Extern Deut FFs OFLT2EX3 29-AUG : 9 52 X B2 Extern Deut FFs OFLT2EX3 29-AUG : 2 52 X B2 Extern Deut FFs OFLT2EX3 29-AUG : 5 52 X B2 Extern Deut FFs OFLT2EX3 11-SEP :51 52 X B2 Flat Field OFLT2EX3 12-SEP :44 52 X B2 Flat Field OFLT2EX3 30-AUG :46 52 X D2 Band 2 Ext D2 Flat OFLT2EX3 30-AUG :14 52 X Band 2 Ext D2 Flat OFLT2EX3 30-AUG :41 52 X Band 2 Ext D2 Flat OFLT2EX3 30-AUG : 8 52 X Band 2 Ext D2 Flat OFLT2EX3 30-AUG :27 52 X Band 2 Ext D2 Flat external source COMMENTS

11 Table 1. STIS prelaunch FWS test data catalog of P flat data (Continued) 11 FSW Entry mode _id cenwave (Å) OSWABSP INTEG (s) SMS Name EXPSTART SLITSIZE (arcsec) MGLOBAL O2P13STBA 22-NOV :16 52 X D2 Mode2.2p13 FF SN= O2P13STBA 22-NOV :13 52 X Mode2.2p13 FF SN= O2P13STBA 22-NOV :11 52 X Mode2.2p13 FF SN= O2P13STBA 22-NOV :10 52 X Mode2.2p13 FF SN= O2P13STBA 22-NOV :10 52 X Mode2.2p13 FF SN= OFT2P17B 30-AUG :30 52 X D2+ND0+DIFFU B2 Extern Deut FFs OFT2P17B 30-AUG :23 52 X D2+ND0+DIFFU B2 Extern Deut FFs OFT2P17B 30-AUG :45 52 X D2+ND0+DIFFU B2 Extern Deut FFs OFT2P17B 30-AUG :44 52 X D2+ND0+DIFFU B2 Extern Deut FFs OFT2P17D 1-SEP :49 52 X D2+ND0 Mode 2.2p17 FF w/ext D OFT2P17D 1-SEP :35 52 X D2+ND0 bad: dropped data lines OFT2P17D 1-SEP :20 52 X D2+ND0 Mode 2.2p17 FF w/ext D OFT2P17D 1-SEP :20 52 X D2+ND0 Mode 2.2p17 FF w/ext D OFT2P17D 1-SEP : 5 52 X D2+ND0 Mode 2.2p17 FF w/ext D2 external source COMMENTS

12 Table 2. Statistics of the Flat Field Images x x x x x x x x0.5 P FLATS (512x512) Poisson (%) Actual sigma (%) Minimum Maximum P FLATS (1024x1024) Poisson (%) Actual sigma (%) Minimum Maximum P FLATS (2048x2048) Poisson (%) Actual sigma (%) Minimum Maximum x x x x2 SF

13 Table 3. Statistics for the Ratio of Flat Fields to 2977Å Flat P FLATS (512x512) Poisson (%) Actual sigma (%) Resid. sigma (%) P FLATS (1024x1024) Poisson (%) Actual sigma (%) Resid. sigma (%) P FLATS (2048x2048) Poisson (%) Actual sigma (%) Resid. sigma (%)

14 Table 4. STIS Prelaunch FSW Test Data Catalog of L-flat Data 14 ENTRY mode _id cenwave (Å) OSWABSP INTEG (s) SMS Name EXPSTART SLITSIZE (arcsec) MGLOBAL external source COMMENTS OD2LFLAT 13-SEP :52 31 X INTERNAL Band 2 L-Flats OD2LFLAT 13-SEP :19 31 X Band 2 L-Flats OD2LFLAT 13-SEP :47 31 X Band 2 L-Flats NOV : 2 52 X D2 M2.1 L Flat Fields S/N= NOV :19 52 X M2.1 L Flat Fields S/N= NOV :39 52 X M2.1 L Flat Fields S/N= OD2LFLAT 13-SEP :33 52 X INTERNAL Band 2 L-Flats OD2LFLAT 13-SEP :45 52 X Band 2 L-Flats OD2LFLAT 13-SEP :58 52 X Band 2 L-Flats OD2LFLAT 13-SEP :11 52 X INTERNAL Band 2 L-Flats OD2LFLAT 13-SEP :25 52 X Band 2 L-Flats OD2LFLAT 13-SEP :40 52 X Band 2 L-Flats OD2LFLAT 13-SEP :55 52 X INTERNAL Band 2 L-Flats OD2LFLAT 13-SEP : 7 52 X Band 2 L-Flats OD2LFLAT 13-SEP :20 52 X Band 2 L-Flats OD2LFLAT 13-SEP :33 52 X INTERNAL Band 2 L-Flats OD2LFLAT 13-SEP :45 52 X Band 2 L-Flats OD2LFLAT 13-SEP :58 52 X Band 2 L-Flats O22LFLTR 12-NOV :59 52 X D2 Mode 2.2 L Flats O22LFLTR 12-NOV : 8 52 X Mode 2.2 L Flats O22LFLTR 12-NOV :18 52 X Mode 2.2 L Flats O22LFLTU 8-NOV :42 52 X D2 Bnd 2L Flats, short waves O22LFLTU 8-NOV :48 52 X Bnd 2L Flats, short waves O22LFLTU 8-NOV :55 52 X Bnd 2L Flats, short waves O22LFLTU 8-NOV :12 52 X D2 Bnd 2L Flats, short waves O22LFLTU 8-NOV :19 52 X Bnd 2L Flats, short waves O22LFLTU 8-NOV :25 52 X Bnd 2L Flats, short waves

15 Table 4. STIS Prelaunch FSW Test Data Catalog of L-flat Data (Continued) 15 ENTRY mode _id cenwave (Å) OSWABSP INTEG (s) SMS Name EXPSTART SLITSIZE (arcsec) MGLOBAL O22LFLTB 8-NOV :47 52 X D2 Band 2L Flats w/ Ext D O22LFLTB 8-NOV :53 52 X Band 2L Flats w/ Ext D O22LFLTB 8-NOV :59 52 X Band 2L Flats w/ Ext D O22LFLTB 8-NOV : 5 52 X Band 2L Flats w/ Ext D O22LFLTB 8-NOV :11 52 X Band 2L Flats w/ Ext D OMIECHCK 9-NOV :45 52 X D2 Verif. of DMA Timeout Fix OMIECHCK 9-NOV :44 52 X Verif. of DMA Timeout Fix OMIECHCK 9-NOV :43 52 X Verif. of DMA Timeout Fix NOV :45 52 X D2 Mode 2.2 LFlat S/N= NOV : 2 52 X Mode 2.2 LFlat S/N= NOV :19 52 X Mode 2.2 LFlat S/N= O22LFLTR 12-NOV :19 52 X D2 Mode 2.2 L Flats O22LFLTR 12-NOV :29 52 X Mode 2.2 L Flats O22LFLTR 12-NOV :39 52 X Mode 2.2 L Flats O22LFLTR 12-NOV :42 52 X D2 Mode 2.2 L Flats O22LFLTR 12-NOV :52 52 X Mode 2.2 L Flats O22LFLTR 12-NOV : 1 52 X Mode 2.2 L Flats O22LFLTR 12-NOV : 4 52 X D2 Mode 2.2 L Flats O22LFLTR 12-NOV :14 52 X Mode 2.2 L Flats O22LFLTR 12-NOV :24 52 X Mode 2.2 L Flats O22LFLTR 12-NOV :18 52 X D2 Mode 2.2 L Flats O22LFLTR 12-NOV :31 52 X Mode 2.2 L Flats O22LFLTR 12-NOV :44 52 X Mode 2.2 L Flats X OD2LFLAT 13-SEP : X INTERNAL Band 2 L-Flats X O27XLFLT 24-NOV : X D2 B2 XDISP FF SN=10 TST X OD2LFLAT 13-SEP : X INTERNAL Band 2 L-Flats X OD2LFLAT 13-SEP : X INTERNAL Band 2 L-Flats X O27XLFLT 24-NOV : X D2 B2 XDISP FF SN=10 TST X O27XLFLT 24-NOV : X D2 B2 XDISP FF SN=10 TST X O27XLFLT 24-NOV : X D2 B2 XDISP FF SN=10 TST X O27XLFLT 24-NOV : X D2 B2 XDISP FF SN=10 TST496 external source COMMENTS

16 6. Figure Captions 700 Wflat G230M 2977flat52X Sflat G230M 2977flat52X BOHLIN: pltwsflt 9-May :54 Fig. 1 Figure 1: Average lamp spectra R L (λ), aka spectral S-flat (below), and W, aka slit width W-flat (above), as collapsed along the slit direction and along the spectral direction, respectively. The mode is G230M at 2977Å with a 52x2 slit. The fiducials are filled via a set of exposures that are dithered by moving the slit along its long dimension. Within each panel, results are shown for the standard 11 iterations and for zero iterations as offset lower by a constant. The title of the plot includes the optmode, cenwave, and slitsize, while 6912 is the FSW number of the first image of the group from Table 1. 16

17 Wflat G230M 2659flat52X Sflat G230M 2659flat52X BOHLIN: pltwsflt 9-May :54 Fig. 2 Figure 2: Same as Figure 1, except for the 52x0.1 slit and 2659Å. The structure of the W-flat is caused by irregularities in the narrow slit and by a smooth interpolation across the two fiducials. The blip on the S-flat trace is anemission line. 17

18 2977flat52X2-6912/(P*L) Hires Row Iterations=11 Iterations=3 Iterations=1 Iterations= x Pixel Fig. 3 BOHLIN: deflat 21-Apr :23 Figure 3: Row 1024 of a 2048x2048 hires image after correction with LP-flats derived with four different numbers of iterations. Since the LP-flats are applied to the same original data as used to define the flat, noise from counting statistics should cancel, leaving only the true illumination pattern and the artifacts of the procedure. As the iterations increase from bottom to top, the results becomes increasingly smooth and better represent the expected smooth spectrum of the continuum deuterium lamp. The curves are offset by 20 counts for clarity. 18

19 RESIDUAL NOISE IN 2977flat52X2-6912/(P*L) Iterations=11, rms(%)= 0.06 Hires Row Iterations=3, rms(%)= 0.08 Iterations=1, rms(%)= Iterations=0, rms(%)= x Pixel BOHLIN: deflat 21-Apr :23 Fig. 4 Figure 4: Central portion of the four curves from Figure 3 after division by a smooth spline fit. The rms scatter of the artifacts of the process drops dramatically with one iteration and more slowly, thereafter. The lower curves are offet progressively by

20 MAMA2 L Flat G230M L2659FLAT52X01.ITER Central Column Intensity & cols 10 & Pixel in 128x128 binned image Bohlin: biglfig.pro 21-Apr :13 Fig. 5a 20

21 MAMA2 L Flat G230M L2659FLAT52X01.FITS 1.01 Central Column Intensity & cols 10 & Pixel in 128x128 binned image Bohlin: biglfig.pro 21-Apr :13 Fig. 5b Figure 5: L-flat for the 0.1 internal lamp spectrum of G230L at 2659Å for a) zero iterations and b) 11 iterations. The lumpiness before iterating is attributable to imperfect removal of the lamp signature from the MAMA detector response, which varies by ~4x between even and odd rows. The line traces overlaying the grey scale images are the intensities in the central column of the 128x128 image and are associated with the abscissa and ordinate coordinates. Pixel zero is at the bottom. 21

22 9pt median, 128x128 bin, Calstis geom MAMA2 Ratio G230M P2659flat52X2-6605/P2977flat52X Bohlin: Pfig.pro 22-Apr :46 Fig. 6 Figure 6: Ratio of two P-flats obtained at Ball in 96Aug-Sep. Grey scale encoding is from 0.98 to 1.02, as indicated on the reference scale at the top. 22

23 9pt median, 128x128 bin, Calstis geom MAMA2 Ratio G230M P2659flat52X /P2977flat52X Bohlin: Pfig.pro 22-Apr :24 Fig. 7 Figure 7: Same as Figure 6, except that the numerator image was obtained at GSFC in 96Nov. The pattern superposed on the statistical noise is the difference between the two flats. 23

24 Figure 8: Quantification of the difference between the two flats ratioed in Figure 7. The ratio image is divided into 8x8 blocks and the one sigma rms residual scatter is computed for three different NUV MAMA picture element sizes in each block. Each set of three numbers is this rms for resolution elements (2x2 lores px), for lores (2x2 hires px), and for hires readout mode from top to bottom, respectively. The rms residual scatter has the counting statistical uncertainty removed and is the net change attributable to the MAMA detector. The orientation of Figure 7 & 8 are the same, so that both figures show the most change in the top left portion. No block exceeds a 1% change per resolution element (2x2 lores pixels). 24

25 MAMA2 P Flat G230M PG230Msuperflat.fits Bohlin: Pfig.pro 22-Apr :53 Fig. 9 Figure 9: Binned 1024x1024 super P-flat, which is the combination of all P-flat quality data obtained in the 96Aug-Sep timeframe. Features visible are fringing due to the nonintegral relation between the microchannel plate pores and the MAMA pixels, a few blemishes, the hexagonal pattern of the bundles of microchannel plate pores, and fine structure in the orthogonal x-y directions. 25

26 MAMA2 L Flat G230M L2659FLAT52X FITS / L2659flat52x05.FITS 1.01 Central Column Intensity Pixel in 128x128 binned image Bohlin: biglfig.pro 9-May :15 Fig. 10 Figure 10: Ratio image of two L-flats at the same wavelength with external illumination but with different slit sizes. Maximum differences are ~0.3%. The line plots from bottom to top and x,y axis labels show the intensity of column 10 (offset by -.01), the central column, and column 118 (offset by +.01) of the 128x128 pixel L-flat. The slit for the image with 52x05 in the name is 52x

27 MAMA2 L Flat G230M L2659FLAT52X01.FITS / L2659flat52x05.FITS 1.01 Central Column Intensity Pixel in 128x128 binned image Bohlin: biglfig.pro 9-May :15 Fig. 11 Figure 11: As for Figure 10, except for the ratio of an internal L-flat to the same external flat used in the denominator of Figure 10. The plots for columns 10 and 118 demonstrate a smooth and systematic difference between the internal and external L-flat of -.8% in the top left corner to +.8% in the top right corner. 27

28 MAMA L Flat G230M G230M_1915_AVG.FITS / G230M_2977_LFL.FITS 1.01 Central Column Intensity & cols 10 & Pixel in 128x128 binned image Bohlin: biglfig.pro 21-May :37 Fig. 12 Figures 12-15: Ratio of core L-flat at 1915, 2176, 2419, and 2659Å to core flat at 2977Å, respectively. Grey scale and line plots are as in Figure 10. This series of four figures demonstrates continuity of change with respect to wavelength. 28

29 MAMA L Flat G230M G230M_2176_LFL.FITS / G230M_2977_LFL.FITS 1.01 Central Column Intensity & cols 10 & Pixel in 128x128 binned image Bohlin: biglfig.pro 21-May :37 Fig. 13 Figures 12-15: Ratio of core L-flat at 1915, 2176, 2419, and 2659Å to core flat at 2977Å, respectively. Grey scale and line plots are as in Figure 10. This series of four figures demonstrates continuity of change with respect to wavelength. 29

30 MAMA L Flat G230M G230M_2419_LFL.FITS / G230M_2977_LFL.FITS 1.01 Central Column Intensity & cols 10 & Pixel in 128x128 binned image Bohlin: biglfig.pro 21-May :37 Fig. 14 Figures 12-15: Ratio of core L-flat at 1915, 2176, 2419, and 2659Å to core flat at 2977Å, respectively. Grey scale and line plots are as in Figure 10. This series of four figures demonstrates continuity of change with respect to wavelength. 30

31 MAMA L Flat G230M G230M_2659_LFL.FITS / G230M_2977_LFL.FITS 1.01 Central Column Intensity & cols 10 & Pixel in 128x128 binned image Bohlin: biglfig.pro 21-May :37 Fig. 15 Figures 12-15: Ratio of core L-flat at 1915, 2176, 2419, and 2659Å to core flat at 2977Å, respectively. Grey scale and line plots are as in Figure 10. This series of four figures demonstrates continuity of change with respect to wavelength. 31

32 MAMA L Flat G230M L2257flat52X2.fits / G230M_2257_LFL.fits 1.01 Central Column Intensity & cols 10 & Pixel in 128x128 binned image Bohlin: biglfig.pro 21-May :00 Fig. 16 Figure 16: Ratio of measured L-flat for G230M at 2257Å to the higher S/N interpolated L-flat for the same wavelength. 32

33 MAMA L Flat X230M L2703FLAT02X29.FITS / X230M_2703_LFL.fits 1.01 Central Column Intensity & cols 10 & Pixel in 128x128 binned image Bohlin: biglfig.pro 21-May :00 Fig. 17 Figure 17: Ratio of measured L-flat for X230M at 2703Å to the higher S/N interpolated L-flat for the same wavelength. 33

34 MAMA L Flat X230H L3010FLAT02X29.FITS / X230H_3010_LFL.fits 1.01 Central Column Intensity & cols 10 & Pixel in 128x128 binned image Bohlin: biglfig.pro 21-May :00 Fig. 18 Figure 18: Ratio of measured L-flat for X230H at 3010Å to the higher S/N extrapolated L-flat for the same wavelength. 34