AUTHORS : Gerhardt R. Meurer, Ralph Bohlin, George Hartig, Randy Kimble, Marco Sirianni, Doug Leviton
DETECTOR : SBC
Present calibration quality (S/N ~ 100 per pixel) Solar Blind Channel (SBC) ground flats. Assess the quality of the flats for blemishes, wavelength dependence, temperature dependence, and stability. Describe the features present in the flat fields, and discuss the use of these flats for treating on-orbit data.
SBC flats were obtained during the February - March 1999 Thermal Vacuum tests at GSFC. For the most part, the SBC was illuminated with the external stimulus STUFF (STimulus for Ultraviolet Flat Fields), using a PtNe lamp illuminating a micro-mirror array. Variations from this norm included flats obtained with a spectralon diffusor instead of the micro-mirror array; internal D2 flats, "flats" obtained through a 4mm pinhole, and "flats" obtained with the coronographic arm deployed. Neither STUFF, nor any of these other sources produces an illumination pattern that that well models a uniform source observed with the HST OTA (Optical Telescope Assembly). Hence, only the high frequency component of the flats is useful.
Because the 1024x1024 SBC is count rate limited, (global limit < 3x105) about 10 hours worth of exposure are required to build up a S/N = 100 flat. The maximum exposure time for an SBC image is one hour, hence typically 10-12 images are required for each flat field. Summaries of the data frames used to make the flats are given below.
Flat field construction was done independently by GRM and RB. GRM's reductions were meant as a quick look reduction of the raw data. The IDL program combine_sbcflt.pro reads all the images of a given flat from the ACS science archive, adds them, and normalizes by the mean value of all pixels which are not "hot". Here "hot" is arbitrarily defined as a count rate of 0.54 Hz/pixel (twice the global count rate limit, if evenly spread over all pixels). The program also produces a data file that lists the following info on each frame: science archive entry numbers, filter, date and time, exposure time, weighted mean SBC temperature, global count rate, and number of hot pixels. A variety of flat field ratio images were made using the program ratfilts.pro.
Construction of the calibration quality high-frequency flats, or P-flats as they are called (for pixel-pixel flats) was done by RB, using the method discussed in Bohlin et al. (1999a). These have the low-frequency component of the observations removed. In addition, useless areas of the detector (i.e. the bad annodes, detector edges and corners) are masked out (set to a value of 1.0), and the remaining good pixels normailzed to have a mean of 1.0.
Table 1 summarizes the flat field data available. Click on the entry in the first column to get a table summarizing the data available for a given setup. Unless otherwise noted, the setup was STUFF with a PtNe lamp and the micro-mirror array as the illuminating source. ExpTime is the total exposure time for the setup. N is the number of data frames. TSBC is the exposure time weighted mean SBC tube temperature. Level Is the mean number of counts excluding the "hot" pixels. FITS Images list the available gzipped FITS images. Those marked "Raw" are the images made by GRM, while those marked "P-flat", are the high spatial frequency flats made by RB. The GIF images are meant for display only. Versions with three binning factors are included: 1x (full resolution), 2x (half size) and 4x (quarter size).
Flat field features : A variety of features are readily identified in the SBC flats. Here we list them, and describe how well we may expect to remove them from real science data with the present set of ground flats.
Wavelength response : The most prominent features in the SBC flats, the annode, MCP, and resulting Moire patterns are due to electron rather than photon optics (i.e. after each incident photon creates an electron splash in the MCP), and hence should be wavelength independent. This is confirmed in table 2 (From Bohlin et al. 1999b), which compares the measured RMS/mean of P-flats with the expectations of Poissonian statistics. The RMS/mean are reported in percent units in the last six columns of the table. The measurements are done directly on the P-flats [Poisonian rms/mean = (mean counts/pixel)-½], and also on the P-flats divided by the F115LP P-flat [Poisonian rms/mean = ((mean numerator counts/pixel)-1 + (mean denominator counts/pixel)-1)½. The residual reported in this table is the quadratic difference in the measured and Poissonian rms values.
Table 2 demonstrates that the residual rms of the P-flats is well above the Poisonian expectations. This is a measure of the variation due to the first three features noted above (the area adopted for the statistics avoids the repeller wire, broken annode and most dust motes). In comparison, the residual rms of the ratio images, are all less than 1%. Most are < ~0.5%. The exceptions are the flats made with the F122M filter (a coated crystal filter rather than a simple crystal filter like the long pass filters), F165LP (which has many dust motes), the F115LP D2 internal lamp (low count flat), and the prisms. This shows that the long pass flats are wavelength independent to better than 1%.
Repeatability/Stability : In order to test the reliability of the flat fields some data sets were repeated. In particular, the F140LP flats were repeated, and the F115LP flats were repeated with one change: the spectralon diffusor was used instead of the micro-mirror array. In both repeat sets the data are composed of three different filter wheel offsets: -1, 0, +1 steps relative to nominal. This was done to test the sensitivity of the flat field to small changes in the filter wheel position. In general, the residual rms (as defined above) of the ratio beween the repeat and original flats is at the 0.4% to 0.7% level, as is the ratio between the individual offsets. So again we see that the SBC flatfielding to the 1% level is achievable. However, in most cases there is significant striated structure at the ~0.5% level in the ratio of flats with different filter-wheel offset positions. An example is the ratio of the F140LP +1/-1 filter wheel ofset images.
Curiously a similar image, the ratio of the F115LP +1/-1 filter wheel offset flats shows virtually no structure. Indeed the residual RMS is only 0.04% of the mean! This is curious because the ratio of the F115LP +0/+1 filter wheel offset flats does show structure at the 0.5% level. It shows a strong residual Moire pattern, probably because there is a 2.6°C rise in SBC tube temperature between the offset 0 and offset +1 data. In contrast, there is only a 1.0°C rise between the offset +1 and -1 F115LP spectralon data. So, as noted above, tube Temperature changes can cause significant changes to the flat field, particularly in the Moire pattern. However, this effect does not explain the differences in the second set F140LP flats taken with different filter wheel offsets.
As a measure of the stability of the flat field we looked at the ratio of the first F140LP P-flat and the F115LP spectralon P-flat, which was obtained nine days later. SBC was powered down for three days in the interim. Furthermore the SBC tube temperature was not stable while the first few frames of F115LP spectralon data were being obtained. The residual RMS (after removiing the Poisonian component) is 0.8% of the mean. Excluding areas where a residual Moire pattern can be seen, the RMS is 0.7% of the mean. Thus typically we expect the SBC flats to be stable to better than 1% over timescales of about a week. However, if data is obtained during more drastic changes of the SBC tube temperature, they may be subject to larger faltfielding residuals (on the ~1% level), as noted above.
Super P-flats : Since some of the strongest features in the flat fields are wavelength independent we have constructed a Super Long Pass P-Flat, and a Super Prism P-Flat. These are P-flats which remove much of the pixel-to-pixel noise from SBC long-pass and prism images. The data sets used to make the Super P-flats are noted in the first column of Table 1 with the the superscripts SLP (Super Long Pass) and SPR (Super PRism). Here is a link to the Super Flats page which shows thumbnails of the two super P-flats and has links to full scale GIF and FITS images.
Dark spots (blemishes) : The Super Long Pass P-Flat was used to determine the number of blemishes in the image. We define blemishes as pixels that are below a threshold level in the normalized P-flat. The "bad areas" (broken annode, edges of detector) are excluded from the analysis. The number of pixels below threshold levels ranging from 0.1 to 0.5 are listed in Table 3 (recall that 1.0 is the average value of the P-flat). Also available is a list of blemishes which lists the pixel positions and depths of the blemishes, sorted by depth. The list was created with the IDL program sbc_defects.pro
The blemishes typically are part of the hexagonal MCP lines or are at the edge of the broken annode, or repeller wire shadow, and are clustered.
There are 15 pixels depressed below 40% of the mean local level, and we expect most of these to still "flat-field" out. The CEI (contract end item) specification is that there be less than 20 dark pixels. hence at the below 40% of the mean level, the SBC meets this specification.
CEI specifications: The SBC flat data allows us to address several CEI (contract end item) specifications. These are listed in Table 4. All of the specifications relevent to the SBC flats are met.
|Flat field non-uniformity||<8% (1 sigma)||5.8% - 6.2%||Spec. met|
|22.214.171.124-007||Flat field stability (per week)||<1%||0.8% over 9 days||Spec. met|
|126.96.36.199-009||Defect level - anode array||<3
|1 broken anode||Spec. met|
|188.8.131.52-011||Defect level - dark spots||<20||15 pixels <40% of
Note: most dark pixels flat field out
Conclusions: We have presented SBC flat fields obtained with sufficient counts to allow most pixel-to-pixel structure in SBC images to be removed to better than the percent level. This is entirely sufficient for most astronomical applications. Residual Moire patterns at the 1% to 2% level are likely to be a problem in images obtained while the SBC tube temperature is changing. Small errors in positioning the SBC filters may result in residual small scale structure at the ~0.5% level. Dust should be cleaned off of the F165LP and F115LP filters, and if possible (practical), flat field observations should be repeated.