HST GO/DD 13067  F763M Test Exposures
(Last update: 15 Aug 2012 - Glenn Schneider)

EXECUTIVE SUMMARY
Based on test exposures acquired in GO/DD 13067 Visit 15, appx 5 weeks ahead of the Jovian transit of Venus (J/ToV2012), we must reduce the preliminary-Planned  F763M commanded exposure times from 0.7s to 0.5s in all F763M exposures in Visits 01-14.  I will coordinate this implementation change with Tony Roman at STScI.

BACKGROUND INFORMATION
On 14 Aug 2012 UT, three test exposures in the F763M (11% wide "i" continuum) band were taken of the full disk of Jupiter in the WFC3/UVIS channel centered (via POS TARG) in the 2K^2 2-C subarray.  The purpose of these exposures were to refine the exposure time estimates defined in the Phase 1 plan using the ETC and bootstrapped - with uncertainties - from spectrally adjacent F727N (1% wide CH4) band archival images from which cloud contrasts to local features were anticipated to be significantly larger but by amounts not well known.  The goal for the test images is to plan the F763M transit imaging to avoid image saturation for the brightest features that might be seen while still proving sufficient total count statistics to a photometric precision (not accuracy) of order ppm.   Legal (commandable) exposure times for WFC3 in the exposure time domain of consideration are in the range of 0.5 to (perhaps) as high as 1.0s in 0.1s granular steps, but omitting 0.6s which is not a exposure time the instrument supports.  The Phase 1 planning suggested an exposure time of 0.7s. The Visit 15 test images were taken with commanded exposure times of 0.5s, 0.8, and 1.0s.  It is clear from these images that to avoid the risk of saturation against the brightest F763M  features that  may appear during the GO 13067 experiment, all F763M exposure times in  Visits 1 -14 inclusive should be reduced from 0.7s to commandable 0.5s.

DATA
The three images acquired, for this purpose, were processed by the OPUS calibration pipeline, and the "flt"* files (with pixel values reported in electrons) were used to assess the final tuning of the Visit 1 - 14 F763M exposure times.  Those data are non-proprietary and are now retrievable from the MAST archive (http://archive.stsi.edu)  along with other calibration reference data files and OPUS products.  The specific files of interest are:

DATA FILE NAME      EXPTIME*     UTC
ic3g15j2q_flt.fits  0.480000s    22:46:35
ic3g15j3q_flt.fits  0.800000s    22:49:37
ic3g15j4q_flt.fits  1.000000s    22:52:39
* OPUS reports a "calculated" exposure time of 0.480000s for a commanded 0.5s exposure, and I confirm this proportionally from count ratios in the ic3g15j2q and unsaturated regions of the ic3g15j3q the  images.

SATURATION AVOIDANCE
"On-orbit observations have shown that on UVIS2, the onset of saturation varies from about 67000 to 72000 electrons per pixel over the CCD, while UVIS1 has a somewhat larger range of 63000 to 71000 electrons per pixel (WFC3 ISR 2010-10)" per the WFC3 IHB for Cy20.  The referenced ISR further informs the UVIS2 detector full well depth from 66850 to 72500 e- across the detector.  Thus, to avoid saturation at any point in the image exposures must be planned to collect less than the least-depth 66850 e- full well, with margin to allow for temporal variation in the brightest cloud features. (N.B.: We are using the UVIS2 detector).


DISCUSSION
A linear-display of the  ic3g15j2q (0.48 s OPUS calculated exposure time) image, with a 0 - 60000 e- display stretch, is shown in Figure 1 (top).  To better visualize where on the disk the brightest features are located (mid-disk, equatorial region), the same data is reproduced with an intensity-mapped "rainbow"color table (bottom).   As John suggested (and good for us), the cloud contrast seen to small local features is indeed smaller in this continuum filter than the adjacent CH4 band.   Note (of digression): The FLT files analyzed and shown have been corrected for image bias (from detector overscan pixels and reference bias image), dark current, and pixel-to-pixel flat field variations, but single-image CR-hit pixels have not been corrected.  These are the isolated white (top) or black (bottom) pixels in Fig 1 and, in our full transit data analysis will have to be considered, but are not considered here in tuning exposure times for the surface brightness of the Jovian disk.

Figure 3 is a full disk histogram  over the image shown in Fig 1 from 30000 - 60000 e-, with the very brightest regions on the disk (excluding the few individual CR-hit pixels) on the disk are 52040 e- (collected in 0.48 s).  This provides a margin of x 1.28 against features brighter than the brightest observed here w.r.t. the smallest well-depth found anywhere on the UVIS2 detector (min full-well 66850 e-).

The next (longer) exposure time possible with WFC3/UVIS is 0.7s (which we had tentatively planned).  All other things being equal, this would have "exposed to collect," at the peak brightness of the disk in an equivalent image,  75892 e- , which is above over the saturation threshold.  Hence, we must reduce our F763M commanded exposure times to 0.5 s.

The second of our test images was taken with 0.8 s exposure times, from which significant saturation would be expected in the brightest regions based on the unsaturated 0.48 s image.  This is seen in Fig 3 where the linear display stretch has been set to an equivalent 0 - 100000 e- dynamic range (i.e., for 0.8s compared to 60000 e- upper range used for the Fig 1 0 - 60000 e- display).  Note in the regions where the equatorial cloud bands are brightest (but well under saturation in the 0.48s exposure) in this 0.8s exposure show image vertical streaky vertical structure, which is a hallmark of charge bleeding from saturated pixels, though virtually identical (modulo small scale structural changes over appx 3 minutes in time) elsewhere in the image.

For completeness, the 1.0s exposure is also shown (Fig 4), and clearly is "way over the top".

SUMMARY AND CONCLUSION
We must reduce our F763M commanded exposure times to 0.5 s.

ADDENDUM: A NOTE ON COUNTING STATISTICS
The 0.48s test exposure produced a count of  1.92E10 electrons over the full disk.
On Sept 20 2012 (during the transit visits) the SURFACE BRIGHTNESS of jupiter (modulo intrinsic changes in cloud structure and phase angle) will be nearly identical because:
  - the Sun_Jupiter distance changes only a very small amount
  - the 1/r^2 change in the brightness of the return signal from Jupiter_Earth is exactly compensated by the r^2 change in the angular area of Jupiter with the Earth_Jupiter distance change
The disk of Jupiter, however, will appear 1.244x larger in angular area, so all else being equal we should expect appx 2.39E10 electrons per image over the full disk.

Because of SAA impacts during the transit we will obtain only 24 (of max 30 otherwise possible) F763M images over the course of the transit, so we will collect appx 5.8E11 e-.  In terms of counting statistics, though WFC3 has an A-to-D conversion gain of 1.55 e- per ADU (Data Numbers off the A-to-D converter), so is "only" 3.7E11 counts root(n) 6.1E5 or appx 1.6 ppm counting statistics.  So our job has gotten a little harder (thanks to the SAA), but was not unforeseen and not show-stopper.




FIG 1.  Top and Bottom, same data: 0.5 sec commanded Test Image 0 - 60000 e- linear display.



FIG 2.  0.5s commanded test image Full disk histogram



FIG 3.  0.8s exposure - saturating in mid-disk equatorial clouds



FIG 4.  1.0s exposure - speaks for itself...