HST WFC3: UVIS ChargeTransfer Efficiency Losses:
Mitigation and Correction
SPACE
TELESCOPE
SCIENCE
INSTITUTE
Operated for NASA by AURA
J. Anderson, S. Baggett, J. MacKenty, and the WFC3 Team (STScI)
0. BACKGROUND
2. WFC3/UVIS CTE MODEL
HST’s low-earth orbit environment damages CCDs, causing CTE
losses. In WFC3, these losses are even worse than expected
when the background is low, exacerbated by the low dark current
(~6 e-/hr in 2014) and the low sky backgrounds typical of many of
its imaging modes (UV and narrow-band filters, more efficient
short exposures, etc.). By ensuring a moderate image
background (e.g. longer exposure times and/or applying postflash), observers can keep CTE losses below ~20%, a regime
where pixel-based corrections are quite accurate.
Hot pixels in dark frames are used to empirically model CTE losses on a pixelby-pixel basis (Massey 2010; Anderson & Bedin, 2010). The model is applied
to science images to restore charge to its original location. In practice, we:
1) Identify bright warm pixels (WPs) in long (800 sec) darks i.e., with
lots of counts and small corrections.
2) Scale down to estimate “truth” in short-exposure darks.
3) Observe the surviving counts in short darks (where trail is too faint
to measure).
4) Tabulate losses as a function of WP size and background.
5) Fit a comprehensive forward model to this data.
6) Invert to obtain correction for science images.
1. MITIGATION OPTIONS
Losses
Truth
1) Place small targets close to the readout amp.
2) Ensure sufficient image background (e.g. broader filters, longer
exposure times and/or post-flash).
3) Apply formula-based corrections to aperture photometry results
(see Noeske et al. poster, this workshop).
4) Correct images using pixel-based image correction (this poster).
÷8
1) Bright WP
in an 800s
long dark
Truth
2) Expected WP
in a 100s
short dark
Losses
3) Observed WP
in the 100s short
dark on no bkgd
4) Observed WP in
100s short dark
dark w/post-flash
Post-flashed biases and darks revealed a new kind of peculiar pixel: those that
trap some of the electrons they collect and, as a result, end up registering low
counts. Dubbed ‘sink pixels’, they appear to contain some kind of charge sink,
likely radiation-induced as their number is increasing steadily over time.
‘truth’ image
No background
1) Identify a range of warm pixels (WPs) on a range of
backgrounds (bkgd).
2) Measure
50elosses as a
function of WP
level and bkgd.
This plot shows
30ethe surviving e−
in WPs that
had 10, 30, and
10e−
50 e at the top
of the detector.
Data were
taken in March
2014.
3) Fit a smooth model to determine the marginal losses
as a function of e− packet size.
6. STATUS/FUTURE WORK
5. SINK PIXELS
4. EFFICACY OF CORRECTION
Traps capture and release
charge during readout,
generating trails of charge
above the sources (middle
panel). A modest amount of
background can significantly
reduce CTE losses.
3. MEASURING THE CTE LOSSES
At bottom left: a ~35 e- sink pixel. Almost a delta-function in high-background
images, CTE loss forms a long trough in lower-background images. At bottom
right, the distribution of pixels in a 100 e− post-flashed dark. While only 0.05%
of the pixels have
sinks deeper than
20 e−, the
associated troughs
impact about 0.5%
of the pixels in
low-background
images, equivalent
to the fraction of
pixels considered
warm/hot.
16e- background
WFC3 image subsections
farthest from the amplifier
before (top) and after
(bottom) application of the
pixel-based CTE correction.
(see also Dark Calibration
poster). Plans are to
incorporate the code into the
calibration pipeline in 2014.
The CTE correction works well but the nature of the
algorithm is such that it cannot completely recover
what was lost, particularly at the faintest levels. To
avoid amplification of read noise, the algorithm is
conservative in its reconstruction at the low
background levels where losses are non-linear.
1) Standalone FORTRAN code available to correct raw / flt
http://www.stsci.edu/hst/wfc3/tools/cte_tools
2) Runs in parallel (up to 40x faster than original version!)
3) Runs on all full-frames and most supported subarrays
(workaround available for unsupported subarrays).
4) Next: finalize characterization and flagging of ‘sink’ pixels.
5) Re-constrain model with 2014 data
6) Implement correction in MAST pipeline (by Dec 2014)
7) Correct for small amount of serial CTE (see ISR-2014-02)
7. CONCLUSIONS
Radiation damage continues to degrade the WFC3/UVIS
detector but our knowledge and characterization of the
detector is improving even faster. By ensuring sufficient
background (post-flashing if necessary), CTE losses can
be kept below the 20% “perturbation” level for years to
come. The pixel-based CTE correction, along with its
implementation in the MAST pipeline later this year, will
help keep WFC3/UVIS at peak performance.
\
WFC3: www.stsci.edu/hst/wfc3
STScI general help desk: help@stsci.edu
UVIS CTE: www.stsci.edu/hst/wfc3/ins_performance/CTE
Massey, R., et al., “Pixel-based correction for Charge Transfer Inefficiency in HST/ACS”, MNRAS 401, 2010.
Anderson, J., & Bedin, L., An Empirical Pixel-Based Correction for HST/ACS, PASP 122, 1035, 2010.