THE NEW CALACS
Pixel-based CTE Correction of ACS/WFC:
Potential Benefits from Charge Injection
David Golimowski1, J. Anderson1, L. J. Smith1, J. MacKenty1, E. Cheng2,
A. Waczynski3, E. Graham3, E. Wilson3, L. Mazzuca3, and M. Loose4
1Space
Telescope Science Institute, 2Conceptual Analytics, LLC, 3NASA Goddard Space Flight
Center, 4Markury Scientific Inc.
Putting the electrons back where they belong
(a)
(b)
(c)
Figure 4: Stacking plots of dark frames obtained with proton-irradiated spare WFC CCD and
radioactive Fe55 isotope whose X-ray emission produces pixel signals of 1620 e− and 1780 e
−. Insets show vertical profiles of X-ray events in each stacking plot. (a) Pre-radiation plots
show nearly constant X-ray values across 2048 CCD rows (i.e., near-perfect CTE of
0.999996). (b) Plots after irradiation equivalent to ~5 yr in HST orbit; CTE has degraded to
0.9998. (c) Post-rad plots after flat-field CI of 15000 e−; CTE has been restored to 0.999995
with increased noise.
(a)
(b)
Figure 5: Measurements of average (a) CTE and (b) noise between charge-injected lines as
a function of CI line intensity and line spacing. The CTE measurements are based on
stacking plots from dark frames exposured to Fe55 X-rays (Fig. 3b), and so reflect the CTE
at a pixel signal level of 1620 e− and very low background signal. LABORATORY INVESTIGATION OF CHARGE INJECTION 2.
ABSTRACT
The use of charge injection (CI) to mitigate charge transfer inefficiency
(CTI) in WFC3ʼs UVIS channel has led us to study the possible use of CI
with ACS/WFC and its potential benefit to the pixel-based CTI correction of
Anderson & Bedin (2010). We have demonstrated CTI mitigation in the
laboratory using both flat-field and line-injection modes of CI with an
irradiated spare WFC CCD. Although the mitigation and noise characteristics of line injection are favorable for scientific use in the regime of very
low sky background (< 10 e− pix−1) as is prevalent in UVIS exposures, line
injection is impractical for most WFC exposures, which have backgrounds
of 40−150 e− pix−1. Also, the short release time of WFC charge traps
precludes uniform mitigation in the interline regions and complicates
science data analysis. However, a continuous flat-field injection of 1000 e−
with ≤ 7 e− noise would limit CTI to 20% at all signal levels and pixel
locations while minimally affecting the S/N ratios of the sources. Such CTI
losses could then be corrected by the pixel-based algorithm to < 5%. 1. CURRENT STATE OF CTI MITIGATION
The WFC is primarily a broadband imager and low-res grism spectrometer
optimized for V, R, and I bands. Thus, most WFC science images have
large sky background signals that fill pixel charge traps and mitigate CTI
effects on astronomical targets during the exposure. Also, Anderson &
Bedin (2010; hereafter A&B) have developed a pixel-based algorithm for
reconstructing WFC images that lose < 25% of their signals to CTI.   What sky backgrounds do WFC images typically see?
In Cycle 19, 83% of science exposures were obtained with broadband
filters whose sky background rates are > 0.05 e− pix−1 s−1 (i.e., > 18 e− for
the most commonly used 339 s exposure).
Figure 1: Distribution of measured sky
backgrounds in all ACS/WFC science
images recorded between May 2010 and
August 2010.
The blue hatched region
denotes short exposures that are used to
enhanced the dynamic range of associated
deep exposures. The red arrow marks the
range 20−70 DN (40−140 e−) typical of
most WFC science exposures.
  How much CTI mitigation does the natural sky provide?
We used the recently revised A&B CTI model and TinyTim PSF models
(Krist et al. 2011) to simulate CTI-degraded WFC point sources for a range
of source brightnesses and sky background levels.
Signal lost from 5x5 pixel aperture
−)
SKY
BACKGROUND
(e
FLUX
0
20
50 100 200
(a)
10 e− 95% 70% 25% 15% 10%
50 e− 80% 60% 20% 15% 10%
100 e− 80% 35% 20% 15% 10%
150 e− 75% 30% 15% 15% 10%
200 e− 65% 30% 15% 15% 10%
(b)
700 e− 55% 25% 15% 12% 10%
Note: shaded columns reflect conditions
for typical WFC exposures
(c)
Figure 2: Simulated WFC PSFs for a 200 e− source located at row 2000 for four
pixelation phases. (a) Zero background with perfect CTE. (b) Zero background with
current level of CTI. (c) 40 e− background with current level of CTI.
  What does the current pixel-based correction do?
Image reconstruction errors are 10−20% of the signal loss from CTI. For
most WFC science images with sky backgrounds of > 50 e−, the signal
losses are < 25%, so the correction is good to < 5%.
June 2009: The unexpected discovery after SM4 that the circuitry of the ACS/
WFC CCDs can be manipulated to inject charge into the image area prompts
the formation of a CTI Mitigation Task Force at GSFC to investigate on-orbit CI
as a means of improving the degraded CTE.
May 2010: GSFCʼs Detector Characterization Laboratory (DCL) demonstrates
stable “flat-field” CI over image areas of three spare ACS CCDs by reversing
the overflow drains and transfer gates (Fig. 3a). Best performance is achieved
with CI signal of ~104 e− and 15 e− noise (Fig. 4).
(b)
(a)
4 e− over image area of spare WFC CCD by reverse-biasing
Figure
3:
(a)
“Flat-field”
CI
of
~10
overflow drain and transfer gate. (b) Line injection every 100th row of image area by charging
and reverse-clocking serial register. The random bright pixels are from Fe55 X-rays.
November 2010:
DCL attempts to lower noise via various techniques:
(1) faster transfer time; (2) pulsed vs continuous CI; (3) average injected signal
by (a) rapid vertical shift, (b) allowing charge to flow across all columns, and
(c) ʻsmearingʼ during line transfer. None were successful, but recent efforts
look promising.
September 2011: DCL demonstrates CI in every nth line of image area by
lowering reset drain to charge the serial register and then reverse-clock the
charge from the serial register into the image area (Fig. 3b). Best CTE and
noise characteristics achieved for CI of ~104 e− spaced by 10−20 lines (Fig. 5).
3. CAN CI BENEFIT ACS SCIENCE?
CONCLUSIONS
Lab tests show that CI mitigates the impact of CTI during the image exposure.
Unlike the pixel-based correction, CI can prevent otherwise unrecoverable loss
of S/N. But does CI improve CTI better than the natural sky background? We
first model the line injection, as it imparts lower noise than flat-field CI.
Figure 6: Simulated WFC PSFs at
row 2000 as in Fig. 2. (a) 40 e−
background with current level of
CTI and no CI. (b) Same image
with 100 DN (200 e−) of noiseless
line injection every 10 pixel rows.
(b)
(a)



150 e-
1500 e-
100 e-
(200 e-)
Line injection provides moderate mitigation of CTI losses (15% from
initial 25%) in typical ACS/WFC science images. However, this mitigation
is non-uniform between injected lines because of the short release times
(< 5 parallel shifts) of the charge traps in the WFC CCDs.
The mitigation suffers non-uniform Poisson noise from the released injected
charge which is worst in the regimes where the mitigation is most effective.
Few astronomers would tolerate such complicated noise structure.
Uniform “flat-field” CI avoids the non-uniform mitigation and noise of line
injection, but science photons must be collected in charge-injected pixels
instead of between injected lines. Consequently, the intrinsic CI noise must
be less than the sky-background noise (i.e., < 7 e−) to avoid handicapping
ACS discovery space. Early attempts to lower the flat-field CI noise
below 15 e− were unsuccessful, but new efforts at DCL look promising.
  On-orbit
implementation of CI as a means of mitigating CTI in
ACS/WFC may yet be pursued.
  The current implementation of CI for WFC3/UVIS is unaffected by our
700 e-
ACS investigation because UVIS enjoys less natural CTI mitigation due to
lower UV sky background, smaller pixels, largely narrowband imaging, and
lower dark current.
40 e-
200 e-
10 e-
Figure 7: Signal in 5x5 pixel box for PSFs of
various intensities, as a function of sky
background and line spacing for 100 DN
(200 e−) of injected charge. The shaded
regions mark the range of typical WFC sky
levels. The benefits of line injection (blue,
magenta, cyan) over natural sky (green) are
greatest for the lowest sky levels, but only
modest for typical WFC sky levels.
Figure 8: Poisson noise from the trails of
released charge between injected lines, for
a 25-line pattern and various CI signals and
sky backgrounds. The release probabilities
are taken from the A&B CTI model. Note
that the noise profiles are the relatively large
and non-uniform at the very lowest sky
levels where the relative benefit of line
injection over natural sky is greatest. REFERENCES
Anderson, J. & Bedin, L. R. 2010, PASP, 122, 1035–1064
Krist, J. E., Hook, R. N., & Stoehr, F. 2011, Proc SPIE, 8127