SPACE
TELESCOPE
SCIENCE
INSTITUTE
Jennifer Mack & Ariel Bowers (WFC3 Team)
ABSTRACT
OBSERVATIONS
White dwarf HST standards were
stepped across the detector in 8
UVIS filters. To mitigate the effects of
CTE, observations from the 2 most
recent calibration programs were
post-flashed with a 12e- background.
The subarray positions were then
mapped back to the full-frame
detector coordinates.
RESULTS
TEMPERATURE DEPENDENCE
To quantify the accuracy of the UVIS flats, a white dwarf standard was stepped across the
detector field of view in a subset of filters. At visible wavelengths (F336W, F438W, F606W,
F814W), the stepped photometry is consistent to 0.7% rms and ±1.3% peak to peak. For the
bluest UV filters (F218W, F225W, F275W, F280N), the ground flats were obtained in ambient
conditions require a correction for sensitivity residuals with temperature. For these filters, the
stepped photometry varies by ±3.3% peak-to-peak, and the residuals correlate with the UV
'crosshatch' pattern. We have developed a model to correct these residuals, and the revised
flats now correct the stepped photometry to 0.7% rms and ±1.6% peak to peak. Further
testing using dithered star cluster observations is in progress, and a new set of UV flats
should be available in Fall 2014.
Filter
Proposal
F218W
13584
F225W
WD Standard
Exptime
Number of
Positions
PostFlash?
G191B2B
2.5 s
44
Y
13096
GD71
3.3 s
44
Y
F275W
12707
G191B2B
1.5 s
41
-
F280N
13584
BD+75D325
2.6 s
44
Y
F336W
12090
GD153
5.4 s
46
-
F438W
12707
G191B2B
1.3 s
50
-
F606W
13096
GD71
2.0 s
20
Y
F814W
12707
G191B2B
3.0 s
50
-
To obtain sufficient signal in the
4 bluest UV filters, the ground
flats were obtained in ambient
conditions (-49C), rather than
Ch1
under vacuum at the expected
flight temperature (-82C). To
get an idea of differences in
response due to temperature,
the F336W ground flats were
Ch2
obtained at both temperatures.
The ratio image in Figure 5
shows a residual ‘crosshatch’
pattern at an amplitude of ±2%.
Similar residual structure is
expected for the 4 UV filters,
albeit with a different strength.
Operated for NASA by AURA
-82 C
-49 C
Ratio (±2%)
Figure 5: A subsection of the F336W ground flats acquired at -82C (left) and -49C
(center) during TV3 testing. The ratio image (right) shows a residual crosshatch pattern
similar to that expected for the F218W, F225W, F275W, and F280N filters. In the plot,
black (white) corresponds to values less than (greater than) 1.0.
-49 C
-82 C
Chip 1
-49 C
-82 C
Differences in the white dwarf flux for the stepped observations
are summarized in the table at right for the current pipeline flats
(red) and for the new UV flats (blue). Flux residuals are also
plotted as a function of wavelength in Figure 10 for the current
pipeline flats and in Figure 11 for the new UV flats. The
‘crosshatch’ pattern is significantly stronger in the top chip
(UVIS1), and we see larger residuals in the photometry (blue
points) using the pipeline flats. With the new UV flats, the
photometry is consistent to 0.7% rms and ±1.6% peak to peak.
Because the residual pattern is weaker in UVIS2, users are
advised to place their target on the bottom chip if possible
to minimize flat fielding errors.
Red=UVIS2
Blue=UVIS1
Chip 2
Filter
RMS
Current (New)
Peak-to-Peak
Current (New)
F218W
1.5%
(0.7%)
±3.3%
(±1.6%)
F225W
1.2%
(0.5%)
±2.2%
(±1.1%)
F275W
0.8%
(0.7%)
±1.7%
(±1.5%)
F280N
1.8%
(0.5%)
±3.3%
(±1.2%)
F336W
0.3%
±0.8%
F438W
0.5%
±1.0%
F606W
0.7%
±1.3%
F814W
0.4%
±0.7%
Red=UVIS2
Blue=UVIS1
Figure 6: Histograms of the F336W flats at
2 temperatures. The broad ‘tail’ at low QE
for Chip 1 (left) corresponds to the dark
crosshatch pattern in the ratio image. This
structure is much less evident in Chip 2.
The dip in QE for Chip 2 is due to the gain
offset in the flats between amps C and D.
UV MODEL CORRECTION
Figure 10: Flux residuals for the white dwarf stepped photometry
are plotted as a function of filter wavelength.
To derive a model for correcting the the UV flats, we plot the
F336W flat ratio (-82C/-49C) versus the -49C flat for all pixels
in chip 1 (see Figure 7) and find that the relation is ~linear.
Figure 1: Stepped positions for program 12707 to
measure the accuracy of the UVIS flare correction and
the L-flat correction derived from inflight observations
of Omega-Centauri.
Figure 2: Stepped positions for program 13096 &
13584 to measure the strength of the UV
‘crosshatch’ pattern in the ambient ground flats.
Black indicates regions of low sensitivity in the flat.
Similarly, we plot the white dwarf flux residuals as a function
of the flat field value for that pixel. In Figure 8, we see that the
slope of this relation is similar for all filters. Removing the QE
offset between UVIS1 and UVIS2, we overplot the flux
residuals for all four UV filters in Figure 9 and note that the
combined fit is similar to that of the individual filters, and that
there is little difference between the fit for UVIS1 and UVIS2.
Thus we used a single linear fit to all points to model the UV
flat field correction.
UV MODEL TESTING
F336W
Figure 7: F336W flat ratio (-82C/-49C) versus QE
for chip 1. The relation is linear, suggesting that a
simple model may be used to correct the UV flats.
UVIS1
UVIS2
PHOTOMETRY
Photometry was performed using both DRZ and FLT*PAM frames with an aperture of r=10 pixels
and a sky annulus from 80-200 pixels. The flux residual maps for both methods agree to 0.2%.
0.391 ± 0.027
0.370 ± 0.021
0.291 ± 0.038
0.406 ± 0.017
UV, ch1
UV, ch2
UV, both
0.385 ± 0.011 0.615 ± 0.011
0.333 ± 0.025 0.666 ± 0.025
0.378 ± 0.010 0.622 ± 0.010
Calibration observations of Omega-Cen were obtained in 2009-2010 to derive low-frequency
corrections to the inflight flat field response. These data may be used to test the accuracy of the model
UV flats. For F225W, relative photometry of 2 images rotated 90 degrees are plotted in Figure 12 as a
function of X-position, where the detector is subdivided into 4 regions along the Y-axis. The left panel
shows that photometry using the pipeline flat is flat, albeit with large scatter (3.6% rms). The central
panel uses the new UV flat and the chip-dependent zeropoint ratio based on white dwarf standards.
An additional offset of 0.036 mag is required to match the cluster photometry across UVIS1 and UVIS2
(right panel). While the right panel shows a smaller overall scatter (3.0% rms), these residuals are
larger than expected based on the stepped WD photometry.
These two points suggest that the ‘crosshatch’ structure may be sensitive to color differences between
the cluster stars and the white dwarfs. Further investigation of any UV flat color-dependence is
underway. Low-frequency residuals in the right panel are also apparent, and work is in progress to use
the full set of dithered cluster data to derive residual L-flats for this new set of model UV flats.
Slope
Intercept
f336w, ch1 0.342 ± 0.001 0.657 ± 0.001
f218w, ch1
f225w, ch1
f275w, ch1
f280n, ch1
Figure 11: Same as Figure 9, but with the 4 new UV flats.
0.610 ± 0.027
0.630 ± 0.021
0.709 ± 0.038
0.592 ± 0.017
Stay tuned for more to come!
Y>3072
Figure 8: Flux difference vs. flat QE for each UV filter.
2048<Y<3072
Circle = UVIS1
Star = UVIS2
Figure 3: Percent variation relative to the mean flux per chip
for the white dwarf standard G191B2B in F814W. Positions
correlate with the map in Figure 1. The photometry in this
filter is consistent to ±0.7%, as shown in Figure 10.
Figure 4: Percent variation relative to the mean flux per chip
for the white dwarf standard GD71 in F225W. Residuals
differing by more than 3% are annotated on the plot, where
red is >1.0 and blue is < 1.0. These residuals correspond
directly to the crosshatch pattern in Figure 2.
UV Model:
•  Start with the -49C flat field for each filter.
•  Remove the QE offset between chips and the gain
offset between amps.
•  Smooth the flat using an 11x11 pixel median filter.
•  Scale the flat using a linear fit to the flux residuals from
all four UV filters.
•  Multiply the pipeline flat by the correction image to
correct the residual ‘crosshatch’ pattern.
1024<Y<2048
Y<1024
Figure 9: Flux difference vs. flat QE for all UV filters.
Figure 12: Relative photometry of Omega-Cen for F225W images rotated by 90 degrees. Plotted are the 5K brightest stars with photometric errors
<2% in an r=10 pixel aperture. The left panel uses the current pipeline flat, and the center panel uses the new model flat with the chip zeropoint ratio
derived from white dwarf standards. The right panel requires an additional 0.036 mag offset to match the cluster photometry across UVIS1 &UVIS2.