Spatial non-uniformity in the photometric response of array detectors can be produced by flat-fielding errors, flexure in the camera, distortion in the optical system, vignetting or a number of other sources. As you may recall, flat-fielding residual tests made with dark-sky flattened data taken the Prototype Camera showed that there could be as much as 15% peak-to-peak variation in the photometric response to point sources across to the array. At least some of the response non-uniformity was linked to the reset decay bias pattern. These results led us to design the unique twilight sky flat-fielding technique used in the survey. Twilight sky flats reduced the photometric residuals to <=5% peak-to-peak in the same ProtoCam data.
Date #fr RA DEC nJ nH nKs
971026n 48 352.727 +38.803 874 618 582
971125n Lost due to tape error
980113n 48 41.461 +38.790 843 822 721
980125n 48 55.209 +38.788 2007 1609 1231
980715n 48 270.220 +10.634 2712 2177 982
980801s-a 48 269.505 -0.103 1154 2191 2408
980801s-b 48 236.322 -39.364 1097 1928 2273
980801s-c 48 269.505 0.100 1167 2068 2525
990915n 48 330.259 +43.060 668 2079 1611
010123s 48 166.567 -54.385 1580 2534 2832
010126s 48 166.567 -54.387 1683 1645 1190
For each unique source, the mean brightness of all apparitions was evaluated, and the flux residual with respect to the average was calculated for each appartion. The 256x256 arrays were divided up into 10x6 bins and the average and RMS residuals for all apparitions falling within each of the 60 bins was then evaluated. The mean residual in each bin gives the relative response of pixels in that ~26x42 pixel region of the array.
To map the true relative response over the full array, we should utilize only sources detected 60 times. In practice, though, this often yields too few sources to produce a statistically useful result. Sources with 54 apparitions were used in some of the tests; this still produces a useful measure of the 2d response residual (see the 980113n summary for an example). The accuracy of the residuals measured in each bin will also be governed by the number of samples and the brightness of each sample. A reasonable number of bright sources are better than a large number of faint sources. Thus, the magnitude threshold was "tuned" for each night and band to the brightest level that provided at least ~10 bright sources per bin.
The links below give the summaries for each night. Each summary provides numerical tables showing the mean flux residual in each of the 10x6 bins, the RMS of the residuals in each bin, and the total number of star apparitions going into each bin. Also shown are images constructed from the residual and RMS tables.
Figures 1-3 show the all responsivity residual maps together, for J, H and Ks, respectively. In each Figure, the maps are arranged top-to-bottom in the order 971026n, 980113n, 980125n, 980715n and 990915n. Figures 4-6 show the J, H and Ks responsivity maps for the southern nights; the three 980801s tests are shown in the top three panels of each figure, followed by the 010123s and 010126s results. All of the responsivity images are shown with a full range of 0.9-1.1, with lighter colors indicating larger flux ratios (i.e. relatively higher responsivity) and darker colors indicating smaller ratios (i.e. relatively lower responsivity).
Figure 1 - J Responsivity Flux Residuals for 971026n, 980113n, 980125n, 980715n and 990915n
Figure 2 - H Responsivity Flux Residuals for 971026n, 980113n, 980125n, 980715n and 990915n
Figure 3 - KsResponsivity Flux Residuals for 971026n, 980113n, 980125n, 980715n and 990915n
Figure 4 - J Responsivity Flux Residuals for 980801s, 010123s and 010126s
Figure 5 - H Responsivity Flux Residuals for 980801s, 010123s and 010126s
Figure 6 - KsResponsivity Flux Residuals for 980801s, 010123s and 010126s
The six in-scan samples made for each star during normal 2MASS data acquisition will effectively smooth out any residuals in the in-scan direction. Therefore, the net cross-scan photometric residuals can be derived by column-averaging the residuals. Figures 7 and 8 show plots of the column-averaged relative cross-scan response residuals measured on all northern and southern nights, respectively. For the northern data, the red lines indicate the data from 971025n, green is 980113n, blue is 980125n, cyan is 980715n, and magenta is 990915n. For the southern data, the red, green and blue lines correspond to 980801s tests a-c, respectively, cyan is 010123s and magenta is 010126s.
Figure 7 - Column-Averaged Flux
Residual for all northern nights
Figure 8 - Column-Averaged Flux
Residual for all southern nights
The following two tables contain the relative column-averaged photometric
residuals from each of the cross-scan test nights. The columns labelled
"C" refers to the ~25 pixel-wide cross-scan bin.
Northern Column-Averaged Flux Residuals Southern Column-Averaged Flux Residuals The residual maps from 971026n appear very different from
the 1998 and 1999 tests. They are dominated by a strong gradient across
the
array, with a ~10% variation from edge-to-edge. This would be expected
to produce large photometric biases when comparing scan-to-scan overlaps
on this night. However, the satellite movie from that night
shows intermittent clouds. The photometric zero point plot for the night
also shows considerable internal scatter within the calibration scan sets,
so we cannot rule out that the apparent responsivity gradient could have
been produced by a changing transparency due to clouds during the test.
The consistency of the 1998 tests suggest that the October 1997 results
may be pathological.
The southern cross-scan tests were made using higher
density star fields than any of the northern tests and thus show generally
smoother residual maps. There is a mild spherical surface visible in
the residual maps in the J and Ks bands, and the column-average
relative residual plots show the same behavior as the northern plots.
The center-to-edge
bias in the southern H and Ks data is small, ~2-3%, consistent
with the northern system. However, the southern J-band shows a much more
substantial center-to-edge bias than the northern J-band data, as much as 5-6%
in the 980801s test. The results of the five southern tests, widely
separated in time are essentially consistent. Because the J bias is
more extreme than the H or Ks biases, there can be
a focal plane position-dependent color gradient introduced into the
photometry.
Because the column-average flux residual maps shown in Figures 7 and 8
show that the cross-scan behavior of the residuals is highly
consistent with time in each of the two hemispheres, except for the one
errant northern night, we construct the survey-mean flux residual
curves for Mt. Hopkins and CTIO. The mean curves were derived by
taking the simple average over all of the good nights of the residual
in each column bin. The tables below gives the northern and southern
mean residuals in each band, and the RMS about the mean. The RMS
is given only for illustrative purposes.
Figure 9 below shows the color biases expected from the
northern and southen differential flux residuals, plotted
as a function of cross-scan bin. The color biases are
calculated simply as -2.5*log(Fres1/Fres2), where Fres1 and
Fres2 are the flux residuals in the two bands defining the
particular color. In Figure 9, biases in J-H, J-Ks
and H-Ks are shown in the top, middled and bottom
panels, respectively. Mt. Hopkins residuals are shown in
blue and CTIO residuals are shown in magenta.
Figure 9 - Survey-mean color biases expected
for the measured cross-scan flux residuals
Figure 9 shows that there can be a maximum J-H color bias of over 0.02 mags
across both the northern and southern focal planes, but the sense of
the cross-scan dependence is reversed north and south. Sources
appear redder on the West side of the northern focal plane, but
in the southern system, sources falling midway between the
eastern edge and center appear redder. Similar behavior is seen
in the J-Ks color biases, but to a lesser amplitude.
The H-Ks color biases are very small, indicating that the
severe bias in the J-H colors is produced primarily by residuals in
the J band.
The distribution of the expected color residuals follows very closely
the color variations reported by Mike Skrutskie in his
Evidence for Flat Field Bias report. Rae Stiening is carrying out
a systemmatic analysis by measuring the mean colors of sources
as a function of cross-scan focal plane position for different time
periods at the two observatories. His
figure showing the mean J-H color as a function of scan edge distance
follows very closely the trends and amplitudes of the predicted color
variations shown above in Figure 9.
R. Cutri - IPAC
971026n 980113n 980125n 980715n 990915n
C J H Ks J H Ks J H Ks J H Ks J H Ks
1 1.0885 1.0683 1.0677 1.0119 0.9963 1.0000 1.0211 1.0108 1.0067 1.0075 0.9920 0.9990 1.0101 1.0148 1.0182
2 1.0813 1.0897 1.0763 1.0115 1.0033 0.9999 1.0118 0.9983 1.0059 1.0097 1.0036 1.0101 1.0070 1.0067 1.0123
3 1.0577 1.0472 1.0623 1.0043 0.9924 0.9986 1.0064 0.9977 1.0031 1.0029 1.0056 0.9958 1.0043 0.9949 1.0049
4 1.0363 1.0443 1.0393 1.0060 0.9916 1.0024 0.9932 0.9916 0.9967 0.9829 0.9895 0.9821 0.9925 0.9891 0.9915
5 0.9981 1.0125 1.0112 0.9890 0.9957 0.9829 0.9933 0.9945 0.9904 0.9758 0.9876 0.9778 0.9914 0.9876 0.9891
6 0.9644 0.9637 0.9611 0.9863 0.9962 1.0055 0.9929 0.9999 0.9935 0.9918 0.9925 1.0038 0.9952 0.9970 0.9910
7 0.9308 0.9351 0.9298 0.9971 1.0039 1.0030 0.9905 0.9933 0.9989 0.9905 0.9918 1.0010 0.9938 1.0034 0.9933
8 0.9235 0.9181 0.9272 0.9982 1.0080 0.9983 0.9958 0.9967 1.0001 1.0018 0.9975 0.9903 0.9998 1.0034 0.9973
9 0.9288 0.9499 0.9660 0.9926 1.0065 1.0176 1.0072 1.0065 1.0058 1.0086 1.0220 1.0124 1.0091 1.0106 1.0119
10 0.9413 0.9615 0.9636 1.0159 1.0079 1.0002 1.0087 1.0217 1.0119 1.0283 1.0060 1.0241 1.0195 1.0193 1.0227
980801s-a 980801s-b 980801s-c 010123s 010126s
C J10_21 H10_21 K10_21 J22_33 H22_33 K22_33 J68_79 H68_79 K68_79 J H Ks J H Ks
1 1.0090 1.0126 1.0172 1.0168 1.0098 1.0180 1.0208 1.0222 1.0230 1.0203 1.0167 1.0163 1.0240 1.0178 1.0138
2 0.9972 1.0069 1.0043 1.0017 1.0024 1.0091 1.0046 0.9990 1.0110 1.0056 1.0096 1.0100 1.0073 1.0094 1.0082
3 0.9905 0.9977 0.9972 0.9912 0.9958 0.9973 0.9926 1.0023 0.9970 0.9982 1.0012 1.0050 0.9993 1.0002 1.0021
4 0.9825 0.9979 0.9911 0.9900 0.9955 0.9977 0.9896 0.9966 0.9965 0.9939 0.9968 0.9988 0.9930 0.9987 0.9970
5 0.9873 0.9939 0.9886 0.9910 0.9955 0.9933 0.9885 0.9989 0.9871 0.9883 0.9949 0.9924 0.9900 0.9948 0.9954
6 0.9906 0.9918 0.9910 0.9952 0.9995 0.9941 0.9889 0.9999 0.9902 0.9929 0.9975 0.9976 0.9922 0.9968 0.9971
7 0.9909 0.9923 0.9909 0.9959 0.9999 0.9911 0.9905 0.9883 0.9915 0.9993 0.9968 0.9959 0.9992 0.9974 0.9991
8 1.0036 0.9987 1.0001 1.0043 0.9982 0.9982 0.9954 0.9983 0.9975 1.0060 1.0004 1.0004 1.0034 0.9997 1.0009
9 1.0173 1.0012 1.0092 1.0155 1.0057 1.0070 1.0115 0.9979 1.0038 1.0128 1.0074 1.0077 1.0125 1.0050 1.0072
10 1.0383 1.0181 1.0167 1.0248 1.0127 1.0138 1.0228 1.0032 1.0082 1.0243 1.0102 1.0093 1.0240 1.0119 1.0100
Discussion
In all tests, the J-band response maps are best determined
because of the larger number of sources available. The H-band response
maps are always noiser than J or Ks, as evidenced by
the RMS maps for each test.
The flux residual maps from the three 1998 and one 1999 northern
tests show only
small residuals across the array. There is a hint of
residual in the in-scan direction,
but it is small. There are peak-to-peak variation across the arrays
of up to ~5%, but most of this is averaged down by the in-scan sampling.
The general pattern of the 2-d residuals in the individual 1998 and 1999 tests
is random, with perhaps a slight spherical surface-like component (primarily
in the J-band) suggestive of an illumination pattern.
The residual
center-to-edge photometric bias implied by the column-averaged plots
indicates that residual flat-fielding errors contribute no more
than 2-3% error in the extremes. Moreover, because the shape of
the residuals are consistent between bands, no appreciable
focal-plane position-dependent color gradient should be introduced.
The shapes of the column-averaged residual plots are consistent
between the 1998 and 1999 tests.
Survey-Mean Flux Residual Maps
/North Mean Flat-Field Residuals
C Jmean Jrms Hmean Hrms Kmean Krms
1 1.012650 0.005916 1.003475 0.011031 1.005975 0.008838
2 1.010000 0.002205 1.002975 0.003475 1.007050 0.005456
3 1.004475 0.001443 0.997650 0.005725 1.000600 0.004155
4 0.993650 0.009480 0.990450 0.001338 0.993175 0.008621
5 0.987375 0.007915 0.991350 0.004358 0.985050 0.005837
6 0.991550 0.003776 0.996400 0.003047 0.998450 0.007265
7 0.992975 0.003160 0.998100 0.006441 0.999050 0.004183
8 0.998900 0.002538 1.001400 0.005319 0.996500 0.004293
9 1.004375 0.007891 1.011400 0.007326 1.011925 0.004829
10 1.018100 0.008149 1.013725 0.007922 1.014725 0.011112
/South Mean Flat-Field Residuals
C Jmean Jrms Hmean Hrms Kmean Krms
1 1.018180 0.005732 1.015820 0.004796 1.017660 0.003376
2 1.003280 0.003961 1.005460 0.004632 1.008520 0.002578
3 0.994360 0.004097 0.999440 0.002652 0.999720 0.003645
4 0.989800 0.004484 0.997100 0.001235 0.996220 0.002989
5 0.989020 0.001469 0.995600 0.001931 0.991360 0.003425
6 0.991960 0.002380 0.997100 0.003238 0.994000 0.003392
7 0.995160 0.004297 0.994940 0.004614 0.993700 0.003655
8 1.002540 0.004120 0.999060 0.000956 0.999420 0.001482
9 1.013920 0.002400 1.003440 0.003838 1.006980 0.001975
10 1.026840 0.006448 1.011220 0.005370 1.011600 0.003545
Last Update - 21 March 2001
