READ1 variable pixels on row 127 and 255:

WAW -- 17 July 1997

Looking at statistics on READ1-dark frames as part of the saturation level determination, we get, in part, the following data, taken from the dawn flat sequence (scans 094,095,096, and 097) on 970608:

DATA FOR K-CAMERA SATURATION CURVE [part] 

  i   File     # OK    X_bar   X_sig    X(i)      Y_bar     Y_sig      Y(i)
  0 k0940001  64475   167.44   56.72   147.57    9980.63   1243.20    9939.0
  1 k0940091  64475   160.53   67.01   142.43    9907.12   1259.62    9862.0
  2 k0940181  64475   154.91   51.04   138.82    9848.13   1241.03    9804.0
  3 k0950001  64475   160.59   53.12   143.83    9889.40   1239.37    9846.0
  4 k0950091  64475   161.27   63.76   144.85    9885.70   1271.57    9842.0
  5 k0950181  64475   165.03   58.73   147.35    9887.60   1242.83    9845.0
  6 k0950191  64475   165.10   55.35   147.41    9891.77   1254.94    9848.0
  7 k0950201  64475   165.17   53.75   147.99    9890.36   1239.49    9848.0
  8 k0950211  64475   164.07   53.24   146.68    9884.13   1241.11    9842.0
  9 k0950221  64475   163.65   52.48   147.15    9883.57   1239.32    9840.0
 10 k0950231  64475   162.29   51.95   145.69    9879.46   1239.30    9837.0
 11 k0950241  64475   158.08  116.49   146.12    9876.49   1243.75    9836.0
 12 k0950251  64475   163.81   67.87   146.16    9865.21   1240.07    9824.0
 13 k0950261  64475   164.39   53.91   147.21    9864.96   1239.39    9823.0
 14 k0960001  64475   169.97   52.93   152.86    9944.91   1240.48    9903.0
 15 k0960111  64475   167.12   55.02   148.56    9942.51   1243.20    9900.0
 16 k0960121  64475   171.04   54.21   153.68    9951.46   1241.08    9910.0
 17 k0960131  64475   167.56   53.89   149.47    9948.75   1240.96    9907.0
 18 k0960141  64475   170.88   53.04   154.27    9958.86   1241.51    9917.0
 19 k0960151  64475   163.49  114.02   148.74    9960.47   1245.81    9921.0
 20 k0960161  64475   175.07   57.36   155.64    9983.97   1242.90    9942.0
 21 k0960171  64475   171.65   56.37   152.66    9993.95   1242.41    9953.0
 22 k0960181  64475   175.68   56.35   156.71   10010.79   1245.00    9969.0
 23 k0960191  64475   172.48   56.75   154.23   10020.89   1254.70    9979.0
 24 k0960201  64475   177.03   55.30   159.11   10039.08   1243.93    9997.0
 25 k0960211  64475   173.80   54.95   156.12   10052.30   1245.78   10012.0
 26 k0960221  64475   173.56  116.00   158.57   10069.95   1248.51   10031.0
 27 k0960231  64475   178.43   71.26   158.96   10093.71   1245.55   10053.0
 28 k0960241  64475   183.91   59.31   162.96   10123.09   1245.34   10083.0
 29 k0960251  64475   181.15   58.62   161.39   10145.18   1246.20   10104.0
 30 k0960261  64475   185.42   58.61   164.77   10182.23   1246.60   10143.0
 31 k0960271  64475   184.47   59.15   163.54   10215.82   1257.65   10176.0
 32 k0970001  64475   204.40   66.08   179.69   10539.51   1255.38   10503.0

The First three columns just give row, data file, and number of good pixels (not masked).

The next three all refer to the X axis of the saturation plot, (READ1-dark). They give the frame average, sigma, and median, resectively. (The median is used as the abscissa in the saturation plot).

The next three columns give the same information for READ2, the ordinate.

All columns have been masked, so the known bad pixels should not affect the numbers you see.

The thing to notice is the X_sig column, in which there are occasional values that are quite a bit out of line. Looking at images of the 0960151 (#19 in the table) and differencing with its neighbors in the table, it seems that the cause is at the quadrant boundary, rows iy = 127 & iy= 255, where a big chunk of pixels (eg, 50% of the entire row) sometimes goes in and out from one frame to another. This does not seem to happen in READ2, only READ1.

The figure plots pixel DN's for the second of the three discrepant X_sig's in the table, K0960151, along a row of constant Y, iy=127, and shows two confusing but unimportant things, and one thing that I think is important.

  1. First, a fair number of pixels have values of about 8370, the frame median. These are actually in the mask -- they have just been set to the median to suppress them in images.

  2. Second, note the prominant odd/even effect, which is a feature of the AD conversion. It causes even numbered pixels to be about 1500 DN higher than adjacent odd numbered ones. The effect is not limited to row 127 or 255, but is general. As a result raw data frame images images have a column structure. However, this is a known phenomenon, and its effects are removed by the flat-fielding.

  3. Finally, the important thing, there are two blocks of about 32 frames which are about 1500 DN lower than their neighbors. These blocks come and go from frame to frame, and seem to be the cause of the large x_sig that got us here in the first place. They appear also in row iy=255, and I believe the four (2 in row 127, two in row 255) come and go together, although I have not verified this carefully.

Post-script plot for 979608, k0960151_1.fits, row 127.

Go To:
Bill's Home | IPAC Home Page | JPL

e-mail: Bill Wheaton
URL: http://spider.ipac.caltech.edu/staff/waw/index.html
Last modified: 18 September 1997