NIS Ge Calibration Report #3

Based on Bell's NIS Ge Calibration Report #2

From: Jim Bell (jimbo@anarchy.arc.nasa.gov)
Date: 2/22/95
Subject: NIS Ge Calibration Report #2

I am continuing to look into the linearity of the NIS flight Ge detectors using the calibration test data obtained at APL. Background and initial results can be found in Report #1.

This report describes the linearity behavior of the Ge detectors at temperatures other than -26 C and in non-vacuum conditions. Report #1 concluded that the detectors were behaving linearly at T = -26 C over the range of illuminations measured, but that more complete tests were needed to better characterize the array.

NEW RESULTS AT T = -20, T = -10, and T = +24

(A) Gain state = 1.

At T = -20 and T = -10 all of the Ge detector elements exhibit linear behavior over all of the illuminations measured. An example is shown in Figure 1, for T= -10 at 1200 nm. The only anomalies found in the test data are a few bad measurements at 800 and 900 nm (e.g., Figure 2), but if the outlier data are discarded as being due to some systematic error in data acquisition, then these anomalies disappear. The reasons for the bad measurements need to be determined.

Figure 1

Figure 2

At T = +24, both in vacuum and in air, there are substantial deviations in linearity in elements 16, 28, and 30 at all wavelengths measured. This is because the signal goes to zero under low illumination conditions in these elements, and the data can no longer be satisfactorily fit with a straight line (Figure 3). This has to do with the "purposely added offset" issue Scott discussed in his previous reports, and does not indicate any inherently non-linear behavior in the detectors themselves. Nonetheless, these tests should be repeated after the purposely added offset correction has been made. Alternately, the Science Team and/or Project may decide that it is not critical to characterize the detectors this stringently at room temperatures because they are not expected to operate at such high temperatures. We should discuss this issue at our next Team meeting.

Figure 3

NIS Ge array testing: Gain = 1 Summary Table

OK  = All elements of array exhibit linearity with R**2 > 0.999
Xnn = All elements EXCEPT numbers nn exhibit linearity with R**2 > 0.999
BAD = More than 10 elements exhibit R**2 < 0.999 in this test
N/A = situation not tested

 Wvl    T = -26       -20          -10          +24(vac)       +24(air)
-----  ----------   -----------  -----------  ------------   ------------
800       X2,3         OK       X20,21,22,29    X16,28,30        N/A
900       BAD          BAD          OK          X16,28,30        N/A
1000      OK           OK           OK           X28,30          N/A
1100      OK           OK           OK           X28,30          N/A
1200      OK           OK           OK           X28,30        X28,30
1300      OK           OK           OK            N/A          X28,30
1400      OK           OK           OK            N/A        X16,28,30
1500      OK           OK           OK            N/A          X28,30
-------------------------------------------------------------------------

(B) Gain state = 10

Very few of the test measurements yield consistently good information on the linearity of the Ge detectors at gain=10. This is because, in many cases, the detectors are saturated at the highest light levels and thus the DN vs. incident flux curves flatten out at the top end (Figures 4a and 4b), yielding a poor fit. In other cases, bad measurements using a specific ND filter act to lower the goodness of fit coefficient and make the detectors appear less linear than they really are (Figure 5).

Figure 4a

Figure 4b

Figure 5

For the few cases where good sets of non-saturated measurements were obtained, the detectors appear to be behaving linearly at T = -20 and -10.

The linearity results at T = +24 (air or vacuum) and gain=10 are erratic because many of the measurements exhibited saturation or zero DN values for incident fluxes below the ND 0.5 value (Figure 6). This is especially true for channels 16, 19, 28, and 30.

Figure 6

The behavior of the Ge detectors at gain=10 needs to be much more fully determined in the next round of tests.

NIS Ge array testing: Gain = 10 Summary Table

OK  = All elements of array exhibit linearity with R**2 > 0.999
Xnn = All elements EXCEPT numbers nn exhibit linearity with R**2 > 0.999
BAD = More than 10 elements exhibit R**2 < 0.999 in this test
N/A = situation not tested

 Wvl    T = -26       -20          -10          +24(vac)       +24(air)
-----  ----------   -----------  -----------  ------------   ------------
800       OK           OK           OK             BAD           N/A
900       OK          BAD           OK             BAD           N/A
1000    X23-27       X23-27         OK        X16,19,28,30       N/A
1100    X23-32       X23-32         OK        X16,19,28,30       N/A
1200     BAD          BAD          BAD             BAD       X16,19,23-28,30 
1300    X23-32       X23-32       X23,24           N/A      X16,19,23,24,28,30 
1400    X23-32       X23-32        X23             N/A       X16,19,28,30 
1500     BAD          BAD          BAD             N/A       X16,23-28,30 
-------------------------------------------------------------------------

CONCLUSIONS

At T=-20, -10, and +24 degrees (in vacuum and in air) and at gain=1, the flight Ge detectors all appear to exhibit excellent linear behavior over the range of incident flux sampled. All of the deviations in linearity that are seen in the test data appear to be caused by either a measurement error or by inadequate "purposeful offset" (bias) in detector elements 16, 28, and 30.
At T=-20 and -10 and at gain=10, the limited amount of usable data from these tests indicate that the detectors are behaving linearly. However, many channels exhibit saturation, and so these tests need to be repeated over a better range of incident flux values. At T=+24 (in vacuum or in air), the detectors that are not saturated or did not register zero DN values at flux levels below ND 0.5 did behave linearly. ALL of the gain=10 tests should be repeated after correcting the bias problem and devising a way to better vary the incident flux in order to confirm the tentative linearity results.
There is not enough well-behaved test data available to fully assess the linearity of the detectors as a function of wavelength. The incomplete data that are available do not indicate any wavelength- dependent linearity variations, but this should be tested for in a more rigorous way during instrument-level testing.

As usual, comments, suggestions, and revisions would be greatly appreciated.