NIS Ge Calibration Report #2

Based on Murchie's NIS Ge Calibration Report #1

From: Scott Murchie

Date: 2/2/95

Here I describe the temperature dependence of dark current from the NIS Ge detector, and a validation of the gain setting option on the detector.

DARK CURRENT VS. TEMPERATURE

Several facts frame this analysis:

  1. In reality the 32 Ge detector elements or channels should be thought of as 32 little instruments. The behavior of dark current is independent in each channel.

  2. The Ge detector elements return unsigned 12-bit DN values in the range 0-4095. A negative DN would be returned as zero.

  3. The total offset in the data from any channel is the sum of 2 components, dark current from the detector element (which may be positive or negative) and a purposely added offset of intended to assure that the sum of the two values remains positive and thus accurately measurable. By design of the Ge detector, the same magnitude of offset is purposely added to ALL channels using a resistor, and its current value is +40.

  4. The instrument has 2 gain settings, 1 and 10.

  5. Dark current is temperature-dependent, negligible at low temperatures and increasing non-linearly with temperature. It may also change with time as the detectors are exposed to the space environment.

  6. The 40 DN added to the data is unaffected by gain setting. The gain setting does however affect the summed dark current and signal measured by the detector

  7. At gain 1, using the narrow slit and for a 1-sec integration , the expected S/N for Eros at zero phase angle (during the flyby) is ~400. However during orbit, when the highest spatial resolution NIS data will be obtained, a higher phase angle will cause the radiance of Eros and S/N to be many times lower.

Adequate determination of the 1-micron band center requires quite high S/N, in the 400 range. Several strategies could be employed to pump up S/N in the orbital data: averaging of multiple spectra (up to 16), use of the wider slit, and use of gain 10.

We expect offset to approach 40 DN at low temperatures at both gain settings in all channels. However, at a given temperature, the deviation of offset in some channels from a value of 40 will be 10 times larger at gain 10 than at gain 1. Because of the possibility of negative dark current, the offset values from each channel must be examined as a function of temperature to assure that total offset is consistently >0 so that dark current can accurately be measured and subtracted out of the data to yield signal. This is especially important at gain 10.

Thermal modeling of the NIS instrument on NEAR indicates that the expected range of operating temperatures is -40C to -32C. In detector-level testing, dark measurements (total offset) were acquired under vacuum at 4 temperatures, 27.1C, -7.8C, -17.2C, and -23.8C. This allows the temperatures dependence of offset to be investigated down to temperatures at the upper end of the expected operating range. If we see offsets approaching zero at temperatures near the operating range, this is a matter of concern, as in increase in-flight of magnitude of a negative dark current might yield a total offset read as zero, in which case we couldn't accurately measure and remove the dark current.

Figure 1 shows offset for the 32 Ge detector elements at these 4 temperatures at gain 1. Individual channels exhibit varying deviation from 40 DN, implying dark currents of different magnitude and sign. Zero offsets (implying high magnitude and negative dark current) are attained at room temperature only in 4 channels, 16, 19, 28, and 30. At low temperatures the offsets approach 40 in all channels, though the greatest deviations are as expected in the channels with the highest magnitude dark currents. And the lowest total offset values are found in the channels indicated above.


Figure 1

Figure 2 shows offset from the detector elements at the same 3 temperatures at gain 10. Note that channel 30 retains very low offset values even at the coldest test temperatures, implying a negative and high magnitude dark current. A small fractional increase in the magnitude of its dark current could yield a zero DN offset at temperature conditions approaching those in-flight. Channels 16 and 28 also retains low offset values at -17C.


Figure 2

None of these channels is situated in the 1-micron band per se, but the two worst channels are at long wavelengths where we're trying to tie the Ge and InGaAs data together. Furthermore, channel 30 is located near where a 1.4 micron OH or H2O absorption would be found if there happens to be any hydrated material on Eros. Given the much lower S/N of the InGaAs detector (which would cover other OH and H2O absorptions), if we want to have a chance at detecting any minor amounts of altered mafic minerals of Eros's surface, we need reliable performance from channel 30 (and 28).

I've therefore asked Hugo Darlington to provide us a list of possible replacements for the resistor used to add offset and estimates of the magnitude which they would add. At that point (a week or 2), a team decision should be made about whether to replace this component to increase the magnitude of purposely added offset.

VALIDATION OF GAIN SETTING

The increase in gain at "gain 10" is achieved by a resistor, which affects the signal and dark current from all channels simultaneously. The properties of the resistor are supposed to be very stable with temperature. Thus we should expect the increased gain at "gain 10" to be nearly constant between channels and independent of temperature.

I verified that this is true in the following manner. I used data acquired using the 900-nm wavelength filter and ND filter 0.5, which peak in value above 3000 for gain 10, and show no obvious evidence of saturation. I derived signals using data acquired at gain 1 and gain 10, at -17.2C and -23.8C, by subtracting the appropriate sets of offset measurements. (At -17.2C offset in channel 30 was 0 and therefore imprecise, so I eliminated data from this channel at this temperature.)

The signals measured by the 31-32 detectors under these 4 sets of conditions (2 gain settings, 2 temperatures) are plotted in Figure 3. Values are on the left axis. Note that for each gain, temperature has very little effect on signal. However the plot is divided into 2 domains, channels 1-22 and 23-32. The latter group of channels is about twice as sensitive, by design, to compensate for the steeply falling solar spectrum at the corresponding wavelengths (>1250 nm).


Figure 3

In each domain and at each temperature, I calculated the ratio (signal at gain 10 / signal at gain 1). This is plotted with values on the right axis, the ratio for -23.8C in purple and the ratio for -17.2C in orange. The means and standard deviations of this ratio are also plotted for each of the two groups of channels and each temperature. The actual increase in gain is 9.93 +/- 0.02, and shows no significant variation with temperature of wavelength. The uncertainty in the gain is also compared with the average 1/DN for each group of channels, which is proportional to the quantization noise. Basically the uncertainty and the scatter in channel-by-channel values of the gain increase can be explained entirely by quantization of the data. These results are entirely to be expected, but it doesn't hurt to check.