To: MSI/NIS Science Team

From: Mike Malin (malin@msss.com)

Subject: NEAR Calibration Update #4

Date: 10-15 January 1995

Status as of 10-15 January 1995

This page contains the results of analyses I performed on the "dirty" CCD calibration data during the past week. I undertook these analyses in order to gain some insight into the relative performance of the Thomson CCD; all of these measurements, and the interpretations and conclusions drawn from them, are subject to complete revision when data from the "clean" CCD are available.

Medium Temperature Measurements

These analyses are based on images contained in directory: /project/near/SDC/MSI/CalData/CCD_medT on computer retro.jhuapl.edu. A description of the acquired data (prepared by Scott Murchie) and a listing of subdirectories containing these images is provided in the file medT_descriptions.html. Various image parameters, which I extracted from the labels of these images, or extracted or inferred from the data themselves, are provided in the file medT_picture_data.html. Included in this table are maximum, minimum, mean, and standard deviations for a 135 X 165 pixel box (22,275 pixels total) located in a relatively clean (not really clean, but as clean as one could find) portion of the image.

Dark Current at -30 deg C

As noted in previous reports, I determined the mean and standard deviation for the 135 X 165 area. I then computed the averages of the means and standard deviations for all images taken during each test. Dark current signal vs. temperature is plotted in Figure 1 and dark current vs. shutter (exposure duration) is plotted in Figure 2.

I'm not sure why dark current seems to increase with decreasing temperature over these small Delta-T's, but this was also observed at -20 deg C. It may be a function of when or when T is being measured. As for the dark current accumulation rate, it's clear that for one set of points the change in temperature over the course of the measurements masked the steady-state accumulation rate, but the second set seems to give a reasonable result. From these figures, it appears that the average dark current signal for NEAR MSI images is about 40 DN, and the dark current accumulation rate at -30 deg C is about 1.3 DN/sec (which is about 1/2 to 1/3 of what was seen at -20 deg C).

Figure 1: Dark Current vs. Temperature at -30 deg C

Figure 2: Dark Current Accumulation Rate at -30 deg C

Neutral Density Filter Transmission at -30 deg C

A more extensive (or at least complete) set of measurements were made of the transmission of the neutral density filters, at three specific spectral bandpass/shutter speed combinations: 500 nm @ 917 ms, 700 nm @ 70 ms, and 1000 nm @ 917 ms. The results of the ND Filter Transmission test are shown in Figure 3. Again, the characteristic curve is an exponential. The same discrepancies noted in Report 2 are seen here: the 700 nm/70 ms and 1000 nm/917 ms exposures are nearly the same DNs despite the difference in exposure time, and the relationships have slightly different slopes and intercepts. Again, I suspect that the 700 nm and 1000 nm filters have different widths. Note that in this example, I have included two fits to the data, one including and one excluding the ND 2.5 values. It appears to me that we will need a significantly brighter light source if we are indeed going to calibrate the 1.5, 2.0, and 2.5 ND filters.

Figure 3: ND Filter Transmission Test at -30 deg C

Exposure Response of CCD at -30 deg C

For completeness, I'll include two analyses I did of the exposure response of the CCD at -30 deg. The first two graphs (Figures 4 and 5) show the exposure response (linearity) and the standard deviation vs. mean brightness. Figure 4 shows the relationship between exposure duration and output DN, after correcting for dark current. With a readout rate of 1 ms, it is clear that the shift away from linearity occurs between 10 and 20 ms exposure.

Figure 4: Mean Brightness versus Shutter Speed/Exposure Duration

Figure 5 shows the two, well-defined branches of the noise relationships: an essentially constant "read noise" component and power-law counting noise component. For reasons I'm investigating, however, this is not the square-root relationship it should be. [NB 1/15/95 - This and other photon-transfer curves are suspect and should be ignored for now. When I get this straightened out, look for a separate report.]

Figure 5: Standard Deviation of Brightness versus Mean Brightness

Raw Spectral Response (i.e., No Lamp Correction) at -30 deg

I had a hell of a time with this data set. There are no less than 7 ambiguous image pairs, summarized in the following table:

Ambiguous Images at -30 deg C
  File ID     Filter  Shutter ND filt	Temp	  Mean     StDev

fumw301.355	500	100	0	-29.4	  40.29	    6.62
fumw302.355	500	100	0	-29.4	 220.74	   14.12

fumw341.355	500	500	1	-29.4	 136.07	    8.89
fumw342.355	500	500	1	-29.4	 217.06	   13.34

fumw541.355	600	500	1	-29.4	 613.55    35.21
fumw542.355	600	500	1	-29.5	1089.66	   62.62

fumw941.355	800	500	1	-29.4	 819.65	   37.18
fumw942.355	800	500	1	-29.5	1467.03	   67.74

fumwb42.355*	900	500	1	-29.5	1091.54	   42.46
fumwb11.355*	900	500	1	-29.4	 613.99	   24.21

fumwd41.355	1000	500	1	-29.4	 203.27	    9.29
fumwd42.355	1000	500	1	-29.5	 339.66	   14.10

fumwe41.355	1050	500	1	-29.4	  85.14	    6.80
fumwe42.355	1050	500	1	-29.5	 121.80	    6.76

Of these image pairs, the ID numbers of the ones marked with asterisks (*) decode differently than the label information: fumwb42.355 decodes to (900 nm, ND 1) while fumwb11.355 decodes to (900 nm, ND 0.1), although both have labels that say (900 nm, ND 1, 500 ms). The rest of these pairs have identical ID numbers (as far as image type is encoded in the ID number), identical labels, but difference means and standard deviations. Owing to these abiguities, it wasn't possible to put together any reasonable spectral curves for the -30 deg C measurements.

Low Temperature Measurements

These analyses are based on images contained in directory: /project/near/SDC/MSI/CalData/CCD_loT on computer retro.jhuapl.edu. A description of the acquired data (prepared by Scott Murchie) and a listing of subdirectories containing these images is provided in the file loT_descriptions.html. Various image parameters, which I extracted from the labels of these images, or extracted or inferred from the data themselves, are provided in the file loT_picture_data.html. Included in this table are maximum, minimum, mean, and standard deviations for a 135 X 165 pixel box (22,275 pixels total) located in a relatively clean (not really clean, but as clean as one could find) portion of the image.

Dark Current at -40 deg C

I computed the dark current signal and dark current accumulation rate as described in previous reports. Dark current signal vs. temperature is plotted in Figure 6 and dark current vs. shutter (exposure duration) is plotted in Figure 7.

Again, dark current seems to increase with decreasing temperature over these small Delta-T's; again, this must be due to some experimental setup relationship, since dark current shouldn't increase with lower temperature. Since I am already suspicious of the 917 ms values since they were taken at lower temperature but give a higher dark current signal, I believe the 100 and 500 ms measurements are more self-consistent. From these figures, it apparent that the average dark current signal for NEAR MSI images is again about 40 DN (~43 DN), and the dark current accumulation rate at -40 deg C is about 0.75 DN/sec (which is about 1/2 of what was seen at -30 deg C, and 1/4 of what was seen at -20 deg C).

Figure 6: Dark Current vs. Temperature at -40 deg C

Figure 7: Dark Current Accumulation Rate at -40 deg C

Neutral Density Filter Transmission

The results of the ND Filter Transmission test are shown in Figure 8. This figure basically shows the same results as seen at -20 and -30 deg C: the detector produces an exponential relationship between the neutral density filters. The 100 ms exposure at 700 nm produces a brightness only about 50% greater than the 917 ms exposure at 1000 nm because the 700 nm filter's full-width at half-maximum (FWHM) is 37.4 nm while the FWHM for the 1000 nm filter is only 9.2 nm.

Figure 8: ND Filter Transmission Test Results

Exposure Response of CCD

Quantum efficiencies and exposure response can be determined "graphically" by use of the "photon transfer curve" (Janesick, Klaasen, and Elliott, Opt. Eng. v. 26, no. 10, 972-980, 1987). This curve plots noise against signal, and absolute quantum measurements can be made if appropriate, calibrated light sources are viewed at appropriate wavelengths. [NB 1/15/95 - There appears to be a discrepancy between this and other papers published by the same author(s), where discussing the value by which to divide the standard deviation: other articles say to divide by sqrt(2) rather than by 2. I'm looking into this and will report back in a separate report discussing the photon-transfer curve.]

Figure 9 shows the relationship between exposure duration and output DN, after correcting for dark current. I've plotted this as a log-log plot to show the loss of linearity at low shutter speeds (low illumination in these examples), since when the exposure rate and readout rate are comparable, there are substantial variations in exposure across the image. With a readout rate of 1 ms, it is clear that the shift away from linearity occurs between 10 and 20 ms exposure.

Figure 9: Mean Brightness versus Shutter Speed/Exposure Duration

I next plot the standard deviation against the mean for each image acquired (Figure 10). These data clearly show the two, well-defined branches of the noise relationships: an essentially constant "read noise" component and power-law counting noise component. Note, however, that in this representation the counting statistics limit curve is not a square-root--this is because I haven't (couldn't) extract fixed-pattern noise. [NB 1/15/95 - actually, I don't know why this curve isn't correct. This and other photon-transfer curves are suspect and should be ignored for now. When I get this straightened out, look for a separate report.]

Figure 10: Standard Deviation of Brightness versus Mean Brightness

Raw Spectral Response (i.e., No Lamp Correction)

The final set of data acquired at -40 deg addressed spectral response (Figure 11). I haven't created lamp- and filter-corrections to these curves...I think people who've done this before should take a try at it (Jim, Mark, Lucy, Joe?). Here is a table of filter locations and full-widths at half-maxima (Scott gave these to me, and said the filters were more or less gaussian in shape):

Spectral Filter Centers and Bandwidths
    Band Center (nm)	   Band Width (nm)

	 408.8			37.1
	 506.7			33.8
	 601.1			35.9
	 704.7			37.4
	 801.3			 9.7
	 902.3			10.5
	1003.0			 9.2
	1052.3			11.4
(see Report #2 for the probably spectral shape of the lamp's light)

Figure 11: Raw Spectral Response at -40 deg C

Unsuccessful Attempt at Creating Photon Transfer Curve

Using images from the -20 and -30 deg C suites that were identically exposed, I tried to create the photon transfer curve as described previously. I used the fumd007.355 as a dark image (for subtraction), the VICAR2 program 'f2' to subtract the dark from each image and to difference the two images (-30 deg minus -20 deg), and the VICAR2 program 'hist' to compute the mean and standard deviations of each 135 X 165 image sub-section (dark-corrected data and difference images). I then divided the standard deviations by sqrt(2) (as per Janesick's CCD short course notes and a telephone conversation with Klaasen). The resulting plot of brightness vs. standard deviation should be permit the graphical representation of the photon transfer curve. Figure 12 is the result; it does not follow display the expected square-root; I'm looking into this but for the time being this and other photon-transfer-like curve should be ignored.

Figure 12: Photon Transfer Curve from -30 deg minus -20 deg C Images