If CTI is small, we expect only a small amount of trailing charge behind each warm pixel. To improve the accuracy of the measurement, we need to examine the average effect on many warm pixels, and use pixels with dark current higher than 200 counts (1000 electrons) in the "imd.fits" images generated from each ACA dark current calibration, to provide hundreds or thousands of measurements per calibration.
CTI affects charge transfer in both the parallel and serial transport directions. To measure the parallel CTI, the charge in several trailing pixels in the same column as each warm pixel is measured. Similarly, serial CTI is determined from the charge in several trailing pixels in the same row as each warm pixel. In each case, the trailing charge is measured above the baseline dark current of the CCD given by the average of the two pixels immediately leading the warm pixel and the fourth trailing pixel behind the warm pixel.
CTI is expected to be approximately uniform across the CCD and thus the amount of trailing charge is proportional to the number of pixel transfers:
W N
----- = (1 - CTI)
W+T
where T is the amount of charge trailing the warm pixel
having charge W, observed after N pixel transfers of the charge packet.
Ten in-flight dark current calibrations have been performed to date, and we use these to study the spatial variation and time evolution of CTI effects on the CCD.
The code and results for this analysis are in /proj/sot/ska/acacal/cti/
Trailing Charge Images
To show the effects of CTI in each calibration, a 7x7 pixel dark current sub-image was extracted around each warm pixel in the calibration. The median image was constructed from the median values for each pixel in the set of sub-images. To extract the sub-images expected to show the maximum effect of CTI, only warm pixels in the innermost 100x100 pixel regions of each CCD quadrant were used, as shown by the red square in the following figure.
The sub-images from each CCD quadrant were appropriately reversed in the row and/or column directions to have the same orientation relative to the readout direction before median filtering. Before median filtering, each sub-image was normalized by the value of the warm pixel in the sub-image. The median images for each calibration are shown in the following figure, where the warm pixel, with relative intensity =1 from the normalization, is at sub-image coordinate (row=4,col=4). The trailing charge is clearly seen in the immediate trailing row and column pixels. The number of warm pixel sub-images contributing to the median sub-image in each calibration is shown.
It can be seen that the charge in the trailing row pixel is increasing with time, which indicates the serial CTI is increasing with time. The charge in the trailing column pixel is larger, but shows less variation with time and is present from the time of the first calibration on 1999:223 (August 11).
Data Analysis
It can be seen from the above figure that the trailing charge is largely confined to first trailing row and column pixels. Therefore, we have separately analysed the parallel and serial CTI in the CCD. For each dark current calibration, we extract warm pixels from 100 pixel wide regions of the CCD, to characterize the CTI as a function of the number of pixel transfers. For CTI in the parallel (i.e. along column) clocking direction, these regions are parallel to the serial readout registers, as shown in red in the left-hand figure below. For CTI in the serial (i.e. along row) clocking direction, these regions are perpendicular to the serial readout registers, as shown in red in the right-hand figure below. In each case, small regions adjacent to the inner quadrant boundary are excluded from analysis, to exclude warm pixels with trailing pixels in another CCD quadrant.
To examine the effect of charge transfer distance on CTI, we combine the data from the paired regions on opposite sides of the central quadrant boundary which have symmetric transfer distances. Thus effectively the CTI is separately analysed in 5 regions in the parallel and serial directions, numbered 1 to 5 in the above figures.
The measured trailing charge profiles for each region in each dark current calibration are given in the following pages. Histograms of parallel and serial CTI measurements from each warm pixel in the calibrations are also shown.
1999:223
2000:326
2001:060
2002:057
2002:137
2002:237
2002:338
2003:116
2003:195
2004:002
The time evolution of the parallel and serial trailing charge is shown in the following
figure. Separate plots of trailing charge against mission duration are shown for each
region.
For the trailing row pixels, there is only weak time evolution of the trailing charge.
But there is a clear dependence on region (i.e. charge transfer distance), as expected.
For the trailing column pixels, the trailing charge is significantly larger, but shows
little change with time or charge transfer distance.
The calculated parallel and serial CTI, calculated from equation 1, is shown in the
following figure as a function of region and mission duration. The parallel CTI is
consistent within the errors with no variation from region to region, and demonstates
only mild time evolution. But the calculated serial CTI clearly varies with region.
Conclusions
The relatively large and constant charge in the trailing column pixel, which results in the anomalous serial CTI measurements, is not consistent with expected CTI variation with time and charge transfer distance. This indicates another mechanism may be responsible for this trailing charge. For example, incomplete transfer of the pixel charge through the CCD quadrant preamplifier into the ADC may be occuring, with the remaining charge slipping into the next ADC sample.
The parallel CTI, currently about 3 x 10-5, is probably consistent with true
CTI effects. However, the readout speed
of the CCD during dark current calibrations is significantly slower than for normal ACA
star tracking operation. The time interval between parallel shifts is about 16 ms,
whereas it is typically 24 μs for normal operation. Figure 8.53 (trap emission time
constants vs CCD temperature) in "Scientific Charge-Coupled Devices" (James R. Janesick,
SPIE Press, 2001) shows there are 2 trap species with emission time constants between
16 ms and 24 μs, for a CCD operating at -15C:
(i) P-V traps have 300 μs emission time constant.
(ii) Divacancy electron traps have 50 μs emission time constant.
Both of these trap time constants are short compared to 16 ms, indicating that CTI effects may be smaller in dark current calibrations than in normal operation. However, the serial shift time in dark current calibrations is 25 μs, which is similar to the parallel and serial shift timing in normal operation. Therefore serial or parallel CTI in normal ACA operation should be smaller than about 6 x 10-5 which is the inferred serial CTI in region 5 from this analysis. Comparison with CTI effects inferred from any centroid discontinuities for stars dithering across quadrant boundaries should provide more insight.
Comments or questions: Aspect Help
Last modified: 12/27/13