Results from the July-September 2005 test of CTI as a function of temperature.
The primary 1999 dataset consists of 8 observations taken during a single orbit on 17-19 September 1999. The focal plane temperature ranges from -109.2C to -69.7C and the exposure time ranges from 3.2 to 7.9 ksecs. Only events from S2 and S3 were telemetered. Otherwise, the operating mode is nearly identical to that currently used for CTI monitoring. An additional data point from 7 October 1999 with a focal plane temperature of -118.9C is included to increase the temperature range. There are no unprotected radiation belt passages between this later point and the earlier data and the time differences is short enough that we do not expect any significant change in CTI.
The 2005 dataset consists of 45 standard CTI monitoring observations taken from 24 July through 5 October 2005. The focal plane temperature ranges from -119.8C through -89.7C and the exposure time from 5.5 to 24.3 ksecs. Both I- and S-array measurements are included but only data from S2 and S3 were analyzed for comparison with the 1999 data.
Fortunately, the particle background level during the 1999 and 2005 datasets is relatively similar so variations due to sacrificial charge should be minimal. The mean particle background levels as measured by the S3 high energy reject rate was 74 and 70 cts/frame in 1999 and 2005 respectively, with a standard deviation of 1.1 and 5.4 cts/frame. The higher standard deviation in 2005 is due to scatter in the data points and from a solar event, rather than a systematic trend. No correction is being made for sacrificial charge. We estimate that the particle background variations should add no more than 1.1% standard deviation to the 2005 cti measurements and less than that to the 1999 measurements.
Both the 1999 and 2005 datasets were analyzed identically. In general, the 1999 dataset has slightly larger eventlists and much higher signal to noise due to the radioactive decay of the cal source which is currently at about 20% of the 1999 rate. This data analysis includes two procedures that are not part of the standard CTI monitoring which are described below.
For the S2 CCD at temperatures above -110C, the Mn-Kα line (5.9 keV) is increasingly blended with with the neighboring Ti-Kα line (4.5 keV) for high row numbers. The standard CTI processing algorithm is unable to reliably fit the extremely broadened line. Instead, the pulseheight versus row number was fit in Event Browser using a linear regression with sigma clipping algorithm with the initial values input by hand. This seems to produce more reasonable and more repeatable results for the high temperature data. For consistency, all the S2 data was processed in this manner. At low temperatures, the absolute CTI is not identical to the standard data processing, but has a small offset of about 6e-6.
The second non-standard procedure is to filter the S3 datasets to include only G0 events. This lowers the background level and allows for more reliable fits of the spectral line. (In the standard CTI monitoring, bad fits are discarded. In this test, each dataset has unique information, so the fitting reliability is more important. I'm considering adding this grade filtering to the standard monitoring scripts in the near future.)
Results of the CTI analysis are shown in the above plots The 1999 data is shown as blue squares while the 2005 data is red diamonds. The dotted line is not a fit but is a three point quadratic interpolation. It is intended only to guide the eye and provide an estimate of the direction of curvature.
In both cases the CTI has increased between 1999 and 2005. The size of the increase at low temperatures is consistent with that measured during regular CTI monitoring. In both cases the shape of the CTI-temperature relation has changed as well. This is not unexpected since the cause of the initial damage (S2: low energy protons in the radiation belts, S3: manufacturing) is not identical to the cause of the slow increase in the six intervening years (low and high energy protons from solar storms and cosmic rays). The change in the CTI-temperature dpendence is most impressive for S3 where the initial CTI was lower so the continual radiation damage is a larger fraction of the total damage. For S2 the change in the CTI/temperature dependence is much more subtle with only a slight change in the curvature.
The above plot is an attempt to understand the change in the S3 CTI-temperature dependence as a combination of the initial damage plus radiation damage. The 2005 S3 data is again shown as red diamonds. The dotted blue line is the 1999 S3 data and represents the initial damage. The dotted red line is the 2005 S2 data reduced by a factor of forty. This represents the additional radiation damage since 1999. The sum of these two components is shown as the dashed line. We do not expect an exact match, since the proton spectrum that caused the S2 damage is much softer than that which is causing S3 damage, but the sense of the data is well represented.