Preliminary Analysis of On-Board Squeegee Mode Test

On Saturday, April 15 a real time CAP was executed to perform a series of observations testing the squeegee readout mode on the S0 CCD. A description of the squeegee mode can be found elsewhere. The squeegee mode has been shown to improve the charge transfer efficiency of a radiation-damaged CCD in the lab. The results from the flight test show that the squeegee mode is also effective in improving performance of the flight CCDs.

S0, the CCD used for the flight tests, has the worst CTI of all the flight devices. Devices with better performance initially would most probably have much better final performance. Two squeegee modes were tested using a 16-row squeegee and an 8-row squeegee. As expected, the 16-row squeegee is more effective than the 8-row squeegee, presumably because it collects more charge. Observation details of the test are listed in the following table along with those of a normal CTI run used for comparison. The high energy amplitude reject rate (cts/frame) is also listed which is an indication of the particle background level during the observations. The comparison dataset may have a slightly lower background but the effect should be negligible.

An image of the overclock-corrected bias frame for the 16-row squeegee run is shown below. The top 16-rows are the location of the squeegee charge which is never clocked out of the image area and is therefore never readout. Any charge in these rows in the bias frame accumulated in the framestore during readout. Below the squeegee is a region of 'spillover' of charge from the squeegee. This is similar to the trailing charge seen behind X-ray events from trap re-emission. A cut along column 400 is also shown which illustrates this spillover.

In the flight experiment a time dependence of the squeegee effectiveness was noticed. This effect may be a residual charge which decays away over 10s-100s of minutes. It is believe this residual can be removed by resetting the CCD voltages in a particular order at the start of a science run. The result of the residual charge is that the bias level is high at the beginning of a run and much lower at the end. The bias frame itself is much higher than any of the science exposures, and has excessive squeegee spillover. The decaying bias level is best seen in the figure below.

Both the squeegee tests have negative corner pixel centroids indicating that the original bias frames taken before the start of the science exposures had a higher level of residual charge. The centroid continues to drop as the residual charge in the image drains away.

Shown above is the centroid of the corner pixel distributions of the bottom and top rows of the CCD for the 16-row squeegee test as a function of time. The size of the corner pixel decay is much larger for the top rows (near the squeegee) than the bottom rows (near the framestore). This differential causes an increase in measured CTI also shown above.

As the bias level, measured by the corner pixel values, decreases, the event island is less likely to produce a split event, thus G0 events become more plentiful and G2, G6 and G7 events, whose numbers have been increased due to CTI-induced charge trailing (also known as grade morphing), become less numerous. This becomes important to remember when trying to measure the FWHM of spectral lines using the summed pulseheight of G02346 events. While the total number of G02346 events does not change during the squeegee test, the grade distribution of these events does. Near the end of the squeegee test, the CTI-induced trailing charge in the top pixel is no longer above the split threshold for many events and is not included in the summed pulseheight. An increase in FWHM for G02346 events is seen at the end of the squeegee test which is due to the short term trap re-emission not being included in the event pulseheight.

Shown above is a comparison of the pulseheight in the top pixel of the event island for standard readout and 16-row squeegee mode for the first 400 frames and the last 300 frames. The dotted vertical line indicates the normal split threshold of 13 ADU. The top plot is from a standard CTI science run with normal readout mode. The Gaussian centered around 0 ADU is the normal CCD noise. The tail extending to higher pulseheights is from both CTI-induced charge-trailing and true vertical split charge. The top pixels in the squeegee run have an additional negative pulseheight tail as a result of the large amount of residual charge contaminating the original bias frame and the strong vertical gradient of that charge. That charge in the top of the CCD decays much more rapidly than the bottom, thus increasing the spread of the negative bias-subtracted top pixel values. If a bias frame from a normal CTI measurement is used instead of the squeegee bias, the negative tail disappears and the top pixel distribution looks much like that for the standard readout.

Results of analysis of the squeegee mode data are shown below. Because of the changing bias level and structure just discussed, it is unclear how best to characterize the change in CCD performance due to the squeegee mode. In the following analysis, the squeegee bias is used, however this may not be the best representation of the true performance of the squeegee mode. CTI is calculated from the center pixels values alone, while FWHM is calculated using the summed pulseheight from G02346 events and a locally derived gain so the FWHM is an overestimate. While both squeegee modes improve the total charge lost to each event (better CTI), only the 16-row squeegee mode has a significant affect on the FWHM @ Mn-K. The FWHM @ Al-K results are not as good, perhaps because of the changing bias level discussed in the previous section. These results can also be seen in the following plots.

Plots comparing 16-row and 8-row squeegee to standard readout

• Mn-K peak versus row number (G02346 events) Postscript PDF
• Mn-K FWHM versus row number (G02346 events) Postscript PDF
• Al-K peak versus row number (G02346 events) Postscript PDF
• Al-K FWHM versus row number (G02346 events) Postscript PDF
• Center pixel pulseheight versus row (all events)
  • 16-row squeegee Postscript PDF
  • 8-row squeegee Postscript PDF
  • standard readout Postscript PDF