NOTCam new clockboards Jan 2010

During an investigation to identify a problem with the Hawaii array it was discovered that the clock voltage levels were only going to ~3.5V where they are specified to go to 5V. It was proposed to redesign the array controller clockboard so to deliver the correct voltage levels with the hope that this could solve the so-called "dark current problem". The "dark current problem" is the strange behaviour seen where long dark integrations have less counts than short dark integrations.

The new clockboards were first tested on December 13th 2009, but we had contact with only two quadrants. This was related to missing contacts on the clockboards which was easy to correct. The next test was made on January 13th 2010. The new clockboards were found to be fully functional. As shown below several improvements were found in the images. The previously strong shift register glow is practically removed, as is a pattern of horizontal stripes. The readnoise for the reset-read-read mode is lower and more stable over time. But the behaviour of counts in dark exposures as a function of integration time remains essentially unchanged.

On the 17th of January a full Detector Quality Control was run with the new clockboards and the results are in all aspects in agreement with results obtained with the old clockboards, except for a notably lower and more stable readnoise for the reset-read-read mode.

In the following we show the differences between the old and the new clockboards.

Dark images

Long ramp-sampling darks dframe 40 8 shown from left to right: 1) old clockboards, 2) new clockboards, and 3) the difference image.

Short reset-read-read darks dark 0 shown from left to right: 1) old clockboards, 2) new clockboards, and 3) the difference image.

Notes on darks:

As shown with the 320s long darks, the new clockboards got rid of the shift register glow around quadrant corners.
The jumps in level seen at y=51, y=252, and y=447 in each of the four quadrants are the same in the old and new images. Looking at the difference image the gradient across each quadrant does not seem to be altered with the new clockboards. The jump positions did not change by changing the array PCB either. They do change with arrays, though.

Dark levels

The count level is plotted versus the integration time on the x-axis. The four quadrants are shown with different plot symbols. For each quadrant the median level of all pixels, excluding the first 100 lines to avoid reset-anomaly problems, is plotted. The upper window shows the ramp-sampling mode and the lower shows the reset-read-read mode. This behaviour is as it has always been for NOTCam darks.

Reset images

From left to right: 1) OLD clockboards reset image, 2) NEW clockboards reset image, 3) difference image.

Notes on reset images:

The faint broad bands, tilted about 20 degrees and marginally seen only in the new clockboard reset image, are the interference features that are believed to be due to pick-up noise.
We notice the presence of narrow horizontal stripes of alternatingly higher and lower intensity in the old clockboard reset image. (These stripes actually became stronger when the array PCB was upgraded in Dec-2007, and the peak-to-peak amplitude went from 7% to 11%). With the new clockboards these stripes have disappeared.
NOTE. Those stripes normally subtracted out well, and there was no presence of such stripes on the reset-subtracted images in most cases. For low-background cases, however, such as darks, there were two exceptions: 1) where there was strong interference pattern (due to pick-up noise?), and 2) along the first few columns of each quadrant. Here the stripes didn't cancel well.

Conclusions

The new clockboards work fine and improve the image quality. Improvements are:

Practically full removal of the earlier strongly present shift-register glow.
Removal of horizontal stripes.
Lower and more stable readnoise, especially seen for the reset-read-read mode.

Practically all other aspects of the detector remain as before. There is a possible marginally higher amount of hot pixels in longer exposures (preliminary analysis shows an increase from 0.8% to 1.1% on 42 s darks), but this needs to be confirmed. Cold pixels remain as before.

Comments to Amanda