Due to an accidental loss of vacuum of NOTCam the vessel was recently opened. Before closing, the detector was checked and found to be working. After closing and starting to pump the detector was checked again, this time one quadrant only gave zeros. After extensive investigations it was found that one of the interface boards that was connected to the quadrant was broken, but replacing it with a similar board from one of the other quadrants did not solve the problem and it was concluded that there must be a second failure closer to or on the quadrant.
Spare interface boards were supplied by Copenhagen University, but replacing the boards there was still no signal from the quadrant. It was then decided to remove the detector array from its mounting board and it was found that the output pin of the quadrant on the mounting board was at zero rather than 5V which indicated that a transistor was damaged. The mounting board was replaced with a spare after which we got signal from all four quadrants again. We believe the transistor of the mounting board was damaged due to the failure of the interface board, and the only thing not clear at the moment is what caused the original failure of the interface board.
In the cause of trying to solve this problem, it was found that the documentation on the interface boards (both here in Copenhagen where they were originally manufactured) were incorrect and part of the design had to be guessed. In fact, after cooling the instrument down to normal operating temperature it was found that part of the detector could not be read properly because a resistor that was installed was wrong by a factor 40. Replacing this resistor it now appears that all works properly.
As reported previously, the science array (SWIR3) in combination with the electronic upgrade gave some problems with the flat-fielding procedure previously used. There is now a more persistent ``reset anomaly'', and this has unwanted effect in differential flats that the dc-gradient does not subtract out well. See report at
for details. As a preliminary solution we considered correcting the effect during the reductions. Now the script "mkflat.cl" in the NOTCam quick-look reduction package has the option of correcting for the dc-gradient. See the above report for an example. This method seems to work well.
The stronger ``reset anomaly'' does not only affect the differential flats. As is well known, it affects all images where there is a sudden change in background levels. This happens when you change filter, camera, or exposure time etc. The typical solution is to take a test exposure to let the detector ``stabilize'' on the brightness level, alternatively, to skip the first image when reducing the data. If there is variable background due to moonlit clouds, however, it will be very hard to achieve nice images.
Another effect discovered over the last months is sudden jumps in the reset levels. It was known that there have been changes earlier. With the engineering grade array in 2004-2005 there was a substantial drift in this level, and we had to apply higher dc-offset voltages on each quadrant. Thus, changes in the level has been seen to happen before. The new thing is the sudden jumps to very low levels, down to 1000 adu or less when default is 6000 adu, and then recovering. The images with low reset levels are almost always useless. The effect is very variable. See for examples and a description
A solution of how to modify the clock-boards was designed and boards were send from Copenhagen University. Modification to the clock driver board have been made to correct the voltage levels. A plan will be made for the testing and implementation of these boards.
Thomas Augusteijn 2009-05-14