NOTCam Array Electronics upgrade Dec 2007


The purpose of upgrading the array electronics was to get rid of the "dark current" problem, i.e. the fact that the measured dark counts are much higher than expected, and the strange behaviour where the dark images have less counts the longer(!) the integration time is ...

The electronic upgrade consisted in exchanging the printed circuit board (PCB) to which the array is connected with an upgraded PCB with recommended voltage settings (from Rockwell) for the Hawaii arrays. The main difference between the old and the new electronic configuration is that while the old setup had no "biasgate" voltages set, the new setup allows the user to adjust these around a default recommended value of 3.5V.

On the 3rd of December 2007 the new PCB was tested out with the engineering grade array cold. The QC for the ramp-sampling mode was run for two different values of the biasgate voltage (default=3.5V and max=3.8V), but it was not possible to do further tests because of a serious problem with the vacuum of the cryostat. (Also, it was not possible to have the QC analysis updating the web because of the web server swap at these dates. We can re-run this and get the plots on the web.)

No major difference was seen between the old and the new PCB from the QC results. The gain was around 3.2 e-/adu and the ron was at 13 e- for the ramp-sampling mode. The 1% linearity limit was at 17000-24000 adu. Also, no obvious undesirable effects were found with the new PCB. But apparently the upgrade did not cure the dark current problem: Still the the measured counts on darks were too high and the behaviour vs. exposure time as before, i.e. many counts on short darks and less counts on longer darks.

In the following we describe the differences between the old and the new PCB as seen on the engineering grade array.

We also compare the behaviour of the new science array with the old arrays.

Darks


Engineering array "dark 0" images from left to right: 1) new PCB (biasgate voltage set to default=3.5V), 2) new PCB (biasgate voltage set to max=3.8V), and 3) the old PCB where we could not adjust this value.


Vertical cuts through the above images at x=274-276 averaging over 3 columns.

Notes on darks:

Reset images


From left to right: 1) eng. array with new PCB, 2) eng. array with old PCB, 3) the new science array with the new PCB.

Notes on reset images:

The reset level


The reset level and its variation with time. From left to right: 1) 13/12-2007, 2) 21/5-2008, and 3) 11/11-2008.

Notes on the reset level:

Skyflats - comparison between the 3 arrays


From left to right: 1) masterflat for the eng. grade array (SWIR1), 2) master flat for the first science array (SWIR2), 3) master flat for the new science array (SWIR3). All are masterflats obtained in the same manner from differential skyflats for WF-cam and the Ks filter.

Notes on sky flats:

Skyflats - SWIR3


The average of all columns is plotted for SWIR3 WF-cam Ks band differential twilight flats. Evolution in time from left to right: 1) 13/12-2007 2) 17/4-2008 and, 3) 11/11-2008. All are masterflats obtained in the same manner from differential skyflats for WF-cam and the Ks filter.

Notes on SWIR3 sky flats:

Domeflats


From left to right: 1) Differential (lamp ON-OFF) domeflat in J obtained from 5 raw frames 4 5 images, 2) J domeflat created using only the "lamp ON" images, and 3) the ratio between the two previous ones.

Notes on dome flats:

Conclusions

We have never tried the new array with the old PCB, but we know, from having used the engineering grade array with both PCBs, that the new PCB increases the "dc-gradient", the gradient across lines in each quadrant with highest amplitude in the first few readout lines, a known effect with the Hawaii arrays referred to as "the reset anomaly".

A consequence of this is that differential flats don't work well any longer for NOTCam. Due to the "reset anomaly" the "dc-gradient" is different for different intensity levels, and therefore the "bright minus faint" subtraction leaves traces (most notably a horizontal bar of lower intensity across the centre of the array).

It should be noted that the reset anomaly for the NICS/TNG instrument is quite similar and makes it impossible for them to obtain accurate differential flats.

Multiple reads gets rid of the "dc-gradient", but multiple reads require of the order of 15-20 seconds exptime, not feasible on the twilight sky. Domeflats are considered, but from earlier experience we know that they are not optimal. This is still under study. Another option is to try to get rid of the remnants of the "dc-gradient" in the differential flats by collapsing a median column and subtract it.

Results of "dc-gradient" suppression in the mkflat.cl script

The option mentioned above, to get rid of the remnants of the "dc-gradient" in the difference images by collapsing a median column and subtract it, has been implemented as an option in the script mkflat.cl in the notcam.cl quick-look reduction package. Every difference image in the stack is corrected individually.


Differential twilight flat for WF camera and Ks band with "dc-gradient suppression" (left) and without (right).

Comments to Amanda


Back to top Last modified: 11-Mar-2009