Search
Text Size:
Smaller Text Normal Text Larger Text

Facebook

Follow us on Twitter

First NOTCam Science Grade Array (SWIR2)

Content of this page:


Detector overview

The Science Grade Array was installed in NOTCam on the 20th of October 2005 and the first cold tests made 25/10 (see the Science Array Commissioning report ). All NOTCam data taken before this date has been using the Engineering Grade Array , for which we refer to that web page. Also, the Engineering Grade Array was re-installed in NOTCam on the 5th of May 2006 upon malfunctioning of the Science Array.

NOTCam is offered with two possible readout modes: the standard reset-read-read mode and a ramp-sampling mode (multiple non-destructive reads during the integration). Read more about this in NOTCam User's Guide.

Array Hawaii HgCdTe 1024 x 1024 x 18.5 micron
SWIR2 P/N: 9-173L A2 #61
Bad pixels ~ 2 % (mainly edges)
Gain 2.2 e-/ADU
Dark current TBD e-/s/pix
Non-linearity < 1 % up to ~ 30 000 ADUs
Saturation starts at 54 000 ADUs
Read out noise 14-15 e-
Read out time 3.6 s
Cross talk not seen
Memory effect < 1 %


Fig 1: Schematic drawing of the detector with the four quadrants and their numbering marked. The arrows show the direction of the fast readout and the corners in which it starts. All four quadrants are read out simultaneously. The lower left corner of quadrant #1 is the location of pixel (x=1,y=1). Note that this is valid for the image as displayed on BIAS DS9, while the stored images are flipped in X since Jan-2006.




Cosmetics - bad pixels

Fig 2: Two dark images obtained with the reset-read-read mode (command dark t). Integration time (t) is 0 seconds (left) and 100 seconds (right). Note the amplifier glow and the higher number of hot pixels on the 100s long dark integration.

Dark-0s Dark-100s

The hot rows seen in the darks of the science array are differently placed than for the engineering grade array. But they are stable and subtract out well. Since they are fixed at a given y-value and equal for each quadrant they are probably intrinsic to the array.

There is a higher brightness in the first few rows in each quadrant where the count level can be more than 3 times the average level in the remaining part of the quadrant. This seems to subtract out well, although the it may take time to stabilize sufficiently that the subtraction is perfect.

In dark images of longer integration times there is clear evidence of amplifier glow along the edges. This effect also apparently subtracts out well.

There are two dead columns [1,*] and [513,*] along the whole detector, contributing with 0.2 % zero pixels. This is inherent to the controller which has problems with the first column of each quadrant. (Note that since Jan-06 the stored data are flipped in X, and therefore the dead columns are now [1024,*] and [512,*].)

Pixels are called bad when they deviate by more than 8 sigma from the mean level. Among the bad pixels we distinguish between hot and cold. The cold pixels include also the zero pixels which show no response at all. The hot pixels may be strongly non-linear. In dark images the amount of hot pixels is 1.4 % (on 42s darks), while in well exposed images the number of hot pixels is < 0.2 %. (Note that this high percentage for the darks is NOT due to the wrap-around effect of very low value pixels in the darks. That effect amounts to only 0.1 % in the cases that have been tested. It is more likely that the hot pixels drown in the higher noise of the exposed dome flats.) The majority of the hot pixels are found close to the edges.

The number of hot pixels increases with exposure time. The table below is taken from the Spectroscopic mode commisioning report which has more information also on clusters of hot pixels.

Percentage of hot pixels (i.e. more than 8 sigma above background) in the central 250x250 pixels of the top-left quadrant (Q3).
Exptime [s] 0 3 9 27 81 243729
Hot pixels [%] 0.1 0.2 0.3 0.6 1.0 1.8 2.5

There is a small bad pixel group (mostly hot and non-linear) of about 8 x 9 pixels large, centred at [784,268], or in images taken after Jan-2006 at [240,268]. The majority of the individual bad pixels (hot and cold) are found along edges and corners. There is a region inside the array at [524:674,800:900] which also has an elevated number of individual bad pixels. There are also a few individual bad pixels spread over the whole array.

The number of dead pixels was believed to increase with every thermal cycle of the array. This is because of the different thermal expansion properties of the layers of an infrared array, which may cause detachment of the bump bonds upon repeated thermal cycles. However, for the engineering grade array which has undergone a number of thermal cycles during the more than 4 years it has been inside NOTCam, we did not see any increase in the number of zero-level pixels. The apparent coming and going of zero-level pixels was found to be due to a drift in the array causing the reset level to vary. This effect is described under the engineering grade array .

The NOTCam detector quality control is monitoring any change in the number of bad pixels:

Reset-read-read mode
Ramp-sampling mode


Dark level

The behaviour of the dark level with exp time is not well understood.


Readout noise

The readout noise is calculated in a representative area within each quadrant.
The readout noise in [e-].

Date Readout mode Quad 0 Quad 1 Quad 2 Quad 3
26-Oct-2005 r-r-r 13 14 14 15
26-Oct-2005 r-s 15 15 15 15


reset-read-read mode
ramp-sampling mode


Please, check the NOTCam User's Guide for a description of the two different readout modes available with NOTCam.


Gain

The gain in [e-/ADU].

Date Readout mode Quad 0 Quad 1 Quad 2 Quad 3
26-Oct-2005 r-r-r 1.9 2.1 2.0 2.2
26-Oct-2005 r-s 2.3 2.3 2.2 2.3


The gain is calculated in a representative area in each quadrant.

reset-read-read mode
ramp-sampling mode



Non-linearity

Non-linearity is an inherent feature of infrared arrays which distinguishes them clearly from the linear CCDs. While the saturation of the detector starts at 54000 ADUs the array is found to be linear to 1% accuracy up to about 32000 ADU on the average.

For each readout mode you can check the non-linear behaviour for each of the four quadrants from the monitoring data:

reset-read-read mode
ramp-sampling mode



Detector flat field

Fig 6: Processed flat field obtained from 10 differential dome flats taken with the WF camera through the Ks band. The differential method (pair-wise subtraction of "lamp on" minus "lamp off") is used to eliminate the thermal contribution from the master flat. No bad pixel correction was attempted, instead the final master flat was median smothed by a 3 pixel box, which almost eliminates the bad columns.

Flat-Ks

The detector flat field looks relatively flat and has few disturbing features. The figure above shows the master flat obtained from 10 dome lamp ON images minus 10 dome lamp OFF images for the WF camera and the Ks filter. The standard deviation in small boxes of 20 x 20 pixels is about 2%. The deviation over the whole field is ± 5%. These numbers are the same for the HR camera Ks flat. A worst case diagonal cut through the flat field above is shown here (eps file). The above detector flat field can be compared to the data sheet with the QE image that came with the array (flip in y-direction).


Memory effect (charge persistency)

For non-saturated pixels there is memory (or charge persistency) only in the first of the subsequent exposures. The memory effect is negative and the level is 1 % or less.

For saturated pixels the memory is more persistent. It is negative in the first subsequent exposure, but positive thereafter. The level is 0.2 % in the 2nd exposure and 0.03 % in the 6th exposure. If you can not avoid saturation, it is recommended to clean the array with a couple of dark 0 commands between each science exposure.

The behaviour of the memory effect is illustrated by a series of images shown below.

  • The upper left is an argon lamp spectrum with two strongly saturated lines (>> 54000 ADU) and a number of lines from < 1000 ADUs to 20000 ADUs.
  • The upper middle shows the first reset read after the arc image. The two saturated lines appear as low level counts in the reset image. While the reset level is at 6000 ADUs, the level in the negative memory is about 3000. Thus, it seems the gain has been transiently lowered by 50% at the location of the saturation. For the non-saturated lines we see no trace in the reset read.
  • The upper right image shows the first (reset subtracted) dark image after the arc. This image has negative memory of many of the arc lines, both the saturated ones and the less bright ones - even lines with only about 4000 ADUs in the original arc image turn out to show some persistency in the first dark thereafter. Because this is a dark image - i.e. it has few counts all over (about 60 ADUs) - the negative traces are of even fewer counts. When the pixel values become negative, they wrap around to > 60000 ADUs (data is saved as unsigned integers), which makes the image look very ugly.
  • The lower left image is the next reset read. All traces of the arc image have now disappeared.
  • The lower middle image is the second (reset subtracted) dark. We still have a memory of the two saturated lines. From now on the memory is positive. The level is 100 ADUs which for a saturated line of >> 54000 ADUs means less than 0.2 %.
  • The lower right image is the sixth (reset subtracted) dark. The positive memory still persists, but at a level of 0.03% only.




    Comments to
    Anlaug Amanda Djupvik

Back to top Last modified: 31-Oct-2007