This page describes instrument related particularities, unwanted features and problems.

2009-10-15 to 2009-11-20: noise and saturation in channel #14 of chip #6

Chip #6 shows in channel #14 sporadic large area saturation events. The effect was never seen in DARK or RESET frames. It becomes apparent in all 5 jitter frames of a standard star calibrations and in science frames. It can shows up also in twilight flat calibrations (e.g. the last three raw frames of the H-band twilight flat stack taken on 2009-11-23). No masking is provided within the calibration recipes of the data reduction pipeline operated at Paranal and at ESO HQ.

Impact:

• no stars can be extracted from that detector region to be compared with 2MASS for reasons to obtain a photometric zeropoint.
• twilight flat raw frames with this problems must not be used to generate a master twilight flat.
• science tiles composed of 6 paw prints match the sky region a second paw print in case the detector error does not show up in the second paw print.
  

The effect is monitored: For each first raw frame of a standard star stack the median flux of part of channel #14 region is measured and ingested into the QC1 database as qc_counts_C14. The same QC measure is applied to science paw prints, in case they are processed by QC in Garching.

2009-10-01 - 2009-11-03: twilight flat and dome flat saturation

The acquisition of photons by near infrared detectors consists of an initial reset, a first non-destructive read out followed by one or more further reads. The signal recorded between the reset and the first read is subtracted by the acquisition system before being written to file. The time between the reset and the first read amounts to 1.0011 sec and is called the minimum possible DIT (= MINDIT). The time between the first and the final read is called DIT. The fits frame contains the counts accumulated within DIT seconds, but the pixels on the detector hosted more counts accumulated within DIT + MINDIT seconds and might saturate the chip. This internal saturation can occur for high fluxes and when a DIT of the order of MINDIT is used. This effect occurred for a number of raw twilight flats in some periods during the commissioning (2008-10-01 - 2009-10-25), science verification (2009-10-15 - 2009-11-03) and dry runs (2009-11-04 - 2010-02-15). Assuming linear response the signal on the detector before reset subtraction can be estimated :

Left figure: fluxes of a stack of twilight flat raw frames with DIT=1.0011 sec for detector #5. The black dots dow the counts[rawframe] values and are well be low the saturation level. The blue dots show the counts[detector] values of which the first frames show saturation effects. The saturation level and the scoring threshold (= 90% of saturation level) are shown. Right figure: The same stack for detector #11 with a higher saturation level.

The upper part of detector #16 shows the following effect when being illuminated:

• In J band and NB118 shows about 75% of the flux when compared to the remaining area of the chip. The level of the flux depression was rather constant within the investigated time interval (2009-11 ... 2010-02).
• The H and K band flats are not affected.
• The Z and Y band images show a remaining flux of 60% to 80% . The flux depression is variable in time and the variability of the flux level is not correlated between Y and Z band.
• The flux depressed pixel area on chip #16 is not homogenous, but the area remains constant in time.
• In addition to the large scale fractal-shape area with flux depression there are a considerable number single pixels (isolated pixels) with unreliable response.
• As a consequence source detection as part of the pipeline science recipe is affected for the upper part of chip #16.

Figure: The QC1 parameter qc_twflat_fluxdep, as obtained from detector #16 master twilight flats. The parameter represents the median flux in the upper part of the detector in the pixel area: 1500,2048: 2048,2048. The abscissa is the MJD-OBS (modified Julian Date).

The behavior of chip #16 is monitored in a dedicated trending plot.

All VIRCAM raw frames show horizontal stripes for all DITs, most apparent in dark calibrations. The stripe pattern is conserved for each group of four detectors in a row, meaning detectors #1, #2, #3 and #4 show the same stripe pattern, detectors #5, #6, #7 and #8 show a distinct pattern but among themselves the same. Detectors #9, #10, #11 and #12 build another group of detectors with the same stripe pattern; detectors #13, #14, #15 and #16 as well. The pattern is changing from readout to readout and it is not reproducible. The frequency of the pattern, the amplitude and the offset is variable. The master dark pipeline products shows the interference of five raw input frame signals.

• Left: median raw dark frame columns and the averaged (black) from the corresponding master dark for DIT=1.0011 and detector #10.
• Right: median raw dark frame columns and the averaged (black) from the corresponding master dark for DIT=120 and detector #10.
• click on the plot to see the entire column

• Left: QC report of a DIT=6sec NDIT=10 master dark frame (here chip #14)
• Right: QC report of a DIT=6sec NDIT=10 master dark frame (here chip #2) with stronger stripes.

Horizontal stripes have the following implications

• Stripes are not corrected in any of the vircam pipeline recipes. Therefore pipeline products, like master dark frames contain the stripe pattern averaged over the N raw input frames (black line in the plots).
• For the instrumental quality control, features in dark frames with amplitudes of the order of or smaller than the typical amplitude of the horizontal stripes cannot be resolved.
• For the dark recipe: When the stripe pattern of the five consecutive raw frames is coherent, the readout noise QC parameter dervied from raw difference frames is rather low, since the pattern mostly cancels out, but in the master dark the pattern is averaged and is therefore enhanced. When the stripe pattern is mostly non-coherent in the five consecutive raw frames, the readout noise QC parameter is rather high, since the raw difference frames enhances the pattern, while in the master dark product the stripes average out and are less strong.
• Analysis of the detector linearity using the ESO detmon recipe has demonstrated, that the interpixel capacitance is contaminated and biased by the stripe pattern.
  For this reason this detector property cannot be monitored.
• For the monthly acquired gain calibrations, which consist of two dark frames and two flat frames, the horizontal stripes bias the measured noise values (photon noise and read out noise). Variations in the stripe pattern amplitude impact the derived gain valaue by up to 20%. Pairs of low stripe amplitude dome flat and dark frames (all with the same DIT) can provide less contaminated gain values, than the raw frames generated by the gain template. For quality control operations, the readnoise QC parameter of the gain recipe is most sensitive to stripes and is used to asses the quality of the gain calibrations.

On 2010-05-12, the detector persistence was tested. First five dome flats were taken with DIT=8 sec to measure the flux in ADU/sec and thereafter 5 dome flats with DIT = 80 sec were taken, which results in nominal flux of about 400000 ADU. Thereafter 12 dark frames with DIT = 300 sec were taken to measure the persistence and its decay. The reference dark level that has been subtracted from the measured persistence was retrieved from a dark taken 8 hours later. In the log-ADU versus log-time diagram the decay looks like a straight line, hence the persistence is fit by the arithmetic function (t in sec):

P(t) = a * t ^{-m}

The best fit parameters a and m are given for the decay sequence in following table, together with the counts for the extrapolated lamp flux f. For the fit of the decay, the first data point acquired after 4 minutes has been excluded from the fit, as it deviates from the trend of all other measurements taken after ~9 min of the saturation.

D	f / ADU	a / [ADU/sec]	m
1	460260	30	-0.77
2	283510	27	-0.74
3	326020	100	-0.87
4	371620	60	-0.90
5	380410	112	-1.02
6	430810	135	-1.02
7	413130	283	-1.10
8	346750	189	-1.05
9	359700	243	-1.11
10	376500	236	-1.15
11	351710	250	-1.13
12	384880	540	-1.18
13	282690	10902	-1.70
14	372010	129	-1.09
15	428480	203	-1.13
16	323570	46	-1.10

Left: measured persistence P(t) of all 16 detectors in ADU/sec. Right: Measured persistence of detector #10 for two stimulus events and best fit and best fit parameters a and m. Best fit results of the first decay (filled squares) are given in the table.

Some of the VIRCAM detectors are subject of radioactive events. Here we report on the number and the character of the radiation-induced charge collections.The following data have been analyzed:

• raw dark frames with DIT=120sec NDIT=1 from 2010-07-24...27 (primary data set)
• raw dark frames with DIT=300sec NDIT=1 from 2010-05-13 (to check against exposure time)
• raw dark frames with DIT=120sec NDIT=1 from 2010-06-29 (to check against detector temperature)
• raw dark frames with DIT=120sec NDIT=1 from 2009-11-03 (to check stability)

The difference between two VIRCAM raw dark frames with long DIT, show charge collections, in a similar manner as is known for HAWK-I chip #78.

Figure caption: Difference of two consecutive VIRCAM dark frames of chip #5 with DIT=300sec

The charge collections itself are variable in the number of counts and in the shape. The following number of events per minute have been measured on dark difference from the primary set (Delta n ~ +- sqrt(n)/2 ) :

• The event rates have been compared with DIT=300 sec dark frames taken on 2010-05-17 in frame of the persistence tests to verify that the number of events increase linearly with exposure time.
• The event rates have also been measured on dark frames acquired on 2010-06-29, when the detector temperature increased by 12 deg. The event rates are not affected by the small temperature increase.
• Finally the event rates have been compared with event rates measured on DIT=120sec dark frames taken during science verification on 2009-11-03 to confirm the expected long-term stability of the event rates.
• For HAWKI #72, the events are homogeneously distributed over the full detector. For VIRCAM, each detector shows a different subregion with higher event rates:
  Chip #4 : upper left quadrant shows more events
  Chip #5 : upper left quadrant shows more events
  Chip #11 : left half shows more events
  Chip #12 : lower left quadrant shows more events

Beside the normal behavior of the events given in the Figure above, there occurs from time to time more peculiar events collected in the following snapshots: left: crowded event cascades, middle: deep impacts, right: trajectory in detector plane:

The dark level values show weak dependence on the ambient temperature.

Left: QC parameter qc_darkmed retrieved from dark frames with DIT=120sec of detector #10 versus the ambient temperature. Data are from 2014-07 to 2015-09.
Right: QC parameter qc_ron12 retrieved from dark frames with DIT=120sec for detector #10 versus ambient temperature. Data are from 2014-07 to 2015-09.
Only chip #10 was uses as the read out noise from chips #1-#8 are often contaminated by the horizontal stripes feature.