Principal:
Deputy:
Analyst:
AORKEYS:
Last Updated:
The following steps were followed in analyzing the data: 1) V2.31 of mips_sloper was used to reduce the data. We used flags -l -j CamG for the stim-calibrated dark data and -l -q -j CamG for the read noise data. The 160um read noise data also included a -y flag, to output the data in 2x28 rather than 3x20 format. The stim-calibrated darks were also run through mips_sloper with the flags -d -C CamG. 2) We used get_rdnoise_new.pro on the 70um data to calculate read noise and dark current. We set n_dce_rej=10. We also used the bad70 flag that Karl G added to the get_rdnoise program to exclude the B-side of the array and readout 4,4 on the A-side. The get_rdnoise routine does not handle the 160um data properly yet, so improvements will be needed. For now, the measured read noise values are about 6000 e at 15mV bias and a factor of 10 worse at 30mV bias. These numbers are tremendously higher than what was seen on the ground. 3) The stim-calibrated dark data were run through the IDL program mips246_stimcal.pro with flag 'skip=10' to determine the mean stim brightness and the stim repeatability. This program determines the array mean for each of the stim flashes, ignoring the first 'skip' flashes, and then determines the mean of these mean brightnesses. It also takes the brightness of each odd-numbered flash, divides that number by the average of the adjacent even-numbered flashes, and then calculates the standard deviation of these normalized values. This is not the best way of assessing stim repeatability, but it does provide a crude first estimate. It is a quantitative representation of the array-average brightness plots presented by John S below. 4) Karl Misselt performed a stim repeatability assessment, using the same methodology used in his earlier report on 70um stim repeatability. Details and plots will follow in a later version of this writeup. 5) The stim-calibrated dark data were run through the IDL program mips246_darkcal.pro. This program determines the array mean for each of the dark DCEs, then calculates the mean and standard deviation of these array means. It is possible to set the number of DCEs to ignore after each stim flash, to minimize the effect of stim latents. In practice, since the cosmic ray filtering is not perfect, I used the skip parameter as best I could to clean out the effect of bad DCEs. A median filter would have been more effective and will be implemented later, when there is more time to reanalyze the data. I will also try using the -q flag in mips_sloper, for 3sigma rather than 5sigma cosmic ray filtering. For now, the dark levels from the read noise data are probably a more accurate assessment of dark levels, even though they are not stim-calibrated. I record the time since the last thermal anneal in the tables below so that an eyeball correction can be made for the effect of cosmic rays. This post-anneal time is also useful in looking at Karl's stim repeatability results. 6) These results can be combined with insights into the effects of bias voltage on the dynamic range of the detector, the linearity of the data ramps, and the behavior of the B-side of the 70um array in deciding how best to set the detector bias.
AORID 70Bias 160Bias 70ReadNoise 70DarkCurrent 160ReadNoise 160DarkCurrent
(mV) (mV) (e) (e/s) (e) (e/s)
7152640 30 15 385 +/- 156 583 +/- 208 TBD TBD
7155200 30 20 381 +/- 132 617 +/- 233 TBD TBD
7152896 35 20
7155456 35 25 644 +/- 217 472 +/- 163 TBD TBD
7153152 40 25 531 +/- 202 718 +/- 270 TBD TBD
7153408 40 30 538 +/- 183 697 +/- 257 TBD TBD
7155712 40 35 721 +/- 266 654 +/- 228 TBD TBD
7152384 45 35 783 +/- 293 615 +/- 210 TBD TBD
7155968 45 15 829 +/- 340 773 +/- 261 TBD TBD
7154688 50 15 871 +/- 351 773 +/- 268 TBD TBD
For each bias voltage, we measured 278 DCEs with a stim flash on the 2nd
DCE and every 6 DCEs after that, for a total of 47 stim flashes. The IDL
routines mips246_stimcal and mips246_darkcal and Karl Misselt's stim
repeatability routine provided the following results:
70um Analysis:
AORID 70Bias 160Bias Mean Stim Stim K.M.StimRep Dark Time Since
(mV) (mV) (DN/s) StdDev (sigma) (DN/s) Anneal (h)
70 160 70 160
7152640 30 15 9106 +/- 296 .0138 8.8 10.5 105 +/- 18 0 2+
7155200 30 20 8872 +/- 78 .00863 7.0 86 +/- 20 1 0
7152896 35 20 9.7 0 1
7155456 35 25 10477 +/- 178 .0103 7.4 103 +/- 37 1 0
7153152 40 25 11404 +/- 388 .00811 9.6 78 +/-158 0 1
7153408 40 30 11664 +/- 161 .0122 --- 109 +/- 31 0 0
7155712 40 35 13102 +/- 305 .00767 6.7 140 +/- 66 1 0
7152384 45 35 13022 +/- 302 .0116 9.3 122 +/- 91 0 1
7155968 45 15 15366 +/- 408 .00770 6.3 10.4 154 +/- 21 1 0
7154688 50 15 14727 +/- 684 .0130 9.0 10.7 140 +/- 81 0+ 1+
A downlink occurred prior to 7154688 and 7152640, which is why the time since
the last anneal is larger than the value listed in the table. I do not know
the duration of the downlink.
John Stansberry has produced plots of brightness vs DCE number for the stim-calibrated dark data. This includes array averages and is very useful for assessing stim repeatability and responsivity drifts. For the 160um data, the repeatability is good for 15mV and 20mV bias and becomes terrible starting at 25mV bias (in great part due to heavy saturation at this and higher biases). For the 70um array, there is no obvious trends in repeatability with bias voltage. These plots are presented below:
Karl Misselt produced plots of repeatability vs bias for both of the arrays. For the 160um array, the repeatability is always poor, and it degrades rapidly with increasing bias. For the 70um array, there is a dependence on settling time but no dependence on bias voltage. The one discrepent point at 50mV in the plot below was actually taken with zero settling time and so should have been plotted in red.
160_RepeatabilityVsBias.jpg
Conclusions
It is pretty clear that we want to reduce the 160um bias to either
15 or 20mV, with 15mV being slightly favored. The main criteria
in this assessment are the ramp shapes and the stim repeatability.
From the John S plots, the repeatability looks to be slightly better
at 15mV, and we think that we can make it better by increasing the
stim brightness by about 30%. We will have Karl M perform his stim
repeatability analysis on the 15mV and 20mV stim data to make this
more definitive. For the 70um array, the responsivity and noise
both grow roughly linearly with bias. This is a bit surprising
since ground tests showed that responsivity grows with the square
of the detector bias, as expected by detector physics. The stim
repeatability is noticeably better an hour after an anneal than
it is immediately following an anneal, but there is no obvious
dependence on bias voltage. The plan for campaign J is to update
the mips-irs patch block to use a 160um bias voltage of 15mV and
a 70um bias voltage of 55mV. We will explore 70um bias voltages
from 55-70 mV in that campaign, in addition to a number of
noise-mitigating strategies for the B-side of the 70um array.
Output and Deliverable Products
None
Actions Following Analysis
None
