We are interested in temperature stability requirements, especially since we plan to use multiple ADRs.

Several years ago, Shaul's student worked out the relationship between responsivity variations and focal plane heat bath variations for their thesis. Pages from Sagiv_thesis2011.pdf

The algebra looks sound, but we should double check the numbers, specifically the loop gain and alpha. The student computed loop gain from a large value of alpha inferred from the data in Fig 7.2. I don't understand the experiment behind that, but the conclusion of 2800 seems very high, and inferred Loop Gain is thus high as well. The isolation of the detectors from focal plane drifts is very sensitive to this number, and smaller values of these parameters may not ensure as high an isolation.

Here are some published figures on Loop gain from other experiments:

See Fig 17 in http://iopscience.iop.org/article/10.1088/0004-637X/812/2/176/meta The histogram shows data from one SPIDER tile, 94 working detectors. I went back and looked at the numbers, and it's LoopGain=22 \pm2.9, although we did not report with this precision in the paper. We extracted these from studies of time-constants of responses to a chopped bias at different DC voltage biases.

Table 5.2 on pg 107 of Erik Shirokoff's thesis (spiderwebs for the original SPT camera): http://kicp.uchicago.edu/~shiro/erik/docs/shirokoff_2011_thesis_final.pdf He reports ~10 very casually, but I don't think this was ever published and reviewed. Similar design as EBEX.

See pg 17, paragraph following Eqn 4.2 in http://iopscience.iop.org/article/10.1088/1475-7516/2017/05/008/meta reports that polabear's loopgain is 2. These were fabricated by Kam Arnold concurrently with Ben Westbrook's EBEX fab.