QL not catching varying CCD amp bias levels

[sbailey]

Testing http://desi-1.kpno.noao.edu:9090 using code from desispec #702 (now tag 0.25.0) and qlf #253, mostly looking at Observing History window and drill-downs.

From the QLF trend plots, it appears that the CCD amp offset (bias) and noise values are jumping around, which appears to be a bug in the input simulations. See desihub/desisim#443. However, quicklook should have caught this as an error -- if the real data were jumping around this much we'd want to know.

[Srheft]

@sbailey great to know. Thank you very much!

• The focus was to cast a wide net for config parameters such that QLF would show green petals for the screenshots for the past review.

• We also assumed simulation data are to be assumed 'all ideal' data and were asking for faulty exposures a while ago. Good to know that the current exposures are not all ideal.

• I completely agree that QL should raise a warning/alarm for steep changes in CCD statics[has been challenging to figure what should be warning and what is alarm without having a faulty exposure]. Please mention any exposure that you try[you did it for issue #704] just so that we have an example of a WARNING[yellow] or ALARM[red]-worthy values for the shifts in noise and bias.

• One other problem that this is a good place to bring up again: science config file currently has separate references for DARK, GRAY and BRIGHT programs but not for b,r,z arms. To be able to be sensitive to jumps in CCD stats, shall the references and ranges also be arm-specific?

[rkehoe]

@sbailey Can you indicate if response from @srheft has resolved this issue? It sounds like buggy sims may be part of what you were observing. Definitely specifications on metrics need to be tightened and tuned. If that is the purpose of this ticket, then we should leave it open. Otherwise, looking to identify issues that are still out there. Thanks

[Srheft]

@rkehoe : thanks for the followup. This one needs multiple runs on the current set of exposures to catch the outliers to the noise and bias trend. Like I sad above, I'll have to set up a full night to catch the CCD stat patterns.