search

simonsobs / scheduler

Scheduler for SO
0 stars 1 forks source link

Sun avoidance not respected by pre-observation tuning #37

Closed ktcrowley closed 8 months ago

ktcrowley commented 10 months ago

It appears the scheduler successfully accounts for sun avoidance when policy.apply is run -- that is, the scheduler will not begin an observation until the sun is further away than the sun-avoidance angular distance in rules -- and it will start it ASAP after that point.

However, at least in SAT cases, the support for tuning readout channels seems to be to move to the initial starting az of the observation at a time before the obs start time, with the magnitude of the delay set by arg det_setup in time_costs. This clearly violates the purpose of sun avoidance, and currently for SATp1 the det_setup time cost is 40 mins, which is enough to not only violate the theta=45 sun avoidance criterion, but also the theta=41 shade criterion to protect the PE window on the SAT.

I would propose that the default behavior should be to shift the start time of the scan to a time det_setup later than the ideal (i.e. shift it to the future to allow space to get dets setup), and make the start time of the detector setup the first sun-safe time.

kmharrington commented 9 months ago

If we're relying on sun avoidance through the scheduler. Be very aware of the much larger issue: https://github.com/simonsobs/scheduler/issues/18 It only looks at azimuth / uses 1D angles.

ktcrowley commented 9 months ago

I could be interested in trying to fix that issue at least, let me see if I can make headway there. It seems like proper wait times to include sun avoidance may be second order.

For other folks interested, I think most of us have been checking by hand that 2D sun avoidance is satisfied during daytime scans. It's probably not been perfect, though.