Weird break in schedule and no biasstep before cmb scan.

[ykyohei]

Sometimes I find weird break in satp3 schedule and also there is no biasstep before cmb scan.

Following is the example generated by auto scheduler
https://site.simonsobs.org/satp3/nextline/db/runs/3151

[mmccrackan]

It seems like a CMB block got cut here and it left a space. It tried to add a relock there, which combined with detector setup and an already short CMB scan, caused that block to take too long. It looks like this was an edited auto schedule based on the logs (started at 18:23). It is probably best to just always regenerate it on the fly so it will always get the state at the start. Still, it shouldn't drop a potential relock but I don't think we have a way to prevent that right now.

[mmccrackan]

On the bias step, that occurs on a set cadence so if there is a scan that is shorter than that cadence, it won't run one on the next one. We should probably just set the bias step cadence to the minimum CMB scan duration which is 10 minutes.

[ykyohei]

Thank you, I found another example and this looks like same issue.
https://site.simonsobs.org/satp3/nextline/db/runs/3139

I tested to reduce biasstep cadence down to 10 min.
With this cadence we will have redundant biasstep after detsetup. because its duration is 20 mins.
20 mins + delta might be the better value. What do you think?

[mmccrackan]

Ah yeah I guess it will run a bias_step again after every det_setup if it is too low. We could avoid that by splitting out the last two steps of det_setup maybe. But yeah lowering to a little more than 20 minutes should be fine and not all short scans will have det setup anyway.

[mmccrackan]

I'll dig more into the gaps in the schedule and try and come up with a solution. You can also see that the scan duration is negative in line 178 in your example which is this problem: #231.

[mmccrackan]

Ah I guess we might be able to solve the skipped scans by passing in the CMB block to the relock step and checking if its duration is long enough to not get cut if relocking occurs. I'll test that out.

[mmccrackan]

Some updates have been done related to removing the gaps, but this addresses a different issue, so will not close with #289 just yet.