Implement apply decision for quantization

[teutoburg]

Closes #306

[codecov]

Codecov Report

Attention: Patch coverage is 76.56250% with 15 lines in your changes are missing coverage. Please review.

Project coverage is 74.85%. Comparing base (6057f73) to head (b943451).
Report is 3 commits behind head on dev_master.

Files	Patch %	Lines
scopesim/effects/electronic.py	76.56%	15 Missing ⚠️

Additional details and impacted files

@@              Coverage Diff               @@
##           dev_master     #396      +/-   ##
==============================================
- Coverage       74.92%   74.85%   -0.08%     
==============================================
  Files              56       56              
  Lines            7777     7818      +41     
==============================================
+ Hits             5827     5852      +25     
- Misses           1950     1966      +16     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[teutoburg]

The Quantization effect needs to be actually included in the IRDB (mostly METIS for now). I did try that locally and it seems to work with the following examples passed to the properties of the corresponding UserCommands of a simple METIS observation:

• "!OBS.exptime": None or "!OBS.exptime": 5 results in floats because DIT & NDIT are auto-determined by the AutoExposure
• "!OBS.ndit": 5, "!OBS.dit": 5, "!OBS.exptime": None results in floats because NDIT > 1
• "!OBS.ndit": 1, "!OBS.dit": 5, "!OBS.exptime": None results in integers because NDIT == 1

In this case, not supplying any DIT and NDIT will also result in floats, because the default DIT set in the IRDB is increased based on mindit by the AutoExposure, thus resulting in "automatically determined DIT". So only when DIT and NDIT are both supplied with NDIT == 1, and pass mindit, the quantization takes place. I guess that's what we want?

[teutoburg]

One slight complication about all this: exptime, dit and ndit all have default values (see IRDB/METIS/default.yaml). There's not really a (non-hacky) way for the code to know if any (non-None) values found for these are coming from the defaults or from user input. Removing the values from the defaults might break things and will then also require the user to always specify either exptime or dit & ndit. Which is probably good, but extends the necessary overhead and reduces the ability to "just quickly run it".

Possibly solution, at least for METIS (or anything else using AutoExposure:
Put the defaults for exptime, dit and ndit to either 0 or None (which is better conceptually??) and include the following check in AutoExposure: If both exptime and at least one of dit & ndit are nonset, issue a warning but use a best-guess to still run the simulation regardless. This could be MINDIT if it's present, or otherwise just 1 s.

Thoughts @hugobuddel ?

[teutoburg]

Possibly solution, at least for METIS (or anything else using AutoExposure:
Put the defaults for exptime, dit and ndit to either 0 or None (which is better conceptually??) and include the following check in AutoExposure: If both exptime and at least one of dit & ndit are nonset, issue a warning but use a best-guess to still run the simulation regardless. This could be MINDIT if it's present, or otherwise just 1 s.

After discussion with @astronomyk we decided to go ahead with this "solution" for now.

[hugobuddel]

Sounds good! I didn't take those defaults into account in my thought process.

Maybe we could set either exptime or dit/ndit to 0/None by default but not both? Then at least users get something interesting by default. All three are 1 now in METIS/default.yaml. I suggest keeping exptime default at 1, and set dit/ndit at 0/None, because then people who do not change any defaults would get the most useful result.

And all of this should be kinda part of a hypothetical ETC-simulator (or actual ETC?), which can then be coupled to ScopeSim somehow.

[hugobuddel]

It seems this MR is the cause of #438; investigating.