How to run an xytest using desimeter

[joesilber]

If we want to test using desimeter transforms (alternate to platemaker) on the instrument, I think the easiest thing might be to run an "xy test" with a few suggested modifications, described below. Nothing is easy, exactly, but I guess this is least pain. The system is pretty intricate --- I'm sure we'd want Duan or Kevin to do this. Anyway I took a pass through the code and think these are the changes needed. All code links below are at current revisions as of writing.

pecs.py

• c.f. https://desi.lbl.gov/trac/browser/code/focalplane/plate_control/trunk/pecs/pecs.py?rev=129541
• Line 62: insert a new case like if fvc == 'desimeter':
  • Might need Klaus here to give us a platemaker-less FVC proxy that just returns images. c.f. https://desi.lbl.gov/trac/browser/code/online/DOSlib/trunk/python/DOSlib/proxies.py?rev=129583
  • Or perhaps David R could give us some code to more directly set up the FVC.
• Line 182: make a conditional like if fvc == 'desimeter':
  • Put in code here that triggers an exposure, sends the FITS file to desimeter.fvc2fp() function, gets back QS values, and then shoehorns them into format that the "meapos" dataframe is supposed to contain. (Who here wishes we were running in a properly typed language... raise your hands! I don't have the heart to trace down how meapos is structured at the moment.)

xytest.py

• c.f. https://desi.lbl.gov/trac/browser/code/focalplane/plate_control/trunk/pecs/xytest.py?rev=129521
• Line 173 - 200: make this compatible with multi-petal input targets file (see format description below)

Input targets file

• The current format is kind of awkward.
  • It looks like: https://desi.lbl.gov/trac/browser/code/focalplane/fp_settings/test_settings/xytest_cmx_063064-0.csv?rev=128750
  • Has 1000 columns
  • Requires a unique file per petal.
  • Petal ID is encoded in file name string.
  • Device location ids are encoded mixed with coordinate system.
• Better format would have headers:
  • TARGET_NUMBER
  • PETAL_LOC
  • DEVICE_LOC
  • uv_type
  • u
  • v
  • For example with 10 petals, ~500 positioners / petal, and 10 test targets, the csv file would have 50,000 rows.
    • You could break this up into smaller files by petal, or smaller files by target number. Seems pointless to me, but you could do that. I prefer one file == one test. If you do break it up into smaller files, you need to change how these are stated in the config file.
  • For description of valid uv_type for move requests, please see posschedule.py.
    • request_target() function, comments in lines 42-53.
    • https://desi.lbl.gov/trac/browser/code/focalplane/plate_control/trunk/petal/posschedule.py?rev=129559
• Easiest place to put the targets csv is the same folder as the xytest config file, i.e. code/focalplane

Test settings config file

• Based off of template: https://desi.lbl.gov/trac/browser/code/focalplane/fp_settings/test_settings/xytest_template.cfg?rev=129521
• We should add a parameter:
  • fvc2fp_mode
  • valid options would be: 'platemaker' or 'desimeter'
  • or perhaps could add further suboptions if you want to try different flavors of desimeter, like maybe 'desimeter_7paramfit' vs 'desimeter_42paramfit'. Whatever.
• Fill in all petals to be tested, like ['01'], ['02], etc
  • mode = 'all' --> will test all positioners on that petal with CTRL_ENABLED
  • alternately to test only a specific list of positioners...
    • mode = 'can'
    • posids = [posid_str1, posid_str2, posid_str3, ...]
• input_targs_file = path to your csv
  • this may require some adjustment (see above)
• target_type
  • I think this should be eliminated and just use coordinate system specified in the uv_type of input file. It's redundant and just one more setting to get wrong at time of test.
• num_corr_max --> probably set this to 3 for purposes of test
• targ_max_radius --> set to None for clarity (it will be ignored, since using csv of targets)
• targ_min_radius --> set to None for clarity (it will be ignored, since using csv of targets)
• pause_interval --> I'm not sure what our latest info on cooling is, but 590 sec is the nominal design limit. We can probably go a bit faster.