standard positioner analysis outputs

[joesilber]

These are my suggestions for broad direction of getting "the analysis stuff currently in PECS" into desimeter.

General requirements

• Analysis runs completely independent from the instrument.
• Past analyses are instantly reproducible, at any time.
• All code in Github, with total visibility to the collaboration.
• Any data cached to disk must be in simple, standard formats (CSV for tables, JSON for structures, etc).

Data selection

Need some script(s) for grabbing move data off the onlinedb, and then caching it. See PR #71

Suggested new analysis tools

1. summarize_pos_errors

   • analyzes positioning performance data for a broad collection of positioners
   • capable of merging night data and multiple xytests etc --- not just one "xytest" (!)
   • generates a table like DESI-5500, summarizing targeting errors over a broad swath of positioners
     • counts of target errors at quantiles: 1σ, 2σ, 3σ, 4σ, 5σ, 100%
     • see script there, xytestdata_summarize_v2.py
   • generates error histogram for each submove
     • Num targets, max err, rms err, mean err, median err printed on it.
   • option to do the summaries in various groupings:
     • by petal
     • by canbus
     • everything-that-was-tested-that-day

2. plot_pos_targmeas

   • generates a plot that overlays targets and measured points for a given positioner
   • does this overlap with anything already in desimeter?

3. plot_pos_arcmeas

   • select arc calibration data for a given positioner
   • do so by parsing notes in posmovedb
   • include ability to find / merge results from multiple calibration runs (!)
   • generate a plot of a given set of arc calibration points for a given positioner
   • does this overlap with anything already in desimeter?

4. fit_posparams_arc

   • same underlying selection function as plot_pos_arcmeas
   • do circle fits + mechanical assumptions to calculate calibration params