Wavelength limits of throughput and resolution about desimodel

[weaverba137]

@dkirkby, I'm assigning this to you so that you can assign it to someone who can actually resolve this.

[sbailey]

The difference between psf-quicksim and thru*.fits comes from the difference between what wavelengths all fibers see vs. the min/max that any fiber might see. psf-quicksim uses the range for fiber 100 which is a "typical" case in between. Quoting from a June 29, 2014 "Big desimodel update" email:

David K noted that the psf-quicksim.fits file doesn't have as much wavelength coverage as thru-?.fits files, and David S has asked about where to find out the min/max wavelength coverage since it isn't in desi.yaml anymore, so some explanation:

Due to the spectrograph optics, different fibers see a different wavelength ranges. Pat Jelinsky provides PSF spots and xy locations sampled at various wavelengths and slit positions (DESI-0334), but these do not correspond to specific fibers or to the exact edges of the CCDs. In desimodel/bin/spots2psf.py I fit these with Legendre polynomials and interpolate to specific fiber positions and calculate the wavelength coverage and put these in header keywords of the data/specpsf/psf-?.fits files:

WAVEMIN : smallest wavelength seen by any fiber
WAVEMAX : largest wavelength seen by any fiber
WMIN_ALL : smallest wavelength seen by all fibers
WMAX_ALL : largest wavelength seen by all fibers

These are rounded down(min) and up(max) to integer Angstroms. HDU 1 of the data/specpsf/psf-?.fits files contain the Legendre coefficients of the y vs. wavelength fits that can be used to determine the min/max wavelength range of an individual fiber.

bin/combine_throughputs.py combines system throughput (DESI-0347) with spectrograph throughput (DESI-0334) and outputs data/throughput/thru-?.fits, which covers WAVEMIN to WAVEMAX for each spectrograph arm. i.e. it covers the throughput range for all fibers, but any individual fiber will see a smaller wavelength range than that.

bin/psf2quicksim.py extracts PSF parameters needed for quicksim (FWHM, Neff) using fiber 100 as a "not the best but not the worst" case. It uses the ranges WMIN_ALL to WMAX_ALL, i.e. all fibers would see this wavelength range. quicksim should treat this as the wavelength range seen by the spectrograph arms, even though individual fibers will see slightly more. I think that is a better choice than tying quicksim to a specific fiber wavelength range.

i.e. the throughput files cover what might happen for some fiber, while the psf-quicksim file covers a typical range that should be used for the quick simulations, even though some fibers will see higher/lower wavelengths.

This has come up multiple times so better documentation could be in order, but I don't think it requires any code or data file changes.

[weaverba137]

OK, so where exactly should the documentation go? In the data model for desimodel perhaps? Or in desimodel's own documentation?

[dkirkby]

There is some explanation now in the DESI config file used by specsim:

# Specify the wavelength grid to use for simulation. For comparison, the
# table below summarizes the ranges of non-zero throughput in each camera,
# wavelengths that can disperse into CCDs (with a 5-sigma cut), and the
# extents of each CCD's active pixels covered by all fibers (as defined by
# cameras.*.ccd.table below). Note that individual fibers will generally have
# slightly more coverage, so we are simulating the worst-case coverage of
# all fibers.
#
#   Range     b(min)   b(max)   r(min)   r(max)   z(min)   z(max)
#            Angstrom Angstrom Angstrom Angstrom Angstrom Angstrom
# ---------- -------- -------- -------- -------- -------- --------
# Throughput 3533.000 5998.000 5564.000 7805.000 7360.000 9913.000
# Simulation 3550.000 ................................... 9850.000
#   Response 3565.400 5952.600 5621.700 7744.300 7431.000 9838.000
#        CCD 3569.000 5949.000 5625.000 7741.000 7435.000 9834.000
wavelength_grid:
    unit: Angstrom
    min: 3550.0
    max: 9850.0
    step: 0.1

[weaverba137]

If this documentation is sufficient, I propose we close this.

[dkirkby]

Ok with me.

[weaverba137]

Closing.