search

desihub / desitest

Testing coordination for DESI code
BSD 3-Clause "New" or "Revised" License
0 stars 1 forks source link

wavelength ranges in the test spectra #38

Open segasai opened 3 years ago

segasai commented 3 years ago

Hi,

I don't know if it's an issue or not with the simulated spectra, but I've just looked at a random spectra-**fits file and saw that the wavelength range (for the B camera for example) doesn't match whats in real data

[3569.3, 3570.1, 3570.9, ..., 5946.9, 5947.7, 5948.5 vs [3600. , 3600.8, 3601.6, ..., 5798.4, 5799.2, 5800. ] Is this intended ?

(I was plannign to use one of those spectra-*fits file as part of my test-suite, but if the wavelength range is different, that won't work for me)

Thanks, S

moustakas commented 3 years ago

Hmm, @sbailey should jump in, of course.

But it does look like the noiseless simulated targets have an odd wavelength range, 70 Angstroms bluer than the default (which is 3600-10000 A) and with an unusual pixel size:

python -c "import fitsio ; ww = fitsio.read('/global/cfs/cdirs/desi/datachallenge/reference_runs/20.12/targets/52/5263/dark/truth-dark-64-5263.fits', 'WAVE') ; print(ww.min(), ww.max(), ww[1]-ww[0])"
3523.0 9923.0 0.19995117

In any case, these become DESI spectra via desisim.simexp.simulate_spectra (via desisim/scripts/newexp_mock), which I think (under the hood) uses the nominal camera-specific wavelengths in the desimodel throughput files.

Meanwhile and separately, the 3600-5800 A used in desispec during production runs are hard-coded somewhere that I can't find right now (hence https://github.com/desihub/desispec/issues/1009 is a related issue), but the simulation infrastructure never sees those ranges.

This definitely needs some tender-loving-care!