interpolate stellar templates

[moustakas]

When passed an input metadata table with "real" TEFF, LOGG, and FEH values for stars (or just TEFF and LOGG in the case of white dwarfs), the stellar templates generated by templates.STAR and templates.WD should -- but do not -- interpolate the templates. This issue can lead to the kind of gridding seen in desihub/desitarget#251 and elsewhere, and may also be affecting the flux-calibration portion of the pipeline.

This issue has been worked on and discussed in many places, e.g.:

• https://desi.lbl.gov/trac/wiki/MilkyWayWG/Interpolation
• https://github.com/desihub/desisim/blob/master/doc/nb/mws-interpolation.ipynb
• https://github.com/segasai/mynotebooks/blob/master/desi_interpolation/Interpolation.ipynb

Recently, @callendeprieto wrote the gettemplate module to call his long-used and well-tested ferre code to do the interpolation. However, using ferre as-is would introduce a dependence on Fortran.

Let's track developments and progress here.

[julienguy]

I think we should not interpolate the standard star templates in the simulations.
Interpolations introduce errors. We would loose the ability to test the accuracy of the fit of standard stars at a level better than the interpolation errors.

[sbailey]

Whether or not we interpolate, the true spectrum is kept in the simspec file and can/should be used for comparisons rather than going back to the original template files. Although flux calibration might use a different interpolation scheme (or no interpolation), and both might not be completely physically realistic, allowing them to be different is more realistic than keeping both the input and the template fitting on exactly the same grid.

[callendeprieto]

How about using a set of finer grids that I recently finished? That will avoid interpolation completely. But we should do this now, since John is reloading the templates as we speak ... John?

…

On 01/25/2018 04:40 PM, julienguy wrote: I think we should not interpolate the standard star templates in the simulations. Interpolations introduce errors. We would loose the ability to test the accuracy of the fit of standard stars at a level better than the interpolation errors. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#315 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AXwt-YOTNtvDvxBfrRFLT4o3KKKLwcF4ks5tOK6SgaJpZM4RpeZm>.

--------------------------------------------------------------------------------------------- ADVERTENCIA: Sobre la privacidad y cumplimiento de la Ley de Protección de Datos, acceda a http://www.iac.es/disclaimer.php WARNING: For more information on privacy and fulfilment of the Law concerning the Protection of Data, consult http://www.iac.es/disclaimer.php?lang=en

[moustakas]

I think it's worth leaving this issue open for now, but I'll note that in #434 we have implemented a much finer grid of stellar templates, which should help mitigate the need for interpolation (which we don't do).

[julienguy]

A note that we have to verify the flux calibration is still working fine with the new stellar templates v3.