[RM-6104] analyze the cpu time gain vs success rate using loglambda, taking into account the expected LSF of lambda

Created on 2020-11-06 09:28:38 by Didier Vibert. % Done: 10

F. Madec has computed LSF profiles for various fibers wrt the wavelength and the position of the spectra on the detector-field.

First all the LSF seems very close to Gaussian shape. But since the instrument focal field is spherical and the detector is flat there is a radial de-focus leading to an increase of the LSF FWHM when going from the central wavelength to the extremities of the spectra, plus a global increase for the spectra at the border of the field compared to the spectra in the middle of the image.

Taking into account this aspect for log-lambda resampling, means selecting a log-lambda step equal, for each spectra, to the better resolving power (smaller delta_lambda/lambda). This can increase by a huge factor the number of sampling points compared to a log-lambda resampling associated to a constant resolving-power.

We thus need to re-evaluate the time performance using a log-lambda sampling adapted to the expected resolving-power.