Comment by Vincent Le Brun on 2023-04-06 10:54:17:
OK if the 2nd pass amplitude is found to be 0, but if it is negative, and we do not switch, an error will be raised ?

Comment by Didier Vibert on 2023-04-06 12:27:02:
Vincent Le Brun wrote in #note-2:
> OK if the 2nd pass amplitude is found to be 0, but if it is negative, and we do not switch, an error will be raised ?

The option 2 is to not raise any error which abort the object (QSO/Galaxy) fittting, but issue a warning and let the code continue with a null continuum (when it is negative, above the negative threshold it is considered null). Keep in mind that in the first pass, around the same redshifts (on a coarser grid) it was not considered negative, so we are in the case were it is at the limit of the threshold, It will probably not change anything at the end. And to introduce a "fromspectrum" continuum switch that is not global needs some refactoring.

Comment by Didier Vibert on 2023-04-06 12:39:46:
Another comment: in first pass, the switch to "fromspectrum" arises if at all z in the full z range the continuum is fitted with a negative amplitude (below the threshold). What happens currently in 2nd pass is to switch if it is fitted negative on all the z-range of one 2nd pass window. Since the 2nd pass window contains some of the z of the first pass grid, the only thing I see that can change the behavior is the use of fft in first-pass only (it is deactivated in 2nd pass because the z-range is too small to have a cpu tiime benefit).

Comment by Vincent Le Brun on 2023-04-06 13:29:32:
OK let's go like this

Comment by Gaelle Daste on 2024-01-29 12:42:40:
J'ai pas trouvé dans les ITs (même full) de cas où on avait ce switch. Pour voir et valider le avant/après si quelqu'un a une idée je suis preneuse

Comment by Gaelle Daste on 2024-02-06 15:00:09:
#8580