Fixed incorrect shift in PSF when calculating atmospheric transmission.

[dobos]

Dear @kiyoyabe , I've found this bug while debugging an issue with SNR in my updated version of the ETC. This causes a shift in the atmospheric transmission curve with respect to the sky continuum which isn't a big issue but I saw numerical differences everywhere until it got fixed. The hard-coded value of 7.5 is only correct when SP_PSF_LEN=16 but it's set to 64 by default. By the way I don't think using such a wide PSF is worth it, at least not at normal resolution.

[kiyoyabe]

Thank you for finding the issue.
Can you do the same treatment in gsGetSNR_Continuum?

spt_ExposureTimeCalculator/src/gsetc.c

Lines 1371 to 1373 in 3007110

	/* Atmospheric transmission */
	gsSpectroDist(spectro,obs,i_arm,lambda,7.5,0,SP_PSF_LEN,FR);
	num = den = 0.;

Yes, we should remove the hard-coded things like:

spt_ExposureTimeCalculator/src/gsetc.c

Lines 804 to 808 in 3007110

	iref = (long)floor(pos-7.5);
	if (iref<0) iref=0;
	if (iref>Npix-16) iref=Npix-16;
	gsSpectroDist(spectro,obs,i_arm,lambda,pos-iref,0,16,FR);
	for(j=0;j<16;j++)

Can you also fix this part?

64 pix is a bit large but perhaps necessary if we want to consider the wing carefully.

[kiyoyabe]

@dobos Any comments on this? ↑

[dobos]

you for finding the issue.
Can you do the same treatment in gsGetSNR_Continuum?

I'll take another look once I get there. But this will be a breaking change in the sense that the output of the ETC will be numerically different. I think this tweak was added to speed up calculations with a narrow PSF because the value of 64 is overkill. I checked the differences and they're in the 1e-3 - 1e-4 range.