This is a flux comparison of a CALSPEC star between pfsSingle and a model spectrum. The pfsSingle is a spectrum flux-calibrated with FLUXSTD in a visit. This star is V=16mag star. The model is LSF-convolved and scaled by x0.25 at 600-610nm.
fluxratio_CALSPEC_C26202_v082906_mod_v2.pdf

The flux ratio shows an offset at the blue side and a negative offset at the red side. The offsets are about 0.2mag.

This is a flux comparison for a DA WD standard. There is an offset of ~0.2-0.3 mag at arm r.

This is a summary of three visits of standards in the Nov run.

Are these all in the same visit? Is one of these the FLUXSTD star, if not, how does it look? If they are all from the same visit, you seem to be saying that the different fibres behave very differently. Is that correct? Are they all observed by the same spectrograph?

These are separate visits. The blue and orange are a CALSPEC standard star in different visits. The green is a DA WD standard star in a other visit. These all are not FLUXSTD stars but known standard stars in science fibers. 

In the last technical telecon, we discussed the offset at arm r relative to arm b, which is seen in the DA WD spectrum above (green). I looked into fluxCal and individual flux calibration vectors of FLUXSTDs in the visit and found no issues relevant to this offset. This issue could stem from the quartz normalization.

These figures are pfsSingle, pfsMerged, and fluxCal around the problematic wavelength range. From top to bottom, I plotted

• (top) the pfsSingle spectrum and a model spectrum of this DA WD star which is normalized at 600-610nm. We can see again the offset around 640nm.
• (2nd) the normalized pfsMerged spectrum and linear fits. One linear fit is performed at 590-630nm, while the other is at 660-700nm. I try to emphasize different continua between b and r, although I admit that the choices of the fitting wavelengths are arbitrary.
• (3rd) the fluxCal used for this visit.
• (4th) the normalization, pfsMerged.norm.
• (5th) the pfsMerged flux without normalization.

In fluxCal, there is no offset around 640nm. On the other hand, the offset already exists before flux calibration, and it coincides with a jump from arm b to r in the normalization.

I updated the plot of flux comparisons. This is an edr-20230331 version which Yabe-san has processed. The results are not significantly different from the previous one.

I don't think this is terribly important, but Takuji Yamashita I wanted to recreate some of the plots above because it seemed like your `pfsMerged.flux / pfsMerged.norm` subplot (2 of 5 here) was off for the DA WD, I'm guessing because of the different calibration set?

I've attached my (mostly) similar plot that shows more sensible values for that subplot. (I did slightly different values for the linear fit just for fun)

Your pfsMereged.flux/pfsMereged.norm spectrum looks very different from the pfsSingle of the DA WD, especially for noise levels. And we can find a strong absorption line at H-alpha and two absorptions at 590 nm (probably they are NaD/HeII). So, I guess your subplot may be from another star than the DA WD, for example, a FLUXSTD?

Looks like my fiberId and fiberIndex might have been swapped due to me not being careful with the notebook. I'll double-check things again. Thanks for the response!

I have checked the flux-calibrated spectra of the CALSPEC and DA WD stars. We close this ticket.