To evaluate the flux calibration accuracy, I compared flux-calibrated spectra of CALSPEC stars with the original CALSPEC spectra. The plot here makes the comparison. In the top panel, the blue is the PFS spectrum of FEIGE22 and the black spectrum is from CALSPEC. The overall shape looks similar, although there is a systematic offset. The bottom panel is the flux ratio as a function of wavelength between the two spectra in the top panel. The blue spectrum is the ratio. The light blue is the same ratio but scaled to unity around 600-800nm so that the relative shape (=color) is easier to see. Plots for other stars are attached to this ticket. WD2149 suffers from bad extraction (??) in the n-arm. P330E seems to show a +/-20% tilt between the blue and red edges. Overall, we are often over-correcting the flux (i.e., PFS spectra are brighter than CALSPEC and the plot included in this comment is the only exception that I have found).

To better understand the over-correction, I compared flux-calibrated spectra (i.e., pfsSingle) of FLUXSTD in a visit and compared with their broad-band photometry form PS1.  The difference between the synthetic i-band and PS1 i-band is shown as a function of position on the focal plane is plotted here. The figure comes with an explanatory caption. The figure also shows that the PFS spectra are too bright at the field center.

The spatial pattern seems to accompany a color change: see this figure. We might be seeing the positional offsets of the fiber centers...?

To better understand the too-bright-flux problem, Mineo-kun suggested to make a similar plot but using pfsFluxReference, which is the best-fit model spectra. This is the difference between the synthetic i-band magnitude and PS1 i-band magnitude. The normalization seems fine. So, something happened in fluxCal.

Additional tickets for the specific issue have been filed, so can we close this ticket? Masayuki Tanaka

Yes, this ticket summarized the EDR2 flux calibration and can be closed now.