I have measured the flux in PFI and SuNNs exposures and compared the value.

I have taken detrended images and, using the detectormap, extracted 7x60 region around each available fiber (7 pixels in x direction, 60 pixels in y direction) at 800 nm.  I have then taken the median value of all extracted fibers to be a representative measurement of flux in that particular visit.

I have used the data taken by SuNNs in March and April 2021. For PFI I have used exposures with visit_id 68878, 68879 and 68881.

In comparison_of_flux_PFI_SuNNs.png I have plotted the mean measured flux as a function of time during the night, relative to midnight Hawaii time. This is because I can cleary see the dependence of flux as a function of time during night (comparison_of_flux_PFI_SuNNs_via_visit_id.png). I have also overploted very naive fit to the SuNNs data - but please note many outliers and spread, which makes this extremely unreliable measurement.

We see that PFI exposures are of comparable brightness to a darker set of SuNNs exposures. They might be dimmer than expected given the time of the night (but also note the largest difference in seasons when PFI and SuNNs data have been taken; and different observations configurations. I did very simple of analysis of moon phase during SuNNs observations, but I see no dependence).

The initial investigation was promoted by my observation that PFI exposures  reach up to 4k flux in the brightest pixels, while rhl seemed to remember that SuNNs observation tended to saturate the detector - which would indicate large discrepancy in the flux levels. On the other hand, the vast majority of SuNNs exposures that I have looked at have comparable levels of flux, and I do not see any saturations. rhl is it possible that your observations that SuNNs observations are often saturated is from earlier data (and I should look at them), or perhaps from some sort of dedicated observations? At the moment, I am a bit confused by this discrepancy between your statement and the data that I am looking at.

There is no code to review, I just wanted to make sure you are informed that I think I am seeing some weird arcLines.flux measurements in SuNNS data. That is probably not surprising given that Paul only recently implement this code for such a difficult problem, but just confirming, we might need to look into that more. Lets keep that in the back of our heads.

I have redone the analysis in a `proper` fashion, using the pipeline, with the addition of more early data and SuNNS data taken in October. My main results stay unchanged. 

I have analyzed the flux in a line at 882.952 nm - well-behaved singular and isolated line. The Figure shows the amount of flux in a line in a second for different observations with SuNNS (dots) - I am showing the flux from SuNNS with a dashed black line. As we can see, PFI observation is consistent with SuNNS observations.  . I am also adding two examples, in one of which SuNNS observation is brighter, and one where SuNNS observation is dimmer than PFI.
I believe that RHL earlier suggestion that sometimes SuNNS observations are saturated is somewhat true, as there are observations where the continuum is super strong, almost to the point of saturating the instrument.
I have also noted some discrepancy between the arcLine fluxes being reported for SuNNS a and PFI data being different by a factor of 2, even though the flux in the line is the same in both exposures. As such, I have done this analysis by using the pfsArm[flux] measurements in a small region around the line, rather than the values reported in arcLines object. This might need to be investigated further.