Tile 25744 has petals 3,4,9 rejected for large STARRMS about desisurveyops

[schlafly]

Linking the QA plot for completeness
https://data.desi.lbl.gov/desi/spectro/redux/daily/tiles/cumulative/25744/20220127/tile-qa-25744-thru20220127.png

[julienguy]

For petal 4 that shows large rms, the calibrated spectra do not present discontinuities at the boundary of cameras for all 3 exposures of tile 25744 and in particular for exposure 00119660. Could it be due to large positioning errors + an unusual seeing ? No. position errors are small for the std stars of Petal 4.

[schlafly]

Transparency was reported to be good on that night, with decent seeing bouncing around between 1 and 1.5". First exposure of the night, but skies were pretty dark. It is the case that the observers reported that the focus was "a bit off", but I seem to think that this usually means tens of microns.

18:47 [01:47] Exp. 119660- Good - First tile of the night. A bit off focus and seeing a bit high (1.43). RMS of first correction move is around 4 (SO)
Tile 25744, Exptime: 844.75, Airmass: 1.20, Sequence: DESI, Flavor: science, Program: BRIGHT

[julienguy]

The guiding PSF has some peculiar triangular shape. Only 44% of light in fibers on average. If we are out of focus, could there be radial changes in the PSF?
https://data.desi.lbl.gov/desi/spectro/data/20220121/00119660/etc-00119660.png

[julienguy]

I suspect we were out of focus.

from astropy.table import Table
for spec in range(10) :
    t=Table.read(f"/global/cfs/cdirs/desi/spectro/redux/daily/exposures/20220121/00119660/cframe-r{spec}-00119660.fits","FIBERMAP")
    s=Table.read(f"/global/cfs/cdirs/desi/spectro/redux/daily/exposures/20220121/00119660/cframe-r{spec}-00119660.fits","SCORES")
    i=t["FIBERFLUX_R"]>10
    x=t["FIBER_X"][i]
    y=t["FIBER_Y"][i]
    ratio=s["MEDIAN_CALIB_COUNT_R"][i]/t["FIBERFLUX_R"][i]
    ratio /= np.median(ratio)
    plt.scatter(x,y,c=ratio,vmin=0.8,vmax=1.2)
plt.colorbar()

[schlafly]

Thanks. It looks like we need to mark this as bad in the offline pipeline. @akremin, can you mark 119660 as bad so that it gets EFFTIME = 0 and reprocess this tile 25744, and then we need to clear its QA status in tiles-specstatus?

I hadn't used the SCORES extension before---that's obviously useful! Aiming to be more self-sufficient in the future, can you confirm that:

• RAW_COUNT -> frame file values
• FFLAT_COUNT -> frame file values after applying fiberflat
• SKYSUB_COUNT -> sframe file values
• CALIB_COUNT -> cframe file values

I suspect that since the GFAs are on the edge, in terms of raw SNR, this tile is good on the edges and very good in the center, but I agree we don't want to deal with out-of-focus images and should just mark as bad.

[julienguy]

yes

[akremin]

The following have been performed:

• 119660 marked as bad
• removed all files associated with tile 25744 on 20220121
• removed all redshift files associated with tile 25744 on 20220127
• reran redshifts for 25744 on 20220127

The observation on 20220120 was flagged by David as low efftime. There were no other observations on 20220121, so after flagging this exposure there was nothing to rerun. The tile was also observed on 20220126, but it was a bad exposure that wasn't used. So the 20220127 is the only exposure remaining for this tile.

I think this is below the efftime threshold, but could be released for more observations.

[schlafly]

Resetting to QA = none. Confirmed that this tile now has EFFTIME_SPEC < GOALTIME in $DESI_SPECTRO_REDUX/daily/tiles.csv, so this should be scheduled for future observation. Closing.