Excess of z=1.6 sources on a few tiles from 2024-12-27 (eg, tile 10720, exp 270555 & 270556)

[dstndstn]

Anand noted strong excesses at z1 and z1.6 in the tertiary44 (XMM DESI-2) tiles in petal 7, and then I noticed a slight pattern in some of the main DARK tiles.

I marked tile 10720 "unsure" for this.

Slightly visible (but I marked "good") are:
tile 10749 (exp 270550)
tile 1350 (exp 270552)
tile 7713 (exp 270557)

Some of these have excess light in the high-wavelength end of the z camera in petal 7.

[sybenzvi]

Here is the screenshot of the sky fiber z distribution from 20241227 posted by Anand:

Checking 20241229, there seems to be a redshift pileup in petal 7 but it's subtle and not nearly as pronounced as two days ago:

I don't see the same pileup in the seven tertiary44 tiles (83532-8) observed on 20241229.

[djschlegel]

Noticed these small pile-ups again at z=0.957, z=0.976, z=1.628 on the dark tiles of 2025-01-05. The majority of these also have a warning set (ZWARN > 0). Of the 35 tiles on that night, I count only 12+25+8=45 spurious redshifts on petal 7 at those redshifts with ZWARN=0. Redrock is typically latching onto residuals right at the 9800 Ang edge of the wavelength coverage and calling it either [OIII}4959 or [OII]3727.

[sybenzvi]

Following our Zoom conversation, here are the QA plots for tile 10720. The pileup at z=1.6 in P7 is subtle.

[araichoor]

I guess there s nothing to do here, but for what is worth, a precision:

for the night first mentioned by Dustin (20241227), looking at the petal=7 maindark fibers which are at z=0.975 or z=1.628:

• it s 10% of the 13x500=6500 fibers (including skies);
• and 3% if you restrict to ZWARN=0.

so it s higher fractions that what David reports for 20250105 (25+8=33 fibers out of 35 tiles, ie 33/(35*500)=0.2%...

code snippet:

night, faflavor = 20241227, "maindark"

t = Table.read("/global/cfs/cdirs/desi/spectro/redux/daily/tiles-daily.csv")
t = t[(t["LASTNIGHT"]==night) & (t["FAFLAVOR"]==faflavor)]

ds = []
for tileid in t["TILEID"]:
    tdir = "/global/cfs/cdirs/desi/spectro/redux/daily/tiles/cumulative/{}/{}".format(tileid, night)
    fns = sorted(glob(os.path.join(tdir, "redrock*-7-*fits")))
    d = vstack([Table.read(fn, "REDSHIFTS") for fn in fns])
    ds.append(d)

d = vstack(ds)
for z in [0.975, 1.628]:
    sel = np.abs(d["Z"] - z) < 0.01
    sel2 = (sel) & (d["ZWARN"] == 0)
    print("{}\t{}\t{}".format(z, sel.sum(), sel2.sum()))

[geordie666]

I think that if we can't distinguish this pattern on 10720 from, e.g., 10749 or 1350 (that Dustin also reported above, but passed as "good"), then we should probably mark 10720 as good, too, and move on.