Scripts and code for DESI Survey Operations
desihub / desisurveyops Goto Github PK
View Code? Open in Web Editor NEWScripts and code for DESI Survey Operations
License: BSD 3-Clause "New" or "Revised" License
Scripts and code for DESI Survey Operations
License: BSD 3-Clause "New" or "Revised" License
This was discussed on Slack and by email, but I am unclear as to the status. The beginning of the night on 20220413 featured engineering observations on main survey tiles. We need to mark some of these exposures as bad.
QA for observations 20619/130091 and 25297/130094 look fine to me, though the night log reports that 25297/130094 did not have a turbulence correction applied. I think we should keep these two observations.
QA for 22170/130095 looks terrible and this exposure needs to be marked bad and the tile reprocessed. This duplicates #43 and #40 .
Night log reports observations of a few other tiles that may have already been marked bad and do not make the nightly QA (e.g., 22202, 22211, 22203, 20638). Because there is no QA, I have not inspected these tiles.
On 20220414 the observers started observing backup tiles to avoid problems like the above. But then they moved to BRIGHT tile 23075 and did more test observations. It looks like we got 4 back to back short exposures there.
https://data.desi.lbl.gov/desi/spectro/redux/daily/tiles/cumulative/20375/20220414/tile-qa-20375-thru20220414.png
QA looks fine to me and I don't see major positioning problems like those we saw for other fast sequencer test observations. It's certainly not normal survey observations to have 4 short exposures on a BRIGHT tile, though, so we could consider marking all of these as bad. The rest of 20220414 looks fine to me.
On 20220417, I don't see anything dubious in the night log. There's a weird exposure of 42426 in the nightqa. It's weird because:
Nothing else exceptionally alarming.
Short version: let's mark 22170/130095 bad and reprocess. Let's discuss 23075, 20619, 25297, but probably keep all of these; their QA is fine. Let's check to see if there's a bug in the QA script where data from bad, excluded exposures are included in the positioner offset plots, as suggested by 42426.
20220219 had bright moon and great seeing, and had a number of dark tiles observed. Some remaining moon-induced sky gradients lead to a bit of signal in the spectra that get identified as z > 5 spectra based ~entirely on the continuum as far as I can tell. This doesn't look to affect LyA rates and I am marking these as good, with a few exceptions. But these would benefit from future reprocessing with a more flexible sky model.
https://data.desi.lbl.gov/desi/spectro/redux/daily/nightqa/20220219/tileqa-20220219.pdf
Affected: 1610, 1611, 1650, 2563, 3143, 6055, 7067, 7217, 7893, 7914, 8801, 10006, 11566.
TILEID 8801 looks especially bad and has a weird clump on P1 and has been left unsure.
https://data.desi.lbl.gov/desi/spectro/redux/daily/tiles/cumulative/8801/20220219/tile-qa-8801-thru20220219.png
This TILEID=3298 has been observed on 20220519 with EXPID=134023.
It has been marked as good.
Nevertheless, there is an accumulation of Z~0.25 for 1890 < FIBER < 1912 (approximate range).
https://data.desi.lbl.gov/desi/spectro/redux/daily/tiles/cumulative/3298/20220509/tile-qa-3298-thru20220509.png
There is a drop in the flux at lambda~4660A (you can see it when a different z is picked), which is often interpreted at a negative oii flux:
tmp-3298-thru20220509-1890fiber1911.pdf
Maybe some fibers in that range could be marked as bad.
Original slack message: https://desisurvey.slack.com/archives/C01HNN87Y7J/p1652218526122329
This TILEID=21809 has been observed on 20220512 with EXPID=134448.
It has been marked as "unsure".
It has ZSKYPDF alert for petal=6,7,8.
The n(z) look a bit awkward.
https://data.desi.lbl.gov/desi/spectro/redux/daily/tiles/cumulative/21809/20220512/tile-qa-21809-thru20220512.png
The sframesky has some brightish lines in the z7 camera, but I can t figure out where that does come from (no bright star in the viewer); the neighboring exposures have nothing like tht.
https://data.desi.lbl.gov/desi/spectro/redux/daily/nightqa/20220512/sframesky-20220512.pdf
The sky monitor also has significantly brighter sky there (it's the outlier at 8:00):
Don't know if that s related, but the cryostat pressure for sm8 (=petal=7) has some fluctuations before that exposure:
https://telemetry.desi.lbl.gov/d/xcYVcw3Zk/spectrographs-cryostats?orgId=2&from=1652403600000&to=1652443200000
Original slack message: https://desisurvey.slack.com/archives/C01HNN87Y7J/p1652480247392209
z3 had CTE issues from 20211126 - 20211202 (inclusive). These are minor and could be largely fixed with the per-sector fits.
Tile 23224 observed on 2022-02-12 rejects a patch of hundreds of fibers in a patch of high Galactic extinction for failing the effective time cut. We had planned to replace this EFFTIME cut to be corrected for the median EBV for the tile, rather than the per-object EBV. That would recover these fibers.
I had marked this tile as QA='good'. If one wanted to back this out, is it sufficient to edit the surveyops/ops/tiles-specstatus.ecsv file to have QA='unsure' before it gets archived?
Most of the signal for tile 25744 comes from the exposure 20220121/00119660. Petals 3,4,9 are rejected due to a large STARRMS > 0.20 (dispersion in the standard star model fits) from the file /global/cfs/cdirs/desi/spectro/redux/daily/exposures/20220121/00119660/exposure-qa-00119660.fits
PETAL_LOC STARRMS
0 0.14335816
1 0.14170471
2 0.07650502
3 0.20699991
4 0.25244614
5 0.13656339
6 0.10696009
7 0.16174984
8 0.12091315
9 0.20128961
Could the cause of this be the standard star fitting being pulled by the very poor data in the previous exposure 20220120/00119533 on the same tile?
Exposure 130095 on BRIGHT tile 22211 22170 (night 20220413) did not have fiber correction move and should be rejected. I didn't catch this until seeing the QA plot, combining those data with exposure 130266 on night 20220414.
That first exposure should be marked bad, and the cumulative tile reductions removed. We'll need another exposure on this tile to complete it.
@djschlegel marked tile 22170 as unsure; recording this here. QA looks pretty horrid; both exposures probably need to be marked bad?
https://data.desi.lbl.gov/desi/spectro/redux/daily/tiles/cumulative/22170/20220414/tile-qa-22170-thru20220414.png
4 tiles (TILEIDs 8808, 11015, 11046, 11542) from 20220326 have fibers near the edge of petal 1 with low EFFTIME:
https://data.desi.lbl.gov/desi/spectro/redux/daily/tiles/cumulative/8808/20220326/tile-qa-8808-thru20220326.png
https://data.desi.lbl.gov/desi/spectro/redux/daily/tiles/cumulative/11015/20220326/tile-qa-11015-thru20220326.png
https://data.desi.lbl.gov/desi/spectro/redux/daily/tiles/cumulative/11046/20220326/tile-qa-11046-thru20220326.png
https://data.desi.lbl.gov/desi/spectro/redux/daily/tiles/cumulative/11542/20220326/tile-qa-11542-thru20220326.png
These tiles are marked as unsure in the QA.
Discussions of this issue on slack:
https://desisurvey.slack.com/archives/C01HNN87Y7J/p1648414581954449
The fixed, stationary locations of the stuck positioners were incorrect for the nights of 20220507-20220513. We could repeat the turbulence corrections for these nights using the now known positions. We would need to think about how we want to incorporate that information into the pipeline to propagate it into the final data. The current original source is the coordinates files in data/YEARMMDD/EXPID.
A similar mechanism could be applied to all pre-turbulence correction data.
Only a few fibers per exposure move by more than 30 microns, though there are a few tens with corrections more than 20 microns.
No obvious impact on QA, but the sframes look pretty bad on r3.
https://data.desi.lbl.gov/desi/spectro/redux/daily/tiles/cumulative/24638/20220329/tile-qa-24638-thru20220329.png
https://data.desi.lbl.gov/desi/users/raichoor/main-status/sframesky/sframesky-20220329-00127995.png
Field acquisition failed on tile 2219 exposure 128339 on April 1. This is a high-Galactic-latitude DARK tile. Conditions were good with 1.3 arcsec seeing in clear, dark skies: https://www.legacysurvey.org/viewer?ra=187.319&dec=36.81&layer=ls-dr9&zoom=8&desifoot=187.3190,36.5810
Steve Kent's response: Yes, low star density. Several other fields in the same region succeeded, so this one was just over the edge. There are several tuning parameters in the star finding and matching code, and I think allowing fainter stars is the easiest fix (the current limit is 18th mag). Note that the parameters were tuned in the days when the GFA CCDs were often in their noisy state and PM still succeeds when that happens. Given how rare the failures have been, I'd suggest waiting for a few more to occur before making any change (which is eay to do - Klaus needs to edit a configuration file but no reinstall of code is needed.)
We could change this configuration file (to a faint limit of 18.5 or 19.0) and try re-observing at this exact location again. I suggest we design BACKUP tile 42546 which has exactly the same coordinates and observe during engineering time. (The corresponding BRIGHT tile has slightly different coordinates due to our small offsets to some tile centers to avoid bright stars on disabled fibers, and we've already observed it anyway.)
David Schlegel has documented the cause of the spurious jump here.
desihub/desispec#1619
This causes the tile to lose a half amp. I plan on marking the tile as good anyway, so we'll reobserve objects in that half amp. Tagging @djschlegel and @sbailey in case they want to overrule.
Expid 113678 is a bad exposure on tile 21281 for which the ETC estimates 0.085 s of effective exposure time. This exposure looks to have been rejected by the pipeline (EFFTIME_SPEC = 0), but it doesn't have a record in the tsnr-exposures file and so the NTS believes that this tile has 0.085 s of effective time, when it should have 0 seconds of effective exposure time. Can we get an entry in tsnr-exposures.fits recording the fact that the pipeline has rejected this tile? Thanks!
The bias variation in b8 causes jump in spectra, which results in some features in sky Z vs. FIBER plot and lower-than-normal success rate of LRGs, and it's particularly noticeable for the night of 2022-02-02:
Sky Z vs. FIBER plot: https://data.desi.lbl.gov/desi/spectro/redux/daily/nightqa/20220202/skyzfiber-20220202.png
There are a few otherwise fine tiles that hang off the edge of the footprint and have a weird redshift distribution. I don't understand what is going on with these. I don't know where we're getting off footprint redshifts? I don't know if these are failures of real targets or just weird targets?
There may be many more cases that never got marked unsure.
Affected tiles: 20682, 21347, 23401, 25506
https://data.desi.lbl.gov/desi/spectro/redux/daily/tiles/cumulative/20682/20211211/tile-qa-20682-thru20211211.png
https://data.desi.lbl.gov/desi/spectro/redux/daily/tiles/cumulative/21347/20220109/tile-qa-21347-thru20220109.png
https://data.desi.lbl.gov/desi/spectro/redux/daily/tiles/cumulative/23401/20211218/tile-qa-23401-thru20211218.png
https://data.desi.lbl.gov/desi/spectro/redux/daily/tiles/cumulative/25506/20211211/tile-qa-25506-thru20211211.png
I'm leaving these as unsure; at least they're all on the edge of the footprint so they aren't so problematic from a depth-first perspective.
z6 CTE or amp offset that could likely be fixed with sector code.
https://data.desi.lbl.gov/desi/spectro/redux/daily/tiles/cumulative/10941/20211106/tile-qa-10941-thru20211106.png
https://data.desi.lbl.gov/desi/users/raichoor/main-status/sframesky/sframesky-20211106-00107649.png
Tileid=21984 has been observed with expids=134791-92 on 20220514.
It has several r- and z-band sky alerts (petals 4-5-6-7-8-9); and a suspicious bgs_faint n(z), despite having nominally ~1.5 more effime than required (efftime=265s):
https://data.desi.lbl.gov/desi/spectro/redux/daily/tiles/cumulative/21984/20220514/tile-qa-21984-thru20220514.png
If relevant, on the same night, all backup exposures from expid=134789 onwards were flagged as "unsure" by the observers in the nightlog.
Besides, these expids=134791-92 were preceded/succeded with backup observations, so conditions might have been variable.
One of the new replacement collimators was installed for the night of 2022-04-13 only, on SP1 (SM10). It has a deep absorption at 420 nm, which is bluer than the dip in the original collimators. Only BRIGHT tiles were observed on that one night. The reduced data do show uncorrected flux-calibration residuals around this new dip, although the BGS redshifts appear robust to this. A couple of representative calibrated spectra attached.
Bright tile 26056 observed in one exposure (expid 123247) on 2022-02-21 shows many spurious redshifts at z=0.14-0.15. This is due to a cosmic landing next to (or on?) the overscan of B1 amplifier D. Cosmic is at (2045,3707) in the raw image. The row-by-row bias subtraction results in large negative (and significant) values at ~5600 Ang, which is eventually fit as H-beta in absorption for most fibers in the range 500-749.
This will have to be solved by using both the bias region and the overscan region to filter out the effects of cosmics independent of bias level jumps.
Another exposure at the very beginning of the night that looks like it was out-of-focus causing petal 7 to mostly be dropped:
Tileid 9934 was observed with exposure 136310, 136311 and 136312 on 20220522.
It apperead as a clear outlier on the Lya plot. The redshifts as a function of the fiberid looks clumpy in some locations.
In the nightlog we can read: "SPLIT on DARK tile 9934, exposure 136311. Survey speed dropped from 85% to 40%, presumed to be seeing related."
The single exposure on tile 22230 on 2022-02-11 (expid 122156) rejects half of P4 (fibers 2250-2499) due to a small CTE from a cosmic effecting the overscan region, resulting in a too-large overscan jump (OSTEPB=6.07).
This has previously been seen in r8-D, z8-D, z3-D, z5-A, as described in the desispec issue #1619 (desihub/desispec#1619). However, it had not been identified previously on r4-B.
Below is the raw image centered on the bright cosmic at location (2219,1036).
alerts on the tile qa page and doesn't look good on the sframe; no obvious impact on redshifts
The guider crashed on exposure 137368 on tile 10589 on 2022-05-30. Night log states:
GUIDE5 GFA image size error followed by second failed automatic revere attempt. Loss of guiding mid-exposure.; Action: Stopped exposure and read out data for DARK tile 10589 exposure [137368] once guiding was confirmed as lost.
This results in no guider cube being written for the exposure. We've been marking such exposures as bad, since we lose the metadata such as the guider-computed effective times. The last such instance was exposure 133140 on 20220503, triggered from a failure of GUIDE7.
Tileid 25103 was observed with exposure 136464 on 20220523. It has multiple sky substraction warnings and a strong redshift clumping around z=0.75 across five petals (although the sky substraction warnings and the clumping might be unrelated).
The sky substraction on petal 7 is definitely problematic
We need to mark EXPID 109653 on TILEID 22381 as bad, and then reprocess TILEID 22381.
This is a tile that we marked "bad" months ago but then forgot about. @akremin, could you get this one, too? Thanks!
This tile got a ton of EFFTIME in pretty poor conditions. Redshift distribution shows clear pileups at z=1.8 and 5.5, as well as SKYCHI2PDF warnings.
https://data.desi.lbl.gov/desi/spectro/redux/daily/tiles/cumulative/7009/20211215/tile-qa-7009-thru20211215.png
I suspect that the sky variation is just too large and things are being misfit as high redshifts. We either need to improve the sky modeling or mark the component exposures as bad so that the tile may be reobserved.
Possibly the dark time equivalent of #4 .
Exposure ID 128784
on tile 23108
was out-of-focus, leading to a high RMS across petals on that exposure. Here is the tile:
https://data.desi.lbl.gov/desi/spectro/redux/daily/tiles/cumulative/23108/20220405/tile-qa-23108-thru20220405.png (this is the bad exposure)
https://data.desi.lbl.gov/desi/spectro/redux/daily/tiles/cumulative/23108/20220406/tile-qa-23108-thru20220406.png
Here's a plot of the relative focus from Eddie:
See also desihub/desispec#1738.
Tile 3533 will soon be marked "good", but we want to handle this case better.
It looks like >90% of the fibers on this petal are masked due to low EFFTIME. In this case, we want to throw the petal out rather than keeping the few fibers that survive the futs.
These are currently "unsure" and typically have lots of significant redshifts at z=1.8. Often have SKYCHI2PDF warning. May just need to have their component exposures marked bad so that NTS will reobserve them.
Affected: 20140, 21321, 23276, 24633, 24634, 26147
sframe files show lots of sky amplitude variation from fiber to fiber that is either a gradient or cloudy?
https://data.desi.lbl.gov/desi/users/raichoor/main-status/sframesky/sframesky-20211117-00109233.png
https://data.desi.lbl.gov/desi/spectro/redux/daily/tiles/cumulative/20140/20211117/tile-qa-20140-thru20211117.png
https://data.desi.lbl.gov/desi/spectro/redux/daily/tiles/cumulative/21321/20220114/tile-qa-21321-thru20220114.png
https://data.desi.lbl.gov/desi/spectro/redux/daily/tiles/cumulative/23276/20211117/tile-qa-23276-thru20211117.png
https://data.desi.lbl.gov/desi/spectro/redux/daily/tiles/cumulative/24633/20211219/tile-qa-24633-thru20211219.png
https://data.desi.lbl.gov/desi/spectro/redux/daily/tiles/cumulative/24634/20211219/tile-qa-24634-thru20211219.png
https://data.desi.lbl.gov/desi/spectro/redux/daily/tiles/cumulative/26147/20211219/tile-qa-26147-thru20211219.png
Tile 4402 on 2022-02-05 expid 121487 which rejects fibers 4250-4499 due to a cosmic appearing at location (2197,3718) of the r8 CCD. This has OSTEPD=5.7 in the frame file. Corresponding spectro issue is desihub/desispec#1619 .
Exposure 130556 on BACKUP tile 42424 observed 2022-04-16 reports a high read noise on amplifier R3B with OBSRDNB = 10.58 e- (or 6.38 ADU). This is apparently due to a bright (but not saturating) star next to the overscan, and hysteresis/CTE pushing down the counts by approx 30 ADU immediately after that star.
Attached is a cut throw a portion of the raw image showing the effect where the star is at column 2197.
Weird bright feature on r8; presumably a hot row or something?
https://data.desi.lbl.gov/desi/spectro/redux/daily/tiles/cumulative/10184/20220106/tile-qa-10184-thru20220106.png
Pretty clear in sframe:
https://data.desi.lbl.gov/desi/spectro/redux/daily/nightqa/20220106/sframesky-20220106.pdf
Leaving this as unsure for now.
There's a dubious set of z=2.5 objects on all petals on TILEID 4403.
https://data.desi.lbl.gov/desi/spectro/redux/daily/tiles/cumulative/4403/20220113/tile-qa-4403-thru20220113.png
My suspicion is that this is the fiberflat / humidity code, but I am not sure.
https://data.desi.lbl.gov/desi/spectro/redux/daily/nightqa/20220113/sframesky-20220113.pdf
It is remarkable how perfect the humidity correction is at the beginning of the night, and then how it suddenly gets worse for the rest of the night. If this is actually the cause, then most of the night would benefit from reprocessing.
I can't mark this tile as good since this presumably affects LyA decisions, so this one is relatively high priority.
We had problematic calibration images on 20211226 and have marked all tiles from that night as unsure until we are able to decide if this night is salvageable. Tiles affected:
2576 3238 3270 3607 6028 6376 6744 8764 11778 22955 23474 24146 25715 41253 41260 41484 41757 41758 41773
(note that some of these are backup 4xxxx series and are less important, but there are a number of dark and bright tiles as well)
Update: Title is incorrect, 20211226, I have decided that we should have included the dashes in the night strings as a matter of policy. Darn.
P4 shows clear pileups of garbage redshifts around z=0.25 and z=0.5.
https://data.desi.lbl.gov/desi/spectro/redux/daily/tiles/cumulative/10357/20211210/tile-qa-10357-thru20211210.png
This doesn't look like it affects LyA and I will mark it good.
I am having trouble seeing what is causing them; the sframe files look pretty good.
https://data.desi.lbl.gov/desi/spectro/redux/daily/nightqa/20211210/sframesky-20211210.pdf
A declarative, efficient, and flexible JavaScript library for building user interfaces.
๐ Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. ๐๐๐
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google โค๏ธ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.