Can we add a few more details? Which lamps, and what do we want to do about persistence?

What does "can be non-simultaneous" mean? That we can take sub-exposures to reach 30k counts (meaning photons not DN?)? Do we need brn to be all taken at the same time?

For item 2 (fiber configuration allHome) how many counts do we need (peak and near the dichroics)? Do we need all of brn, or can we do br and rn separately if needs be?

Maybe the best way to be sure that we know exactly what we need would be a list of exposures, albeit with unknown exposure times.

I believe "non-simultaneous" means that b/r/n exposures may be taken at different times. The goal of item 1 is measurement of the profile in each arm, for which we do not need to be able to compare fluxes across arms.

Item 2 is for the normalisation measurement (hence the need for simultaneous exposures so we can compare fluxes across arms). This behaves in our pipeline like a flat-field (divides every science exposure), so it needs to have good signal-to-noise everywhere we care about. For a starting proposal, how about a minimum total across exposures of 10k electrons (S/N ~ 100) at the dichroic crossover points, with a minimum of 3 exposures?

Item 2 must be simultaneous b/r/n. Please do not do b/r and r/n separately, because that would produce separate normalisations for r (maybe we could deal with that in software if we had to, but in my opinion it's not worth the loss of time or loss of confidence in the result).

We should also take data to see how the flat non-uniformity depends on the instrument rotator. In particular, we expect the pattern from the ring lamps and the JEG lamps to be fixed in dome coordinates.

It'd be good to investigate the twilight sky as a function of rotator angle too, but that's harder. Maybe take different rotator angles on different nights?

To characterise the backgrounds, we want to observe a variety of input spectra and measure the background.

• With both fiber configuration mod32 (constructing a background model):
  • Observe with all available spectrograph arms (including both r and m; non-simultaneous across arms is OK; different sources may be observed at different times through the run):
    • Separately: ring lamps, Gunn quartz, twilight (set exposure to 30k peak in each arm individually)
    • Separately: each of the arcs, dome fluorescent lamps, night sky with varying amounts of moon (set exposure so that >~ 95% of the lines are unsaturated in each arm individually)
    • Separately: over-expose each of the arcs (set exposure in each arm individually so that bright lines are very saturated but not bleeding)
    • Any other input light sources that might be available that I haven't thought of
• Repeat the above with fiber configuration allHome (to test application of the background model on regular data)
  • Do not repeat the over-exposed arcs, just regular arcs

There are no timing or sequence requirements on any of the above observations, except that it would be good to get an allHome observation for every corresponding mod32 observation (but they needn't be back-to-back, just take some effort to try to get the same input spectrum).

For commissioning n3, we will need a full set of calibs:

• Dark (may as well get for the other arms at the same time).
• Dithered flat (I've heard that n3 is cleaner than n1, so we might be able to limp along without this). If it doesn't introduce extreme additional overheads, why not do all other arms at the same time? Exposures may be non-simultaneous across arms.

For all arms (brn and m), we need

• Full set of arcs and quartzes to fit the detectorMap. This is required for n3 and r3/m3 post-refocussing, but let's get them for all arms so we can confirm that we have good detectorMaps.
• Fiber profiles, as outlined in the ticket description. We especially need this for n3, r3/m3 post-refocussing and for m1/m3 (because we don't have any for those yet), but let's get them for all arms.

May as well get some good biases for b/r to track any changes.

For dithered flats, we want 3 exposures at each dither position (so we can remove image artifacts), observed back-to-back. Apart from that requirement, dither positions may be obtained at different times.

I'm curious about the repeatability of the hexapod. Can we take an arc at the home hexapod position, move the hexapod around a bunch, then return to the home hexapod position and take the arc again?

price about hexapod repeatability, it's an interesting study but we can use DCB / IIS for that matter.
So I did update the plan, the only problem I see, is dealing with persistence, taking interleaved darks would multiply the time cost by two basically and I'm not sure we can afford it.

I'm not worried about persistence. While I'm sure it's an important effect, it doesn't seem to be debilitating (I didn't worry about it last time) and so I'm tempted to suggest we simply skip the interleaved darks so we can get a product which will be good-enough until we figure out how to deal with the persistence properly.

arnaud.lefur Could you confirm that data in red and orange are must-to-take at the beginning of the run?
Exposure time in the last run was as follows:

• br – HgCd: 30s, Ne: 10s, Kr: 30s, Xe: 45s, Ar: 60s
• m – Ne: 20s, Kr: 60s, Xe: 90s, Ar: 120s

Following the previous run, are 3 frames per line OK? We may need to optimize exposure time for n band.  In regard to quartz for detectorMap, it it OK to use 60s? Or should we take longer?

That's right, highest priority.
Yes, those exposure time looks fine, thanks and if we can afford it, 3 repeats would be very good.
For the quartz, I guess we could increase slightly the exposure time maybe 90s ? We need a lot of quartzes so the time budget can quickly explode.

Sorry if you got an alert, I pressed some key command that assigned this to me by mistake but switched it back.

I don't think we need twilight m-band. The only difference from r is the grating, so we can measure the throughputs using quartz in rm and r-band on-sky data.

Actually, I don't think we need twilight data in more than one band, but it comes for free. And, of course, comparing the three sets of arms will be interesting as a sanity check.

Run12 is now over, calibration data can be found here .