*Calibration:* Redo if wavelength solution with ThAr+Ne is not good (>~ 5e-5
Å ; < 500 lines)

Basically, advice and practice of Johny Setiawan:  one should get below 4e-3
Å rms on the object fiber and below 5e-3 Å on the sky one; if not, one can
redo the calibration and there's a fair chance that one gets a good
solution. Long-term monitoring is displayed locally in a web browser on the
computer that  does the reduction; it allows to check whether the solution
(rms, number of lines) is better or worse than usual.

Some history: The ThAr+Ne lamp was much better (< 3e-3 Å rms) than the
historical one ThArNe (~ 5-6e-3 Å rms), but has become more unstable (rms
varies from calibration to another, between 3 and 6e-3 Å).  On average it is
still better, but the difference has decreased. This unstability appears to
have been triggered during two shorts (week?) events that display on the
graph rms vs. time as a very large scatter (instrumental problems I guess).

*Coordinates*: To be checked twice and put into a recent frame (p2pp expects
J2000)

Both the PI and the service astronomer should check object coordinates and
make sure ICRS or J2000 are used.

In two cases we lost time pointing at wrong coordinates because of typos (a
wrong digit and a missing sign).

I also corrected some coordinates by a few minutes, enough to put the star
off the field of view. It might have been B1950 coordinates (they're off by
much more than the FOV of the pointing) or a problem with the accounting of
high proper motions from that date (~ 1-2"/yr for stars of the programme
like HD 285968 or GJ 1065).

*Magnitude* *& Priority*:  Arrange for the magnitude and user's priority to
be systematically displayed in the listing of p2pp

Magnitude really helps to decide which observations to be carried out, as
one can more or less deduce the integration time from V, and decide to wait
for better conditions for faint objects.  It also helps to know whether a
target is feasible. For instance, there is a faint brown dwarf in the list
which I ignored (2MASSJ03393521-3525440): with V=16 it was (maybe) feasible,
with V=18 clearly not, but I had no knowledge of that, as SIMBAD only says:
B=20.2, I = 13.3.

Therefore I suggest to normalise the field "name" in the "constraints set"
to "P<user_priority> V=<magnitude>" where user_priority is 1, 2, or 3, and
magnitude is given with a leading for V<0 zero to allow sorting. Examples:
"P1 V=08.3" and "P2 V=11.2" and "P3 V=04.8"

*Finding charts: *Should be given with 5 x 5' FOV and a pointing arrow,
unless trivial

Obvious OB should have the comment " clearly brightest object within 5' " if
this is the case, which spares some minutes downloading the Aladin map to
check it (on the very slow machines of the control room, it may be quite,
quite long.)

In all other cases a finding chart  5 x 5' in size with an arrow pointing at
the object.  I decided to skip some OBs as I wasn't sure of picking the
right object; I lost also some minutes here and there because of the
dimensions of the chart (it becomes misleading, and at 5 am one thinks
slightly slower...)

For those OBs that do not fulfil this, the service astronomer could do that
on the first day of the run using Aladin/SIMBAD and contact PIs when there
is still a doubt.

*Exposure times:* Try to guess them right and adapt them to conditions.

Some bright objects of the SPHERE programme were either saturated or had to
be edited to stay below 60 000 counts.  On the other hand, faint objects
require long enough exposure to get a minimum of ~ 800 (that's low SNR ~ 20
to 50 depending on wavelength) and I had to edit one to get this minimum (if
not, data would not have been really noisy).

Basically, when V <~ 5 or V >~ 11, the support astronomer should check the
integration time (as it appears that problems arise for faint or bright
targets.) With constant conditions during the night, the counts can be
estimated from the first exposures of the night: That's proportional to
10^(-0.4*VMAG)*DIT, but I imagine experienced FEROS people do not need to
compute any more...

According to Johny Setiawan, if I got him right, the ESO exposure calculator
is not reliable. A small table giving the range of DIT for each Vmag  could
be done to guide proposers.

*README or comment:* Precise sample and/or period for variability or radial
velocity studies.

Many programmes have repetitive OBs in order to detect radial velocity
variations.  It is of great help to have a hint on the period and/or the
desired sampling.  For instance, TW Hya of period ~ 4 days can be observed
each night or twice per night ; objects with periods > 1 month can be
observed at larger intervals.  Hopefully, Johny knew most of these
programmes and could tell me when to observe the targets.