Feedback from Michel from Nov. 2013 - and replies by Stephen in red/green (also input from Mark, Andrea, Cosimo)
-The NTT points very well (within a few arcsec). As a result, the object of interest always falls on, or very near to, the bad column in the center of the chip. This is not good for accurate photometry. One should therefore systematically point slightly off the nominal position (e.g. 20" E or N, depending on the orientation of the rotator). There are several possibilities for that:
-Change the coordinates of the object in the OB: not recommended, as you want to keep a trace of which object you were aiming at!
-For imaging only: use the acquisition template with dithering, but enter just one position (e.g. 20" West), for single imaging.For multiple exposures, the ditehring pattern is anyway to be preferred (see below).This will require to modify all the template OB's for imaging.
The bad column is at x=486 (in binned, unprocessed pixel coords) is known. This is about 30 pixels from the centre of the chip (x=515) or about 8 arcsec. This can be a problem, as the telescope pointing is not necessarily the CCD centre, and pointing drifts can lead to the science object falling close to the column. To be safe, an offset should be adopted.
ALL imaging OBs should add a shift by 20 arcsec EAST, by default.
This only affects follow-up imaging, not acquisition images.
ACTION - Cosimo to change.
-For spectroscopy: you cannot use a dithering acquisition template, as you have to point and then move to slit.
This is already known - objects are placed at x=545. This is offset enough from the bad column such that we do not have problems.
But you can tell the operator to move manually the telescope before acquiring the guide star. This solution is the easiest, as it allows you also to check the position of the rotator, but you have to remind that at every pointing!
We will adopt the standard 20" offset EAST, not this ad-hoc approach, which will be forgotten,
Imaging with EFOSC
When multiple exposures are required for the same field (with one, or multiple filters), one should avoid to take them always at the same position. A dithering pattern allows a better subtraction of the background and helps to remove fringing, or defects (if your object falls, by accident, on a defect, it will be there for all the exposures if you do not dither...). It implies a slightly more complex reduction procedure, but I suppose it should not be a problem? An exemple of OB is attached.
Correct that the if the object falls on a defect then it is a problem. But the downside to large dithers is that if there are a small number of comparison stars on the FOV, then some of them may be shifted off the FOV, or too close to the detector edge. Some science targets may have a strict requirement for comparison stars.
Fringe removal - with N dithered frames, a sky frame can indeed be made with the other N-1 frames. But a careful assessment needs to be made on the added noise level of the custom frame compared to the higher signal-to-noise library fringe frame. It's not obvious that the N-1 sky frame is better than the library frame. But it is a good point to look at further.
Currently we trust PIs to balance these issues for their targets.
ACTION - no changes to default strategy right now. All PIs to consider the advantages and disadvantages of both (this only affects deep imaging targets really). The DRAQC team will be kick started in early 2014 to consider the fringe problem in more detail and make a recommendation. The decision to dither or not is up to the PIs at all times (as there are reason to dither and reasons not to). PIs can provide specific OBs if they want.
Slit positions and Standard Stars
The various slits in EFOSC do not necessarily fall on the same Y position. By chance, the 1" and the 1.5" were close in position, 1 or 2 pixel difference, but this is nevertheless 1/4 ", so is not negligible with respect to a 1" slit. The 5" slit was much further away...So an OB for standard stars doing successive gratings with several slits, without repointing, is not good, you may loose quite some flux! OB's for different slits should thus be made separate, but you can then ask the operator to skip the pointing part of the new OB, and just take a verification image before re-centering on the new slit.
We don't use the 5" slit, so that particular one is not an issue in itself.
The difference in position of the 1" and 1.5" slits is certainly a concern, although if the seeing is greater than 1.5" and the uncertainty in position is 0.25" then the losses are marginal.
However, it might make sense to have the 1" and 1.5" standard OBs separate.
ACTION - DRAQC to look at this.
The 1" slit is often marginal for the usual seeing (0"8 at best). There exists a 1.2" slit, which seems to me optimal in most cases. To be evaluated whether using this slit (instead of the 1") would significantly degrade the classification possibilities with Gr#13 in most cases due to the slightly poorer resolution (but you would definitely gain in S/N...). To be investigated. (adding a different slit should not be a big problem for data reductions!)
Gains are marginal. Slit throughout for a 1.2" slit compared to 1" slit in 1" seeing are approximately 85% and 75%. Therefore 13% gain in photons, and sqrt(1.13) = 6% gain in S/N (and this approximate calculation assumes shot noise is dominant, when sky background, or galaxy background is significant, the gains decrease).
It's not worth perturbing the observing procedure and having 3 slit choices for this marginal gain.
The list of standard stars looks to me too short...about 5h interval in alpha between them does not allow an optimal choice during observations (and some of them are quite far north...). A few more (good ones) could be added (e.g. LTT1020 around 1h30, etc...): the cost is to introduce them also in the reduction pipeline, but this should not be so difficult...or is it?
Not essential - but we will look at adding another to the pipeline.
ACTION - DRAQC to look at
This is a real problem in some cases. Particularly for LSQ candidates, where the size is too small (not to talk about the "wrong" orientation...) to always allow a clean identification of the target. Same problem for PSN objects...
For OGLE sources, it is much better. One should systematically add a finding chart with about 5'x 5' FOV, and orientation properly indicated on the Marshall (we extracted systematically a DSS image to secure the identification,but it was not always enough...) For follow-up objects, I would be tougher! No proper finding chart = no follow-up...! This is simply to avoid loosing time, or exposing the wrong object, as happens sometimes...
Agree, the finders should always be of high quality. But LSQ finders are always 5x5' and the orientation is clearly marked. I don't understand what you mean by small and "wrong orientation". There is an issue with the LSQ finders, but it is neither of these. The scaling is often not optimal which means SNe close to bright hosts are not clearly visible (LSQ13ddu is a good example, see the link below). But automatic scaling is difficult to get right in every case. Manually it is much easier.
ACTION - Mark/Stephen to look at how to make the LSQ finders better. We previously designed these for blind-offsetting which we don't actually. So we can invertigate making them more optimal for EFOSC2.
See here for an example : https://sites.google.com/a/pessto.org/wiki/private-area/pessto-followup-targets/lsq13ddu
ACTION - ALL follow-up targets must post a finder based on an EFOSC2 acquisition image. There's no reason why this cannot be done, since the targets must have had previous EFOSC2 images and the acquisition images are perfect finders. ACTION ON ALL PIs.
I understand that the OB's have been made such as to spend the minimum time on classification...but this is also at the expense of S/N, of course, so classification, at the end, is not always possible. For instance, it seems to me that between 900sec for 18.5 <mag< 19.5, and 1500 sec for 19.5 <mag< 20.5, there is a margin for an intermediate exposure time of, say, 1200 sec, around 19.5, etc... I would suggest one should systematically take the upper exposure time, rather than the lower one, for a given magnitude, specially as often the magnitude of the candidate is not recent, and is likely to be underestimated... But to be discussed again, of course.
Classification is possible ~80% of the time, unless I am missing something. Now within that 80%, we may be missing some subtleties (e.g. definitive 91T or 91bg types, distinguishing between normal Ibc at z>0.05), but I would argue that is not a reason to increase the exposure times. For the other 20% we tend to go back and get better data - these tend to be blue featureless spectra objects which evolve into type II, IIn, SLSNe Ic, CVs. We usually do get them classified, but a few do manage to avoid classification. The question is - are we missing scientifically interesting targets with this strategy . It's not clear that we are.
Andrea also thinks there are times that the classification spectra SOMETIMES (but not always) lack enough S/N.
ACTION - the simplest solution here is for the observers ALWAYS to take the longer exposure time option when the magnitude is uncertain, or it is close to a boundary. e.g for a ~19.3 target - take the 1500sec one.
The current magnitude should be systematically indicated...this is indispensable for a proper evaluation of the required exposure time...but is not always given...and it is not enough to say "I want a 45mn exposure" as this has to be justified, in view of the pressure on follow-up observations! I have suggested already that one needs some better scientific justification of the requirements for follow-up, to be able to make the appropriate selection/decision at the telescope, when there is conflict between several targets.
There is no need for PIs to give science justification in the Marshall. The simple CRITICAL, IMPORTANT, USEFUL should be enough. Observers should not request science reasons, they should just carry out instructions. It does not appear to me that there was a severe conflict of RAs in targets during the last few months such that all CRITICAL and IMPORTANT targets could not be done. Rather it seems that some issues raised by PIs were to do with choices between these categories and the balance between follow-up and classification.
Personally I (Stephen) think the labels are OK if the observers stick to them. And when there is a genuine conflict then Mark/Andrea/Stephen can help decide.
ACTION : none needed, unless genuine conflicts can be shown to us
The choice of gratings for follow-up is also important, and should be better justified also (my feeling is that, at the moment, this choice is just made routinely for follow-up): choosing Gr#11 + Gr#16 is more time consuming (and there is also a large overlap between the two spectra) than simply taking Gr#13 again (as was noted by Andrea already). Generally speaking, the observers need to have all the elements in hand, to be able to make the right decision at the telescope (they have to make the ultimate decision when there is conflict!) It would also be usefull to be able to update directly the Marshall (e.g.classification, or magnitude), not only via the comments, but directly in the main body (e.g., the current magnitude in bold and big fonts!)
The observers should NOT be making decisions about what gratings they think are better. They should follow the marhsall requests.
But there is an issue for PIs - generally they should request Gr13 unless there are reasons to request Gr11+Gr16.
ACTION : More importantly - PIs should specify the exposure times should be given, OR the magnitude, more explicitly - to ensure the correct OBs are chosen by the observers.
Observers - should follow exactly the instructions from the PIs.