ULTRACAM observing tips

This page is an accumulation of hard-won knowledge from using ULTRACAM. It intention is to be a fairly short reminder of what to do to get the best data and best practice in general. Let me know of any others to add.

Biases Make sure that lights in the dome are off. 'makebias' will ignore the first and last frame so you may want to add a few more. Take them as identical to the data as you can. Check them for peculiar offsets which can occur especially after power-ons. Re-do if necessary; mark bad ones as junk in the comments.

Drift mode Be careful of the lower edge of the imaging region which often seems to have odd features. Make sure that the drift mask is not causing steps in the sky background. Make sure that the drift mask is in!

Flats Make sure the telescope is moving while taking sky flats so that stars can be medianed out of it. Always combine with median, not clipped mean rejection.

Darks Probably only really worth taking at night when weather is poor, but please always try to take some if only so that we can monitor the prevalence of hot pixels. If you have to do so during the day, then just note the fact that the dark is probably nor reliable because of light leakage.

Bad pixels Try to create bad pixel mask and transformation files early in a run and use them for acquisition. It is not at all difficult to do. I have had observations wrecked by bad pixels. There is an especially nasty bad column in the right-hand side of the green chip which can be hard to spot. Why rely on luck?

Naming of targets This is currently the only way of identifying what we are pointing at so get it right. Also please check the current target list to see if a target has been observed before and try to use a consistent name.

Log comments It is very useful to know about the observing conditions, whether the targets moved during the run and whether a run is essentially junk. Please use the comments for such things, and say 'Good data' or some such if all is OK. Can't have enough comments in the log in my opinion. If observing in drift mode, it is common to have just 2 targets, then say 'target in left-hand window' or whatever to help later identification. This is useful in all cases especially in sparse fields.

Noise calibrations We are woefully short of these which can give us the values of readout and gain. Especially we don't know how they change from fast to slow etc. If you can set up a good internal flat, then please try to take a series of frames with count levels starting from almost 0 above bias and rising to a large value. Try to do so while keeping exposures short so that there is not much dark, i.e. by removing neutral densitiy filters. Try to run the results through 'noise'. You will need a good set of biases to remove from the data without adding significant noise (i.e. a run of 100 at least). Its tedious to take these and analyze them, but it is necessary.

Tom Marsh, Warwick

Biases	Make sure that lights in the dome are off. 'makebias' will ignore the first and last frame so you may want to add a few more. Take them as identical to the data as you can. Check them for peculiar offsets which can occur especially after power-ons. Re-do if necessary; mark bad ones as junk in the comments.
Drift mode	Be careful of the lower edge of the imaging region which often seems to have odd features. Make sure that the drift mask is not causing steps in the sky background. Make sure that the drift mask is in!
Flats	Make sure the telescope is moving while taking sky flats so that stars can be medianed out of it. Always combine with median, not clipped mean rejection.
Darks	Probably only really worth taking at night when weather is poor, but please always try to take some if only so that we can monitor the prevalence of hot pixels. If you have to do so during the day, then just note the fact that the dark is probably nor reliable because of light leakage.
Bad pixels	Try to create bad pixel mask and transformation files early in a run and use them for acquisition. It is not at all difficult to do. I have had observations wrecked by bad pixels. There is an especially nasty bad column in the right-hand side of the green chip which can be hard to spot. Why rely on luck?
Naming of targets	This is currently the only way of identifying what we are pointing at so get it right. Also please check the current target list to see if a target has been observed before and try to use a consistent name.
Log comments	It is very useful to know about the observing conditions, whether the targets moved during the run and whether a run is essentially junk. Please use the comments for such things, and say 'Good data' or some such if all is OK. Can't have enough comments in the log in my opinion. If observing in drift mode, it is common to have just 2 targets, then say 'target in left-hand window' or whatever to help later identification. This is useful in all cases especially in sparse fields.
Noise calibrations	We are woefully short of these which can give us the values of readout and gain. Especially we don't know how they change from fast to slow etc. If you can set up a good internal flat, then please try to take a series of frames with count levels starting from almost 0 above bias and rising to a large value. Try to do so while keeping exposures short so that there is not much dark, i.e. by removing neutral densitiy filters. Try to run the results through 'noise'. You will need a good set of biases to remove from the data without adding significant noise (i.e. a run of 100 at least). Its tedious to take these and analyze them, but it is necessary.