Quote:
Originally Posted by Somnium
SO i have read both that you need to take bias images and you don't need to take bias images . given that you are going to take darks, some suggest that the dark images already have the bias data included ... but, isn't the bias data also included in the flat field images ? so when you process the subs, you are subtracting the bias with the darks then adding it in again with the flats therefore you would need a bias only image to subtract it again. with that rabbit hole opened, wouldn't you need to take your flats at the same sensor temp as your subs, darks and Bias'
thoughts ?
|
In a perfect world, the things that you need to do are:
1. remove the bias+darkcurrent from the lights, so that you only have signal that came from the sky - you can do this by subtracting master dark data taken under the same conditions as the lights.
2. remove the bias+darkcurrent from the flats, so that you only have signal that came from the flat source - to do this you subtract a master flat-dark derived from data taken under the same conditions as the flats (but not necessarily the same conditions as the lights etc).
3. divide the subtracted lights by the master flat obtained from the subtracted flats.
However, there are occasions where you may not have appropriate darks: eg you use sky flats where each flat has a unique exposure, or you do not have any darks taken under exactly the same conditions as the lights. Then Peter's caution re dark non-linearity applies (to all CCDs, not just SBIG) and bias needs to be considered separately.
If you have varying sky flats, it may be OK to accept that the results will not be quite perfect and simply subtract the bias (taken at the flat temperature) from the sky flats in place of step 2 above. This assumes that the dark current in a sky flat exposure will be negligible (which is not too bad an assumption most of the time). It also helps if you clean up the worst hot pixels with a bad pixel substitution, after you do the flat correction on the lights.
If you have darks taken under different conditions from the lights, you may be able to scale the darks for temperature and time, accepting that the dark non-linearity will give you a less than optimum outcome. To scale the darks in any way, you will need to remove the bias (taken at the dark temperature). after scaling the dark data, you will need to subtract it and an appropriate bias from the lights (in place of step 1 above). Alternatively, PI implements a dark scaling based solely on statistics and chooses a scale that minimises the noise in the outcome - this approach should let you get an optimum result using not-quite-matching darks and it is fully automatic (I assume that it performs a bias subtraction, so you will need bias data at the dark temperature).
The PI scaling approach may be usable on variable sky flats as well, but it would possibly be stretching the friendship to try to extrapolate from normal long-exposure darks. It would probably be better to use a set of darks taken at short exposures, more closely related to the sky flat exposures.
At least I think that is the case and I would happily be corrected...
Edit: the more I think about, the more sensible it seems to take as much cal data as possible at the same temp as the lights. It isn't strictly necessary, but it can cut down on the amount of cal data you need in some circumstances (eg, you can use one bias set throughout).