Question about bias and flats

[RichLD]

Recently, when trying to correctly calibrate my subs in PI I made the mistake of not including a master bias. The outcome was that my flats overcorrected quite severely, with very obvious reverse vignetting on the calibrated frames. Redoing the calibration with my usual master bias remedied the situation. I don't tend to use dark frames, running Cosmetic Correction in PI to deal with cold and hot pixels. I also calibrate my flats using flat darks (the flats are 8s in duration using a EL panel on the QHY9, HA filter).

Try as I may my poor brain can't seem to work out why the lack of bias in the light calibration routine would cause the flat overcorrection. To satisfy my curiosity only (given that I don't have an actual problem), would any of the experienced imagers out there know why?

Many thanks in advance for any responses, it's driving me mad! 

All the best

Rich

[sharkmelley]

If you omit the master bias then the software will assume the bias is zero.  This normally results in overcorrection because flats are generally well exposed.

The reason for this can only be illustrated by long and boring arithmetic, so here goes:

Let's assume your optics vignette at 50% and your bias level is 1000.  Your lights might have sky fog levels of 4000+1000 in the centre and 2000+1000 in the corners, giving pixel values of 5000 and 3000.  If your flats are well exposed then the pixel values may be 30000+1000 in the centre and 15000+1000 in the corners, giving 31000 and 16000.

When the master flat is applied, the bias is subtracted from both, then the light is divided by the flat so, in this example, you get (5000-1000)/(31000-1000)=0.133 in the centre and (3000-1000)/(16000-1000)=0.133.   Both values of 0.133 would then normally be scaled up by the maximum value in the flat but we'll ignore that.   We are just interested to see that the centre and the corner have the same value.  This is proper correction.

If the bias is omitted when the light is divided by the flat then in this example you get 5000/31000=0.161 in the centre and 3000/16000= .1875 in the corner i.e. the corner now has a higher value than the centre, which is the overcorrection effect you are seeing.

Mark

 

[Astrofriend]

Long time ago I did this tutorial:

http://astrofriend.eu/astronomy/tutorials/tutorial-pre-calibration-flat-dark/tutorial-pre-calibration-flat-dark.html

 

Maybe explaine something of the different steps of calibration.

But the main problem to omit the bias is that it give wrong compensation as Mark told above.

/Lars

[Scooot]

6 hours ago, Astrofriend said:

Long time ago I did this tutorial:

http://astrofriend.eu/astronomy/tutorials/tutorial-pre-calibration-flat-dark/tutorial-pre-calibration-flat-dark.html

 

Maybe explaine something of the different steps of calibration.

But the main problem to omit the bias is that it give wrong compensation as Mark told above.

/Lars

Thanks for posting the link. I found it very informative. 

[Astrofriend]

Hi Richard,

Thanks for the comment. I took a look at it agian and find some uggly misspellings. It's not easy when writing in a foreign language. I have corrected and did a complement with some links. Maybe a little bit clearer now.

BR

Lars