Master Dark and Pixel Rejection

[AstroScience]

When I'm using large number of files to create Master Dark, do I need to follow the same rule as for lights and use rejection algorithm according to the numbers of frames used?  Fore example if I integrate 70 Dark frames (DSLR), should I use Linear Fit Clipping?

If I do use Linear Fit for Master Dark, with it's default Sigma clipping (Low-5, High-2.5) I notice that I reject around 4% in total. Is it OK, or do I need to fine adjust rejection too?

[MikeOates]

Have you seen this page:
Master Calibration Frames: Acquisition and Processing By Vicent Peris
http://www.pixinsight.com/tutorials/master-frames/en.html

Mike

[AstroScience]

Thousand times, Mike,
and it doesn't explain why I should use Winsorised Sigma Clipping specifically and why Sigma 4 and 3 for high and low.

Is this valid the same for 10 darks as for 100 darks? Does it matter if this is a DSLR darks or CCD darks?

We all know that you select rejection algorithm dependently on the amount of subs you going to integrate and then fine tune High and Low sigma's.
But Darks or Bias are not Registered as Lights when they integrated, they just stuck up on each other, so is there a point to use different rejection algorithm or there is a point to adjust High and Low Sigma's? 

I truly appreciate every suggestion that come from Team, but I would like to know why this, or that, being suggested.

[TobiasLindemann]

IMHO "linear fit clipping" makes no sense at darks or bias. All darks should look like about the same. Linear fit clipping only makes sense when the signal varies a bit between the frames. For Example, at dawn there is some light added to the signal of the object, compared to midnight. When you select "winzorised sigma clipping" you loose the signal of the dawn images because there is too much difference between the midnight and the dawn images. So "linear fit clipping" is the better algorithm for this case.
I hope this is right 



Greetings
Tobias

[pfile]

on the other hand linear fit clipping kind of degenerates into sigma clipping as the slope approaches 0.

probably when integrating darks the important thing is to reject cosmic ray hits, etc. these will be pretty bright compared to the rest of the pixels so it should be easy to reject no matter what method is used.

rob

[AstroScience]

Ok,
I wanted to know what will really happen and decided to experiment.

I have Integrated 50 Darks from my DSLR to 3 Master Darks:

1. Master Dark with NO pixel rejection at all - (?K = 1.354e-003)
2. Master Dark with Winsorised Sigma Clipping and sigma's of 4 and 3  (High and Low Range clipping at default too)   (?K = 1.352e-003)
3. Master Dark with Linear Fit Clipping and default sigma's of 5 and 2.5  (High and Low Range clipping at default too)  (?K = 1.342e-003)

Then I took one matched Sub frame and Calibrated it with all 3 Master Darks. For scaling to work I have also used Master Bias.
Debayering and Noise Evaluation of Light Frame showed:

1. When calibrated with Master Dark with no pixel rejection :

Dark scaling factors:
k0 = 1.069

Gaussian noise estimates:
s0 = 1.136e-003, n0 = 0.561 (MRS)
s1 = 1.113e-003, n1 = 0.710 (MRS)
s2 = 9.848e-004, n2 = 0.536 (MRS)

2. When calibrated with Master Dark that had WS rejection 4:3 :

Dark scaling factors:
k0 = 1.060

Gaussian noise estimates:
s0 = 1.137e-003, n0 = 0.560 (MRS)
s1 = 1.113e-003, n1 = 0.709 (MRS)
s2 = 9.875e-004, n2 = 0.536 (MRS)

3. When calibrated with Master Dark that had Linear Fit rejection 5:2.5 :

Dark scaling factors:
k0 = 1.068

Gaussian noise estimates:
s0 = 1.142e-003, n0 = 0.563 (MRS)
s1 = 1.116e-003, n1 = 0.711 (MRS)
s2 = 9.978e-004, n2 = 0.540 (MRS)

Now if we compare the results of WS against LF , the differences in numbers are minute, but WS showed just a bit lower noise levels in calibrated sub than with LF.
What is interesting, is that sub calibrated with master dark that had no pixel rejection at all, showed noise level that are lower than the other two.

Of course, having no pixel rejection can leave cosmic rays in master dark, but maybe we don't have to be that restricted with Sigma's Low and High?