image integration Question !

[pscammp]

Hello All,
    I have a question regarding Image integration if you don't mind......

When doing the final integration I use Linear Fit Clipping and I also use a preview to play with the High and Low sigma sliders to get the best noise result I can.

The question I have is not so much to do with the final integration but more to do with integrating the Biases and Darks etc.

Based on this tutorial:

http://www.pixinsight.com/tutorials/master-frames/index.html

The tutorial suggests using Winsorized Sigma Clipping for the darks and biases, is there any real benefit from also playing with the high and low sliders using a
preview to get the lowest noise in these calibration files, or just let it run at default values   

Many Thanks People
Paul

[bulrichl]

Hi Paul,

my understanding is that we use pixel rejection in dark frame integration in order to exclude cosmic ray hits which are high outliers. In the case of bias frame integration the probabilty of such artifacts is extremely low.

I don't know of other reasons for pixel rejection in these cases - perhaps someone else does?

Bernd

[ngc1535]

Hi Paul,

my understanding is that we use pixel rejection in dark frame integration in order to exclude cosmic ray hits which are high outliers. In the case of bias frame integration the probabilty of such artifacts is extremely low.

I don't know of other reasons for pixel rejection in these cases - perhaps someone else does?

Bernd

Just to add a subtle point. The cosmic rays detected in a bias typically occur during the charge transfer time (downloading image crudely). Of course some cameras are faster than others in terms of this process.

[bulrichl]

I must admit that I never looked for cosmic ray hits in a bias frame because I supposed that the probability is proportional to the exposure times. With my DSLR camera I chose 1/4000 s and with my CMOS camera 0 s. What is your experience: did you actually sometimes detect a cosmic in bias frames?

Bernd

[RickS]

What is your experience: did you actually sometimes detect a cosmic in bias frames?

I have certainly seem them.

Cheers,
Rick.

[ngc1535]

Yes, with older/slower cameras I observed cosmic rays in the biases. Granted, there were not many- but just to make certain I was not making this up, I went back and blinked a set of 11 unbinned biases from an STL11000 camera. This camera took quite of bit of time to getting the image from instrument to the computer (something like 20+ seconds). Some part of the time is just shifting the charge off of the chip (this is same regardless of exposure time) I think. If I am correct... I should see a gradient in cosmic ray hits that increases towards the bottom of frames.

Attached is the snippet from Chromey   "To Measure the Sky: An Introduction to Observational Astronomy" Note the last sentence in the first paragraph.

-adam

[bulrichl]

Hi Adam,

OK, this is new for me. Also, the cited texbook seems to be very interesting, thank you for referring the title.

In the first paragraph, Chromey enumerates cosmic ray hits, local radioactivity (which is often imprecisely included in the term 'cosmic ray hit') and electronic interference. What is the effect of electronic interference? Will it produce a sort of banding or are single pixels affected? And back to the original question: Is this kind of artefact also removable by pixel rejection? (I guess that it will not be removable.)

Bernd

[ngc1535]

Bernd:

The original poster wondered about the "real importance" of statistical data rejection of darks and biases. That is why I chimed in with the subtle reason for the use of rejection for bias frames.
To answer the original question, the same logic that is used to determine the rejection type and sigma levels with light frames is basically applied here. Windsorized Sigma Clipping is just fine. Specifying 3-4 sigma for the low and 2-3 for the high is just fine.

Since bias frames (and darks) are noisy- in an ideal situation with no transient events or electronic changes- the mean of the data is desired. So for years I have been quite draconian about the rejection with darks and biases. I used asymmetric min/max clipping. With 50 biases, for example, I would min/max clip (3,5). This means I would be averaging only 42 values of the set. But so what! Biases are cheap! Easy to take and with that kind of clipping- no sigma conditions need to be kept- just find the lowest three and highest 5 values in the set and cut them away!

Concerning electronic changes (interference)- if it is a random (or semi-random, like a phased effect)- rejection will not do anything and it just becomes part of the noise term. If, however, the fluctuations occur in only a few frames in a set... sure rejection will take care of this.