Noise evaluation script

[edif300]

Hi all,

I am evaluating the noise of the master dark and master bias from my ccd cam by noise evaluation script. They aren't definitive dark and bias since it were taken at only -15ºC (heatsink at 38.5ºC). 6 individual darks of 5min each, 10 subframes for master bias.

Can anyone explain me what I must read from these results?

?K = 3.619e-005, N = 16497152 (98.33%) for master bias.

?K = 5.136e-005, N = 10642547 (63.43%) for master dark.

Regards,
Iñaki

[Juan Conejero]

The meaning of these numbers is as follows (for the first line):

- The standard deviation of the noise is 3.619x10^-5, assuming a Gaussian noise distribution.

- The noise evaluation algorithm (multiresolution noise evaluation, unless otherwise stated) has found 16,497,152 of pixels pertaining to the noise component of the image, or a 98.33% of the total pixels. The remaining pixels correspond to the signal component. Noise estimates are ingeneral not reliable when the fraction of noise pixels is very small; that's why we provide this value.

As for an interpretation of these noise estimates, there's not too much to say IMO. Both values are pretty normal for bias and dark frames.

[edif300]

Thank you very much Juan!!

[drmikevt]

So, would it be possible to use this script to compare dark frames integrated with different settings (winsorized sigma vs linear clip, different sigma high settings, etc) to see which one is 'best'?  Assuming that the lowest noise image would be generally best...? 

If this is not the right tool, how do you determine the best dark frame in a group? 

Thank you for all input.  I just finished a workshop with Vince and I'm trying to look at things with a finer lens now.

Mike

[mschuster]

This is possible, but results are somewhat corrupted by fixed pattern noise in the darks. This fixed pattern is what an ideal dark should be subtracting, but it can be seen itself as noise by the script.

On my setup I just assume all darks are good, except for spurious noise (eg cosmic rays). So rejection is set conservatively to reject just them. And then I dither to reduce the remaining problems.

Thanks,
Mike

[drmikevt]

Thanks, I understand. 

But still, if we use the instructions here:  http://www.pixinsight.com/tutorials/master-frames/index.html we can get very different results than if we use Linear fit, which usually results in a higher noise value (# of pixels), but perhaps that is what we want...?  I'm guessing the tutorial outdates Linear Fit so is that the correct setting for integration (assuming 50 frames) now?

I'm sure this discussion has happened before, but there must be a way to determine an optimal dark or bias from a set of iterated images.

Again, all input is greatly appreciated
Mike

[mschuster]

I am guessing here:

Linear Fit might help with illumination or gradient changes across lights, but it might be less helpful for biases and darks. For bias and darks Linear Fit might end up just fitting noise and as a result might effectively provide only a less strict rejection.  If so, this is OK but maybe it is not accomplishing anything useful.

Thanks,
Mike