Flats frames are not properly calibrated by biases

[CIZIRF]

Using Sharcap to create flats, and saving individual flat frames I noticed that flat-bias frames created by Sharpcap do not work as well as flats frames calibrated using the Siril "Synthetic biases" method ( https://siril.org/tutorials/synthetic-biases/ ).
It makes a real difference: Using the flat-bias frames, darks spots / vignetting are still visible on the final stack, while calibrating the flats using "synthetic biases" completly removes all optical defects (dust and vignetting).
Also had such problems using livestacking for lucky imaging, sometimes flats are over compensated (dark spots appear as light spots on the final stack), sometimes under compensated, but in most cases the flats unperfectly remove dust and vignetting from the final stack.
This is not related to the latest release (I'm currently using 4.1 64 bits), but something I could see also from earlier 4.0 versions.
Regards Patrick

PS: I'm using a Uranus C pro camera most of the time at -10C, flats sre created using a flat panel. Flats are created at the same temperature, but obviously not at the same gain (flats correct exposure requires Gain to be set to 0) then adjusting the exposure time to get around 50% on the histogram (between 350ms, and 3s (depending on the filters used). Of course the offset level remains the same (generaly using 50).

An image better than explanations: https://flic.kr/p/2p6PKuV

[admin]

Hi Patrick,

do you have some unprocessed images that you can share that I could investigate? At least one (or more) individual flat frame(s), bias frame(s), light frame(s) and dark frame(s) along with the settings file SharpCap saves next to each would give me a starting point.

Under or over correction is usually a sign that somewhere the settings of flat vs bias (or dark flat) and light vs dark do not line up, meaning that in one side of the calculation or the other the amount of signal in the data due to the light being collected (as opposed to the camera's black level) is being miscalculated. I'd like to check for that and rule it out before looking at the different bias approaches.

cheers,

Robin

[CIZIRF]

Hi Robin,
Thanks for your reply.
Here are the requested unprocessed files, together with the camera settings for each one https://transfert.free.fr/xrmnTSp (link valid 7 days)

PS this is the "bias" data statistics that I used to determine the multiplier factor to be used for synthetic biases (Uranus Cpro @-10C bias exposure 0.01ms with telescope covered)
image chan mean median sigma min max noise
1 0 6.431545e+02 6.400000e+02 9.768679e+00 5.440000e+02 7.840000e+02 9.717567e+00
2 0 1.283150e+03 1.280000e+03 9.756432e+00 1.168000e+03 1.408000e+03 9.709022e+00
3 0 1.923160e+03 1.920000e+03 9.758938e+00 1.792000e+03 2.064000e+03 9.710732e+00
4 0 2.563122e+03 2.560000e+03 9.770448e+00 2.416000e+03 2.688000e+03 9.713920e+00
5 0 3.203150e+03 3.200000e+03 9.759678e+00 3.072000e+03 3.328000e+03 9.709022e+00
6 0 3.843167e+03 3.840000e+03 9.763130e+00 3.712000e+03 3.968000e+03 9.714151e+00
Image numbers correspond to offset levels of 10, 20, 30, 40, 50, 60 So using median values the factor is clearly 64 so the value to be used for an offset of 50 is 50*64=3200, this is the value I used for my synthetic biases which gave me far better results than stacking the flatbias frames as I initialy did...

[admin]

Hi,

thanks for sharing those files - I will have a look and see if I can work out anything useful.

Robin

[admin]

Hi,

OK, I think I see the problem.

It looks like the colour balance of the camera was set to auto for the flat/bias capture, which is not a good idea. It has brought the colour peaks into line for the flat frame, but for the bias I think it has shifted the whole image brightness up considerably and also left the three colour channels out of alignment with each other as it attempted to balance the colour of a basically black image. If you load the bias frame into SharpCap's 'Test Camera 1 (Deep Sky)' and look at the histogram you will see that it is *not* the sort of histogram you would expect for a nearly dark frame.

The fact that the bias level is much higher than expected will lead to incorrect flat frame correction (since the wrong portion of the signal in the flat is being calculated as due to the illumination).

Additionally, the white balance settings are different between the light frame and the dark frame (although both are on manual). That will also lead to issues with the dark correction not being quite right.

Best practice is to leave the colour balance settings on default values and on manual. If you do wish to adjust to get a more natural image colour straight off the camera then the settings need to be exactly the same between dark and light frames and between flat and flat bias / dark flat. It's probably best to keep them the same for all four to be honest.

It's difficult to be sure exactly how much better the flat correction would be with these errors corrected, but it definitely would be improved. I suspect that the synthetic bias approach helps avoid the problems caused by the odd bias frame and hence gives better results.

cheers,

Robin

[CIZIRF]

Thank you Robin,
Thought that it was a good idea to align the colors for the biases...
OK will try leaving the colour balance the same for lights, darks and flats/biases,
Thank you
Patrick