›

OSC and flats adjustment per channel

Mark Anderson
in PixInsight
Hi Adam and Forum, I'm imaging with the RASA 1100 V2 and the ASI6200 MC PRO. I am having ongoing problems with flats not calibrating the lights correctly. I saw the video you made on adding a constant to the flats to get it to calibrate correctly. I tried separating the flat to separate RGB channels and found that I had to add (with Pixelmath) 0.3 to the R, 0.16 to the B and 0.13 to the G to get them to adjust the master light separate RGB channels to an even field. This makes sense as I noticed a definate three peak histogram when capturing the flats in "Voyager". I thought the WBPP tick box "Separate CFA flat scaling factors" (on the calibration tab with the flat selected) would have adjusted for this automatically, but this doesnt seem to be happening as I'm still getting quite dark corners on the final master light.

Is there something I am missing?

Is there a way I can calibrate my flat with the required channel constants and use this adjusted flat back in WBPP given its already debayered and channel adjusted? Can you add a flat for each channel as a "master flat R" "master flat G" "master flat B" in the flats tab and have it use these when it gets to the calibrate step? Is there a method to "bayer" the flat back to a single file (or Pixelmath in the adjustments) so I can still use WBPP?

PS. WBPP wants to calibrate the lights before it has even debayered the image so this seems out of order for this methodology as I seem to want to debayer before I separate and then calibrate the separate RGB channels.

It almost feels like a totally manual process if I can't "pixel math" to get an adjusted flat.

Thanks in advance,
Love from Australia,
Mark Ando

PPS. I think I've talked myself into the adjusted flat. Anyone have any suggestions as to how to formulate adjusting (adding a constant to) the bayer array in pixelmath per channel? (RGGB)

Comments

  • BlockHeadBlockHead
    You should definitely not adjust the flat as I show as a course of action. My demonstration simply shows the problem is that flats are not being calibrated correctly due to the fact some constant value is either not being subtracted or being over-subtracted. The flats are not being calibrated correctly. 

    There are many reasons... bad flats, using different software to acquire the calibration data and light frame data, mis configuration issues... using dark frame optimization (in general or incorrectly)... 

    What you need to do in each case I mention above and others I have not mentioned is demonstrate each CANNOT be the issue because you looked at images and the values. 

    The fact that the histograms do not align does is not diagnostic. Your sensor sensitivity (and transmission through the filters) will not be the same in each color. There can also be a straight-up non-neutral color to your flat generation.

    Regarding the calibration order. No... the order is correct. The calibration of the image for OSC is done without respect to color information at all! Biases and darks do not deal with light. Flats illuminate the chip in the same way (at least that is what we are trying to do) they do when you take pictures of the sky. There is no color information... the pixels are just counting photons regardless of the wavelength and flats adjust (calibrate) this count. We *subsequently* map these pixels as having a color based on the CFA. 

    So... my suggestion is you need to patiently do a little more work with your data. Please consider the FastTrack mentality. Look at everything and write back what you find.
    (e.g. Is the value of your bias or dark frame for the flat larger than the dark frame for the light frame?)

    -the Blockhead

    Separate CFA scaling factors will not fix this issue.




  • Mark Anderson
    Hi BlockHead.

    K value with a 15 pixel probe size and a 5 decimal readout

    Bias 0.001sec = 0.00766 (might see the occasional 0.00767)

    dark for  30sec light = 0.00700 (might see the occasional 0.00701)

    dark for 60sec light = 0.00772 (might see 0.00771 or 0.00773)

    I dont have a dark for the 2sec (sky) or 6 sec(panel) flats. I was using the Bias.

    So to me, these look ok. 



    So now looking at the same spot for each below in a typical background no star or nebula position

    R: 0.0162 G:0.0394   B:0.0333 Raw light  60sec debayered
    R: 0.0065 G:0.0289 B:0.0231 Calibrated light 60sec debayered
    R: 0.0097 G:0.0105 B:0.0102  Difference

    So about 0.01 removed from the raw to get the calibrated at this position. 

    Debayered flat at this location shows
    R: 0.1654 G:0.6704 B:0.5547 with some changes to the 3rd decimal place on a light hover

    Looking now at the extracted channels of the calibrated sub. The red looks like it has calibrated correctly over the whole image even with the boosted STF applied, whereas the green and blue have a dark section on the left of the image top to bottom, more pronounced in the corners (looks radial) (bottom right looks slighly dark )( top right looks slighly light) (which is what you see in the integrated image)

    So the red is about a quarter of the signal compared to the green and blue. No real conclusions here just the observation. 

    The camera cables are in slightly different positions sometimes but I wouldnt expect to see this problem because of it. I might have to fix the cables in place to eliminate this variable.

    I will also try rotating the flats panel and my sky flats ND panel to see if there is any difference between the flats. 

    I also just did a comparison between my master flat and itself flipped horizontally. The left of the flat is definately darker that the right and has a red hue to it in the darker areas. This pattern looks similar to one I saw yesterday when I took fresh darks on the scope and covered up the objective. When I stretched the new darks, I saw a similar pattern and worked out that the ASI6200 has a little red light on the top of the camera. Now that im taking my flats with a longer exposure time. This light might be causing the problems.  

    Fresh flats coming with the light turned off or blocked with some blutak. 





  • BlockHeadBlockHead
    If you used those biases to calibrate the flats it would be a problem.
    The value of your bias is larger than your much longer darks... how is this possible?
    This was the point of my video on this matter.
    0.0007 is 45 ADU for a 16bit camera. That is quite a bit.

    On another note... in general I would not use a large probe size as you are doing... unless you understand all of the consequences. A probe size of one pixel always gives you the expected answer. Larger probe sizes can trick people sometimes because they do not realize the values of some of the pixels in the probe. In this case...for a bias, it is certainly just fine. 

    So I guess I disagree with your initial impression that things look OK.

    -the Blockhead
  • Mark Anderson
    My apologies. The k for the 30 sec was 0.00770 (with the occasional 0.00771)

    Noted on the probe size. As I was reviewing darks, I thought it would be ok.

    I have taken a fresh set of flats with the light blocked. The self compare, with horizontal flip was quite symetric compared to the old master flat. Once ive processed ill report back. 

    Thanks BlockHead.
  • Mark Anderson
    edited May 2023
    Just reporting back with the flats issue. This has now been resolved with the covering of the little red light on top of the ASI camera when im taking flats. If your using the RASA and obtaining you flats by sitting the flats panel on top of the dew shield beware of the camera power light.

    It's so obvious now that this was the problem. 

    Single 1min Sub manually calibrated is attached :)
    RUNNING CHICKEN SUB.jpg
    9576 x 6388 - 19M
Sign In or Register to comment.