Flat frames for OSC images

[sbenz]

Hello all,

I am shooting my images with an OSC ("one shot color") camera (ZWO ASI071MC Pro).

I have some doubts whether I am applying my flat frames correctly.
White images coming directly from the camera have a strong greenish color, as they are not white calibrated yet, so also all my flats (taken with a pure white flat field box) are quite greenish.

I have now three options:
- take the flat frames as they are (greenish) and apply them in Pixinsight. I doubt that this will lead to good pictures.
- color calibrate the flat frames before applying them. (how?)
- shoot flat frames in black&white mode (this is possible e.g. in Sharpcap)

So far I used option 3, but I do not known if this is a good idea to apply a black&white flat frame to a color image.

Anyone any ideas?
How do you process your OSC images with flat frames?

Thanks!
Stefan

[pfile]

the color cast in the flats does not matter. just make sure the weakest channel is well-exposed.

after calibration / registration / stacking of the lights, you need to do ColorCalibration anyway, and this will fix the color cast.

rob

[sbenz]

Hi Rob,

thanks for the super fast answer!

I do not know though if I understand "weakest channel is well-exposed" correctly.

I attached the histogram of my image.
Red and blue are well in the middle, but green as at the edge.
Was my exposure time too long?

Thanks
Stefan

[pfile]

yes, if that is the histogram of an unstretched, still-linear flat then the green channel is very overexposed. the exposure goal with flats is to make them as bright as possible without being so bright that the pixels have departed their range of linear behavior (meaning, that the value of the pixel is linearly proportional to the number of photons hitting it; past some number of photons most sensors start to react non-linearly to additional photons.)

by the way, is the flat coming out of your capture software debayered already? or did you debayer it first in PI? (or is it a so-called RGB Bayer image with 3 planes?)

regarding the weakest channel: because it is not unusual to have color casts in OSC flats, even with a white light source, you will invariably have one channel which is dimmer than the others. if you have to set the exposure such that one of the channels has very low signal, then your master flat may also have poor SNR in that channel. in your case if you reduce the exposure so that the green channel is mostly in the middle of the histogram, then maybe the blue and red channels will be kind of weak. if you can not change the color of the light source to compensate for this, then the alternative is to take more flat subexposures to make sure the SNR in that weak channel is acceptable.

rob

[sbenz]

Hello Rob,

yes, this is the histogram of an unstretched, still-linear flat that was debayered in Pixinsight.

Thanks for your remarks.
So, next time I have to make sure that green is in the middle, with red and blue being off the left edge. Right?
I usually shoot 20 flat frames for a master flat, so in the future maybe more.

Thanks!
Stefan

[Niall Saunders]

Hi Stefan,

All excellent advice from Rob.

One thing that I do with my OSC flats is to learn how to monitor what is going one without having to deBayer first - you already see the three 'peaks' in the histogram that correspond to the RGB channels after deBayering, but thos three peaks (more likely four in reality) will also be present in the original non-deBayered image(s). IN other words you might have one peak for Red and one peak for Blue, and maybe even a 'double' peak for the two sets of Green CFA pixels.

But, you still need to follow Rob's advice - try to get the exposure times such that the 'highest' peak isn't too high, and the 'lowest peak' isn't too low. And take quite a few Flat Frames (for which you may also need a corresponding number of FlatDarks, of the same exposure time, and may even need to apply Bias Frames as well - it all depends on your camera). By the way, if you do need Bias Frames, then they are just the Bias Frames that would be used for all your other calibration requirements.

Personally, I take my OSC Flats, and Flat Darks (not that I really need them as my dark current at -20°C is virtually zero) after I have finished imaging for the night and am packing up all of the rest of the equipment in the observatory. I aim for about 100 of each, and then finally walk away leaving the imager to take Darks (again, not that I really need them) once I head back home and fall asleep.

The next day, all being well, I should then have all the data I need, and I transfer this to my big i7 12-core number-cruncher PC before deleting it and looking for property for sale at around 5000m above sea level

[sbenz]

Hello Rob, Niall,

thank you very much for your advices!

One more thing: I am not really sure whether my light source is ideal (I am using a Geoptik flat field box, illuminated by LED).
What are you using as light source for your flat frames?

Thanks
Stefan

[Niall Saunders]

Hi Stefan,

I have a choice of two - one of which has been dismantled pending a full rebuild.

The one actually in use at present is nothing more than an A3 sheet of white foam-board, hung on two screws on an observatory wall about 1.5m away from the end of my 10"/250mm RC OTA. This sheet is illuminated by a 'Warm White' 11W low-energy spiral CFL lamp in a standard E22 or BC27 socket).

The colour of the illumination is not ideal - in terms of 'pure white', but it works quite well as it seems to have a lower Green spectral output and a higher Blue Spectral output. This leaves my four 'raw' peaks quite closely grouped, making it easier to select an adequate exposure time to give ~50% ADU levels.

The second liht source is actually a back-illuminated sheets of matte-finish translucent acetate. The source of illumination was a circular array of 100 RGB leds, and the spectral content of the RGB peaks could be controlled using three potentiometers, allowing the peaks to be prettly closely matched in the raw flats.

However, I want to experiment with the same 100 leds mounted in a ring of foam-board that could be slipped over the end of the OTA, allowing the other flat board to be illuminated from the OTA itself (which I should have explained, can be easily pointed perpendicularly to the card - without affecting any previous pointing model that had been established).

I hope thiis helps.

[STEVE333]

Taking flats with OSC's can be tricky as you have found.

I've had good success using a thin colored cellophane film between the light box and the telescope. In my case the Red channel is much weaker than the Green and Blue channels (because of a filter I'm using), so, I use a slightly red (almost pink) cellophane and it brings the G/B channels histogram peaks much closer to the Red so that all three channels can have good signal.

For you situation where the Green is too strong I would think a Purplish cellophane would help to reduce the Green thus bringing the Green histogram closer to the  Red and Blue channels.

Just a thought. Maybe it will work for you too.

Steve

[Tertsi]

Hi,

I may have similar flat problem with my new ASI071MC-pro. Image histogram looks super weird after calibration. (My previous camera was unmodified Canon EOS 500D and I'm pretty familiar with PixInsight OSC workflow. I only use flats and bias frames in calibration).

First image is single light frame and I have manually debayered in PI (point is to show image histogram) and histogram

Second image is single flat frame and I have manually debayered in PI and histogram. Suggested exposure time  was obtained with AstroPhotography Tool's CDD flats aid. My target ADU was 30000 and exposure time 0.30781s. I have also tested various flat frames with different exposure times.

Third image is single light frame calibrated with master flat and master bias and debayered and histogram is shown as well. This histogram is weird!

If I stack images without calibration, I get nice histogram, but clearly calibration does something funny now.

This results in image without color in the end, which is the main problem.

Does anyone has idea where seems to be the problem or am I doing something wrong? Any help is appreciated!

Sincerely,
Tertsi

[Tertsi]

Addition to my previous port; I used same frames and here is the final image after calibration (flats and bias) and image stack with Deepskystacker 4.2.2. I used STF to show Autosave.tif. Image and histogram looks totally fine.

This suggests that there is a problem in PixInsight calibration, but what is it?

Edit: second image is calibrated and stacked in PixInsight (version 01.08.06.1457 Ripley) with same input files (light, bias and flats) as with Deedskystacker. Difference is huge between DSS and PI. 

I guess my post is under wrong topic now, because this might be a problem in PI calibration(?).

[pfile]

something else is weird here; two of the channels appear to have exactly the same signal - there are only 2 curves visible in your histogram.

btw you should set the histogram tool to something greater than 8-bit resolution...

does this camera have an electronic shutter? what is the light source for the flat? you say you experimented with different durations, but did you try 3-5 seconds rather than sub-1 second values?

rob

[Tertsi]

Hi Rob,

Many thanks for replying!

ASI071MC Pro has electronic shutter. My light source is this:

https://www.teleskop-express.de/shop/product_info.php/info/p3867_Illuminated-flat-field-foil--220mm-active-diameter--mounted-in-frame.html

I have used it for years with my DSLR (Canon EOS 500D).

I did experiment with 0.01s, 0.05s, 0.1s, 0.15s, 0.2s and 0.25s subs. Not over 0.3s subs. So I have not experimented with 3-4 second subs.

Edit: Yes, two channels are exactly same after calibration! But why DSS calibration and stacking works just fine? It should work in PI too, I guess.

Sincerely,
Tertsi

[Tertsi]

Hi Rob,

I also noticed that after PI calibration, calibrated image is horizontally flipped. It's like a mirror version of the original image. That should not happen in calibration, right?

Sincerely,
Tero

[Tertsi]

Hi,

I think I solved the problem! Flipped image was clue to this mystery! I checked "Up-Bottom-FITS" option ON from the BatchPreProcessing script and now calibrated images are just fine and histogram looks as expected. I attach final stack image from PI with Up-Bottom-FITS ON!

Sincerely,
Tero