O3 filter calibration issue - Forums

Topic starter

Jan 16, 2022

Since I had my QSI690wsg-8, all my filters got calibrated very well except the O3 filter, It left uneven illumination on my images with weird artifacts making image processing challenging with some targets. I don't have this issue when I swap the filters to my QSI683wsg-8. I don't know if it sensor related issue or something else ?

Thanks,

Abdullah.

AstroDarkSky

Jan 16, 2022

If I may ask, are you manually creating the flat exposure length or are you using software to estimate the exposure length (like a flat wizard type thing)? The reason I ask is that I've noticed some software wizards really make the Oiii flat exposure length far longer than any other filter. I've found that cutting down the target ADU level to 35-45% of the max ADU has helped with my Oiii filter for those type of wizards.

I'm guessing that what you are seeing is an overcalibration of the Oiii filter caused by higher than needed exposure length for the flat. So basically it adds artifacts to the final image rather than reduces them. Kind of like when overstretching any image to the extreme ends will reveal matrix filter patterns and all sorts of stuff that you normally wouldn't see in average exposure levels.

andrea tasselli

Frequent contributor

Jan 16, 2022

I'll second what #AstroDarkSky said. My take is that the OIII exposure should be no longer than it needs to give a mean ADU about equal to the worst case scenario in your light frames (i.e., the one giving you the highest average ADU).

Abduallah Asiri

Topic starter

Jan 16, 2022 · Edited: Jan 16, 2022

Thanks for your reply, for flat, I'm using SGP (flat calibration wizard). I usually set the target mean between 23000 - 22000; the recent O3 flat that I used was 1.28s . as I understand from you that I may need to tweak my settings.

AstroDarkSky

Jan 16, 2022

I have not personally used that camera, but the specs look like it is 16 bit and your ADU's sound right for calibration. This is a bit puzzling and would lead me to think it could possibly be the algorithm used for calibration or the lights are getting over calibrated so to speak so it gets more added to the final signal than intended. Like the pedestal ADU is lower for lights than the calibration's ADU, but someone else on the forum could comment beyond my experience on that factor. Not sure if flats were taken at the same time as the lights or with the same settings for bit depth.

I didn't see what software you are using to stack, but maybe try DeepSkyStacker as a quick experiment. I haven't seen situaions where that magically fixed an issue I was seeing with PixInsight stacking for example, but some people have seen differences and it is a quick/free experiment to try.

Option 2, and this sounds crazy, but try without the Oiii flats and worst case you should see just the two wide dust donuts but none of the other flat patterns added in. That would conclusively tell you that the flats are contributing instead of correcting vignetting to the final image.

Last idea, I couldn't tell if you had flat darks or biases. In general, if you take darks with the same length as your flats(1.28s as you indicated), it may work better for your particular camera than a bias+ regular darks. That's a frequently debated topic as well and can get into a pretty passionate debate on both sides so I'm just throwing some ideas to try

Good luck on experimenting and hope more people can thow some ideas your way!

andrea tasselli

Frequent contributor

Jan 17, 2022

One instance when I had results similar to those is when I mis-calibrated my flats by not removing the bias (or bias+dark) beforehand. So you might want to look into that too.

Abduallah Asiri

Topic starter

Jan 17, 2022 · Edited: Jan 17, 2022

Thank you all for your input. I'm using Pixinsight (Batch processing), and I'm getting the same result if I did flat calibration using the traditional manual workflow.
I have no experience before with (dark-flat), but I will try to see what I can get with this method. Also, I will try to reduce the exposure length to less than 1.28s; it's a good idea, it could be calculation error by (flat calibration wizard).

Pablo Gazmuri

Jan 17, 2022

I recently struggled with this myself. From what I can see, your flats are "overcalibrating" your lights (for lack of a better term). In my experience, when you calibrate flats with very weak signal (or try to stretch the lows too far), you can get this effect.

I don't completely understand it, but I suspect that when there's a dynamic range "mismatch" between the background levels of the lights and flats then it's difficult to calibrate properly. It's best corrected by collecting more Oiii data such that you are working with a stronger signal to begin with.

What I've found in trying to resolve this issue on my end:
1. Focus point matters. Make sure you have the right focus set for the right filter when taking flats.
2. Calibration of flats themselves (with dark flats or bias frames) only affects the resulting noise. They will make no difference when it comes to this kind of "overcalibration", so don't worry too much about that.
3. Sky flats work best IMO, though I find that t-shirts, cardboard sheets, led backlights, etc… all work pretty well.

You could try lengthening your exposure times or perhaps trying a different Oiii filter (I had your same experience with the Baader high speed Oiii filter myself).

Would love to hear others' thoughts on this though, I am just relaying my own experience here…

Andy Wray

Jan 17, 2022 · Edited: Jan 17, 2022

FWIW: I tend to try and get my flats at around 13500 to 15000 ADU which seems to work well. It's worth spending time taking flats at different exposures and then using image inspection tools to work out the median ADU of the flats. Separate flats at different exposures for each filter with an LED panel set at its lowest setting. Dark flats at the same exposure times as the flats plus darks and definitely no biases.

Pablo Gazmuri

Jan 18, 2022

Thinking about this further, I would second Andy's suggestion to try a lower ADU level.

I believe that what causes the dynamic range mismatch I mentioned earlier is the non-linearity of sensor response. So capturing a few photons in an empty well, vs a half-full well, will result in the same amount of increase in pixel values in a perfectly linear sensor. If the sensor is not perfectly linear (and none are really), then as photons are captured, they will result in smaller and smaller increases in pixel values as the well fills up. Going from 0% to 1% full takes far fewer photons than going from 90% to 91% full.

So the differences between the highs and lows on your flats, taken with higher ADUs than your average subexposure, will be mismatched when calibrated, if that makes sense. So lowering the ADU target should help, as long as the darkest portions of your flat image are registering well above 0.

Please let us know if Andy's suggestion improves your results…