CMOS Dark library rebuilds - when?

7 replies187 views
Linwood Ferguson avatar
I have a ASI 6200MM Pro, which has quite low noise to begin with. Quite a few people have told me not to bother with dark subtraction because of the low noise, but I am doing it anyway just to be more safe, and it doesn't seem likely to do harm.

However….

My dark library is about a year old.  It's made from 50 subs each, so takes several days to build (I do narrow band so build at 300s and 600s as well as shorter, and all at two temps). 

People (ironically sometimes the same ones who say don't bother at all) also say you need to rebuild it periodically.  When?  "When it no longer corrects well."

But honestly it's tough by eye to see that it corrects at all, so telling if it's correcting slightly less well is beyond my eye. 

So… 

How do you decide when it is time to rebuild a dark library?   

Every X months? 

Try to count hotter pixels and see if they are growing in number or some statistically technique? 

Never?

I may of course be over-thinking the question, since it's rainy season and rare I get to image, the mind wanders

Linwood
Engaging
Bruce Donzanti avatar
I would take 2 sets per year, one for summer at -5C and one for fall/winter/spring at -10C.  Now with the ASI2600 that I use and your ASI6200, the read noise is so low that we can get away from not using them but I still do them as a very accurate and easy way to correct hot pixels since they are simple to make.  Also, due to the lower read noise, I only make one set a year at -5C as I find no difference between -5C and -10C with these cameras.  Why once a year?  Just a habit to do so and I have never identified a reason to make them more frequently.
Helpful
Kevin Morefield avatar
I use the same sensor (QHY version) and I would first say you should use darks.  Two reasons; 1)  there is certainly dark noise and while the hot pixels will be rejected in the stacking, this will reduce your SNR vs correcting and keeping the pixels.  2) if you look closely at the subs after aligning and resampling you often see a ring of dark pixels around the hot pixels.  So you have multiplied the pixels lost to rejection by about 8x.  

To your actual question, twice a year should work.  And as Bruce said it’s a good opportunity to adjust temps for Summer if needed.  

I personally use 20 darks and that seems to work fine.  Maybe reducing the number of dark subs would lessen the pain of re-shooting?

I also run cosmetic correction on the subs post calibration to fix any remaining hot pixels. I use the master dark and a 3 sigma threshold.  I found the auto correction actually makes things worse.
Helpful Respectful Engaging Supportive
Linwood Ferguson avatar
Thanks @Bruce Donzanti and @Kevin Morefield .

To Kevin's "twice" and opportunity for two temperatures, well I built it initially at two temperatures.   Even in the summer some nights I can get to -10c, some nights I can't, so I want a 0 and -10 all summer.

I guess the question though comes off of Bruce's last sentence "never identified a reason to make them more frequently".... 

How do you know to re-make them that frequently, as opposed to every 2 years, or never?

Despite bringing it up, I am not suggesting I would not use Darks, just that it is hard enough to "see" their impact, to see if their impact begins to degrade is even harder.

What's the rationale for once a year as opposed to twice, four times, or every four years?   Or more to the point, is there some test I can run and look and say "oh, yes, it's time". 

(And yes I'm being a bit lazy -- but the combination of not being able to reach -10c at ambient indoor with keeping light leaks out is tedious.)
Kevin Morefield avatar
Well, the darks are never perfect and you will have remaining hot pixels even with brand new darks. So it’s kind of whenever the calibrated subs have more remnant hot pixels than you can stomach.
Linwood Ferguson avatar
So I decided to just go ahead and start building, rearranged things, got some cables to put the camera in the fridge, discovered there's light inside a fridge (the initial lens cover was a bit translucent) and fixed that, and did one set of darks.

There's a LOT of new hot pixels.  My one year old dark for 120s at gain zero had 13 hit pixels, the new one has 146 (how you count is probably a bit arbitrary but I counted these the same way). 

The mean didn't vary much (501.346 up to 502.371) and the variation wasn't to different (0.662 to 0.730).  Really all the statistics went up very slightly, but the big difference was hot pixels.  And if I do a preview stretch (this is of the master dark) they are quite visible, little spots all over the place.

So… I think definitely worthwhile.

I guess that fridge stays tied up a day or two.
Helpful Engaging
André Bremer avatar
Kevin Morefield:
Well, the darks are never perfect and you will have remaining hot pixels even with brand new darks.

Indeed. I do keep darks in my back pocket for the times I forgot to dither, but even then they don’t catch everything and I still end up needing to do cosmetic correction - regardless how recent they are.

I use a both ASI2600MC and MM and found with dithering and a healthy amount of subs, there is no discernible difference with or without darks. The hot pixel junk simply averages/rejects out. It’s way less of a hassle.
Helpful Concise
Stjepan Prugovečki avatar
It is always good to go to the basics. Both hot and cold (dead) pixels are property of the particular sensor chip itself .  It comes with some pixels that are hot or cold always. Others become hot or cold under certain condition (temperature, number of incoming photons, whatever…smile .  One thing is for sure, the number of always hot or cold pixels will just be equal or higher from the day "0" as the time goes . All together, dark library will never correct it 100%.  So there is no "good" frequency of rebuilding it . Perhaps the most opportunistic approach is to do it when correction is no longer acceptable. With modern sensors , dithering/stacking takes good care about it anyway.
Helpful