Dark frames - will this work?

[Astro Noodles]

I am using an uncooled DSLR, so it is impossible for me to get dark frames at exactly the same temperature as ambient (unless I want to spend 50% of my imaging time taking photos with the lens cap on - which I don't).

If I take dark frames which are always at a higher temperature than ambient, what would be the drawback in doing this?

[scotty38]

Can't answer the question but brilliant darks and no lights, well you get the picture lol....

I'd be tempted to take whatever lights you can while you have chance and then tomorrow, if it's not clear, set up at about the same time and take your darks. Unless we have a heatwave I'd say that'll be far better than nothing.

[Richard_]

Are you capturing images from the DSLR using an intervalometer or from something like a computer/laptop/ASI Air?

I attached my Canon 600D to the ASI Air and saved images to a USB stick, but I could also include sensor temperature in the filename. That way, I could capture a bunch of darks over different nights and group them by temperature (+/-1C), ISO and exposure time, and build a small dark library. I'd do the same for my lights and then select the set of darks which bent matched the average temperature of my lights. This is seasonal of course due to temperature changes. 

I never did any testing to see if this approach was better/worse than shooting darks each night and using those instead, but it seemed the best way to control it. 

[barbulo]

I don’t think they help at all since the temperature of the lights is not constant and the temperature of the darks cannot match by any means. I don’t use them. 
When one enjoys 1 or 2 clear nights per month one can’t afford taking darks. 

[Astro Noodles]

2 minutes ago, barbulo said:

I don’t think they help at all since the temperature of the lights is not constant and the temperature of the darks cannot match by any means. I don’t use them. 
When one enjoys 1 or 2 clear nights per month one can’t afford taking darks. 

I've just read on a different post that Darks may not be really necessary  for Canon DSLRs. @vlaiv @alacant

 

 

[scotty38]

4 minutes ago, barbulo said:

I don’t think they help at all since the temperature of the lights is not constant and the temperature of the darks cannot match by any means. I don’t use them. 
When one enjoys 1 or 2 clear nights per month one can’t afford taking darks. 

Yes that was my point about setting up again at the same time (if it's not clear), the idea being the temperature range could hopefully more or less match what was done for the lights, well as good as can be expected.

[The Lazy Astronomer]

I'll caveat this by saying I've never done astrophotography with a DSLR, and am very, very far from being an expert, but generally I see the advice to not bother with darks with a DSLR, and just use bias frames, and dithering between lights.

[Astro Noodles]

1 minute ago, The Lazy Astronomer said:

I'll caveat this by saying I've never done astrophotography with a DSLR, and am very, very far from being an expert, but generally I see the advice to not bother with darks with a DSLR, and just use bias frames, and dithering between lights.

Dithering would require guiding. I'm not at that stage yet. 🙂

[vlaiv]

3 minutes ago, Astro Noodles said:

I've just read on a different post that Darks may not be really necessary  for Canon DSLRs. @vlaiv @alacant

 

 

For a period of time I advised that one tries dark optimization for cameras that don't have set point cooling and have bias that works.

Recently I experimented with Canon darks and bias files and I noticed that they can't be made to work with above algorithm for some reason. In fact they behave rather oddly. I suspected that Canon does something "behind the curtains" but I never bothered to dig deeper into that.

It turns out that Canon does their own "dark calibration".

On one hand - that is bad, as it changes the data without letting you know in which way. On the other hand - it does it (according to some sources on the net) - in rather good way as far as dark current is concerned. You don't have to worry about temperature mismatch because this dark current offset removal is done at exact same temperature camera was on when taking light sub.

However this method does not remove bias signal, so there is benefit in using master bias. There is also one additional down side to this method Canon uses - it introduces horizontal banding in each sub.

because each row is "dark calibrated" on its own.

In any case - for DSLR, don't use dark optimization, don't use darks, use bias as both darks and flat darks and dither between subs

[david_taurus83]

2 minutes ago, Astro Noodles said:

Dithering would require guiding. I'm not at that stage yet. 🙂

If you use a goto mount and use APT then you don't need guiding to dither. APT will dither for you.

[vlaiv]

3 minutes ago, Astro Noodles said:

Dithering would require guiding. I'm not at that stage yet. 🙂

Not necessarily. If you don't guide - it is likely that you'll have "natural" dither. As long as you don't have perfect mount and it moves pixel-two every few subs - it's effectively dithering.

[barbulo]

10 minutes ago, The Lazy Astronomer said:

and just use bias frames, and dithering between lights.

Don’t forget flat frames. 

[happy-kat]

I use the same camera and get better results with less noise if I use flats and dark flats only when stacking using DSS, median on the dark flats works better than average for me. YMMV

[alacant]

22 hours ago, Astro Noodles said:

Darks may not be really necessary  for Canon DSLRs. 

Hi

Hands on tells us that...

For any DSLR we've tried, (Canon, Nikon, Panasonic and Sony come to mind) dark frames of any sort have served only to make processing the image more difficult.

If you really must, the dark optimisation routine in the latest version of Siril is the only way we've come close. But even then...

Cheers 

[ONIKKINEN]

I was curious about this conversation about darks being not helpful with Canon DSLRs, and i admit i never thought about it or measured my own darks when i shot with a 550D. I took some test darks and found that a 1s dark has a median ADU of 2049 while a 600s one has a median ADU of 2047? So it seems Canon is doing something to make all dark signal be somewhere around 2048 ADUs regardless of exposure time.

[david_taurus83]

11 minutes ago, ONIKKINEN said:

I was curious about this conversation about darks being not helpful with Canon DSLRs, and i admit i never thought about it or measured my own darks when i shot with a 550D. I took some test darks and found that a 1s dark has a median ADU of 2049 while a 600s one has a median ADU of 2047? So it seems Canon is doing something to make all dark signal be somewhere around 2048 ADUs regardless of exposure time.

The bias is 2047. The Canons have very low thermal noise in general but if imaging in warm weather it does become more apparent. 

[alacant]

Haven't tried with a 6d but with modern sensors, there isn't need to take bias or dark flat frames. Subtraction of the median works fine. E.g. on a 700d, this value is 2048 (round to the nearest power of 2). Use this value on your light and flat frames and away you go.

In fact this method is recommended and implemented in the latest version of Siril.

Worth a try?

Cheers

Edited January 6, 2022 by alacant

[david_taurus83]

13 minutes ago, alacant said:

Haven't tried with a 6d but with modern sensors, there isn't need to take bias or dark flat frames. Subtraction of the median works fine. E.g. on a 700d, this value is 2048 (round to the nearest power of 2). Use this value on your light and flat frames and away you go.

In fact this method is recommended and implemented in the latest version of Siril.

Worth a try?

Cheers

That suggestion started a bit of a debate on another thread!

[vlaiv]

1 hour ago, ONIKKINEN said:

I was curious about this conversation about darks being not helpful with Canon DSLRs, and i admit i never thought about it or measured my own darks when i shot with a 550D. I took some test darks and found that a 1s dark has a median ADU of 2049 while a 600s one has a median ADU of 2047? So it seems Canon is doing something to make all dark signal be somewhere around 2048 ADUs regardless of exposure time.

I found the same results on my Canon 750d and after a bit of research - it turn out that Canon does "internal" dark subtraction of sorts and then adds 2048 constant offset.

Internal dark subtraction, as far as I understood, works like this. There are some 30 or so pixels in each row on the edge of sensor that are masked off - covered with something, so that they can't register light but are "exposed" together with rest of the sensor and in fact gather dark current during regular exposure.

For each row these are measured and their mean value is then subtracted from all other pixels in the row (ones that have been properly exposed). This sort of removes dark current per row.

Good thing about it is that it removes temperature dependence. Bad thing about it is that it does not remove bias variation signal and it introduces small level "per row" noise in each sub. Each of those pixels has some read noise and "stacking" them reduces that read noise about x5 times, so it is very small amount of noise - but it is embedded equally on whole row - that creates very visually recognizable pattern.

Here is animation of several dark subs I took with my DSLR - stretched:

These have been scaled and stretched of course, but animation shows how there are some features that are the same between each dark sub - that is bias signal in each that is not removed, and one interesting feature is also visible - that is horizontal banding that is result of this internal dark calibration - each sub has different per line random offset that is visible as dark or light band for each row.

Here is master dark created out of these darks. This is actually master bias signal as it does not contain dark current - that has been removed by internal calibration.

As comparison - this is stretched master bias for same camera:

stretched a bit harder as I did not pay attention to level of stretch, but you can see that two are in fact the same - they contain bias signal.

From this, we can see that often given advice is very sound and is in fact best calibration procedure for Canon DSLR-s (maybe others as well):

1. Don't use darks or attempt dark scaling - it won't work as dark current has already been removed

2. Calibrate with master bias as both darks and flat darks (don't just remove constant of 2048 if you want best results - since there is actual bias signal / signature that you want removed instead showing in your image)

3. Dither to additionally spread that horizontal line issue from internal calibration

[Swillis]

@vlaiv any idea if this applies to any Canon DSLR? I have a 350d, not sure if that's considered a bit 'vintage' 😅

Thanks

[vlaiv]

28 minutes ago, Swillis said:

@vlaiv any idea if this applies to any Canon DSLR? I have a 350d, not sure if that's considered a bit 'vintage' 😅

Thanks

I think it does apply, but you can check it out.

Take one bias sub and one dark, say 60s long. Measure their mean ADU value and if they are the same - say around 2048, or it can even happen that bias has mean ADU value slightly higher than dark - then it probably applies.

Dark needs to have higher mean ADU value because dark current accumulates and adds to pixel values. If that is not the case - then dark current has probably been removed already.

[Swillis]

@vlaiv thanks

Not sure if I'm looking at the right thing. I opened the files in SIRIL but I'm not familiar with it yet. If I go to image information>statistics I get 257 mean for bias (2403 max) and 257.3 mean for dark (4095 max). Not sure if these are ADUs or not?

[vlaiv]

On 07/01/2022 at 13:37, Swillis said:

@vlaiv thanks

Not sure if I'm looking at the right thing. I opened the files in SIRIL but I'm not familiar with it yet. If I go to image information>statistics I get 257 mean for bias (2403 max) and 257.3 mean for dark (4095 max). Not sure if these are ADUs or not?

Yes, those should be ADU values and it looks like you have same values of 257 for both bias and dark (with small variation due to noise - that is ok).

It looks like your camera behaves the same - hence bias only and dither.

[Swillis]

Thanks @vlaiv, and sorry @Astro Noodles for hijacking the thread