Dark Flats with a DSLR

[Sarek]

How important do people think dark flats are?  I have to admit I hadn't twigged they were different than darks. 

[AMcD]

As I understand it, it is important to subtract a flat dark frame (i.e., a dark frame that is the same length, temperature, gain etc. as your flat frame) from the flat frame before the flat frame is applied to calibrate the light frame.  Otherwise, there is a risk that the flat frame will not work.  Flat darks appear to be preferred for this purpose to bias frames for CMOS cameras, although there are those who have equal success calibrating their flat frames with bias frames on a CMOS camera. 

As with all these things, some people produce amazing pictures without using the 'standard' procedure and some produce amazing pictures using their own, idiosyncratic, methodologies.   I think the best approach is what works for the individual, which can best be established by experimentation.

[Sarek]

7 minutes ago, AMcD said:

As I understand it, it is important to subtract a flat dark frame (i.e., a dark frame that is the same length, temperature, gain etc. as your flat frame) from the flat frame before the flat frame is applied to calibrate the light frame.  Otherwise, there is a risk that the flat frame will not work.  Flat darks appear to be preferred for this purpose to bias frames for CMOS cameras, although there are those who have equal success calibrating their flat frames with bias frames on a CMOS camera. 

As with all these things, some people produce amazing pictures without using the 'standard' procedure and some produce amazing pictures using their own, idiosyncratic, methodologies.   I think the best approach is what works for the individual, which can best be established by experimentation.

Yes that makes sense. They seem relatively strightforward to take so I'll probably add them to my workflow next time. 

[ollypenrice]

The problem with darks-for-flats with a DSLR is that you can't ensure they will be temperature-matched with the flats. If they're not, they are likely to do more harm than good.

I would try uncalibrated flats first, to see if you have a problem.  How important do I think darks-for-flats are? Well, with CCDs I used as bias as a dark-for-flats and that worked. Except when it didn't - which was for a long time with one particular scope-camera. In the case of this problem rig, nothing worked and, believe me, I tried everything. Flats over-corrected.

So just see if uncalibrated flats work. If they do, you're good to go. If they don't, the bad news is that darks-for-flats may not solve the problem!

Olly

[alacant]

Hi

Dark frames of any type on a DSLR are bad news. Forget. Don't use... etc.

DSLRs employ sizeable offsets. To prevent the flat frames over-correcting, this should must be removed.

Don't know what camera you have but on e.g. recent eos models this is easily achieved by simply subtracting a constant bias of 2048. Don't forget to do this for the light frames too. No need to shoot separate bias frames.

No nonsense, open-source method for this? Siril.

Cheers and HTH

Edited December 12, 2022 by alacant
spell

[Sarek]

On 11/12/2022 at 19:52, alacant said:

Hi

Dark frames of any type on a DSLR are bad news. Forget. Don't use... etc.

DSLRs employ sizeable offsets. To prevent the flat frames over-correcting, this should must be removed.

Don't know what camera you have but on e.g. recent eos models this is easily achieved by simply subtracting a constant bias of 2048. Don't forget to do this for the light frames too. No need to shoot separate bias frames.

No nonsense, open-source method for this? Siril.

Cheers and HTH

I have to admit you've lost me here....

I have a Canon 700D astro modified camera.  I've not yet tried Siril so I don't know how to subtract a constant bias?  I had shot seperate bias frames. 

I shot 20 dark frames after all my lights had been taken . Everything I've read says to do this and that darks are an essential step in removing noise?

 

 

[alacant]

1 hour ago, Sarek said:

I shot 20 dark frames

Apart from the eos circuitry which introduces its own artefacts when combined with conventional dark frames, the impossibility of matching the varying temperature of your dark frames with your equally temperature variant light frames introduces more noise rather than help eliminate it.

1 hour ago, Sarek said:

subtract a constant bias

700d? Same here, so simply enter:

=2048

in the field where you would normally enter the location of your master bias file.

1 hour ago, Sarek said:

Everything I've read

...  about dark frames on a 700d? Really? I'd guess the authors had never been near a DSLR. In this game theory, whilst maybe interesting to some, is of little use to those wishing to retain what remains of their hair.

Cheers and HTH

Edited December 12, 2022 by alacant

[Ratlet]

I use a t3i (600D) for widefield stuff and have taken zero dark frames of any description. Just use =2048 typed into siril for the bias and just apply flats.  I do either every 5 or so subs.

The beauty with siril is you can very quickly and easily stack with and without darks.  Image quality was improved not using darks and using a constant bias.

 

[Sarek]

11 hours ago, alacant said:

Apart from the eos circuitry which introduces its own artefacts when combined with conventional dark frames, the impossibility of matching the varying temperature of your dark frames with your equally temperature variant light frames introduces more noise rather than help eliminate it.

700d? Same here, so simply enter:

=2048

in the field where you would normally enter the location of your master bias file.

...  about dark frames on a 700d? Really? I'd guess the authors had never been near a DSLR. In this game theory, whilst maybe interesting to some, is of little use to those wishing to retain what remains of their hair.

Cheers and HTH

 

11 hours ago, Ratlet said:

I use a t3i (600D) for widefield stuff and have taken zero dark frames of any description. Just use =2048 typed into siril for the bias and just apply flats.  I do either every 5 or so subs.

The beauty with siril is you can very quickly and easily stack with and without darks.  Image quality was improved not using darks and using a constant bias.

 

Thanks both - well that's something else to think about. The beauty of having the data stored is being able to experiment with and without.

I did read somewhere that taking darks immediaitely after each light frame would reduce temperature induced variations but I like the idea of doing away with them all togerher!

[WolfieGlos]

Personally, I found dark frames were helping me:

https://www.cloudynights.com/topic/830048-opinions-on-not-taking-dark-frames/?p=12137897

But that thread showed up that I was actually suffering from walking noise by not dithering, something which I am still testing and trying out with APT. I've read in many places that dark frame temperatures can be +/- a few degrees, but clearly they are best when perfectly matched.

I'm certainly no expert, but I find the best thing is to take some darks (i.e. when clouds roll half way through your session...) and run two stacks; one with and one without.

[dph1nm]

Probably depends how long your flats are! I only ever take flats in the region of 1 sec or so.  I would be very surprised if a flat-dark that length on a DSLR is in any significant way distinguishable from a 'bias' (basically a 1/4000 sec exposure on my Canon). You should subtract something though, be it flat-dark, bias or constant, otherwise the flats won't work correctly.

NigelM

[TiffsAndAstro]

On 12/12/2022 at 21:13, alacant said:

Apart from the eos circuitry which introduces its own artefacts when combined with conventional dark frames, the impossibility of matching the varying temperature of your dark frames with your equally temperature variant light frames introduces more noise rather than help eliminate it.

700d? Same here, so simply enter:

=2048

in the field where you would normally enter the location of your master bias file.

...  about dark frames on a 700d? Really? I'd guess the authors had never been near a DSLR. In this game theory, whilst maybe interesting to some, is of little use to those wishing to retain what remains of their hair.

Cheers and HTH

bit late to the party, sorry. also new to siril and basically use the pre provided scripts. how would i try this out via the scripts? copy an existing one, rename it and replace the bit of the script that shows my master bias location as =2048 rather than an actual location?

[alacant]

46 minutes ago, TiffsAndAstro said:

try this out via the scripts

JTOL

By the time you have worked out how, tested the new script... and it still doesn't work... Perhaps better to use Siril manually. This puts you in control rather than being constrained by a script. Quicker too.

Cheers 

Edited April 10 by alacant

[TiffsAndAstro]

16 minutes ago, alacant said:

JTOL

By the time you have worked out how, tested the new script... and it still doesn't work... Perhaps better to use Siril manually. This puts you in control rather than being constrained by a script. Quicker too.

Cheers 

Jtol? You're totally correct about I should learn to use siril rather than just how to click a few buttons to get a result. That will take me some time and more useful when I have decent data to plug onto it.

I'm hoping the "dialing it in" stage will end at some point.

[alacant]

44 minutes ago, TiffsAndAstro said:

will take me some time

It's much quicker than you may think. Be cautious though. A search for help will throw up all manner of material of varying usefulness...

Perhaps go with the official start guide. If you've a decent computer, it should take around 30 minutes. Less on subsequent runs. A multi core Ryzen with 32Gb ddr5 under Linux is the most instantaneous I've tried so far.

Cheers 

Edited April 10 by alacant

[TiffsAndAstro]

3 minutes ago, alacant said:

It's much quicker than you may think. Be cautious though. A search for help will throw up all manner of material of varying usefulness...

Perhaps go with the official start guide. If you've a decent computer, it should take around 30 minutes. Less on subsequent runs. A multi core Ryzen under Linux works particularly well.

Cheers 

Er, why would my pc performance make a difference on reading a guide?

Also what is jtol ?

[Ratlet]

43 minutes ago, TiffsAndAstro said:

Er, why would my pc performance make a difference on reading a guide?

Also what is jtol ?

I think he means it'll take about 30 minutes to process your images the first time and once you get the hang of it, and with a decent computer you can do it manually pretty quickly.

I tried figuring out how to run scripts, but in all honesty doing the manual siril guide linked is easy and gets updated.  It also tells you why you are doing the steps as well as alternatives.  You get really fine control so you can mix and match different processing, stack sizes etc. and get the best out of your data.

[TiffsAndAstro]

4 minutes ago, Ratlet said:

I think he means it'll take about 30 minutes to process your images the first time and once you get the hang of it, and with a decent computer you can do it manually pretty quickly.

I tried figuring out how to run scripts, but in all honesty doing the manual siril guide linked is easy and gets updated.  It also tells you why you are doing the steps as well as alternatives.  You get really fine control so you can mix and match different processing, stack sizes etc. and get the best out of your data.

I'm reading through it as I get a chance but will take a while. My pc is ok but getting old