DSLR imaging - Do you really need Dark frames?

[Alien 13]

Isn't it immediately after the light? I would have thought that this would mean the sensor is warmer for the dark.

Sent from my iPad using Tapatalk

I dont know if this would be the case if you shot 20 subs without any delays i would assume the temp would be quite close.

Alan

[D4N]

True enough but isn't it wise to give a little cooldown time between subs? I would have thought using the time taken for the dark may even be more effective than just the dark itself in some cases.

Sent from my iPad using Tapatalk

[Alien 13]

Good point it might be worth running a test between them as i normaly never use any delays.

Alan

[freiform]

Looking at my stacks, I also don't have the impression that darks help the SNR. They do eliminate hot pixels though. Is there another way to get rid of those when not using dark frames?

Having said that, I'm using an entry level Nikon D3100 and as I'm not guiding, I'm taking lights of 180s at the most.   

[Psychobilly]

You can remove them in the Post processing using the cosmetic tab in DSS but you can sometimes end up with "holes" in the middle of stars as it can remove the centre pixel at times...

I used to get a fractional pixel drift between images which meant that over a few frames the hot and cold pixels would have moved relative to the stars in the images and by having enough frames and using an "outlier" rejection stacking method - like Kappa-Sigma - they would be rejected and repalced during stacking... Dithering (makiing small movements ) the mount by a few pixels between subs has the same effect...

Peter...

[Astrosurf]

This is an interesting thread. I tried processing with and without darks on one image and the non-darks one was awful. But lately I've been getting bad noise in my images and wondered if the darks were to blame. I'll try without darks and see what happens.

I don't really understand that DSS tutorial on darks!

Please can someone explain how to do dithering?

Alexxx

[mikehab]

Fascinating thread :-)   Like most other people above - I 'prefer' the second image (no darks) ... but it's a very close run thing.   I would also be interested to see an equivalent comparrison between with and without flats and /or bias - i.e. an image stacking just the lights only vs one with lights + bias, vs one with lights + flats + bias.   I'm guessing there would be a much more noticable difference in those cases?

Mike

[Bizibilder]

Flats most definitely have a huge effect! Mainly the removal of vignetting, but they also get rid of the odd dust bunny and other "funnies" in the optical train that find their way to the sensor.

[Physicist13]

It depends on what is dominating your noise. Darks aim to reduce the spatially varying random noise due mainly to the sensors temperature (this noise is a strong function of the temperature). It has to therefore be a very good representation of this noise at the temperature you took the light frames. Some software will scale the darks to do this ( and it can be done very well!). If the scaling is wrong (typically becase the difference in temperature betwwen darks and lights is large) then dark subtracting can actually increase the noise level. Also if the noise is read-noise limited or image noise limited (typically LP) then dark subtraction may have little effect. Its good to know where the noise comes from in your images!

P

[michaelmorris]

To me, the second one looks a llittle less noisy.

[lensman57]

Hi,

I attach two files, the full size is a single Dark frame from a modded Canon 1100d @iso 1600, sensor temp 25c. The crop is from the centre of the file @100%.

The file has an STF stretch applied to it in PI. You can make your own mind up wether a DSLR needs darks or not, and just as amatter of interest I also attach a 100% crop of a 900s master dark frame from an Atik 428EX one shot colour camera @ -5C for comparison. Hope this helps.

Regards,

A.G

[tingting44]

here is another comparison

the iris nebula, both versions were processed differently and only a tiny of ammount of data.... but you get the idea, darks do make a nice difference (sensor temp for darks and lights were on average around 17oC)

4x600sec subs with no darks (100% crop) as you can see lots of RGB blobs

Iris stack - PS - work in progross - (RGB blobs!) PSD - 100% crop by tingting44, on Flickr

 

4x600sec subs with darks (100% crop)

stack with darks & bias (800) - PS - tiff - 100%crop by tingting44, on Flickr

 

again both above images are from the same data and stack

[Astrosurf]

I did do a stack and stretch of a DSO with and without darks and the one without was horrific. I do get confused!

[wxsatuser]

Darks from a 60Da at 23°C

Would you want that lot in your images?

Best to try and remove most of it.

600secs

300secs

180secs

[Psychobilly]

There were a few fairly long threads from back around 2009-2011.  I will try and dig them out...

Just posting them "reference" for anyone who wants to have a read through them  rather than to resurrect them...

http://stargazerslounge.com/topic/97130-dslr-horizontal-banding/page-1?hl=+dss

and a short one with sample 15 and 30 min subs from my 1000D

http://stargazerslounge.com/topic/49701-how-noisy-are-900s-and-1800s-subs-at-iso800/

Grab everything you need for "full" calibration and have a play ... that's what I used to do with my cameras when I got them...

Peter...

[Knight of Clear Skies]

Interesting thread. In theory, darks should certainly help. In practise, if the noise patterns don't match the light well then they won't. There are all sorts of things that might have an affect on the outcome. The more subs you are stacking the more the noise will average out anyway, but if you're only stacking a few subs darks might matter more. The more the data is stretched the more critical it is to eliminate noise. If it's a choice between taking more lights and taking darks I'd lean towards going for lights, but only some really controlled testing would resolve this.

I'm now officially 'off darks', at least for the time being until I have some more important issues sorted out (more consistent tracking, LP filter and better focus). I'll revisit them another time, and try stacking the same subs with and without darks for a side-by-side comparison.

[Psychobilly]

My approach has always been to try differnt things out and see what happens ....

One sure fire way to improve the SNR is to get more signal ... and if you can use dithering and statistical stacking methods as well as a result  it all helps...

Peter...

[ollypenrice]

It would be madness to swap half your imaging time for dark-taking.

I don't think the stretch/cutback is hard enough in the sample images to allow darks to show their worth but, even with set point cooling and CCD, I no longer use normal darks. They can indeed introduce problems of their own. I use a master bias and a bad pixel map. AstroArt supports this approach and gives instructions on how to make the Defect Map.

If Bizibuilder's PA is imperfect (and perfection is hard to acheive...) there is almost certainly a little natural dither creeping in due to rotation. WIth a nice deep stack like this the effect will be worth having, surely?

Olly