New camera=new darks library?

[BrendanC]

Hi all,

I just got my camera repaired (actually, replaced), and even though it's the same model (EOS1000D, modded), it's obviously a different sensor, different thermal characteristics etc.

Am I right in thinking this means I need to rebuild my darks library? I think it does, I'm just wondering whether there's any way I can around it!

Thanks, Brendan

[andrew s]

Yes best to. Regards Andrew 

[Phillyo]

I would say yes, but also I would say no? Unless your DSLR is cooled then making a dark library may only make things worse. Darks need to be captured at exactly the same temperature (and exposure time) as the light frames. Cooled astro cameras can do this but DSLRs can't. I personally wouldn't bother adding darks to the calibration for a DSLR. Just use Flats and BIAS frames.

Phil

[ollypenrice]

I'm not a DSLR imager but I've read plenty of good DSLR imagers who say not to bother with darks but to use a master bias as a dark and a large dither (12 pixels or so) and that makes sense to me since uncooled darks may be all over the place on real temperature.

In any event, noise is specific to the sensor.

Olly

Edited October 26, 2020 by ollypenrice

[BrendanC]

Ah, here we go again! I've done a ton of research on calibration files and what I've found is, there is no consensus! So here's a perfect example: I have a 'yes', a 'no', and a 'yes and no'! The perfect balanced sample...!

(Don't want to give the impression I'm ungrateful though, thanks for the replies all).

Totally understand the need for darks and flat darks to be at the same temperature, and that's why I referred to a library: with my previous camera I spent quite some time taking loads of darks and dark flats by cooling the camera, letting it heat up, cooling again etc until I could bring sufficient numbers of them all together at each temperature (according to what the camera's sensor told me). I did this because I thought I was supposed to! Also because packages like DSS and APP tell you to, and have specific features for these calibration files such as matching them via groups etc. I just thought it was best practice. They were all shot at ISO800 because that's what I use for everything, with the darks matching the lights for duration too.

I've also come across the argument that darks aren't needed at all. This would be more convenient, but I did find a web page (can't find it again) that outlined why you need both. So, given that there was no consensus, and some people argued that all calibration files are needed given different circumstances, I did both, figuring that it was a bit of a pain to begin with but once done, it was done.

So, if I just use biases, and a large dither (which I do anyway), is it definitely the case that I don't need darks or flat darks (or even dark flats - yes, I've read the thread that goes on and on where people even discuss which phrase to use)?

And if so, another question: do the bias frames have to be at the same temperature too? Again, research yields varying takes on this. 

[Post-edit - I think it was this page, which did my head in, not least the comment that 'a good way to think about it' was Pre-processed Image = (Light - Dark) / (Flat - Dark Flat), which someone then elaborated on by clarifying it as Subcalibrated = (Subraw - Bias - (Darks - Bias)scaled ) / (Flats - FlatDarks): https://www.cloudynights.com/topic/658692-why-do-you-need-bias-frames-if-you-do-darks/]

Edited October 27, 2020 by BrendanC

[BrendanC]

Anyone? 

[bobro]

Interesting thread. I too imaged with a Canon EOS1000D, experimenting with temperatures of over 20 degrees C. The amount of obvious pixel noise at higher temperatures was high, though I aimed for 20+ subs with dithering and used kappa-sigma stacking to eliminate a great deal of the pixel noise. Darks weren't taken.

More recently I'm using an IMX183 mono uncooled camera. Due to the amp-glow darks are essential, with the darks eliminating the amp-glow and not being important in temperature for removing it. A couple of days ago I experimented with stacking an image using darks that were taken about 4 degrees C cooler than the image. The result was pretty good in eliminating pixel noise though there were a few obvious pixels that still showed up as not being corrected. Dithering and kappa-sigma stacking would easily have corrected these pixels, though darks at the correct temperature did the job.

I'm by no means an expert imager, so won't be able to comment on detailed questions. Perhaps my conclusion is: whilst theory is great, efforts to comply with theoretical opinions aren't always so obvious in the resulting images, so it can be a good idea to find out what works for your setup and the final result by experimenting. 🙂

[Sp@ce_d]

Hmm.. ok so short answer..this stuff used to keep me awake at night too... {no pun intended!} yes of course you need to rebuild it.. its electrically changed & uncooled.. it's a nightmare to calibrate... BUT... I used to get really stressed about this, however (ok having not used my "self modified" DSLR for.. well.. literally years...)   Olly has a point, his experience is well worth taking into serious contemplation. It's not as if we're not short of clouded out and moonlight skies to do [removed word] all else right now is it? How do your latest "non darked" processed images compare to the "darked" ones anyway? no seriously, I am interested, uncooled DSLR imaging is a challenge.. I took the easier route to cooled ccd.. I don't even bother with darks.. sleep much better now.. 

[rnobleeddy]

My take is that I've seen more people argue for no darks plus dithering than for darks, but aside from anything else, scrapping darks saves a lot of time. 

I hadn't got around to building a library before I started dithering instead, but without an option to correct for temperature in the software I use, I'm not sure I'd have the patience to capture 32x dark frames every few degrees, especially because temperature is something that's hard to control. 

The 450d I use has clear banding running horizontally across the image on single subs. By the time I've stacked 20 dithered frames it's gone.

[BrendanC]

Very interesting feedback, thanks all.

Regarding the time it takes, well to put together a darks library takes, what, a day or two, in which I wrap the camera up safely in a towel, put it in the fridge, and let it take many, many shots as it cools down. Then I take it out, and let it continue as it warms up again. Then put it back. Then take it out. Etc. Think of it as the GBBO, but with a DSLR. Oh, and not actually baking. OK, so it's nothing like the GBBO, but the end result is, loads of darks at varying temperatures, or as close to those temperatures as I can estimate given the accuracy of the camera's sensor. So, next thing is, I just group them all by temperature and store them in my library. Same with dark flats, and biases (but they take a lot less time obviously). I aim for 50 biases and darks, and 25 dark flats, at every temperature I can manage. Then, of course, I take the light flats during the session, which again takes very little time. Then, in DSS, I group the lights, dark flats, darks and biases by temperature, and I use the same light flats across all of them in the main group (they wouldn't be temperature-matched because I would have to take loads throughout the session to get the right temperatures, plus it probably doesn't matter too much with such short, light exposures).

So, it's labour-intensive and time-consuming to build it, and then you have to store it all too, but once it's done, it's done. I was just wondering whether I needed to or not, given that I have a replacement camera.

The summary seems to be: yes, if I need to use darks; but I might not need to use darks.

I have experimented in the past without using darks. I couldn't see much difference, but then who knows whether it was just that particular image? Maybe a nebula with a lot of space around it would need darks? Maybe a cluster close-up wouldn't need darks? I'm certainly not going to try different combinations every time I process an image! And I'm minded not to do a load of experiments when I could just rebuild the library and have done with.

I'm tempted to rebuild it. It seems that, if I do it all 'properly', I've covered all the 'well, you could not use darks if A' or 'you could just use biases if B' contingencies.

Again, thanks for the input!

Edited October 28, 2020 by BrendanC