Why I DON'T take Dark Frames (and what I do instead) - Astrophotography

To be clear my video is ONLY valid for COOLED astrophotography cameras WITHOUT Amp Glow!! It also assumes good dithering to deal with hot pixels!! Also, yes I have compared with and without, and I can't see any difference.

Also ERRATUM! When comparing the dark current I shouldn't have squared it because the dark current is the mean rather than the standard dev! So for -5C, we'd have sqrt(0.36 + 1.4 * 1.4) = 1.52!!

-----------------------
My equipment
-----------------------

Follow me!

[davidschaeffer]

I thought that dark frames will remove hot pixels as well. Am I wrong In thinking that? Do bias frames do the same with hot pixels?

[grucha]

Thanks Cuiv. Another winner!!!
I've gotten back into the hobby after being away for years and am amazed how much the game has changed from the "old" STL-11000M days to these new CMOS sensors. I honestly never thought they would be anything other than planetary/guide cameras. Gotta love how much lighter and smaller these cameras are. The SBIGs back in the day were tanks that caused monstrous strain on the imaging train.
And now not really needing to bother with dark frames? Another nail in the CCD coffin as far as amateur astrophotographers are concerned.

[jonrbryan]

I stopped taking dark frames as soon as I got a ZWO ASI533MC camera. I use flats with 3-5 second exposures (calculated by my ASIAir+), and the same exposure time for my bias frames (dark flats). This gives me excellent results. Per Adam Block, I also do a Cosmetic Correction when stacking in PixInsight's WBPP, and I drizzle. I find that NoiseXterminator then gives me very clean images.

[patrickstevenson]

My first dedicated OSC camera was an ASI 294MC. I was immediately aware of the amp glow problem. I am an old engineer that has worked with the fabrication and application of solid state circuitry (IC's, or Chips). From the very beginning of solid state circuits back in the '60's, we were aware that the architecture of our circuit boards could have a major effect on the output of certain IC's. As a result, AFTER the electronic design gave us our desired output, we then applied the architecture of the circuit boards so we could minimize any effects caused by the closeness of the chips to one another. Since we were doing this over sixty years ago I was surprised to see a modern camera with amp noise simply because the heat of an IC (usually an amplifier) was too close to a video sensor. Rather than mess with calibration frames designed to correct the error I looked for a camera that didn't exhibit the poor design of the 294. I found that the ASI 533MC was essentially the 294 with a proper circuit design that moved the heat-producing chip away from the video sensor, thereby removing the amp glow problem altogether. That was about three years ago and I have not wasted the time taking Darks ever since. When asked why I didn't use Darks I found that explaining why usually resulted in blank looks. Your explanation is more exact and mathematically accurate, which I appreciate. The simple answer is, of course, don't get a camera that, by its design, injects defects. Thanks for the elegant explanation.

[tiagosnightskies]

Nice video Cuiv. I also own a 2600mc and I completely stopped doing dark frames as well, as I found out that using them in my calibration step for WBPP actually introduced unwanted artifacts (since a master dark frame is inherently a statistical analysis of thermal noise and will never be 100% accurate). The test dark frames I did when I first got the camera (and even on my 533mc) are so "clean" that I felt dark frames would be unnecessary.
One thing that might be worth mentioning is that on older sensors (especially DSLRs), dark frames can be invaluable to evaluate sensor health. My first camera was a used Canon 60D and dark frames were essential, as the thermal noise of the sensor was all over the place, introducing patterns (not just lines) that flat frames did not cover. For those cases, dark frames can be essential.
Happy New Year!

[raygorley]

Thanks Cuiv. As an even more minimalisti approach to calibration files Siril allows for the option of replacing the master bias file with a "synthetic bias" which is a number proportional to the camera offset value. For my Altair imx571 (with APT) it is just the offset value itself. You can also use the $OFFSET keyword if it is defined in the fits header. So just lights, flats and synthetic bias for me.

[deep_space_dave]

Hi Cuiv! I agree with you on this as with the newer CMOS sensors that are both back illuminated and rolling shutter no longer need dark frames especially if the sensor is cooled. Every image sensor, be it a CCD or CMOS, has what is known as dark fixed-pattern noise, a pattern that is the result of the manufacturing process. If you take a master bias stack and use Pixinsights SuperBias process on it, you will actually see that structure hidden in all of the randomness (I wouldn't use it though). Bias frames are basically very short dark frames as you still need darkness to exclude random photons (shot noise) from the calibration. But the structures are so relaxed and faint on modern CMOS that like you said, there isn't too much difference anymore unless you take a crazy long exposure like 10 minutes or more then the dark current will start to build up thus now needing dark frames. Dark frames did have a good use though and that was removal of hot pixels but now the CosmeticCorrection process in Pixinsight can take care of those pesky hot pixels. If you dither well, the stacking algorithm will also remove the hot pixels. So there is no more use case for dark frames on the latest sensors unless you intend on doing exposures over 5 minutes then you may start to see a little amp glow. As always great video!

[UltrasoundJelly]

Really like that you made this video. After watching your last one, and with a 2600 MM on the way, your comment about no darks really piqued my interest. Revisiting the RMS equation and looking at the low dark current contribution was a nice way to explain it (at least for my engineer mind). I've heard the argument "well why not just create a dark library and use it anyway, because it's easy." Good point on possibly introducing artifacts to counter this argument. For monochrome imagers, one fewer thing to think about is nice. We already have to take 3-4 sets of flats, and not worrying about keeping up a dark library will be muy bien.

[wesleydonnelly]

Cuiv this vid is by far the best most easy to understand explanation for the reasons we do calibration frames, I have seen to date. Calibration frames have always been my weak point in astro ( circa 15 months doing astro ) and tbh I just didn't have the motivation to try read pages and pages of technical info to learn about them. Thank You for educating me my friend!

[thomasmastrocinque]

Trust but verify. That is why you are my go-to Lazy Geek! Thanks again my friend.

[christianvontotth]

Interesting Video Cuiv, as always. Glad to see you're safe.

[aethyr]

thanks for doing the math and proofs. you just helped the efficiency of this hobby overall!

[hymanmj]

Superb video Cuiv. I will certainly be taking this onboard. Thank you.

[bbroastro]

A few things I just want to point out. This hugely depends on what software you use to stack. Some software *requires* you to load a master dark if you're using a master flat, otherwise your stack just won't calibrate properly. It can result in massively overcorrected flats, even if the flats are perfectly good.

Also, as others have pointed out, even with correct dithering (by using formulas to work out how many pixels you need to dither by for your FL and guide focal length) you may still have *some* hot pixels left as your sensor ages. You either have to start dithering overly aggressively (and at my FOV that just isn't a practical option) or using aggressive sigma clipping. Ultimately if you're using a cooled camera and shooting a master dark library is a one time event every 6-12 months, I'd still just rather use them and take potential stacking problems out of the equation. 

Obviously bias frames are shorter so take less time, but running a dark session on a cloudy night is super easy. On a current gen cooled camera, you're essentially just doing super short dark frames with the bias, so really you're just swapping one out for the other. FWIW - If you have light leaks in your darks, then you likely have light leaks in your lights too, so it's worth actually chasing down those light leaks in general.

I do think when you *know* what calibration frames are doing you can make this decision yourself, but I think a lot of beginners will see this video and then just assume they don't need darks even though you've explained why you don't necessarily need them for *every* camera. There's lots of confusion in the community already, so I hope people do their own experiments.

[rickbattle]

Very insightful video with valuable information. I have a ZWO ASI 533 MC Pro and I am following you down the path of fewer types of calibration frames. Happy New Year!

[andrebremer]

I haven't been using dark frames since switching to IMX571 and related sensors either. At least for my situation, they also tended to add unwanted artifacts (especially involving the removal of hot pixels). To alleviate this, in addition to dithering, I also curate my own defect maps which I apply in WBPP. This way you can target each bad permanent pixel without collateral damage.

[VisionCommunications]

Thanks for sharing this! I only use dark frames for my DSLR on warm days to deal with hot pixels. For my astro camera, I don't use dark frames. The thing is people lose their minds if you say you don't use dark frames. There's a lot of religion and ritual and actual photography! :-)

[jackwmes]

thanks for this cuiv!! i'm also using the player one implementation of the imx571 and have been spending far too much time trying to get good darks and dark flats - but no longer!

[RynBat]

Even with my old DSLR (Canon 70D), I have tried processing with and without darks and have found negligible differences. When I upgrade to a cooled astro camera, I really don't think I'll need them. It's nice to hear that I'm not the only one questioning their usefulness.

[Stargazer_Astro]

Cuiv, have you tested this? Could we see the difference between calibrating with Darks and then only Bias? Im curious if we would be able to visually see a noticable difference. If you stretched the image enough, would the thermal noise eventually become visible?