What You Don't Know About Astrophotography Calibration Frames | Darks and Flats Deep Dive
🔠Mastering Image Calibration in Astrophotography 🔠In this video, I break down the essentials of image calibration for astrophotography, cutting through common myths and simplif...
Very helpful, Dear Luca! Especially the 'flats first' tipp. i naively thought 'how can my sensor have such a large dark spot' and only realized that it must be on the filter...
Great explanation! The rice analogy is excellent. I like how you have correctly separated the dark frame into the dark signal (true bowl weight) and the dark noise (measurement uncertainty of the the bowl weight) and considered how that uncertainty is average out, and how it can be added to your calibrated image if you don't average the darks enough. One issue I found with flats when I started using them, especially with uncooled cameras & EAA, is that since the dark current (and camera offset) in your light image is also boosted by the flat, any dark current that is not accurately removed with good darks appears as a brighter negative image of the flat frame, with bright edges and bright dust rings. With cooled cameras the dark current becomes much less significant and more repeatable so its easier to remove, but with uncooled cameras, the amount of dark current and the errors in the dark frame due to temperature changes cause big issues when using flats. It can be hard to calibrate out accurately since the temperature changes during the imaging session. This is especially the case when exposures are short (EAA) so the read noise is high compared to the wanted light. The read noise is then multiplied by the many exposures that are stacked in EAA, then boosted again by the flat. Of course this is not how good quality images should be taken, but it is how people may get into EAA and astro imaging (like I did), and this is an issues that crops up with flats. Unfortunately, without guiding and with cheap mounts, long exposures are not practical. But it's the ability to stack these short exposures with basic equipment that opens the door into EAA and starts the journey into more advanced imaging. So it's good to know what is going on here to help resolve this issue and gain valuable experience. This frustrated me at the time. I was convinced there was something wrong with my flats, but years of experience later I now understand what was actually going on. Accurate darks make a big difference for uncooled cameras and the short exposures typically found in basic EAA setups.
Thank you very much for this excellent and informative video. Explanation of why divide by flats really made sense!
I tried flat panels, white t-shirts over the front of the tube while pointed at a white wall, or pointing the tube with a white t-shirt stretched over the front at a clear daytime sky. The third choice seems to work best. No problems with the rolling shutter.
I watched many of the calibration videos. Your video is the best with all the details. Thank you.
thank you I'm glad it's helpful
Late to the party . As a newbie I found this tutorial excellent! Thank you !
Wow!!! That was great!!! Luca, how do you know so much about these topics? Are you an astro-physicist in real life? Crazy!!!
Thanks for such excellent & detailed explanations .
22:21 Bias frames
Tanks.!!!!!
With respect, random noise can be measured, it can be modeled, it can be removed (with lesser or greater success, depending on how we measure and reduce/remove it). A simple example is a mean. It is a naive way of cancelling noise. If there was no random noise in dark frames, for example, we could use just a single dark frame. We take dozens of flat frames, dark frames, bias frames, etc. because there are random variations in pixel values. More added with read noise.
we're saying the same thing here. This is why I said if you're going to do dark frames on modern sensors, do them well (take many frames). My point is, is there even a point to this, when the very read noise we have on the dark frame trumps any dark signal we would potentially try to remove? If not, you can argue that you can do a superbias of 1000 frames and have a much better average than, say 30 darks? Surely you won't take 1000+ dark frames of 600sec.
First time I've 'heard' anyone talking about 'Bias' and especially, 'Flat' Frames as the 'must have' calibration, and, most likely aid to Astro photgraphs. Reflects, confirms my own observation, with Dark frames, not adding, but if anything, taking away from 2min 30sec exposures with my OMD EM1mk2 . For some time, have been neglecting Flats and especially Bias frames, largely because everyone seemed, seems to bang on about Darks primarily. Have, since last shoot, with integration not reflecting apparent gains in tracking, exposure time, been dissecting and exploring process, searching for optimum. Bouncing theories, thinking and trouble shooting off 'Deepseek' very useful. You, have something to say. Interesting take on process, well explained. Subscribed on that.
thanks Paul! Flats are really the only ones you *absolutely* need, no matter how good your sensor is. The bias is just needed to be able to use the flats. If you are using an uncooled sensor, I would still consider darks, unless you are doing a LOT of shots and you dither a lot.
​@the_space_koala Just reprocessed data of M101 taken recently, with Flats and Bias (same iso) included, 20 frames each. I did include the original Darks still. Quality is markedly improved, There is more detail, background galaxy definition star colour and depth, in the first sub set of 24 lights, than there was in the total 72 frame integration before. Another change made was to shuffle the lights, hopefully effectively increasing dither in, to the 3 sub sets. This to also to combat some walking noise in the original second set. As I have to manually dither, interrupting, slowing data progress, I was averse to doing that enough, maybe. Yet to see whether that solves it overall. But your emphasis on Flats and Bias, has made a very noticable difference regardless. I can now see the benefits of the 2min30 exposures over previous 90sec. Going to run it without Darks too. Thank you! and happy star gazing Luca.
Splendid video. Top notch didactics!
Oh, what "THEY" won't tell me huh? So it is a conspiracy
I meant people/guides in general, but yeah, the deep state and their astrophotography secrets! :D
@the_space_koala well, i find myself in a common human trope of duality where i liked the information in your video and will likely watch another when it pops up, while simultaneously hating you for the click bait title and wishing that both sides of your pillow are hot & sweaty for 7 business days. Muahahahaha
I understand everything you said until you opened PixInsight. You need to continue teaching with the same detail, what are doing with the program, what are you selecting, Where does it say "subtracting darks" "dividing by flats", etc. I understand photons, I don't understand astrophotography programs.
Thank you so much for this video. I can learn so much from you. Incidentally, I'm saving up for the Celestron EdgeHD 8 on your recommendation, with everything that goes with it. Also filters. When I have the equipment, I'll take the liberty of sending you a message on Instagram if that's okay with them. Until then, happy Easter despite the lousy weather in the south.
I disagree with you, dark current is noise, but repeatable noise. Noise by definition is signal we don't want. Further more dark frames can help with walking noise and and hot pixels that are hard to remove as they always align unless you dither. There is also diminishing returns, 20 frames usually is enough. To half the noise again relative to the repeating dark signal, you need 40 frames, past that it gets silly
"noise by definition is signal we don't want" - this is not the definition. Signal we don't want is signal, and noise is the random variability we get within any signal. If noise was repeatable, you would never need to stack anything, you'd just have to remove whatever value you think corresponds to the noise (it would be enough to take a single dark frame)
@the_space_koala we stack the dark frames to eliminate the variable non repeatable aspect of the noise or at least average it out. The dark current is a noise pattern we can consistently reproduce at a specific exposure length, gain and temperature. We can't sample it with one dark frame because it's drowning in random noise. So we stack 20 or more so that the random noise averages out and we get a relatively clean sample of the repeatable noise patterns. We can also use rejection algorithms to further refine our elimination of the random aspect. It's not a big deal, I'm not saying your video is wrong, just saying that any signal that we don't want is noise. Dark current just happens to be noise that is reproducible and easy to calibrate out.
if signal we don't want is noise, based on that logic light pollution would be noise, and the light of the Moon would be noise. Or when you are shooting with a narrowband filter, the part of the continuum that seeps through would be noise. It is clearly not the case, it is simply unwanted signal. It is important to understand the difference in my opinion, because noise *is* always random, hence the need for stacking.
There is indeed a lot of information on the web about calibration frames. The latest general opinion seems to be that with (the latest) CMOS sensors, there is no need for bias as they were intended for CCD sensors. Older USB 2 CCD cameras can take a long time to take calibration frames (as in getting the frame physically on the imaging computer) so having a reusable master bias instead of an equal number of dark flats saves precious time. So far, I have failed to find a truly authoritative (and perhaps academic) documentation that is peer reviewed and accessible for free. ADU values for flats also seem to vary as well as the optimal length of a flat frame exposure for a modern CMOS camera. You have provided something for discussion.
it's true about ccd cameras - it was possible to take bias frames, while with modern CMOS cameras we can only ever get close to a true bias, as an exposure time of 0 is not possible. However a good bias makes it so easy to calibrate flat frames we just keep using them, as dark current is truly negligible at these short exposure times
Clear and documented presentation I'm working at -5° for 5 min... I'll try at 0°, it'll extend battery time. I find Sky Flats more practical than those with my panel: better calibrations. Near field, far field, reflections???? Don’t know Creating flats: level between 1/3 and 2/3. Easy with a mono camera. But with color camera ASI6200MC ? And equipped with SO or HO filter? It's not possible. Do you have any advice? thanks.
I use the auto flat option most of the time. As long as your light source is good and even (for a sky flat it is) you will most likely be fine. The exposure of the flat is not as important as people make it out to be, I like to be on the brighter side, just because this way I have a better SNR. Stay away from the far ends of the histogram and you'll be fine. Since we do a division, the absolute values of the reading are not important, just the relative ones. I hope this is somewhat helpful.
@the_space_koala Thank for your answer, I’m not so far Auto flat is mandatory during dusk, with telescope stopped on « null point » avoiding gradient Analyzing my RGB masterflat, 30 flats /ASI6200MC: Two colors RG are between 20,000 and 28,000 ADU The last one B is only at 11,400 ADU, it's not so important Hope you enjoyed your second day at NEAF. I'm looking forward to your second video!
You state on flats your peak is about 2/3. Others have said to use 1/3 to 1/2. Is this because it is a dslr or is it actually better? As always thanks for your great videos and feedback!
The histogram of the flat is actually not that crucial - the homogeneity of the light source is a much bigger factor. You don’t want to saturate any pixels but a brighter exposure will have a better SNR so it’s better for your calibration in my opinion
@ thanks for always being so helpful. One last question. I have made a large investment in equipment but just purchased a cheap light box for flats. In your opinion , if I can afford it should I purchase a high end light box? As always, thanks! Hope you had a great time at NEAF!