𝐘𝐞𝐬, 𝐚𝐧𝐝 𝐈 𝐜𝐚𝐧 𝐩𝐫𝐨𝐯𝐞 𝐢𝐭. (𝐀𝐧𝐝 𝐬𝐨 𝐝𝐨𝐞𝐬 𝐠𝐚𝐢𝐧).

this video is sponsored by brilliant hi Dr Detail sh there's no time I'm finishing an experiment oh cool what are you testing something that's been bothering me for a while what's that whether exposure time and gain matter for your subframes come here let me show you the results I even have statistics oh that's really cool I think people will be surprised welcome to deep Sky detail hi so you may have heard that sub exposure time doesn't really matter in terms of SNR or signal to noise ratio but is this correct I don't think it is well it is and it isn't and I'll show you why in fact I won't just show you like I'll actually do an experiment with real measurements and real numbers and real statistics really really really also and I'm going to spoil a little of the results here but you may be surprised to find out that just looking at two stacked images with your eyes to see a difference might deceive you these type of comparison videos examining long and short exposure times might give you inaccurate beliefs about which exposure time is best before I begin let me start out by saying I did a video on whether stacking increases signal or decreases noise in that video I showed with a perfect camera and with perfect conditions the sub exposure length doesn't matter 10 1 second Subs are with a perfect camera equivalent to 100 second sub but I also said that it is an approximation and that the two are not equivalent in real life why well we don't measure things with a perfect camera we have to deal with read noise light pollution noise and other things then a couple of weeks ago I saw this presentation on the Astro Imaging Channel by Tim Hutchinson on noise it's really good I'd recommend taking a look he did say some things that I'm not 100% sure about though more on that at the end of the video I'd like to experimentally test what he said over a couple of videos this is one of the things he said consider read noise and shot noise as our exposure gets brighter I.E as we take a longer exposure our read noise becomes less and less significant here it's not changing our final number at all but anyway the takeaway here is not so much about read noise but more about what our minimum exposure needs to be right so what Tim is getting at here is that if you have really short exposures the read noise from your camera on each subframe is going to be greater as a proportion of your overall signal than with a longer exposure at this point I think I need to make something really clear definitions matter in this video signal means the average light coming from a deep Sky Target noise is anything that messes up that signal nois is random variation due to your camera's read noise variation from the deep Sky object itself random photons from light pollution Etc and the big thing we need to talk about is signal to noise Ratio or SNR in some videos people talk about signal and signal to noise ratio interchangeably I'll try not to signal to noise ratio is just the ratio of the signal divided by the noise it is not the signal it is not the noise stacking increases the signal to noise ratio what stacking does to signal and what it does to noise is Up For Debate depending on your philosophy I did a three video series on it you should watch it it's really good if I do say so myself but here's the critical thing stacking either increases signal and increases noise or it averages signal and decreases noise stacking does not and cannot simultaneously increase signal and decrease noise watch the three part video series it explains the math to sum up those videos If you think stacking increases signal then it also increases noise but the noise does not increase nearly as fast as the signal the signal to noise Ratio or SNR increases if you personally think that stacking averages signal then your signal gets more constant and your noise decreases subsequently the SNR increases the math in both instances is exactly the same the philosophy you choose is just you arbitrarily deciding what you think should be in the numerator and denominator of the SNR calculation okay tangent over so the focus of this first video is to determine if sub exposure length matters additionally I want to see if gain and read noise matter in this video Tim also said this we already proved a few slides ago that we don't care what the read noise is so push the gain down expose for a longer period of time is this true I'm not sure it is for every single camera look at the read noise graph for my ZW 294 mm camera it has two Gain modes once you get to Unity gain the read noise drops and the dynamic range increases back to 13 stops even though the full well depth is lower I'm thinking that technically for my camera at least read noise might matter a little bit thus Unity gain will give you better SNR I 100% agree with Tim though that there is a minimum exposure time you need in order to get the best are that you can so I'm going to do a classic 2x2 experiment to find out if I'm correct my hypotheses are this using Unity gain at least for my camera is better than low gain settings and having too short of a sub exposure will result in less SNR than a longer one keep in mind that choosing a good sub exposure length also depends on how good your tracking is and things like that so life is complicated and that's okay we can still test things create models and learn new stuff even if it isn't a complete picture that's science so real quick what equipment am I using I've got a 6in rid creation at a focal length of 1,370 mm attached is my zwo 294 mm Pro camera on an Orion serious Mount I'll be using an off-axis guider my target is mesier 101 how will I do the test well as I said before it's a classic 2x2 experiment meaning I have four different types of subexposures two of the types of subexposures will be 120 seconds long and the other two will be 15 seconds additionally two of the sub exposure types will have a gain of 10 and two others will have a gain of 121 almost Unity gain so we have four total conditions and each condition has a unique combination of sub exposure length in game note that 15 seconds at the F ratio of the RC should not be adequate to allow the signal from M101 to overcome the read noise eight times longer at 12 seconds should be at least better so how did I take the sub exposures did I do four nights of Imaging with each night a different type of sub exposure longtime viewers of this channel know that the answer is no that would be silly how could I control the humidity on different nights how could I control the high level clouds each night well I can't what I did did was systematically alternate between 15c Subs 120 second Subs 10 gain and 121 gain each Imaging session for example I take 122 sub at 10 gain then 122 sub at 121 gain then 8 15c Subs at 10 gain then 8 15 second Subs at 121 gain then I take 12 second at 10 gain and then 8 15 second Subs at 10 gain Etc systematically alternating this way should help minimize the effects of atmosphere on on the different types of subframes and it should help minimize the effects of guiding Gremlins and and other things this is the way I need up shooting over 10 nights and got about 40 minutes of integration time for each type of sub exposure so what did I find all right I ran some statistics on this and I trust the statistics I'm going to show you the results visually first you might be surprised that what your eyes see might not match the statistics here are the four Stacks with each stack having four 40 minutes of exposure time each tack has only had an auto stretch applied with no other processing can you tell the difference well this one looks the worst and it's not even close the others look kind of similar but which one has the best signal to noise ratio and which image belongs to which condition I'll come back to matching the Stacked images with the signal to noise ratios later to test the signal to noise ratio I took took the core of Messier 101 and divided it up into 121 equal sections or samples I then measured the signal to noise ratio for each sample and each of the four Stacks each of the four Stacks were aligned in serial you know this took a lot of work you might consider subscribing if you haven't yet I'd appreciate it with only 40 minutes of exposure and a relatively slow scope like the 6in RC if I average all the Stacked image samples together the SNR is about six but we don't care about the overall sample we want to know how sub exposure length affects SNR and gain and whether there's some sort of synergistic effect between exposure length and gain on the signal to noise ratio to test all that stuff I put the data into a thing called a basian model it's just some fancy statistics that can help sort out the individual effects of gain and sub exposure time on SNR in other words I'm trying to predict SNR of M101 with my setup based on the sub exposure length and the game settings so what did I find well before we get into that let's talk about statistics a little bit with Statistics and statistical models you're always comparing different hypotheses it's a different way of thinking about things one thing that get started in statistics is with this video sponsor brilliant in this video I'm basically comparing ing several statistical models using basian methods I really like basian thinking because you can quantify your uncertainty based on what you already know and Brilliant there is a course called basian probability as someone who has had a lot of experience creating basian models I think brilliant does a great job at introducing us to the topic you will learn key concepts related to information Theory basian inference and calasian networks you'll get hands-on experience working on rela problems not just memorizing formulas and flowcharts takes time to learn EMB Brant can help you practice a little bit every day using fun examples before you know it you'll develop a powerful learning habit if baz is a bit too advanced don't worry there is an introduction to probability course that leads Rod into it to try everything brilliant has to offer for a full 30 days visit brilliant.org decy detail or click the link in the description you'll get 20% off an annual premium subscription so for my particular setup sub exposure length really mattered like it's not even a question the statistical model is screaming you got more SNR with 122 subs and 15sec ones basically the 15sec subs averaged about 4 SNR while the 1202 Subs averaged 8 SNR pretty wild sub exposure time really does matter if it isn't long enough the read noise will be too strong in each sub exposure and you'll get a worse image but what about gain well the model isn't exactly screaming at us here but it is saying something gain does matter it matters for both long and short sub exposures but its effect isn't super big the subs with 10 gain had about 5 and 1 half SNR while the subs with 1221 gain had about 6.8 SNR was there any synergistic effect of gain an exposure length on SNR no not according to the model so basically the effects of sub exposure length and gain were kind of independent of each other and that's good because I don't actually have an a priority reason for why there should be some weird interaction the between the two variables so what's the key takeaway too short of an exposure length gives you worse signal to noise ratio not using Unity gain with a 294 mm Pro Camera in bin 2x two mode is a bit worse too but not nearly as bad as having a short exposure let's look at the Stacked images again not surprisingly this image that looks the worst is the 15sec 10 gain exposure and here are the exposure times and gain settings for the other three they actually look quite similar to each other yet these longer exposures are definitely better based on the statistics so what's going on well first of all let's get this out of the way sometimes your eyes aren't the best judge of things especially when comparing two images side by side here's an example this left square has 10 dots the bottom left one has 20 dots you can definitely tell without counting that there is a difference in the number of dots this top right square has 110 dots and the bottom right one has 120 but I couldn't tell you which one has more just by looking at it it's the same with these four Images we can clearly see that the top left one is worse than the others but these other three look similar the two on the right will get us a decent signal to noise ratio much more quickly than the bottom left one will if you get 10 hours of data instead of 40 minutes you might end up with a clear winner the 122 Subs might look much better the difference might get exaggerated as you stack but you can't can't necessarily tell that by looking at a stack of just 1 hour of data and that's where the power of doing an actual experiment with actual statistics shines I don't have to give a handwavy answer at the end of this video that says they look similar I'll let you decide which one is the best no the answer is clear having an exposure length that overcomes the read noise is much better than shorter Subs but are 10minute Subs better than two-minute ones even if the two-minute ones have sufficiently overcome the read noise Tim Hutchinson also said in the astroimaging channels video that and I'm paraphrasing what I think his opinion is that a long sub exposure like 10 minutes will be better than a two-minute one even if the two-minute one has overcome the read noise right we need to make sure that we're above those minimums but the real question is how long can I go because remember the longer we expose the greater impact we have on our shot noise for every single frame and then as we stack those frames drive that noise down away from the signal that we've captured right so in my opinion I would I do not follow the advice of some that suggests that hey these camos cameras are so great that we can take really short exposures and get a great image you can get an image but I suggest I humbly suggest that you won't get a great image right and the best images will come when we when we make that Target signal as bright as we can in all honestly I don't think this is entirely correct I like experiments like this one I did on how filters affect signal I think I'll do one on very long exposure as vers normal ones but that's a video for another day and thanks for watching a

Channel: Deep Sky Detail

Views: 6,429

Keywords: zwo, exposure length, astrophotography, does sub length matter?, gain, read noise, subframe, sub frame, astronomy, M101, experiment, GSO, Orion, Sirius, guide camera, guiding

Id: v6I7nTu6VIc

Length: 16min 31sec (991 seconds)

Published: Sat Jun 22 2024