Spent the past two days researching and trying to understand this topic in more depth and your video was the catalyst! Great presentation and nicely explained. Created the spreadsheet and calculated the values for my mono and OSC cameras. I will give it a try tonight! Again, thank you.
Sharpcap Pro 4 sensor analysis and smart brain histogram functionality does this for you by measuring sky brightness and working out your optimal sub exposure/gain/offset levels automatically.
This is such a great resource. Linda, you have boiled down the math and the practical application into a 20 min presentation in a way that I never have seen before. Thank you!
Very informative. I can tell you that for Nikon and Canon DSLRs in N.I.N.A., the ADUs are not scaled and the 2^16/2^bits term at the end of the first formula is not neede.
Your timing on this video was perfect. Also the choice of camera made my life much easier :) Now I have some ball park numbers and now know why they are what they are. Thanks!
Really well done...so informative and clearly presented. A wealth of practical information presented with a most pleasing, almost hypnotic voice. Thanks for posting this.
As a beginner (6 months into this hobby), this was very informative, despite that I don't have these software yet. The clipping of stars and blowing out the color has been an issue and I don't think I'm alone in this (just personal opinion browsing some images online). Light pollution filters I have experimented with is something I don't particularly like after a few tries, they seem to ruin the natural colors. I started to pay more attention to those ADU numbers and while significantly had to lower my exposure time without any LP filters , the stars seem to be so nice and sharp and all the colors from dark red to blue.
That's a good question. Largely it comes down to answering three questions: 1. How much patience do you have? 2. How bright is your sky? 3. How much patience do you have? The darker your sky the less time you need. I have an image that I got in 90 minutes in a Bortle 2 sky that would have probably taken at least 8 hours in my Brotle 7 sky. Generally, here in my bright suburban sky I try to get at least 8 hours and more if I can. The fainter the object the more time you will need but I don't have any quantitative numbers to suggest. If you manage four hours one night and you process and want less noise then get an additional four hours. If still too much noise then get eight more. Eventually you'll decide you've had enough or lose the object for the season. :)
Thanks much for the clear and calm explatations. Now myself using a ASI2600, my numbers would be from 50 to 105 ADU (at gain 101, offset25) . Normally I expose some 3 to 4 minutes long but if I calculate correctly, some 45 second might be ideal ! Wow that is short. I will have to try it out. Just to be clear here.. Optimal means 'with as little star clipping as possible ?? ' Am i right? Dennys, Montreal/Canada.
Technically "optimal" here means getting good stacking efficiency but as a side effect of that it does tend to get you less star clipping since we're finding the shortest exposures that get you to around 95% stacking efficiency (that is to say the stacked image is within 5% SNR of what a single exposure of that length would provide).
Linda this video is fantastic. Thank you so much for the clear and concise explanation. I have read a good deal about this process but didn't feel as though I had a firm handle on the process until now.
Dear Linda, thank you very much for very useful video. I would like ask you for answer about Offset. I use for my astrophotography astromod DSLR. I know read noise, I know ISO/Gain in eADU, but I do not know how to evaluate Offset. Do you have some idea how to do it ? What value should I put into equitation ? My thanks in advance.
Sorry for the delay in responding. If you are using a ZWO camera I think they default to an offset of 50 but really the goal of the offset is to ensure you don't get any clipped pixels after calibration. You can look at the calibrated sub and see how many pixels have a value of 0. If there are some it probably makes sense to raise the offset enough to avoid that zero clipping.
really good video. i will have to watch it like 10 times to get most of the info committed to mem. one thing that makes it harder is you keep switching between cameras. So following and being able to get a base line understanding is harder.
@@LindasAstronomyAdventures np i was wondering i am using a asi294mc pro. I took it out to shoot orion and had -10 temp at 180 sec and 194 gain. so ok it came out way over exposed after stacking. so i found your video and after doing the calc i was left wondering is it better to play above 120 gain or below, because the cam has that weird divide where read noise drops like a rock at unity. So i am stumped as what to do. any input? yes i am brandy new to this so please forgive me. lol
Brian, this doesn't tell you the exposure time. but rather the exposure level. That is the the median value that your exposures should be. Exactly how long that will take will depend on how dark your sky is and what sort of filters you are using. For the ASI294MC Pro, at gain 120, the optimal value is 344DN (DN just means data number). Odds are that is only going to take 10-30s at most unless you are in very dark skies. The choice of what gain to use comes down to a tradeoff between read noise (higher gain gives lower read noise) and dynamic range (higher gain has less dynamic range). On your camera, though, there is that sudden step where read noise drops and dynamic range goes back up. So gain 120 gives you almost the same dynamic range as gain 0. Odds are pretty good that unless you are using some narrowband filters you can stay at gain 120 and stay in the sweet spot for read noise vs dynamic range. However, if you find that you are ending up with lots of 10s exposures you can drop to gain 0. That would raise your optimal level to 777 DN which would double your exposure times. This stuff can be a bit confusing (that's why I made the video) but the key point is it tells you what values you want to see off the sensor not how long to expose for. How long will depend on things outside of the sensor like filters and sky.
Aloha Linda, Just came across you GREAT video (I guess better late than never) and watched it about 5 times! I'm also using an ASI 1600 mono camera and SGP and PixInsight and sometimes FITs Liberator (to view a sub-exposure). I got my spreadsheet built using your formulas (thanks!). I pretty much understand the formula, but I'm still getting confused and wrapped around the axle about all this scaling up and down business between 12 bit cameras and software (like SGP) displaying 16 bit values . So, based on the first example in the video (gain = 15, offset = 50, bit depth = 12, etc), the calculated min DN value is 920. It seems to me SGP scales up (from 12 bit to 16 bit) and the Rista formula 'scales down,' but SGP is still displaying a scaled up median value. The scale factor here of course being '16.' So, when I open up an ASI 1600 sub-exposure in SGP and look at the median value (which is being reported in 16 bits), does that median value in SGP need to be manipulated again? Or, can I just use the displayed median value (as is) in SGP, and compare that to the formula's min dn 920? Kind of stuck on this (and sorry about the long thread here)... Thanks in advance for any words of wisdom/clarification and Happy New Year! John
John, thanks! The camera manufactures report specifications in terms of the native bit depth of the camera, so 12-bits in the case of the ASI1600. However, the camera software scales this up to 16-bits so we never see the native values reported in software like SGP. We only see the already scaled 16-bit value. The formula does that same scaling for us so we can just think in terms of what the camera reports rather than what is really going on at the hardware level. That means you don't need to do any further conversions. You can use the values supplied by the spreadsheet directly. Hope that helps!
@@LindasAstronomyAdventures Aloha Linda, ok, thanks and thanks for the quick reply! I think I understand what you're saying (and I do remember you mentioned that the asi driver was scaling up [to begin with]). Not having to do anymore conversions is a BIG RELIEF. I was thinking SGP's mean value needed to be divided by 16 (and then compare that new value with the min or optimal dn value generated by the formulas). I guess I can always 'cheat' and use the histogram (as a backup) :) Just spent about 6 hours looking at all those detailed Cloudy Night threads on all of this, but your video really helped. Due to possible quantization errors (caused by ADC converters), I noticed Mr Rista prefers to work with the analog signal (which I guess means electrons), so may need to review that approach again. And, I need to look at your offset video again to. Thanks again!
I just realized, thanks to you and your detailed video, that I have much to learn and study before I can even start understanding what I am doing. It has been 18 months or so, make a sequel please. Respectfully, Henri-Julien Chartrand, Montreal, Quebec, Canada.
I'll see if I can come up with anything useful to say...there are some ideas floating around that I've been thinking about. Hopefully I can get them to line up into something coherent.
This is a great video that I just now discovered. Many thanks. I do have a couple of questions. At 19:25 you mention taking a few frames that don't need to be under the stars and measuring the camera's performance level. I'm not clear on whether these are like flats or what sort of exposure if it is daylight and what would be the means of measuring, PI Statistics? Also, I have two OSC cameras, 294MC and 2600MC. If I load a sub (2600MC gain 100 offset 50 300s under Bortle 8) into PI it is labeled gray, presumably because it is not debayed. PI Statistics shows a median of 14656 on the gray image and 9557, 15784, 11086 (RGB) on the debayered image. Are these numbers commensurate with the DN numbers you talk about in this video? I think your images are all mono. Thanks again.
Apologies for the super long delay in responding. I'm woefully behind on youtube. The frames are darks and flats. You want to expose the flats as you would normally (to about 50% on the histogram). And, yes, those are the DN numbers.
Thank you for the video. Sorry, I am very new to this. The formula calculated DN (?), where I can see seconds/minutes of exposure? How I convert min and ideal DN values to exposure time? Thank you.
No problem. The DN value is the value of the pixels in your image, not the time. This procedure doesn't tell you the amount of time to expose but what exposure level to aim for. If you are using an acquisition program like SGP or NINA you can take some test exposures at different lengths and use the median value for the exposure in the image statistics to determine whether you need to increase or decrease. For example, if the formula told you to aim for 1,000DN and you tried a 60 second exposure and your median was 853 then you know you need to try a longer exposure. It's a bit of trial and error but usually within 4-6 tests you have the amount of time you need figured out.
@@LindasAstronomyAdventures Very good, thank you. I will be looking for the VALUE of exposures in SGP (what I am going to use) and through that I can come to time of the exposure. Like in your example, it has to be more than 60 sec.
You have more experience than I, however if a star is not over-exposed in the camera, then all you need to do is preserve what you have. If it is destroyed by processing, thats a processing error.
Yes, there's generally no harm in going longer as long as you don't cause too much clipping. However, it doesn't really improve stacking efficiency much to go (Robin Glover, author of Sharpcap has a video that talks about this in more detail). So, while it doesn't hurt to go longer, it doesn't help as much as an entirely new sub would.
The value I aim for comes directly from the formula I presented. But, yes, for my ASI1600 I usually end up in the 800-1200 range depending on the exact gain and offset I'm using.
Linda's Astronomy Adventures i have to listen and watch again. Cant wait for my tripod to arrive so I can get back in the game. THANK YOU! Have a wonderful thanksgiving
Dustin, here's a read-only link to the icloud numbers spreadsheet I used: www.icloud.com/numbers/0T96cjAMDXFZBLcgxCVqC7UZA#Optimal_Exposure_Level_Calculator. You can copy that to your own spreadsheet. If you can't access it for some reason or would prefer to create it in excel or google sheets here is what the cells are: b1: read noise b2: gain b3: offset b4: bit depth b6: min dn = (3×B1^2÷B2+B3)×2^16÷2^B4 b7: optimal dn = ((10×B1^2÷B2)+B3)×(2^16÷2^B4)
Hi Linda. Thank you for producing and posting this tutorial which is very clear. I like technical stuff but can't help wondering why can't we just figure it out in the field by trial and error, watching for when stars become saturated. Also, does this need to be done per filter? I am assuming so. I was reading somewhere else (www.lightvortexastronomy.com/measuring-your-camera-sensor-parameters-automatically-with-pixinsight.html) and they are saying readout depth is actually that of the image bit depth and that is usually 16 bits. You seem to be using the same bit depth as for ADC. The difference in results of the readout noise is considerable.
The optimal level isn't about avoiding saturation (though that's important). It's about burying the read noise of the sensor with enough signal from background sky (which is mostly light pollution for most of us) to hide that read noise. You can go beyond that optimal exposure but then you run the risk of saturating more stars than you needed to. All that being said, what you described would probably work fine for the CMOS cameras that have very low read noise. It might be a little trickier on a high read noise CCD to know for sure whether you went long enough.
Hi Linda, great video thank you. I'm a little confused about my results though and was wondering if you can help please? I use an ASI533 and from the website I can see that my figures should be read noise of 1.5, gain of 1, offset for this camera is default 70 and bit depth is 14. That gives me a Min DN of 307 and Optimal of 370. When I look at some of the images I've taken, my mean ADU value is around 3000 which is almost 10x the optimal value...and that's only a 30 second exposure. I live in a bortle 5/6 area so not super bright or super dark, but does that mean my optimal exposure times should be almost 10x less than 30 seconds? So like, 3-4 seconds lol. I'll have no space on my PC! Surely I'm going wrong somewhere? Thank you :)
That sounds about right. With the gain at 1e-/ADU and the very low read noise it doesn't take much to bury the read noise in the sky glow even at Bortle 5. You have a couple of choices. You can just live with it. There's no harm in exposing longer so long as you aren't clipping more stars than you are comfortable with. The other thing is you can lower your gain down to 0 from 100. That raises the read noise to about 3.8 to 3.9 e-/ADU according to ZWO's graph. That would let you expose longer at lower gain so risk less clipping. However, either way you are capturing the same number of photons. The difference is in the read noise and the dynamic range. With that camera at unity gain you don't appear to be sacrificing much dynamic range at gain 100 so you should be ok exposing longer but let your stars guide you. If you are clipping too much then lower the gain.
Hi Linda, with this calculator I've got min 77 and optimal 141. In my 300sec exposure with PI statistics the median is 0.00923 which is way off. But my photo looks okay. Where would you suggests should I look at to determine what's going on please? Thank you.
Hello Linda, I used your video to determine my optimum exposure. I am imaging M101 Pinwheel Galaxy. I have the same camera ASI1600MC cool (but colored) so I assumed that the specifications will be the same as in the video (Optimum DN = 1290). I am in a rather high light pollution area and have a Baader LP filter attached to the camera. A single sub-frame of 45 seconds with gain = 25 and offset of 10 and temperature setting at -10C. This is giving me a median of 1136. If I try to increase the exposure time, I will go over the Optimum DN. But I get a sub-frame that has very little luminosity even after stretching, I can't see much of the galaxy. I took a total of 100 sub-frames for a total of 4500 seconds with relatively good guiding. I have not tried stacking it yet. I am using PixInsight. Will I actually get something out of these sub-frames after calibrating with darks, flats and bias and later stacking? I have never tried this approach and don't really know if I am doing it correctly.
If you are using an offset of 10 then your optimum exposure level will actually be 640 DN lower than mine in the video. That's because your offset is 10 instead of 50 which will lower the optimum value by 640 (16 * 40). You are actually going much longer than you need to. There's no harm going higher though you might clip a bit more data than you might otherwise. As to the data looking thin...yes, it does. You don't need much signal to bury the read noise since the ASI1600 has such low read noise even at low gain. But, when you stack up a bunch of them suddenly the signal pops up out of the noise.
@@LindasAstronomyAdventures Thank you for your reply. I did some image integration on PixInsight and was able to show the galaxy but I think I will need to re-image to get better results. I will watch your next video on the optimum Offset.
@@LindasAstronomyAdventures - I had a chance to do some imaging last night and followed your method. I was imaging at Optimum DN of around 1290 at the beginning but as the night went on I noticed that the value start to increase to 1480++. In that case, what should I do? decrease gain or decrease exposure time or simply leave it as is? After calibrating with darks, flats and bias with PixInsight, I still get a fair amount of noise and amp glow. Can you still advise me what to do? By the way, I am in the inner city limits and I would say Bortel scale 8. Hence making it more difficult to capture images. I would like to try getting some images in really dark locations to compare...Your thoughts will be appreciated.
At gain 120 where it switches over to high conversion gain mode, I'm getting optimal at 226 DN so I don't think you've done anything wrong. With only around 0.8e- of read nose it doesn't take very much signal to overcome the read noise. You can expose longer if you want, but it won't help your stacking efficiency much.
@@LindasAstronomyAdventures So hypothetically speaking we're talking like 5 sec exposures? Am I in the ballpark or dreaming with these numbers? lol I guess since I'm using Hyperstar a 5 sec exposure is equivalent to about a 1 minute exposure at say f/5?
It certainly could be depending on the sky you have. Between f/2 and low read noise it doesn't take much. Going from f/4 to f/2 gets you 4 times as many photons/second so yes, your exposures could be insanely short.
Super video. Thanks. I learned more about PI. There's a minimum and a maximum exposure for any given target. The minimum should be whatever exposure just swamps the read noise. The maximum should be where a linear, unstretched image shows very few if any stars. A fully stretched image should then show the structure. N.I.N.A. has an exposure calculator built in. I've determined that seeding with a 1 second exposure produces a recommendation for an optimum exposure. Getting the camera statistics from SharpCap Pro provides the necessary information for the N.I.N.A. calculator. And of course, the PixInsight method shown in this video will work. But, which method is less complicated? For me, the SharpCap-N.I.N.A. process was easier and quicker to arrive at. Using that process provided me with a perfect optimum recommended exposure to capture M13. I tried other exposures, and none of them produced as good an image as the recommended exposure did.
This might be a duplicate reply as I thought I responded to it earlier today but don't see it htere so just in case... Yes, sharpcap is a fantastic piece of software and nina's ability to consume it's exposure info makes for a great calculator. Robin Glover, the sharpcap author, has some great videos on exposure that are worth viewing. In particular, I'd recommend this one: ua-cam.com/video/3RH93UvP358/v-deo.html. He points out there there isn't really much benefit to going beyond the optimal subexposure length since it doesn't help fhe final stack much. The other drawback is the potential to clip more stars (or with some cameras contending with more amp glow or more odds for tracking error). Also, depending on the faintness of the target, it might be quite subtle or even invisible in an optimally exposed frame but will pop out when you do the final stack so looking at detail in a screen stretch may not be an accurate way to judge. Of course, if your target is very faint and you're willing to sacrifice more stars to clipping then it may still make sense just to get more signal you actually care about but that's likely an unusual situation for mos of us. But, your point is well made...there is a fairly broad range where you can be in the optimal zone and it needn't be complicated to get there. Thanks for your comment!
Paul, the only two permissions I can share it with are "view only" or "make changes". The latter would let people change the formulae which could mess up the spreadsheet for others. You should be able to save it as a copy so you can change the values. Let me know if iCloud isn't letting you do that and I'll see what can be done.
nope couldnt do it lol, i do have some info though if you could put me in the ball park my gain is 100, offsert10. (readout noise 1.575 and gain 1.024 pixinsight )
I'll look for an alternative way to share it. But, in the meantime, assuming your camera is a 16 bit sensor then I get 34 DN as an optimal value and 17 as a minimum. If you have a 12-bit camera like the ASI1600 then it's 548 optimal and 276 minimum.
It's been quite a while since I did that video so I'm not sure what you are referring to. Do the filenames say something like "offset_1.00" in them? Or are you referring to something else? If you can point me to it, I might be able to respond more effectively.
@@LindasAstronomyAdventures Essentially, the offset part of the equation is the median bias ADU. Can you take a look at a thread on Stargazer's Lounge? It's called 'Swamp Read Noise' by Pitch Black Skies, posted April 19th. You will see what I am talking about.
The median bias contains both the offset and the read noise. I took a look at he thread and it seems to be saying the same thing I am. Is there a specific post in that thread you are referring to. It's possible I missed it.
@@LindasAstronomyAdventures For instance, at Gain 100 the 533mc-pro camera has an offset of 70. But we don't insert 70 into the equation. That figure translates to ~2800ADU which is a 16-bit number (the bias median ADU). The offset part of the equation is therefore ~700 which is 2800 converted to 14-bit. I've noticed a few posts down that some other 533 user's are questioning the low number result in their calculations. They used 70 as the offset figure which is incorrect but you didn't mention it to them. All I'm trying to say really is that the offset figure isn't the default offset of the camera or the one we change it to, rather it is that number converted to ADUs.
For a 14-bit camera everything gets scaled by 4 so the offset of 70 ADU would translate into 280 DN since the camera reports in 16-bit values. If I have the right vales from ZWO's graphs, then it should be (10 x 1.5*1.5/1 + 70) * 4 to get the optimal value or about 370 with a "swamp factor" of 10. You could get away with as low a value as 307. The 533 at gain 100 has a very low read noise of only 1.5e- so it really doesn't take much to bury that read noise in the sky glow.
@@LindasAstronomyAdventures could you please explain how I'm supposed to use the DN (ADU) number exactly? If I simply take a 0.001s bias with the 2600mc pro, I get an avg ADU of 500, which is already way over the 77-140 calculated above. Am I supposed to add that 77-140 on top of my baseline ADU from a Bias frame and aim for that in my sub exposure?
Something isn't correct with one (or more) of the numbers you reported) most likely. That formula takes all your settings into account so that's what you want your median ADU to be near in an absolute sense. If you are getting 500 on a bias then something is off somewhere. How did you come up with the numbers you got?
@@LindasAstronomyAdventures the 500 ADU on bias frame is from Asiair Pro histogram's AVG reading (min 367, max 700, avg 501, Std 6). The other numbers for the DN calculation were taken from the ZWO webpage for my camera (asi2600mc). I eyeballed the values from the graphs they provide. I haven't done it in Pixinsight using BasicCCDparameters script yet, but will try that tomorrow.
Assuming the ASIAir is telling the truth then are you sure you are using offset 50 and didn't accidentally put another 0 on the end? Beyond checking that I'd definitely run BasicCCDParameters to see what it says.
Hey Linda, very clear and concise explanation! Gives me something to do under all this cloud!! So, question - on my Atik CCD camera, I don't have an offset value - at least, not one that I can change. What would you recommend/suggest I substitute here? Thanks!!
I have been trying this formula for the ASI533. Unity gain rn1.5, gain 1.0, offset 70 bit depth 14. I am coming up with 370 at swamp level 10. That seems very low
I get the same answer. With very low read noise it doesn't take a lot of signal to keep the read noise under control. You can expose longer but this should maximize your dynamic range. If you try it, let us know what happens.
The offset figure is incorrect. What you need is the ADU median of a bias frame. In my case it is 2796, but we need to divide this figure by 4 as it is already padded to 16bit. Remember, we are multiplying by 4 in the equation as we are using a 14bit camera, hence we don't want to do this twice. X10 SF for me is 2886 DN.
Hi Linda, such a helpful video. I do have one question I'm a bit unsure about. I'm using a QHY183m, QHY specify their gain and offsets differently to ZWO so I'm not sure the right value to use for offset/bias. If I am to use unity gain which is set to 11 in the driver, this obviously is 1e/adu for gain. However I'm not sure what number to use for offset. QHY specify 8 for their offset but if I use this in Johns formula, I get a pretty low DN of 488, i.e. ((10 x (1.5^2 / 1) + 8) x 2^16/2^12 I thought generally you would want a DN around 1000, depending on your sensor. Is this correct? Or should I be converting the offset into something else?
Daniel, if 8 is really the offset then that sounds right. I'm not familiar with that sensor or QHY cameras in general, but at unity gain, a bias frame ought to be around 9-10 ADU depending on whether the read noise quantizes up or down (or 151-160 DN in 16 bits). Is that what you are seeing? If it is, then the offset really looks like it is 8.
@@LindasAstronomyAdventures OK Linda I'll run a bias frame through PI and see what the ADU is, I assume I should be looking for the average value here? Should this just be an individual bias from or should I use a master?
Look for the median. I actually just posted a video today on estimating the offset with two methods: ua-cam.com/video/J1xi24yxvR0/v-deo.html. Let me know what that provides for you! I'd love to get some feedback on it.
Thank you for this very informative tutorial. But I have a question please: If the calculation resulted in an optimal background level of 1200, should I aim for this number for every filter whether broadband or narrowband?
I am here to review, as I had done as you described many a year ago. Today is 18-Apr-2024
Thank you Linda. 👍
Spent the past two days researching and trying to understand this topic in more depth and your video was the catalyst! Great presentation and nicely explained. Created the spreadsheet and calculated the values for my mono and OSC cameras. I will give it a try tonight! Again, thank you.
Joel, thanks! I'm glad you found it helpful!
Sharpcap Pro 4 sensor analysis and smart brain histogram functionality does this for you by measuring sky brightness and working out your optimal sub exposure/gain/offset levels automatically.
It does and Sharpcap is a great tool.
This is such a great resource. Linda, you have boiled down the math and the practical application into a 20 min presentation in a way that I never have seen before. Thank you!
Thanks! I'm glad you found it helpful!
Very informative. I can tell you that for Nikon and Canon DSLRs in N.I.N.A., the ADUs are not scaled and the 2^16/2^bits term at the end of the first formula is not neede.
Thanks Michael! And thanks for the info about the Canon and Nikon DSLR's!
Your timing on this video was perfect. Also the choice of camera made my life much easier :) Now I have some ball park numbers and now know why they are what they are. Thanks!
I'm glad it helped!
Really well done...so informative and clearly presented. A wealth of practical information presented with a most pleasing, almost hypnotic voice. Thanks for posting this.
I'm glad it was helpful!
Nice video Linda! Thanks for sharing!
Thanks, Mike!
I really appreciate learning how to determine exposure levels for gains/ adu.
Thanks, Jeff!
As a beginner (6 months into this hobby), this was very informative, despite that I don't have these software yet. The clipping of stars and blowing out the color has been an issue and I don't think I'm alone in this (just personal opinion browsing some images online). Light pollution filters I have experimented with is something I don't particularly like after a few tries, they seem to ruin the natural colors. I started to pay more attention to those ADU numbers and while significantly had to lower my exposure time without any LP filters , the stars seem to be so nice and sharp and all the colors from dark red to blue.
Glad it was helpful, Frank!
Thanks I watched it entirely I am going to try it. How do you calculate the number of subframes you need?
That's a good question. Largely it comes down to answering three questions:
1. How much patience do you have?
2. How bright is your sky?
3. How much patience do you have?
The darker your sky the less time you need. I have an image that I got in 90 minutes in a Bortle 2 sky that would have probably taken at least 8 hours in my Brotle 7 sky. Generally, here in my bright suburban sky I try to get at least 8 hours and more if I can. The fainter the object the more time you will need but I don't have any quantitative numbers to suggest.
If you manage four hours one night and you process and want less noise then get an additional four hours. If still too much noise then get eight more. Eventually you'll decide you've had enough or lose the object for the season. :)
Linda's Astronomy Adventures we just need a good old fashioned blackout, on a clear night with no moon!
Very informative. Cleared up a lot for me. Than you!
Thanks, Elliott!
Thanks much for the clear and calm explatations. Now myself using a ASI2600, my numbers would be from 50 to 105 ADU (at gain 101, offset25) . Normally I expose some 3 to 4 minutes long but if I calculate correctly, some 45 second might be ideal ! Wow that is short. I will have to try it out. Just to be clear here.. Optimal means 'with as little star clipping as possible ?? ' Am i right? Dennys, Montreal/Canada.
Technically "optimal" here means getting good stacking efficiency but as a side effect of that it does tend to get you less star clipping since we're finding the shortest exposures that get you to around 95% stacking efficiency (that is to say the stacked image is within 5% SNR of what a single exposure of that length would provide).
Linda this video is fantastic. Thank you so much for the clear and concise explanation. I have read a good deal about this process but didn't feel as though I had a firm handle on the process until now.
Bryan, thanks so much for your comments! I'm glad it was helpful!
Great thanks. I never thought of that.
Glad you found it helpful! I can't take credit for thinking of it either but I combined the work of several people.
Dear Linda, thank you very much for very useful video. I would like ask you for answer about Offset. I use for my astrophotography astromod DSLR. I know read noise, I know ISO/Gain in eADU, but I do not know how to evaluate Offset. Do you have some idea how to do it ? What value should I put into equitation ? My thanks in advance.
Sorry for the delay in responding. If you are using a ZWO camera I think they default to an offset of 50 but really the goal of the offset is to ensure you don't get any clipped pixels after calibration. You can look at the calibrated sub and see how many pixels have a value of 0. If there are some it probably makes sense to raise the offset enough to avoid that zero clipping.
Thank you for this valuable info. I learned something new.
Glad it was helpful!
Very helpful, thank you!
really good video. i will have to watch it like 10 times to get most of the info committed to mem. one thing that makes it harder is you keep switching between cameras. So following and being able to get a base line understanding is harder.
Brian, thanks for the feedback. Sorry it proved difficult to follow because of that.
@@LindasAstronomyAdventures np i was wondering i am using a asi294mc pro. I took it out to shoot orion and had -10 temp at 180 sec and 194 gain. so ok it came out way over exposed after stacking. so i found your video and after doing the calc i was left wondering is it better to play above 120 gain or below, because the cam has that weird divide where read noise drops like a rock at unity. So i am stumped as what to do. any input? yes i am brandy new to this so please forgive me. lol
from the calc you gave us it seems 120 gain at 300sec subs is optimal right or did i do that wrong?
Brian, this doesn't tell you the exposure time. but rather the exposure level. That is the the median value that your exposures should be. Exactly how long that will take will depend on how dark your sky is and what sort of filters you are using. For the ASI294MC Pro, at gain 120, the optimal value is 344DN (DN just means data number). Odds are that is only going to take 10-30s at most unless you are in very dark skies. The choice of what gain to use comes down to a tradeoff between read noise (higher gain gives lower read noise) and dynamic range (higher gain has less dynamic range). On your camera, though, there is that sudden step where read noise drops and dynamic range goes back up. So gain 120 gives you almost the same dynamic range as gain 0. Odds are pretty good that unless you are using some narrowband filters you can stay at gain 120 and stay in the sweet spot for read noise vs dynamic range. However, if you find that you are ending up with lots of 10s exposures you can drop to gain 0. That would raise your optimal level to 777 DN which would double your exposure times. This stuff can be a bit confusing (that's why I made the video) but the key point is it tells you what values you want to see off the sensor not how long to expose for. How long will depend on things outside of the sensor like filters and sky.
Aloha Linda,
Just came across you GREAT video (I guess better late than never) and watched it about 5 times! I'm also using an ASI 1600 mono camera and SGP and PixInsight and sometimes FITs Liberator (to view a sub-exposure). I got my spreadsheet built using your formulas (thanks!). I pretty much understand the formula, but I'm still getting confused and wrapped around the axle about all this scaling up and down business between 12 bit cameras and software (like SGP) displaying 16 bit values . So, based on the first example in the video (gain = 15, offset = 50, bit depth = 12, etc), the calculated min DN value is 920. It seems to me SGP scales up (from 12 bit to 16 bit) and the Rista formula 'scales down,' but SGP is still displaying a scaled up median value. The scale factor here of course being '16.' So, when I open up an ASI 1600 sub-exposure in SGP and look at the median value (which is being reported in 16 bits), does that median value in SGP need to be manipulated again? Or, can I just use the displayed median value (as is) in SGP, and compare that to the formula's min dn 920? Kind of stuck on this (and sorry about the long thread here)... Thanks in advance for any words of wisdom/clarification and Happy New Year! John
John, thanks! The camera manufactures report specifications in terms of the native bit depth of the camera, so 12-bits in the case of the ASI1600. However, the camera software scales this up to 16-bits so we never see the native values reported in software like SGP. We only see the already scaled 16-bit value. The formula does that same scaling for us so we can just think in terms of what the camera reports rather than what is really going on at the hardware level. That means you don't need to do any further conversions. You can use the values supplied by the spreadsheet directly. Hope that helps!
@@LindasAstronomyAdventures
Aloha Linda,
ok, thanks and thanks for the quick reply! I think I understand what you're saying (and I do remember you mentioned that the asi driver was scaling up [to begin with]). Not having to do anymore conversions is a BIG RELIEF. I was thinking SGP's mean value needed to be divided by 16 (and then compare that new value with the min or optimal dn value generated by the formulas). I guess I can always 'cheat' and use the histogram (as a backup) :)
Just spent about 6 hours looking at all those detailed Cloudy Night threads on all of this, but your video really helped. Due to possible quantization errors (caused by ADC converters), I noticed Mr Rista prefers to work with the analog signal (which I guess means electrons), so may need to review that approach again. And, I need to look at your offset video again to. Thanks again!
I just realized, thanks to you and your detailed video, that I have much to learn and study before I can even start understanding what I am doing. It has been 18 months or so, make a sequel please. Respectfully, Henri-Julien Chartrand, Montreal, Quebec, Canada.
I'll see if I can come up with anything useful to say...there are some ideas floating around that I've been thinking about. Hopefully I can get them to line up into something coherent.
This is a great video that I just now discovered. Many thanks. I do have a couple of questions. At 19:25 you mention taking a few frames that don't need to be under the stars and measuring the camera's performance level. I'm not clear on whether these are like flats or what sort of exposure if it is daylight and what would be the means of measuring, PI Statistics? Also, I have two OSC cameras, 294MC and 2600MC. If I load a sub (2600MC gain 100 offset 50 300s under Bortle 8) into PI it is labeled gray, presumably because it is not debayed. PI Statistics shows a median of 14656 on the gray image and 9557, 15784, 11086 (RGB) on the debayered image. Are these numbers commensurate with the DN numbers you talk about in this video? I think your images are all mono. Thanks again.
Apologies for the super long delay in responding. I'm woefully behind on youtube. The frames are darks and flats. You want to expose the flats as you would normally (to about 50% on the histogram). And, yes, those are the DN numbers.
Thank you for the video. Sorry, I am very new to this. The formula calculated DN (?), where I can see seconds/minutes of exposure? How I convert min and ideal DN values to exposure time? Thank you.
No problem. The DN value is the value of the pixels in your image, not the time. This procedure doesn't tell you the amount of time to expose but what exposure level to aim for. If you are using an acquisition program like SGP or NINA you can take some test exposures at different lengths and use the median value for the exposure in the image statistics to determine whether you need to increase or decrease. For example, if the formula told you to aim for 1,000DN and you tried a 60 second exposure and your median was 853 then you know you need to try a longer exposure. It's a bit of trial and error but usually within 4-6 tests you have the amount of time you need figured out.
@@LindasAstronomyAdventures Very good, thank you. I will be looking for the VALUE of exposures in SGP (what I am going to use) and through that I can come to time of the exposure. Like in your example, it has to be more than 60 sec.
You have more experience than I, however if a star is not over-exposed in the camera, then all you need to do is preserve what you have. If it is destroyed by processing, thats a processing error.
Yes, there's generally no harm in going longer as long as you don't cause too much clipping. However, it doesn't really improve stacking efficiency much to go (Robin Glover, author of Sharpcap has a video that talks about this in more detail). So, while it doesn't hurt to go longer, it doesn't help as much as an entirely new sub would.
Awesome. What do you aim for mean ADUs? I've heard anywhere from 800 to 2000
The value I aim for comes directly from the formula I presented. But, yes, for my ASI1600 I usually end up in the 800-1200 range depending on the exact gain and offset I'm using.
Awesome stuff!!!!
Thanks! Glad you liked it!
Linda's Astronomy Adventures i have to listen and watch again. Cant wait for my tripod to arrive so I can get back in the game. THANK YOU! Have a wonderful thanksgiving
Do you have a link to your calculator you were using?
Dustin, here's a read-only link to the icloud numbers spreadsheet I used: www.icloud.com/numbers/0T96cjAMDXFZBLcgxCVqC7UZA#Optimal_Exposure_Level_Calculator. You can copy that to your own spreadsheet. If you can't access it for some reason or would prefer to create it in excel or google sheets here is what the cells are:
b1: read noise
b2: gain
b3: offset
b4: bit depth
b6: min dn = (3×B1^2÷B2+B3)×2^16÷2^B4
b7: optimal dn = ((10×B1^2÷B2)+B3)×(2^16÷2^B4)
@@LindasAstronomyAdventures Gosh thank you so much for this!!!!
Hi Linda. Thank you for producing and posting this tutorial which is very clear. I like technical stuff but can't help wondering why can't we just figure it out in the field by trial and error, watching for when stars become saturated. Also, does this need to be done per filter? I am assuming so.
I was reading somewhere else (www.lightvortexastronomy.com/measuring-your-camera-sensor-parameters-automatically-with-pixinsight.html) and they are saying readout depth is actually that of the image bit depth and that is usually 16 bits. You seem to be using the same bit depth as for ADC. The difference in results of the readout noise is considerable.
The optimal level isn't about avoiding saturation (though that's important). It's about burying the read noise of the sensor with enough signal from background sky (which is mostly light pollution for most of us) to hide that read noise. You can go beyond that optimal exposure but then you run the risk of saturating more stars than you needed to. All that being said, what you described would probably work fine for the CMOS cameras that have very low read noise. It might be a little trickier on a high read noise CCD to know for sure whether you went long enough.
Hi Linda, great video thank you. I'm a little confused about my results though and was wondering if you can help please? I use an ASI533 and from the website I can see that my figures should be read noise of 1.5, gain of 1, offset for this camera is default 70 and bit depth is 14. That gives me a Min DN of 307 and Optimal of 370. When I look at some of the images I've taken, my mean ADU value is around 3000 which is almost 10x the optimal value...and that's only a 30 second exposure. I live in a bortle 5/6 area so not super bright or super dark, but does that mean my optimal exposure times should be almost 10x less than 30 seconds? So like, 3-4 seconds lol. I'll have no space on my PC! Surely I'm going wrong somewhere? Thank you :)
That sounds about right. With the gain at 1e-/ADU and the very low read noise it doesn't take much to bury the read noise in the sky glow even at Bortle 5. You have a couple of choices. You can just live with it. There's no harm in exposing longer so long as you aren't clipping more stars than you are comfortable with. The other thing is you can lower your gain down to 0 from 100. That raises the read noise to about 3.8 to 3.9 e-/ADU according to ZWO's graph. That would let you expose longer at lower gain so risk less clipping. However, either way you are capturing the same number of photons. The difference is in the read noise and the dynamic range. With that camera at unity gain you don't appear to be sacrificing much dynamic range at gain 100 so you should be ok exposing longer but let your stars guide you. If you are clipping too much then lower the gain.
@@LindasAstronomyAdventures thank you for your reply. If the clouds ever go away I might try some experiments and see how I go.
I feel your pain about the clouds. This summer and fall have been brutal here. Good luck!
Philip are you sure your calculations are correct? I get optimal at 2890..
@@richards3192 pretty sure, but I could be wrong.
Hi Linda, with this calculator I've got min 77 and optimal 141. In my 300sec exposure with PI statistics the median is 0.00923 which is way off. But my photo looks okay. Where would you suggests should I look at to determine what's going on please? Thank you.
Okay, the problem was that I didn't use 16bit in Statistics.
Oh, good. Sorry for the delay in responding but that's the right answer.
Hello Linda, I used your video to determine my optimum exposure. I am imaging M101 Pinwheel Galaxy. I have the same camera ASI1600MC cool (but colored) so I assumed that the specifications will be the same as in the video (Optimum DN = 1290). I am in a rather high light pollution area and have a Baader LP filter attached to the camera. A single sub-frame of 45 seconds with gain = 25 and offset of 10 and temperature setting at -10C. This is giving me a median of 1136. If I try to increase the exposure time, I will go over the Optimum DN. But I get a sub-frame that has very little luminosity even after stretching, I can't see much of the galaxy.
I took a total of 100 sub-frames for a total of 4500 seconds with relatively good guiding. I have not tried stacking it yet. I am using PixInsight. Will I actually get something out of these sub-frames after calibrating with darks, flats and bias and later stacking? I have never tried this approach and don't really know if I am doing it correctly.
If you are using an offset of 10 then your optimum exposure level will actually be 640 DN lower than mine in the video. That's because your offset is 10 instead of 50 which will lower the optimum value by 640 (16 * 40). You are actually going much longer than you need to. There's no harm going higher though you might clip a bit more data than you might otherwise. As to the data looking thin...yes, it does. You don't need much signal to bury the read noise since the ASI1600 has such low read noise even at low gain. But, when you stack up a bunch of them suddenly the signal pops up out of the noise.
@@LindasAstronomyAdventures Thank you for your reply. I did some image integration on PixInsight and was able to show the galaxy but I think I will need to re-image to get better results. I will watch your next video on the optimum Offset.
@@LindasAstronomyAdventures - I had a chance to do some imaging last night and followed your method. I was imaging at Optimum DN of around 1290 at the beginning but as the night went on I noticed that the value start to increase to 1480++. In that case, what should I do? decrease gain or decrease exposure time or simply leave it as is?
After calibrating with darks, flats and bias with PixInsight, I still get a fair amount of noise and amp glow. Can you still advise me what to do?
By the way, I am in the inner city limits and I would say Bortel scale 8. Hence making it more difficult to capture images. I would like to try getting some images in really dark locations to compare...Your thoughts will be appreciated.
Great, detailed video. I learned a lot, thanks for sharing!
Thanks! I'm glad it was helpful!
For the ASI294mc Pro I'm getting Min DN 160 and Optimal DN 253. These seem extremely low.
That does seem on the low side. What gain and offset are you using?
@@LindasAstronomyAdventures Unity Gain which is around 0.8e and offset 50 is giving me Min DN 260 and Optimal DN at 400. Still seems low to me.
At gain 120 where it switches over to high conversion gain mode, I'm getting optimal at 226 DN so I don't think you've done anything wrong. With only around 0.8e- of read nose it doesn't take very much signal to overcome the read noise. You can expose longer if you want, but it won't help your stacking efficiency much.
@@LindasAstronomyAdventures So hypothetically speaking we're talking like 5 sec exposures? Am I in the ballpark or dreaming with these numbers? lol I guess since I'm using Hyperstar a 5 sec exposure is equivalent to about a 1 minute exposure at say f/5?
It certainly could be depending on the sky you have. Between f/2 and low read noise it doesn't take much. Going from f/4 to f/2 gets you 4 times as many photons/second so yes, your exposures could be insanely short.
Super video. Thanks. I learned more about PI. There's a minimum and a maximum exposure for any given target. The minimum should be whatever exposure just swamps the read noise. The maximum should be where a linear, unstretched image shows very few if any stars. A fully stretched image should then show the structure. N.I.N.A. has an exposure calculator built in. I've determined that seeding with a 1 second exposure produces a recommendation for an optimum exposure. Getting the camera statistics from SharpCap Pro provides the necessary information for the N.I.N.A. calculator.
And of course, the PixInsight method shown in this video will work. But, which method is less complicated? For me, the SharpCap-N.I.N.A. process was easier and quicker to arrive at. Using that process provided me with a perfect optimum recommended exposure to capture M13. I tried other exposures, and none of them produced as good an image as the recommended exposure did.
This might be a duplicate reply as I thought I responded to it earlier today but don't see it htere so just in case... Yes, sharpcap is a fantastic piece of software and nina's ability to consume it's exposure info makes for a great calculator. Robin Glover, the sharpcap author, has some great videos on exposure that are worth viewing. In particular, I'd recommend this one: ua-cam.com/video/3RH93UvP358/v-deo.html. He points out there there isn't really much benefit to going beyond the optimal subexposure length since it doesn't help fhe final stack much. The other drawback is the potential to clip more stars (or with some cameras contending with more amp glow or more odds for tracking error). Also, depending on the faintness of the target, it might be quite subtle or even invisible in an optimally exposed frame but will pop out when you do the final stack so looking at detail in a screen stretch may not be an accurate way to judge. Of course, if your target is very faint and you're willing to sacrifice more stars to clipping then it may still make sense just to get more signal you actually care about but that's likely an unusual situation for mos of us. But, your point is well made...there is a fairly broad range where you can be in the optimal zone and it needn't be complicated to get there. Thanks for your comment!
Thank you
Glad you found it helpful!
I must be missing something, when i click on the link i cant input any of my settings such as readout noise gain etc
Paul, the only two permissions I can share it with are "view only" or "make changes". The latter would let people change the formulae which could mess up the spreadsheet for others. You should be able to save it as a copy so you can change the values. Let me know if iCloud isn't letting you do that and I'll see what can be done.
@@LindasAstronomyAdventures thank you Linda will try now
nope couldnt do it lol, i do have some info though if you could put me in the ball park my gain is 100, offsert10. (readout noise 1.575 and gain 1.024 pixinsight )
I'll look for an alternative way to share it. But, in the meantime, assuming your camera is a 16 bit sensor then I get 34 DN as an optimal value and 17 as a minimum. If you have a 12-bit camera like the ASI1600 then it's 548 optimal and 276 minimum.
@@LindasAstronomyAdventures its the 533 osc so its 14 bit but behaves like a 16 bit? does that make sense ?
Fantastic!!!!
I'm a newbie and you clarified soooo much
Thank you for taking the time to make the video
As always
Clear Skies,
God Bless
This seems to be the number one question that we all have. I'm glad it helped clarify things!
Why is your offset 1.00 in your frames? You stated you used 15 offset for the ascom driver but your files show 1.00 offset. ?
It's been quite a while since I did that video so I'm not sure what you are referring to. Do the filenames say something like "offset_1.00" in them? Or are you referring to something else? If you can point me to it, I might be able to respond more effectively.
@@LindasAstronomyAdventures hi, and yes ,if you look closely at the file names you put into pixinsight,they say offset 1.00. So I was confused.
I think you are misunderstanding the offset part of the equation.
The figure is meant to be the median ADU of a bias frame.
No apologies needed. Can you clarify what you are referring to? The bias frame includes both the offset and the read noise.
@@LindasAstronomyAdventures Essentially, the offset part of the equation is the median bias ADU. Can you take a look at a thread on Stargazer's Lounge? It's called 'Swamp Read Noise' by Pitch Black Skies, posted April 19th.
You will see what I am talking about.
The median bias contains both the offset and the read noise. I took a look at he thread and it seems to be saying the same thing I am. Is there a specific post in that thread you are referring to. It's possible I missed it.
@@LindasAstronomyAdventures For instance, at Gain 100 the 533mc-pro camera has an offset of 70. But we don't insert 70 into the equation.
That figure translates to ~2800ADU which is a 16-bit number (the bias median ADU). The offset part of the equation is therefore ~700 which is 2800 converted to 14-bit.
I've noticed a few posts down that some other 533 user's are questioning the low number result in their calculations. They used 70 as the offset figure which is incorrect but you didn't mention it to them.
All I'm trying to say really is that the offset figure isn't the default offset of the camera or the one we change it to, rather it is that number converted to ADUs.
For a 14-bit camera everything gets scaled by 4 so the offset of 70 ADU would translate into 280 DN since the camera reports in 16-bit values. If I have the right vales from ZWO's graphs, then it should be (10 x 1.5*1.5/1 + 70) * 4 to get the optimal value or about 370 with a "swamp factor" of 10. You could get away with as low a value as 307.
The 533 at gain 100 has a very low read noise of only 1.5e- so it really doesn't take much to bury that read noise in the sky glow.
I'm calculating between 77-140 for the asi2600mc pro at Gain setting of 100 (RN=1.5, Gain=0.25 e-/ADU, Offset = 50, ADC =16). Is that accurate?
That's what I get also.
@@LindasAstronomyAdventures could you please explain how I'm supposed to use the DN (ADU) number exactly? If I simply take a 0.001s bias with the 2600mc pro, I get an avg ADU of 500, which is already way over the 77-140 calculated above. Am I supposed to add that 77-140 on top of my baseline ADU from a Bias frame and aim for that in my sub exposure?
Something isn't correct with one (or more) of the numbers you reported) most likely. That formula takes all your settings into account so that's what you want your median ADU to be near in an absolute sense. If you are getting 500 on a bias then something is off somewhere. How did you come up with the numbers you got?
@@LindasAstronomyAdventures the 500 ADU on bias frame is from Asiair Pro histogram's AVG reading (min 367, max 700, avg 501, Std 6). The other numbers for the DN calculation were taken from the ZWO webpage for my camera (asi2600mc). I eyeballed the values from the graphs they provide. I haven't done it in Pixinsight using BasicCCDparameters script yet, but will try that tomorrow.
Assuming the ASIAir is telling the truth then are you sure you are using offset 50 and didn't accidentally put another 0 on the end? Beyond checking that I'd definitely run BasicCCDParameters to see what it says.
Hey Linda, very clear and concise explanation! Gives me something to do under all this cloud!! So, question - on my Atik CCD camera, I don't have an offset value - at least, not one that I can change. What would you recommend/suggest I substitute here? Thanks!!
Darren, I'm working on a followup to answer your question. I hope to have it out in a few days.
Darren, I just posted the response to your question: ua-cam.com/video/J1xi24yxvR0/v-deo.html. Let me know if this was helpful.
I have been trying this formula for the ASI533. Unity gain rn1.5, gain 1.0, offset 70 bit depth 14. I am coming up with 370 at swamp level 10. That seems very low
I get the same answer. With very low read noise it doesn't take a lot of signal to keep the read noise under control. You can expose longer but this should maximize your dynamic range. If you try it, let us know what happens.
I got the same answer! whew thought i did something wrong......I'm going to try this tonight on Andromeda.
The offset figure is incorrect. What you need is the ADU median of a bias frame. In my case it is 2796, but we need to divide this figure by 4 as it is already padded to 16bit.
Remember, we are multiplying by 4 in the equation as we are using a 14bit camera, hence we don't want to do this twice.
X10 SF for me is 2886 DN.
Great video! Thanks for the clear info
Thanks, Kyle! Glad it was helpful!
Hi Linda, such a helpful video. I do have one question I'm a bit unsure about. I'm using a QHY183m, QHY specify their gain and offsets differently to ZWO so I'm not sure the right value to use for offset/bias. If I am to use unity gain which is set to 11 in the driver, this obviously is 1e/adu for gain. However I'm not sure what number to use for offset. QHY specify 8 for their offset but if I use this in Johns formula, I get a pretty low DN of 488, i.e. ((10 x (1.5^2 / 1) + 8) x 2^16/2^12
I thought generally you would want a DN around 1000, depending on your sensor. Is this correct? Or should I be converting the offset into something else?
Daniel, if 8 is really the offset then that sounds right. I'm not familiar with that sensor or QHY cameras in general, but at unity gain, a bias frame ought to be around 9-10 ADU depending on whether the read noise quantizes up or down (or 151-160 DN in 16 bits). Is that what you are seeing? If it is, then the offset really looks like it is 8.
@@LindasAstronomyAdventures OK Linda I'll run a bias frame through PI and see what the ADU is, I assume I should be looking for the average value here? Should this just be an individual bias from or should I use a master?
Look for the median. I actually just posted a video today on estimating the offset with two methods: ua-cam.com/video/J1xi24yxvR0/v-deo.html. Let me know what that provides for you! I'd love to get some feedback on it.
Thank you for this very informative tutorial. But I have a question please: If the calculation resulted in an optimal background level of 1200, should I aim for this number for every filter whether broadband or narrowband?
Yes, if you are using the same gain and offset for broadband and narrowband then your target ADU level is the same for both.