Jared Willson:
Generally speaking, there is enough dynamic range in most astronomical subjects that you are not going to be able to simultaneously swamp read noise and avoid clipping the stars. You will have four choices: expose for the shadows and lose the stars, expose for the stars and lose some shadows to the higher noise floor, find a compromise exposure that balances the two, or take separate exposures for stars and subject.
With narrowband subjects, I would recommend separate exposures for stars and main subject since most people prefer RGB stars along with the HOO or SHO main subject. For broadband, I would recommend the "compromise" approach of not clipping too many stars, while still choosing subexposure durations that let you get deep into the shadows with a minimum of integration time. The particular compromise may vary depending on the subject and how important the faintest areas of the subject are to the overall quality of the image. Also, keep in mind that additional integration time will help even if you don't have "optimal" exposures to swamp read noise.
The usual recommendation is to take subs that are long enough to reach either 5x (read noise)^2 or 10x (read noise) ^2. Under Bortle 2 skies that may be a fairly long subexposure depending on the gain you choose. I would recommend on settling on either 0 gain or 100 gain. 0 gain will (technically) have the highest dynamic range, allowing you to clip the fewest stars while reaching deep into the shadows. Likely, you will find your read noise is somewhere around 3.2 e- and that your gain is somewhere around 0.78 e-/ADU. If that's the case (you'll have to measure it), then 5x rn ^2 would be around 50 e- which should be around 64 ADU above the offset. So, if you take a bias frame (or a dark frame--they will be about the same) and look at the median ADU, you'd want an exposure that has a median around 64 ADU above that value in empty areas of sky at, say, 50 degrees altitude (a compromise value that likely represents the average for your subjects across the night). If you have a monochrome camera, I would measure this with all three filters. If I were shooting with a relatively wide field of view, as you are, I would go with the 5x (read noise) ^2 rather than 10x simply because you are likely to have more bright stars in the field that are going to saturate.
How many pixels are saturating with that subexposure duration on gain 0? If it's thousands and thousands, you may want to consider separate exposures for stars and nebulosity, a la narrowband imaging. If it is 1,200 or fewer (an arbitrary line), I would just go with this as the standard, and only deviate when there is a reason to.
You can run the same calculations at a gain setting of 100. For that, you will likely find a read noise somewhere around 1.3 e- and a gain of around 0.25 e-/ADU. You'll have to measure your camera. So, you would want to expose till you had a mean ADU of around 34 or so above your bias frame median. That would be 5x (read noise) ^2 assuming your camera roughly matches ZWO's specifications. This will result in a much lower subexposure duration to "swamp" read noise, but will likely have slightly more clipped stars (65,535 ADU or so--not every camera quite reaches that value).
Which is the better choice? Gain 100 or gain 0? Assuming similar numbers of clipped stars? I generally prefer the lower gain since dynamic range is usually marginally better, and I'm happier having a bunch of five minute exposures rather than a whole lot of one minute exposures--less storage and easier processing. You'll have to answer that one for yourself, though. I think more people standardize on gain 100 than gain 0.
Don't feel bad if you find that 5x (read noise) ^2 results in too many clipped stars. That isn't an indication of a problem--it's an indication of really, really dark skies! That's a good thing. If you need to cut back on subexposure duration to avoid clipping too many, so be it. You may not be swamping read noise in each exposure, but that's only because you have so little shot noise! Definitely good. You'll have to use your judgment on how many clipped pixels is too many. It will depend on your taste, processing, and on your subject. You can always take separate shorter exposures for the stars if you want to, but at least figure out what the "optimum" exposure is for the dark portions of the image. That will give you a baseline of, "well, I probably will never need longer exposures than this."
Hope this helps.
- Jared
Jared,
Thank you so much for the detailed response. While I don't quite get all the math, you've definitely put my mind at east a bit. It's been frustrating reading so many different things. I'm used to that with this hobby. But when Joe, Bob, and Sally have three different answers, at least they're usually in the ballpark of each other. Upon further reading this morning, I've found answers ranging anywhere from 200 to 2% which would equate to about half a million on my camera. Understandably, I'm not quire sure what to do with that information.
I mentioned that I don't quite get all the math. But I have run the sensor analysis in SharpCap and applied it to the exposure calculator in NINA. So, I should be ok there. I've only tinkered with it so far. But all the numbers are in there. All I have to do is point and click to get a number.
Ideally, I prefer 0 gain for the same reason you do. I'd never tried it until the last few weeks. I'm not experienced enough to say for sure that it's not the target itself, but the first thing I noticed in that shot were the different shades of red I'd never seen before. I also saw the same thing in the Elephant Trunk shot I just did this morning. There seems to be a lot more detail and contrast. That being said, there's a reason I still use 100 whenever possible. Shooting broadband, I only shoot from a dark site with no moon. That limits how often I can shoot and, therefore, limits my integration time. I have no desire to spend months on one target. So, whatever I get that single night is it. Of all the shots in my gallery, I think the highest integration time is about 5 hours. With so little integration time, I need all the help I can get in reducing noise. 100 gain does that beautifully. But I've been toying around with broadband from my Bortle 6 backyard. For that, I can't imagine not doing 0 gain. The only reason I did Elephant Trunk at 0 gain the other night was because I was saturating the stars at 100 gain and didn't want to reduce to 30s exposures. The less subs to stack, the better.
If I recall correctly, my first test shot at 100 gain/120 seconds saturated about 6,000 pixels. 100 gain/60 seconds saturated about 2600. And 0 gain/60 seconds saturated about 1300; which is what I went with.
I did consider the separate exposure idea. But my recent experience with Sadr and one of the bright stars in Cygnus led me to realize the problem I described in an above post. And that is the large halos those bright stars leave in the long exposure starless image. Although not ideal, it still looks "normal" putting that bright star back in there. But putting a small star back into a large halo wouldn't look quite right. I know there are ways to fix that. But with my current processing skills, I think that would present a bigger problem than the one I'm trying to solve. But I'm definitely considering it for the other 90% of targets I plan on shooting.
I'm unfortunately restricted to doing all of my processing on a rather slow laptop. So, I'd love nothing more than to be able to set every sequence to 300 seconds and let it go to save processing time. But I'm also to a point where I think my images are finally getting good and I don't want to sacrifice that just because me and/or my laptop are lazy. If I have to take 20 second exposures every night to keep the saturated pixels under 1200, so be it. But if there's a compromise that allows me to go longer without sacrificing quality, then that's the route I want to go.
