« on: January 29, 2015, 05:48:00 PM »
Great stuff, guys! Your really progressing!
One piece of advice: Try to get longer exposures, rather than stacking tons of short exposures. Noise does average out, if it is the right kind of noise. Thing about read noise, it adds in a certain way, and it does NOT average out like dark current noise or photon shot noise.
Here's the math (simplified a bit, it ignores other factors that I don't want to get into with you guys yet):
SNR = (n * S)/SQRT(n * (S + D + R^2))
n = sub count
S = signal
D = dark current
R = read noise
Notice here...R is SQUARED in the square root term. That basically makes it a constant. While everything else is scalar, R is squared, so it compounds as you combine more and more frames. It's the one kind of noise you have to watch out for, as if you combine lots of frames (115, for example) the read noise starts to add up, while all the rest of the noise averages out. The signal is combining as well, and the signal grows faster than the read noise, but read noise does not average out like everything else, so you have to be careful with it.
So, with the math out of the way, what does it mean? It means that getting the highest signal per sub is best. More specifically (and this involves more complicated math), you want to get your signal strength high enough in each sub that you totally swamp read noise with your "background sky". When you do that, you become photon shot noise limited, and read noise is effectively dealt with by a simple offset (i.e. when you stretch your data, you just shift the black point up, and that reduces the read noise contribution to zero.) Doing this with a DSLR doesn't actually require that you know the math...really, exposing long enough that the histogram for your images peaks at the 1/3rd mark (from the left-hand edge on most cameras and in all software, where the black point is) is generally enough to swamp read noise unless your combining more than a couple hundred frames.
I therefor strongly encourage you guys to get the longest subs you possibly can, before you start getting elongated, eggy, or trailed stars. That generally means exposing for much longer than 1 second, and that requires some kind of tracking mount. But the difference is massive. Even if you go from one second to five or ten seconds (which, without a tracking mount, if you have large enough pixels and/or a short enough focal length, that could be possible), that would be a very significant difference in per-sub SNR, which would result in significantly better integrations.