Jon, I know, this question has been asked many times, opinions are divided though. Your vision would be very valuable to me.
Does the pixel size affect total exposure time? Does the sensor with smaller pixels need longer exposure to collect the same amount of signal (presume all other properties are equal)?
It seems, the IMX183 is a very sensitive sensor and its tiny pixels are great for resolution, but how about exposure times? This question emerged once again to me after seeing a single 300 sec frame of ZWO294 camera.
Thanks in advance.
On a non-normalized basis, yes. Absolutely, the pixel size affects exposure time. If you are going for resolution, then it will be a factor. This is actually a part of the reason why I am using 10 minute subs, vs. the 3 minute subs I used with the ASI1600. The smaller pixel size collects fewer photons per unit time, so to get the same pixel SNR you need longer subs. The other part of the reason is the higher read noise of the ASI183. It's 2.4e- vs. 1.3e- (for my narrow band work), so that, too, required me to increase exposure times.
Note that on a normalized basis, the differences in Q.E. are not as large (or rather, they remain unchanged as it's an instantaneous factor). They mostly boil down to the read noise difference. Also note that Q.E. affects exposure time, but the Q.E. difference isn't even going to account for a single stop in this case...84% vs. 60% (~1.4x), whereas the strait up read noise difference is a stop...2.4e- vs. 1.3e- (~1.9x), and the area-normalized read noise difference is over a stop...3.8e- vs. 1.3e- (~3x). The read noise, on a non-normalized and normalized basis, is the more significant factor.
Factoring in the 1.4x BENEFIT to the ASI183 due to Q.E. against the 3x DETRACTOR from the ASI183 due to read noise, plus accounting for glow and dark current, I came up with 7 minutes as the necessary exposures to get the same kind of SNR per sub as I was with the ASI1600 at 3 minutes.
I ended up shifting from a moderately high gain of 150-200, down to gain 53, in order to balance out my dynamic range as well. At gain 200, the ASI183 had less DR and more read noise (by a small amount, ~1.5e- vs. 1.3e-) than the ASI1600. While at a higher gain, I needed maybe 5 minute subs, I was clipping stars a bit too much. Moving to gain 53, I used 10 minute subs, but did not clip stars as much. It's always a balancing act.
I shared a single 10 minute Ha sub earlier in the thread, in a comparison of calibrated vs. non-calibrated to show amp glow correction. Note that even though it is narrow band, I have extremely high LP, plus moonlight, plus haze...so it isn't the best of subs (few of my subs are any better these days....it's amazing how bad the LP has gotten around here, even 3nm filters suffer too much background contamination IMO). I did see the ASI294 single exposure, and I am pretty sure that guy is at a fairly dark site, deep yellow zone, maybe green zone. So take that into consideration. I have many 5, 7 and 10 minute single sub images from my green zone dark site that look just as good. The power of a dark site is superior to just about any camera technology, by a long shot.
Jon, thank you for the detailed answer. I do not want to hijack this thread, but this question is directly related to the ASI183 too.
To compare apples to apples, let's consider we have two hypothetical cameras with the equal sensor size, QE and the area-normalized read noise, but with different pixel size and therefore full well capacity. The pair of CCD's - ICX695 and ICX814 are in good approximation to this criteria.
3e RN per 4.54 µm pixel for ICX695 http://www.flicamera...eets/MLx695.pdf
2e RN per 3.69 µm pixel for ICX814 http://www.flicamera...eets/MLx814.pdf
So, on one hand, the camera with the smaller pixels will collect less photons per time interval. On the other hand, the camera has less full well capacity and different gain setting, therefore we should get almost the same ADU distribution histogram and SNR (less RN!) at the end. Does this mean that the both cameras will reach SN ratio in equal time, but the one with smaller pixels would have higher resolution as a bonus?
I know, something is wrong with this conclusion, there are a lot of examples that demonstrate advantage of the larger pixels for collecting the signal faster. But what?