|
I would have put it that the signal is appearing above the read noise sooner.
the total number of photons striking the sensor for the really dim stuff is independent of the sub length - it depends only on the total viewing time. It is quite OK to have some of the subs with no target photon in them - in floating point average stacking, you can end up with a signal equivalent to a fractional photon - this is the probability that a photon would be detected in any given sub and it is a perfectly valid signal, even though there is no such thing as a fractional photon. The only advantage of long subs is that the combined read noise from the multiple subs will be reduced if there are fewer of them - you don't somehow magically increase the detectability of faint signals by using long subs and you certainly do not need to have at least one target photon in each sub for the signal to be valid.
Take the mentioned example, where you have on average 1 detectable photon per hour and image for (say) 10 hours. With a a traditional camera using 10x subs 60 minutes long, you would detect ~10 photons in total - maybe a sub or two would have no photons, many would have 1 photon and some maybe more than one. With a CMOS camera using 100x 6 minute subs, you would still detect ~10 photons - most subs would have no photons, but ~10 of them would have 1 photon and occasional ones might have more. After stacking 10 hours of data, the signal result is the same for each camera (~10 photons detected). The read noise adds in quadrature, so the total read noise increases with the square root of the number of subs. For the chosen examples, let's assume that the 10x60 minute subs are taken with a traditional camera having 10e read noise - after stacking, the total read noise for these 10 subs will be 10xSQRT(10) = 33e. If the 100x6 minute subs are taken using a camera with 2e read noise, the total read noise will be 2xSQRT(100) = 20e. ie for this example dim target, the lower read noise camera produces the same signal and roughly half the noise, even though it is using very much shorter subs.
There is always a slight chance that a photon will come along in the downtime between subs - just image for a bit longer.
Last edited by Shiraz; 28-09-2016 at 04:15 PM.
|