Quote:
Originally Posted by RickS
I think I need to go do some research on how image sensors work and think about this some more...
|
OK, I did a bit of reading. There are several sources of noise, but there are effectively two types:
- noise independent of exposure time (e.g. readout noise generated when the pixel values are read from the sensor, amplified and converted from analog to digital). Let's call this NI.
- noise dependent on exposure time (e.g. thermal noise aka dark current). Let's call this ND.
We also need to know that the ISO value doesn't affect the image sensor at all - it determines the amplification used when the data is read from the sensor (and amplifies noise as well as signal).
The reason that a higher ISO in conventional DSLR photography is more noisy is because of NI noise. As an example, let's say this noise is 10 electrons per pixel. If a pixel in the sensor is receiving 100 photons a second and we assume 100% efficiency:
- an exposure of 8 seconds at ISO 100 results in 800 photons being collected and 800 electrons being generated. The signal to noise ratio is 800:10 or 80:1.
- an exposure of 1 second at ISO 800 results in 100 photons being collected. The signal to noise ratio is 100:10 or 10:1.
In the second case the output is amplified 8 times more to give the same exposure, but the noise is a much higher proportion of the result.
The longer the exposure the smaller the impact of NI. I believe that's why Luigi suggested exposing for as long as possible.
Assuming that ND increases linearly (which appears to be approximately correct) exposure time won't affect it's contribution to the signal to noise ratio.
Does that all make sense? I've used a very simple model and the real world is more complicated, but I think this offers a useful insight. I'm also happy to be corrected if somebody more knowledgeable comes along!
Quote:
|
But....would an 8 sec exposure at 800asa have more noise than an 8sec exposure at 100asa?
|
The sensor would produce exactly the same data. The only difference is that for ISO 800 the camera would amplify the data by 8 times more which would increase both the signal and the noise by the same amount. The signal to noise ratio would be the same. Of course, in practice this might cause a digital overflow/overexposure, but we all know how to avoid that
Cheers,
Rick.