I am not a scientific person, but I would appreciate some information about low light from one or more of FMs more technical experts.
My question may be confused, but sincerely asked. I love low light photography, and want to understand more about the nature of low light.
When you look through a light intensifier scope the image looks noisy. When you shoot a DSLR in low light and at high ISO it looks noisy. When I look around a darkened room at night my vision looks noisy. I think I've read in astronomy magazines that very faint light from distant objects arriving in large astronomical cameras is "lumpy" - presumably another kind of noise?
My question is about the various kinds of noise we experience in low light, what noise is artificial (caused by the eye/sensor) and what is caused by the inherent qualities of low light? What is low light like, is it really discrete and discernable "chunks" or is its lumpiness only visible to lab quality equipment?
Sorry for the rambling question, but I'm striving to understand something here!
The question is very good. There are different sources of noise. Some have to do with the equipment. But one -- shot noise -- is in the nature of light. Think of light as being like rain falling. Sensors are lots of buckets (one per pixel). Not every bucket will catch exactly the same number of drops. The variation from bucket to bucket is random but follows a nice statistical distribution. The "standard deviation" of this distribution is the noise. For the kinds of stochastic processes we're talking about (like rain falling into buckets--a random process and specifically a Poisson process), the standard deviation increases with the intensity of the signal, but not proportionately! Rather, the noise increases as the *square root* of the signal. So in other words, the *ratio* of signal to noise gets better (fast!) as intensity increases, and gets worse as it decreases. The physics of light dictate that the signal to noise ratio is low (noise is high) when the intensity is low. If for example intensity is 4 then noise is 2 and S/N = 4/2 = 2. But if intensity is 100 the noise is 10 and S/N = 100/10 = 10. You've got 5 times better signal to noise ratio at intensity 100 than at intensity 4!
So now along comes Jonny with his DSLR. The light intensity is really low. He takes a shot at ISO 12800 and is then disappointed when the result is noisy. The exposure as presented in the camera-produced JPEG is fine, why is the image noisy? Well, what the camera did is capture an image with a low signal to noise ratio (i.e., a noisy signal, like intensity = 4, S/N = 2), and then it *amplified* the result to make it "bright" enough (100). But amplification does the same thing to the noise as to the signal, preserving the ratio. So Jonny ends up with an image that bright enough but that still has a S/N ratio of 2! If instead the light had been bright enough to require no amplification (ISO = 100 or 160 perhaps), then Jonny would have gotten a JPEG with a S/N ratio of 10. That is pretty much how it works.
Note: If you use bigger buckets, they'll average out rain over a larger area and the variation will be reduced. In cameras, larger more sensitive per-pixel sensors produce better signal to noise ratios. This is why point and shoots are lousy: tiny buckets mean high variation from one to another: low signal/noise ratios.
Final note: Crappy electronics are also noisy. So less expensive or advanced cameras with lesser electronics tend to produce less clean images. External phenomena such as heat or electromagnetic fields can also introduce variations.
Electronics can be improved and per-pixel sensors can be enlarged, and this is how to get lower noise cameras/sensors.
