Originally posted by jolepp The higher ISO values are implemented by turning up the gain of the analog amplifiers in the signal path. A side effect of this is that the noise inherent in all electronics gets amplified too and the finest differences in pixel brightness levels are lost or replaced with random data. Incidently the result is like the film grain of high ISO film where fine detail is lost too.
This is correct to a point. . . The lower light levels have something to do with this also. . . If you can picture light as photons (particles), and the individual sensels as buckets, the lower the light level, the fewer particles there are to fill the buckets.
A good analogy would be to take a small quantity of salt and sprinkle it on a dark surface. If you drop it from a relatively high point, you get a good distribution of salt, but there are areas in between the grains that are not covered. The more salt you dump, the more dark area that's covered, until the whole surface looks white.
Sensels produce electric current in proportion to the number of photons that hit them. At very low light levels, there are not enough photons to fill the buckets, so even in an area of the image where the light should be even, there are small variations in the bucket fill. This means the sensels don't create equal amounts of current -- the deviation is small, but when you amplify the signal, the differences are also amplified. This appears in the image as the grain you see. There is also a corresponding falloff in contrast and color saturation and accuracy when the buckets don't fill.
In the case of the Sony Exmoor sensors that Pentax is using in the Kx, Kr, and K-5, there is less light absorbing material that the light has to pass through, so considerably more photons reach the buckets for a given amount of light exiting the back of the lens.
You might ask how this is practically applied to photography. . .
A correct exposure at a high ISO setting in good light and the same correct exposure in low light at the same ISO setting will give you different levels of noise.
This is because in good light, you are limiting the number of photons that pass by using smaller apertures and/or faster shutter speeds -- there are more than enough photons to fill the buckets. In low light, you are shooting wide open with slow shutter speeds just to get enough light to the sensor, so you get incomplete bucket fill. In either case, you're amplifying the signal more than you would with a lower ISO setting, so the base noise is going to be more due to the amplification of random electonic anomolies that occur because you're talking about very minute amounts of current. There is no free lunch -- but it's getting closer. . .
So with better high ISO noise performance, you have the option to shoot at higher ISO in good light at faster shutter speeds to stop motion and/or at smaller apertures to gain deeper DOF and better resolution without sacrificing much in the way of detail and general IQ. Higher ISO is not only good for low light work. . .
Also "Exposing to the right" (which refers to a histogram shift towards overexposure --towards the right side of the graph -- without blowing out highlights) also results in lower noise levels for a given scene at a given ISO setting.
A small point, but relevant.
Sorry if I'm confusing the issue by adding minutae. . .
Scott