Things I’ve heard and read about ISO film speed in digital cameras that are wrong:
- makes the sensor more sensitive to light.“
- makes the sensor collect more light.“
- groups pixels together and adds them up.“
In reality, ISO film speed in a digital camera is nothing more than signal amplification, like turning up the volume on a HiFi sound system: Analog signal gain inside the sensor up to some level and digital multiplication after analog-to-digital conversion (ADC) beyond that.
For example: A camera’s sensor can apply 10X signal gain to its output, i.e. ISO 1000. Setting it to ISO 1600 will perform an additional 1.6X digital multiplication on top.
This also makes it clear why raising ISO lowers dynamic range: The ADC maps certain voltages coming out of the sensor to floating point numbers between 0 and 1. Amplification on voltages that were in the high regions to begin with will end up beyond the conversible range, thus the captureable dynamic range shrinks at the same rate as ISO speeds up the film.
Digital multiplication beyond what can be amplified is less affected by this problem, as it’s obviously possible to handle and save floating point numbers greater than 1, but any such number might still appear white when viewed with a program that expects 1 to be the brightest possible value.
Digital multiplication, however, will blindly amplify anything that’s there, including noise, while there is some room for optimization with analog amplification done in the sensor. This interaction between analog and digital amplification and where the cutoff between the two lies in a particular camera explains why some cameras are more or less ISO (in)variant than others.