>>3611374>If you're looking at an equation that does this (and not simply saying "equation" to sound smart) then throw it away because it is observably wrong.Nah I'd rather throw your baseless opinion away. This is not up to debate no matter how hard you try to twist it into one. Picrelated is the SNR equation for image sensors.
i_ph = photocurrent, i.e. the current ONE (1) PHOTOSITE induces when hit with light at a rate of some photons/sec. This incorporates quantum efficiency (what percentage of the photons can the sensor actually convert into current) and photon shot noise (how many photons actually arrive at the sensor in the first place, compared to the expected value).
T = exposure time
i_dc = "dark" photocurrent, i.e. the current 1 photosite induces when *not* hit with any light, a form of noise due to silicon imperfections, affected also by temperature. Again, i_dc is PER PIXEL.
σ^2_V = the variance of the random variable V(T) representing circuit readout noise, including quantisation noise. PER PIXEL. That random variable (obviously) has mean 0 cause it's random noise.
σ^2_C = variance of the random variable C representing reset noise (of the circuitry, since reset happens before every capture), as well as fixed pattern noise. This random variable, since it's random noise, also has mean 0. Fixed pattern noise has 2 components, dsnu (dark signal non uniformity) and prnu (photo response non-uniformity). Many authors write that as just dsnu and σ^2_dsnu because photo response is assumed in the model to be linear, hence prnu is not a factor.
It's clear that pixel size is built right into the equation, since i_ph is dependent on pixel size (for the same quantum efficiency and the same photon shot noise, a larger pixel converts more photons into electrons and thus produces higher photocurrent i_ph).
