The population standard deviation of the pixel values of the first input raw bias is computed. This is the combination of structure, fixed pattern noise, and read out noise. The read out noise and the fixed pattern noise contributions are then quadratically subtracted.