the reason to use bias frames is if you want to scale (or "optimize") your master dark.
the idea is this: the dark current in any sensor is a function of temperature and time. since we usually use cooled cameras, we can hold the temperature constant between the darks and the lights. it turns out the dark current is a linear function of time, so it is possible to linearly scale the dark before subtraction from the light, where the dark and light durations do not match. so for instance you might sometimes do 20 minute lights for narrowband and sometimes 10min lights for LRGB. you can make a dark out of 20 minute subs and then scale them to 10 minutes by multiplying them by 0.5 before subtraction, instead of maintaining two separate master darks.
but there is a problem - the bias signal is not a function of time. so if you want to scale your dark, first you have to subtract the bias signal, and then scale the dark. if you don't do this, the calibration will be completely wrong. so when optimizing darks you either have to load a master bias and tell PI to both calibrate and optimize the dark, or you need to make a master dark which has the bias signal pre-subtracted. in that case you *always* need to load the master bias (whether or not you optimize the dark) since the bias signal is missing from the master dark. in this case you need to be sure to never check "calibrate dark" as that will amount to a double bias subtraction, which is also bad.
as an aside, PI doesn't actually care about the dark or light durations. what it does is iteratively scale the dark by calibrating a small portion of the light with different dark scaling factors and choosing the scaling factor that minimizes the noise in the calibrated result.
as a 2nd aside, this generally works for dedicated astro cameras. DSLRs play all sorts of dark current suppression tricks in the camera firmware and these tricks can not be turned off. so sometimes dark optimization does not work well for DSLRs.
rob
