Flats are important, but biases are not so important.
Flats are important because they compensate for areas that are not illuminated evenly across the image.
The easiest example is a little dust bunny. Imagine you have a piece of dust on the sensor. It blocks most light from getting through, and leaves a dark spot on the sensor, and therefore in the image. Flats take a picture of that dark spot, and then when applied to the real exposure, mathematically compensate for the dark area.
(Let's say the perfectly clean exposure should have been 150 exposure values. The flat shows the exposure everywhere across the image at 150 except under the dust bunny, where it is 120. The dust bunny made a diffference of 30 exposure values. The software adds 30 ev's (exposure values) to the Light frames (your real celestial exposures)in those pixel locations, thereby bringing them back up to the values they would have had if the dust bunny not robbed them.)
After compensating with the flats, there is no dark shadow area where there might have been.
The uneven illumination could come from light fall-off due to an undersized secondary, or a focal reducer, or a baffle, or any one of a number of other things (including dust bunnies. A famous place is on the bottom of a frame of a Canon camera (and perhaps others). Ever notice that the exposures on about the bottom tenth of a Canon is significantly underexposed compared to the rest of the sky? Flat fielding can remove much of that.
Bias is not nearly so important. Most cameras ready their sensors for receiving data by having a floor of some exposure value. Let's say it is 10. All exposures will run from 10 to whatever (65K for 16 bit sensors), not 0 to whatever. The 10 value is subtracted from the rest of the image automatically in most cases before processing. However, in some cases you need to know it is there. For instance, if you are using darks with one time value on lights from another time value, you need to cscale the darks to know how much noise there really is. And to do that, you have to subtract the "floor" from each..
(For instance, let's say that the amount of electronics and read noise is perfectly linear: If you expose a dark for two minutes, your dark will have exactly twice as much noise as if you expose for one minute. ANd let's assume in a minute you get 30 ev's worth of noise. Therefore 60 in two minutes. The actual ev readings on these two exposures would not be 30 and 60, but 30 PLUS the bias value, and 60 PLUS the bias value. (If the bias value is 10, then you would get 40 and 70.) Your "bias" frames tell your software how high the floor was so it will know how much to compensate.
The only time it really gets to be important is if you want to scale your darks. That is using darks taken at one time with light frames taken at another. (Or if you know thetemperature characteristics of your sensors from one temperature to another.)
I believe if you are using darks taken at the same time and temperature as your lights, you can skip worrying about bias frames.
Alex
