A commonly used technique is to expose the subframe to be shot noise dominant vs read noise dominant. This is done by exposing a frame till the sky background is some factor of read noise squared - usually a factor between 3 to10 or so. To do the calculation you need to know read noise, gain and offset of the camera.
By exposing to be shot noise dominant, the SNR improvement by stacking approaches that of a single long exposure with the same integration time but with reduced saturation issues.
 

Read More...