|
I wanted to post one final graph which in completely unsurprising news looks very much like RickS's graph.
This has helped me determine the most accurate bias frame possible.
The time on the X axis is not to scale, but it allows you to see the way the camera operates at different exposure lengths.
You can see that the shortest possible exposure does not yield an exposure that is indicative of the kind of read noise you will get in longer exposures.
I verified this by doing integrations of 100 frames for each time, constructing a superbias of each and comparing them. And they were very different. The STF brings them all up to the same level, of course, but you can see that the noise patterns are quite different.
0.0001s has a glow on the left of frame.
0.001s develops a dark gutter on the right, but corresponds to the lowest actual noise levels on the graph. (but for the STF, this image would look darker).
At 1.5s there is a glow to the right that corresponds to amp glow, which continues along the top and bottom edges of frame.
Only once exposures exceed 2 seconds do the bias frame stabilise, as seen in the graph, where all values over 2s are in a straight line.
So in essence, bias frames less than 2 seconds long are unlikely to be representative of the actual bias pattern noise found in longer exposures.
If, for example, you shoot flats for NB that are a few seconds long, and you rely on scaling of a bias to subtract read noise from the flats (rather than actually shooting flat darks that are the same length as the flats themselves), you could be doing yourself a disservice.
Cheers
Markus
|