CCD Imaging for Dummies

[psonice]

Originally Posted by flyingcadet

The CCD chip is composed of thousands of light detecting transistors. Transistors have an inherent flaw that will never be removed. In actuality, removing this flaw would break the laws of physics that allows them to work. The flaw is called Dark Current. Using the word "dark" has nothing to do with light, but its use was originally applied to thermal dynamics, or heat. There is also a leakage current that occurs. The leakage current is from the fact that you can not stop a small number of electrons from moving through the transistor despite it being in a "non transmitting state." The leakage current is directly related to the thermal energy of the transistor. Increasing the heat also increases the leakage current, and reducing the heat reduces the leakage current. Thus, we have dark current.

In a CCD, the initial leakage current for each pixel is going to be different. Once the chip begins to heat up, some pixels will begin to register false data more than others due to impurities and will become predictable.

Dark Frames come in to remove the affects of the dark current. If you take 20 two minute exposures with 2 minutes between each exposure, than you dark frame needs to be at least one two minute exposure so the chip will heat up as much as the other 2 minute light exposures.

This interests me, as I'm interested in perhaps experimenting with some novel ways of doing image processing and stacking (I do some graphics programming).

If the error levels get worse as the chip heats up, then it makes sense to keep the ccd cool. But you seem to be suggesting that the error level for each pixel is fairly stable, as it's possible to build up a dark frame to correct for it?

If that's the case, does it not make more sense to run the CCD continuously so that it's in a stable "hot" state? Then you can build up a much longer exposure for the dark frame, which would be more accurate. Then, if you have a dark frame for a known-length exposure, and you're running the ccd in a stable state, you can subtract the dark frame from each captured frame in the exposure instead of applying it at the end. That should give a more reliable result, because if the ccd doesn't warm up at the same rate as when the dark frame exposure was taken, the dark frame wouldn't match the actual noise levels in the image. You wouldn't need a fixed-length exposure either - you could just watch the image build up until it looks good, then stop capturing.

Does that make sense? Or is the "cold" state so much better than hot that it's worth living with the warm up?