Dark frame characteristics

The dark counts are composed of two components: the bias (which is essentially the null-exposure readout noise) plus the charge built up during the exposure which is called the dark current even though it doesn't move.

The bias is not a function of exposure time, the current is.

Both components are functions of temperature.

So:
Dark(T,exp) = Bias(T) + Current(T,exp)

The current increases linearly with the exposure time (Figure 1).

Figure 1

The rate of increase with exposure-duration is a function of temperature (Figure 2).

Figure 2

For every 5C or so, the rate doubles. Indeed, the rate is well fit by:
<dcts/dt> = (0.223±0.001) cts s^-1 10^{T/(23.9±0.5 C)}
which doubles every 4.58±0.03 C

Note that this is always less than the 1e^-1/s advertised for the ST-8E, so long as it is cooled.

This rate is a mean rate, averaged for the base-line (not inclusing hot) pixels on the chip. Pixel to pixel variations are significant so the equation should only be used as a guideline for determining minimum cooling temperatures for a given exposure.

A mean bias frame appropriate for the temperature must be subtracted to properly reduce the contribution of the dark current.

These values were determined over the range of temperature of -20C to 10C.

The mean cts/pixel in the bias is a function of temperature (Figure 3), going as
<cts/pixel> = 97.8±0.7 + (11.2±0.5) 10^{T/(21±2 C)}