For the HF noise, the analysis of pixel signal fluctuations leads to
a view similar to SW, but, in addition, a significant low frequency
noise is detected in the stable regime (note also that the
digitization noise can be noticeable with the lowest gain
1).
In addition to the HF noise, the LW array shows time correlated gain fluctuations (LF, low frequency noise) which are not correlated from pixel to pixel. This LF noise results in an uncertainty for the determination of the flat-field which increases slowly with time. Best fits to measured fluctuations are obtained with a logarithmic growth of fluctuation amplitudes integrated over increasing time scales. The formulae given below apply to a very well stabilized signal. Flux transients considerably amplify the time correlated gain fluctuations. A few hundred images after the flux step, this amplification factor still amounts to 2 or 3. An accurate description of the time relaxation of this effect is still under study.
Switching the array off, then back on, induces random changes of the dark and of the flat-field.
It is expected that for a given orbit the standard observatory calibration will provide flat-field frames with an accuracy of 1% and dark frames with an rms accuracy of 0.3 ADUs.
Using the same notations as above, we have for LW:
For short sequences of images, the noise of a measurement with the LW array is completely described by combining this HF term with the dark noise and the flat-field noise which are presented in the following sections.
For long sequences of images, the LF noise term analyzed in the following sections must be taken into account.