In this section, we proceed on the assumption that, by the time CCDs are installed into scientific instruments, they will have been subjected to a good grilling in the lab and have had their controlling voltages and timings adjusted optimally. It is now up to us to characterise them in actual use, optimise the way we expose them and finally to provide feedback to the lab for new control aims and specifications. We shall discuss ways of testing them by lamps or stars. Please report back (TINBERGEN@HRDKSW5) any improvements you may invent.
Shutter stability CCDs are not shuttered electronically in astronomical applications. A physical shutter controls the exposure. Since this is a key point in some of the tests and in some observations, a shutter test should be part of any pre-run tryout of the system. The shutters were not designed with millisecond stability in mind. They are mechanical, have been known to stick, they are controlled pneumatically, therefore are sensitive to gas supply pressure, finally are operated via 4MS and Ethernet, with possibilities of delays.
To test shutter stability, configure ISIS so that you obtain a well exposed (but not saturated) CCD frame with the tungsten lamp in about 10 seconds. Let the tungsten lamp (Quartz-halogen type) stabilise for at least half an hour. Then take a number of say 1 second and 10 second exposures and determine the rms spread of intensities in one or more pixel groups (using the DMS). Our tests indicate that the rms spread should be well below 0.1 % at 10 seconds, which is a minor miracle.
Pixel sensitivity stability Having obtained flat fields on daytime or Moonlit sky, one transfers these to observations made many hours later. It is not possible to use lamps for flat fields, but one may use them to monitor pixel sensitivity ratios. Do a normal polarimetric observation for Q or U, and determine the G ratio from it (subsection 3.2). If these ratios stay constant, all is well; if not, the variations tell you the extent of your trouble, but not which of the 2 pixels involved in the ratio is causing it.
Linearity Having ascertained shutter reliability, you are ready to test linearity. Set up as described above, except that full scale output should take a 100-second exposure.
Expose for various lengths of time and divide output by nominal exposure time. For exposures of more than 10 seconds, departures from the overall mean (that are significant given the rms spread) should be attributable to the CCD; below 10 seconds, the shutter mechanism could cause non-linearities or field dependence.
A purely polarimetric way to determine non-linearities, not depending on shutter stability but very time consuming, would be the following: allow the tungsten lamp to stabilise, insert the calibration polarizer and take pairs of CCD frames with halfwave plate at angles X and X+45 degrees. Use some 10 to 20 different values of X in a total range of some 120 degrees. Do the normal staring-mode reduction, then fit the Stokes parameter values for different X by a sinusoid in 2X. Deviations from a pure sinusoid will be due to non-linearities in the high or low intensity parts of the data. Repeat the test with peak intensity halved; if the effect is reduced, it must arise at the high intensities.
Finally, remember that most of polarimetry is relative photometry of nearly equal intensities. As long as non-linearities are similar in corresponding pixels, they are not very important, since it is the intensity ratio that counts.