Effects of CCD Clocking Modes on Detector Peformance

 Calibration analyses are based upon data from a timed exposure mode of CCD operation. While this may become the preferred mode of operation it is extremely likely that other operating modes will be used, perhaps even extensively so. Also, most of the analyses, especially of spectral properties such as gain and line shape, are based upon averages of near uniform "flat field" exposures. This type of illumination is not be typical of many sources with the expected telescope point spread function. Illumination and operation mode are closely coupled, the former often dictating the later. Users should consider how operation mode will affect application of the calibration results. For this reason a sampling of representative operating modes was tested and the small differences in performance are now described.

Serial 2 pixel summation reduces the gain by approximately 5 percent owing to the necessary shorter integration pulse widths. Because of the peculiar shape of the integrated waveform the relationship between gain and pulse width cannot be predicted. For example, an adjustable gain feature cannot be reliably invoked via the pulse width - halving the pulse width does not halve the gain (it reduces the gain.) Serial summation also modifies the conventional meaning of event grades by promoting some lower grade events (3 pixel and serial split) into higher grades (single split and single pixel.) The pileup correction will also require modification.

Parallel summation of course alters the conventional meaning of event grades, especially of the parallel split events.

Continuous clocking does not seem to require any special consideration.

The window readout shows a peculiar result worthy of mention. This is a mode wherein a fraction of available rows in a timed exposure are actually clocked through and should not be confused with software window sampling wherein all available rows are clocked through but a software mask excludes events not within a virtual window of rows. For testing purposes a window of 100 rows is used with the result that independent of where the 100 rows occur upon the CCD or what monochromatic x-ray energy is used the measured line widths are always narrower than for the full 1024 row readout. Typically the differences is no more than 5 percent. The cause of this change in line width remains a mystery. A leading candidate is the "Elvis-effect", a multilevel disturbance in the bias level which occurs and endures for times comparable to the 100 row readout time. The cause of the Elvis-effect is unknown and it is unclear if it occurs with the flight configured instrument. This is one of a class of bias distortions which would not affect spectra from spatially localised events, but would impact spectra from flat field type illuminations.

The effects of various operating modes are small and occur through the electronics or the meaning of event grades, gain, noise, etc. The description of the CCD model is not affected by choice of operating mode.