As was discussed in section 3.2.1, ``The Instrument'', as light falls onto the SWS detectors it generates a current, which charges up a capacitor (one per detector). The voltage across this capacitor is read-out non-destructively 24 times a second and digitised into a bit value between 0 and 4095. At a set interval the capacitor is discharged with a destructive readout (basically short-circuited). This destructive readout affects the readout electronics, which then take some time to return to normal.
An example of this can be seen in figure 5.1 (an example of a glitch). It shows the ERD output of detector against time. A reset time of 1s is used, therefore there are twenty-four samples per reset.
In this example, the first sample read out from the amplifier after a reset is the one located at a bit value of about 2800. The second is at about 2700 and the third is at 2000 - due to the rate at which the capacitor is charging the third sample appears to be an extrapolation of the previous ramp. When a slope is fit to the 24 Hz readouts the first few readouts are ignored, because they are affected by the destructive readout. The number of detector samples to ignore is given in Cal-G file 3 . During PV it was found that the default number of samples to ignore after a reset was different from that thought pre-launch, so the contents of this calibration file was changed.
After the third sample the output of the amplifier seems to have stabilised and the capacitor seems to be charging up linearly, resulting in the increase of the bit value against sample (or time). This is not quite true however. If you look carefully at the figure you can see that each of the ramps has a slight curve in it. This causes errors when trying to fit a slope to the ramps.
During PV it was also found that the entirety of the first second had a slightly different slope to that of subsequent seconds (for reset intervals of 2 or 4 seconds). This was traced to different RC constants for the different seconds, and may be corrected in future versions of the OLP software by having different RC constants for the different seconds. RC constants are held in Cal-G file 2 .