Tek 1K RC time constants

[roger smith x294]

Simon,

The P-Spice simulations we did were for a Tek 1K CCD, and were directed
primarily at understanding crosstalk between phases.  We split the
capacitance (given in the data sheet) among 10 RC's for each parallel
phase, then (assuming symmetry) we modeled only half of these.  We
added phase to phase capacitance between neighbouring phases at every
node.  The resistance values were derived from measurements of an
actual CCD taking advantage of the fact that there are connections to
both sides of the CCD.

We experiemented with various assumptions about the proportion of the
electrode capacitance to substrate versus inter-phase.

We tried driving this quasi-distributed model via our normal slope
control filters and directly, with edges of varying speeds, and varying
filter output impedances.  Of course the results, are only as good as
the model which was based in part on educated guesses, but the CCD's
contribution to settling time (to 10%) was only about 15 us.

Thus, I would have expected the 30us maximum slope time of your
controller to dominate.  Since you say the dark-pattern in your AB
clocked frames was unchanged when you increased the slope time to 30us,
I begin to doubt my hypothesis that the the variation from edge to
center is due to edge smoothing by the distributed RC of the CCD
electrodes.  An intersting test would be to leave one side of the CCD open
while driving the other, and compare the waveshapes.

In the Spice model we assumed that the resistance of the metal busses
for the parallel phases is negligible.  Maybe this resistance can still
explain the vertical variation you see.  Maybe our model was out by more
than a factor of two.

I'm curious to know how you set/modify the slopes in your controller?

Roger