BBD-MN3010 and MN3011- is a MN3101 clock necessary?

[René Schmitz]

At 19:21 04.05.00 +0200, jhaible wrote:
>> An earlier message in this thread suggested that very brief
>> overlap might be better than no overlap in maximizing BBD performance.
>
>Must have missed this message - who said that, and what would be the reasons
>for this ?
>
>JH.

I think it refers to my message. I was a bit fuzzy on the term
non-overlapping, didn't make clear when I was talking of the
dead-time-non-overlapping clock, and when not. Sorry.

Once again: 

I said that a dead time between the clocks might have actually the opposite
effect on feedthru performance. IMHO/untried of course.

I think the mechanism, how the spikes at the output (=differenciated clock
signal) are generated doesn't have to do with overlapping of the clock
phases. 

Lets look for the differenciating cap. I suspect its the gate capacitance
of the output transistors, in series with the last two bucket caps. 
Its the only capacitive path I see. Where else could the differenciation
take place?!

If all transistors/caps were equal, both clocks change state
simultaneously, and the 0-1 and 1-0 transients are symmetrical, 
all would cancel. In a real device that is not the case of course. 
At least the manufacturer included two outputs, which allows to use 
a pot to trim this to a minimum. In tendency we have the sum of a 
pos and a neg spike.

Now when we have a gap between the two transitions of the clock signals,
the spikes are generated at two different time points. They can't cancel out. 

Now that overlap, what does it do? 
My feeling is that overlap must be minimized because otherwise the buckets
can leak into each other. This would attenuate the signal which would not
help to improve the SNR either. 

My approach would be a totally different one: I'd sample the output with a
good/fast S&H in the middle of the clock pulse. I.e. for a 10uS clock
period I'd take a sample at 2.5-3.5 and 7.5-8.5uS. Then the spike has
already decayed down, the result should be a cleaner signal. Of course this
could then be combined with a dead-time clock, ensuring minimum attenuation
of the signal.

Bye,
 René





transconductance | uzs159 at uni-bonn.de
isfutilepreparet | http://www.uni-bonn.de/~uzs159
obeassimilated.. | http://members.xoom.com/Rene_Schmitz