[sdiy] FW: fast opamp with low offset

[David G. Dixon]

(on-list this time)

> There are a couple of clues in the H&H text. First the glitch transfer 
> area under the curve is related to the voltage swing.
> Minimize the swing, minimize the glitch.

Well, here I'm pretty much constrained by the pinchoff voltage of PN4391,
which can be as low as -10V.  Hence, the comparator sends -15 to the switch
during ramp charging, and then just grounds the switch during reset.  Thus,
15V is about the smallest delta V which still guarantee reset.  Plus, this
is the standard method a la Electronotes/Terry Michaels/Ian Fritz, so I'm
loathe to alter it.  I'd much rather make my setup work within it.

> The second hint is that the mechanism can be roughly modeled as a 
> fixed capacitive transfer of the voltage swing...
> 
> <<C_GC is the gate-channel
> capacitance, typically around 5pF.  Note that the amount of charge 
> transferred to the channel depends only on the total voltage change at 
> the gate, not on its rise time. >>
> 
>  so if you match the C_GC with a another matched (= gate-channel
> capacitance) cap (to the channel side) and supply a matched reversed 
> polarity swing (= -[control signal being applied to the gate]), they 
> should (theoretically) cancel each other out....

Yes, I did understand that part.  However, this would complicate the circuit
somewhat, as I'd have to invert the comparator output.  Plus, the capacitor
would have to be tightly matched to the JFET's C_GC, which is dicey as JFETs
can be quite variable.  I still think that if filtering works, that's what
I'll do.  I'm also interested in trying to reset the cap directly to ground
on the output side with the JFET as an alternative, although I'm having
difficulty wrapping my head around how this would actually work (and
simulation suggests it won't).  Resetting with a CMOS switch is another
option I'd try as a last resort if filtering doesn't work.  However, I'm
hoping that it's the high voltage magnitude of the spike which is flustering
the 2164, and not the actual amount of charge (the area under the spike).
If it's the latter, then I'm screwed.