[sdiy] fast opamp with low offset

[David G. Dixon]

All good ideas, Jerry.  I think I've tried 'em already, but I can't really
remember.  Probably worth trying them again!  I was also thinking that it
might be good to increase the current flow through 2164 (by decreasing the
size of the resistor on the input and increasing the size of the integrator
cap).  Of course, the higher cap charge would require a longer discharge
time, but it could be 2164 current biases that are messing me up.  Another
thing I thought of was to put a resistor between the JFET + cap and the
summing node, but to bring the 2164 directly into the summing node.  This
way the 2164 will be somewhat isolated from the switch, but the resistor
won't load the 2164...

(I know neither of these will work, but I'll try 'em anyway... (defeatism at
it's best!))

The darndest thing is, I actually achieved about 0.03% tracking from 31.25Hz
all the up to 32kHz with this beast on the breadboard under one condition,
but couldn't reproduce it on the PCB.  There is some combination of stray
breadboard capacitances which is allowing this thing to behave as it should,
but I haven't figured out how to reproduce them...


> Just a couple of thoughts,
> 
> The action of the JFET discharging the cap may be inducing some current
> flow
> at the summing node that the integrator is responding to with high gain.
> You might as an experiment add some resistance in series with the cap/JFET
> and see what this does to the spike. Of course the resistance will effect
> the wave shape, but this may help isolate/identify the mechanism
> generating
> the spike. Once you have the spike generation mechanism fully understood a
> solution may present its self. Another experiment to try would be add
> resistance into the 2164 current source feed into the integrator, the 2164
> may be overreacting the shift in load, the resistance will have it
> operating
> at a higher voltage for the same current and if it is the spike source you
> should see a change in the spike characteristics.