[sdiy] fast opamp with low offset

[David G. Dixon]

Thanks List, for all the excellent suggestions.  However, for my
application, even 0.1mV offset would require trimming :(

This all relates to the 2164 Expo VCO.  It turns out that the difficulty I
am having in achieving good tracking is caused the current injection by the
PN4391 JFET, which creates 1V to 2V pulses at the summing node of the saw
integrator.  I didn't have this problem with the 5485, probably because of
its much lower input capacitance.  This leads to major instability in the
2164, which requires +/-0.1V output compliance.  Without a Franco resistor,
and keeping the reset timing cap as small as possible, this problem is
minimized but not eliminated, especially at high frequencies.  Adding a
Franco resistor increases the pulses enormously and causes the sawtooth to
take on a J shape with severe negative voltage excursions and complete loss
of tracking.

I was toying with the idea of converting the 2164 output current to a
voltage and then back to a current with an I-V converter followed by a
resistor.  I breadboarded this, and the sawtooth is perfect and completely
insensitive to what's going on at the integrator summing node (as you'd
expect).  However, tracking is terrible because of the input voltage offset
of the I-V converter (and, to a lesser extent, the saw integrator).  I
wanted to avoid a trimming circuit, and thought that I could use an ultralow
input offset opamp, but I'd need an opamp with a decent slew rate, and even
with only 0.1mV offset, the low-frequency tracking would still be pretty
bad.  I'm not even sure that trimming would be completely effective, as the
trimming circuit would have to be very precise, and would complicate VCO
tuning enormously.

In my experience, the 2164 puts out a very precise current.  To then convert
this to a voltage and back with an opamp, even a precision one, seems like a
real step backward.

In the meantime, simulations suggest that all I have to do is put a passive
LPF on the 2164 current output to filter the spikes to well within the
compliance range.  This would be a very convenient fix, since it only adds
one resistor and one cap to the circuit.  Of course, with this method, one
has to use a very small resistor so as not to create a voltage outside of
the +/-0.1V compliance range at the 2164 output.  With a 100R resistor, I
can still be within the range all the way up to about 180kHz.  This and a
cap between 1nF and 10nF should filter the spikes almost completely.  Of
course, I'm away from home and haven't had a chance to test it, but the
simulations are very promising.  If this works, then I am finally done with
this blasted 2164 Expo VCO circuit!



> I vote to AD712, too!
> Japanese audio freaks love it very much.
> 
> And you'd better not to forget LM4562 and LME49860.
> Input offset voltage is 0.1mV(typ.), with good AC performance, you know.
> 
> I love NJM2082, though its relaxed DC specs (just like LF412 but
> improved AC & audio).