[time-nuts] Replacement A9 boards for the HP 5065A
bill.iaxs at pobox.com
Mon Feb 26 23:21:42 EST 2018
After 40 years of doing PID control for industrial processes, I'm used
to an error tolerance of 10E-3. So I couldn't understand an integrator
with a 10 K resistor and a 5 mfd capacitor.
But this is time nuts, and the tolerance is more like 10E-13.
An integrator as a controller takes any deviation from zero error
voltage and moves the output in a direction that will return the error
In this case, 10 K is the practical lower limit to the input resistor
for the desired time constant, with 5 mfd as a practical upper limit.
Any current flowing in that resistor changes the value of zero error,
which causes the output to move when the actual error is zero. This
makes the frequency wander. The current can come from the opamp bias or
capacitor leakage when the output is not zero.
Similarly, a change in the opamp zero offset causes a false error which
makes the output move when it shouldn't.
So I withdraw my comment about aluminum electrolytics, which was made
without a timenuts perspective.
Determining maximum error currents and offsets is simply a matter of
mathematics, which is left as an exercise for the student.
More information about the time-nuts