Page 5 of 5

Re: Op amp matching

Posted: Wed Apr 17, 2019 2:07 pm
by mediatechnology
Pulling numbers out of air:

If I have 100 uV of preamp output noise above the servo cutoff frequency what good is 10 uV DC precision?

Try running experiments in the real world with real components rather than simulation.
If you connect a DC voltmeter to the output of a servo'd high gain preamp and measure it you'll see the readings bobble tens of microvolts despite the DC servo being far more precise.
Why?
Because its sampling the noise.
With a perfect servo you still never be able to pin, during "silence," that output to exactly 0V.
It's just not going to happen.

Re: Op amp matching

Posted: Thu May 02, 2019 4:08 am
by terkio
I realize Ios and Ibias can be as important as Vos about the quality of an integrator.
The capacitor isolation resistance IR must be hight, easy with film caps, but that is not enough, the capacitor should not leak via the op-amp input pin.
The OPA 2277 has a 3nA Ios and Ibias. About this specific point, a FET input op amp would be better.

FET inputs op amps have ultra low Ios, but what is available with very low Vos ?

Re: Op amp matching

Posted: Fri May 03, 2019 12:05 am
by montemcguire
I think a more fruitful approach is to not expect low leakages from the PCB assembly and components, and instead orient the design toward lower node impedances while still achieving the desired time constants. The benefit of this is that you can easily use chopper (auto zero) op amps, which have fantastically low offsets and drifts, but relatively high input offset currents, which are mainly from input stage switching artifacts. Chopper amplifier input noise looks like "leakage" at a first glance, so having a design that is relatively immune to "leakage", because low circuit impedances are used, will work well. However, higher node impedances will degrade the amplifier's offset performance given extra leakage, so this is why low node impedances are so important.

As a practical example, I'm using a DC servo circuit built with a non-inverting integrator using an ADA4522-2 amplifier with 100nF 3216 package COG caps with 66K5Ω integrator resistors. If you calculate the time constant, it's obviously too fast for an audio circuit, but the servo output is attenuated by a factor of about 10 to 1 into the DC servo force node so that the overall loop time constant is around 1-2Hz. So, the actual integrator time constants are not all that long, and the circuit's node impedances at the servo amplifier inputs is not all that high, reducing the circuit's sensitivity to leakage, but the loop time constant requirements are easily satisfied because the servo's force output is attenuated into the servoed amplifier. I've built a dozen or two of these circuits with different force resistor and integrator scalings to accommodate different amplifiers with different gain resistor sizes, and they have all worked flawlessly without any care at all for PCB cleanliness - leakage is not an issue.

Re: Op amp matching

Posted: Fri May 03, 2019 12:33 am
by montemcguire
I forgot to mention, but the 'end of the day' offset using this DC servo on actual amplifier circuits is about 7µV in simulation, and I can't honestly measure anything better or worse on real circuits with a Keysight 34461A meter. I get numbers around 5-10-20µV, but when looking at the numbers I get with the probes shorted, it's not much different. So, this servo circuit is essentially "perfect", to the limits of my measurement abilities.

I attribute this mainly to the low offset and drift of the ADA4522-2, but also to the use of low node impedances in the circuit that prevent any drawbacks to the use of the auto zero 4522-2 amplifier. Contrast this with the use of legacy BiFET amplifiers with 1-3mV of offset voltage and unknown drift and you can understand why I don't think it's worth using BiFET amplifiers for DC servos anymore.

Re: Op amp matching

Posted: Fri May 03, 2019 5:24 am
by terkio
At the Integrator ( dual, non inverting ) with OPA2277 I have 30K 10µF for the time constant I need.
From the data sheet OPA2277 Vos max is 50µV, Ios and Ibias max is 3nA
I am not sure I understand correctly these figures, a bit confusing in the data sheet.
With these figures: 3nA x 15K gives a 45µV input voltage bias.... consistant with 50µV Vos.
I was expecting a smaller cap, using 1µF 300K, I see this clubbers the input offset voltage.

Chopper op-amps is an alternative.
JFET op-amps is another alternative.
Alternatives, with their own issues.

Low Vos JFET op amps: I found OPA827 at Vos max 150µV a single and SMD only. Is there any better ?