A Different Opamp Compensation Technique.

[DF96]

Interesting. Hysteresis may mean that your loop gain Bode plot has a little loop in it somewhere near the (-1,0) point so you may even be only conditionally stable. Reducing the op-amp gain will bring you down into the unconditional stability region. You might find that there is a critical value of resistor between inputs which turns it into an oscillator, not because it is adding capacitance but because it drags the Bode plot loop right on top of the (-1,0) point.

[Mooly]

I set up two OPA134's as shown in the circuit below.

The beehive trimmer cap was adjusted for good compensation of the first opamp (top scope trace in all following shots) at 40 Khz and 10mv pk pk output.

The resistor Rx on the second opamp was adjusted to give an identical waveform compared to the conventionally compensated opamp above. The value of Rx was 1950 ohms. The traces could be overlaid with no difference showing at 10 mv pk-pk output.

These shots show the output at 40 khz and 10 mv pk-pk and also at 1Mhz and 10 mv pk-pk. As can be seen the results appear identical.

[Mooly]

And here is the large signal performance at 40 khz and 10 volts pk-pk.

The non linearity in the C compensation can be seen. The second shot is a detail of the non linearity. This was clearly seen by eye even at much lower frequencys where the eye seems to winkle out detail that the camera doesn't resolve. At 40Khz it's obvious though.

The compensation (is that the right word ?) when done by Rx doesn't give this error.

[jcx]

the usual way of applying "noise gain compensation" is to use series RC so the loop gain reduction doesn't cost as much low frequency (audio) performance

the RC time constant sould be set so that the closed loop (noise) gain flattens out before the gain intercept

looking up the series: http://www.analogzone.com/acqt0814.pdf could useful

[Mooly]

Thanks for the link jcx

Searching and I turned this up from the late Bob Pease.

[jcx]

a warning on non-inverting mode - you may need a input bypass C to AC gnd the noise gain network to prevent input cmrr AC perfromance from interacting with loop stability

this can be part of a RF/EMI rejection filter

[Mooly]

Originally Posted by jcx

a warning on non-inverting mode - you may need a input bypass C to AC gnd the noise gain network to prevent input cmrr AC perfromance from interacting with loop stability

this can be part of a RF/EMI rejection filter

I wouldn't like to say one way or the other on that...

Have a read at page 13,
http://www.ti.com/lit/ds/symlink/opa1611.pdf

The resistor on its own is the recommended technique for testing with ultra high performance opamps.

[jcx]

thats for deliberately making audio frequency perfromance worse to see anything at all with your distortion analyzer

if you want to use the technique to stabliize decompensated op amps or add stability margin with C load there's no reason to want to ruin audio frequency performance by shunting away gain all the way down to DC

[Mooly]

Originally Posted by jcx

thats for deliberately making audio frequency perfromance worse to see anything at all with your distortion analyzer

if you want to use the technique to stabliize decompensated op amps or add stability margin with C load there's no reason to want to ruin audio frequency performance by shunting away gain all the way down to DC

Yes I realise that... and that doing this degrades the noise and distortion properties... but the big question is whether it's subjectively better.

Did you read Nelson posting in the link at the top,

Post in question is here,
opamp inverting input sounds better?

[vacuphile]

JCX was ahead of me; the use of Rx allows to set for distortion gain in test setups. THD+N has come to the level in opamps where most distortion analyzers cannot cope anymore.

@Mooly
The question whether it would be subjectively better is fallacious imo, without defining for what purpose. In order to create an effects box, it might be interesting, e.g. for distorting a guitar, or to emulate a WAVAC without spending 350K. In audio, I think it is a safe bet that all distortions that are not there make it better.

vac