Op amp oscillation with capacitive load

[composition4]

Hi all

I'm designing a high-gain pedal and I'm getting some high-pitched oscillation after the last op amp (TL072) stage.

It's fine on the breadboard if I bypass the three bridged-T filters, but as soon as they are in the circuit it oscillates, presumably because of the capacitive load. The main culprit appears to be the 330nF cap to ground in the first filter, because as I replace it with lower values the oscillation starts disappearing - but of course that changes the filter.

The op amps have bypass power supply capacitors which make no difference in this case, and I've tried putting a small value caps in the op amp feedback loop but this just rolls off the treble response without helping.

I'm assuming I could impedance-scale the bridged-T networks but would prefer not to if possible, or maybe put a non-opamp buffer after the opamp (would this work?).  Just wondering if there's a more elegant solution that I'm missing, as I'd prefer not to add another active stage if possible.

Thanks!
Jonathan

[anotherjim]

The text book approach is to fit a small resistor (often around 100R) in series with the amp output and a feedback capacitor.

Here, Riso is in the feedback loop so the amp compensates for the slight signal voltage drop caused due to load current. In practice, it can go after the loop since we don't usually need computational precision.
Cf also helps. That will be small value since it makes a LPF with Rf. 10pF to 100pF range.

[composition4]

Thanks for the reply... unfortunately the series resistor on the output makes no difference, had tried that already. Had tried even up to 10k and it didn't help.

Anyone else got any suggestions?

[anotherjim]

What about a feedback capacitor?
...rest of circuit...
If Vref is not sufficiently stable, the "rest of circuit" could be causing a feedback loop via Vref.

[Rob Strand]

I'd put a 100ohm resistors in series with the output of the first stage - outside the feedback loop, not inside.

The T circuits you have don't present a capacitive load they present a resistive load - short the caps and you just get 10k's to gnd.    So suspect the problem might be elsewhere.  Layout, supply decoupling.   Maybe the next stage has something to do with it.

The small cap across R28 like anotherjim mentioned is another good idea; that can make it tougher against many evils.

[composition4]

Hey thanks guys,

Turns out it was an artifact of the breadboard - grounded the backing plate (never needed to do that before even with high gain designs) and in combination with a 1k resistor after the opamp, the oscillation disappeared. Doing one of these things without the other made no difference.

Thanks as always!

Jonathan

[Rob Strand]

To some degree, a 1k series resistor will affect the frequency response of the T networks.

[PRR]

> presumably because of the capacitive load

What capacitive load?

Draw a path from opamp to ground which does not pass through a 10K resistor. You can't.

Yes, you do have 6 paths through 10Ks, so could have ??nFd + 1.6K. If the resistance is not much lower than the opamp can comfortably drive, there is no problem. The usual stomp-pedal opamps drive 2K with ease and 300r without getting goofy. This network is not a problem.

> artifact of the breadboard

Yes, that will happen. Breadboard encourage bad layouts and have dubious contacts.

This is an interesting case. You have a high-ish gain preamp and a LOT of phase shifting in the 6R 6C network, and maybe more gain past that? The least little sneakage from output leads to input leads (both typically hay-wire on breadboard) is liable to meet the Barkhausen Stability Criterion (loop gain equal to unity and phase shift is zero). This would explain "high pitched" whereas capacitive loading usually sings in the MHz.

[composition4]

So I *just* figured out my REAL problem after I ditched the bridged T filters and the random oscillation noises still persisted.

I only used half of an op amp, the other half was sitting there on the breadboard. In a rare moment of clarity , I realised that the unused half of the op amp had nothing connected to its inputs or outputs.  My guess was that the high-impedance +ve input was picking up stray fields, and as the feedback loop was totally open, the opamp was trying to amplify these at full open loop gain, and the signal was high enough to capacitively feed through to the other half of the op amp.

So I connected the unused +ve input to +4.5v, and shorted the -ve input and output terminals to each other...  all unwanted noises gone!

It seems like a stupid oversight now that I look back on it.  I hope this helps someone else at some stage if they are as absent-minded as I am.

Jonathan

[Rob Strand]

I hope this helps someone else at some stage

Thanks for closing it off.