Oscillating opamps in the input section (newbie asks)

I was upgrading the opamps at the input of my power amp but I ran into a problem:
Stock op-amps are NE5532AP's which I tried to replace with OPA2134AP, both channel inputs (XLR, also transformer balanced) are mounted on separate PCB's.
Supply voltage is +-15V.

The situation looks like this:
L.ch - 2x NE5532, R.ch 2xNE5532 - no problems, this is also the stock config
L.ch - 2xOPA2134, R.ch 2xNE5532 - no problems
L.ch - 2xOPA2134, R.ch 2xOPA2134 - oscillation on both channels, gets worse when opening up the volume knobs so definitely coming from the input section.

I even tried LM4562NA, crazily enough these don't seem to oscillate and that makes me think what's wrong with the 2134, it's JFET input? I tried 6 different ones and all behave the same way.
The opamps are socketed but then I'd expect the 4562NA to oscillate easier than the 2134 and why is the r.channel making the left one crap out too?

p.s don't look at the 5534 markings on the schematic, it's probably from an older revision of the amp. This one uses 2x dual OP's.


  • ss+(2016-03-12+at+12.15.23).png
    104.3 KB · Views: 67
Last edited:
The slew rate of the OPA2134 is 80% higher than the 5532, and the LM4562 is worse. In my experience the faster the slew rate the higher the chance of oscillation.
The circuit has 100 p capacitors across the feedback resistor, which is good. That is a bit extreme, I used 33 p to kill oscillation in my circuit.
I can't see your resistor values to determine the gain.
The other tried and true method is to make sure .1 uf disk caps are connected to the two supply rails, within an inch of each input op amp. I got away with one for two op amps, and I didn't have to separately have -15 to analog ground and +15 to analog ground. I just had one disk cap -15 to +15 and that killed my oscillation.
If that doesn't kill the oscillation, you can fool with different values of the feedback capacitor to change the delay. If that doesn't do it you may have a layout problem with long parallel runs coupling input traces to output traces. No fixing that without going off board.
5532 is no slouch on noise. if I wasn't mastering at 90 db signal to noise ratio I'd probably leave them alone.