Hybrid amp phase shift oscillation?

[Duo]

This is a really nasty problem and I can't seem to solve it.

I recently made a small hybrid amplifier using LM358 and CFP output stage with a CCS.

The amplifier rails are +-19V and it sounds excellent, except that it has a terrible oscillation problem which I can't seem to stop.

I've added capacitances to the BC junctions of both CFP drivers, filtered and bypassed the heck out of the power supply, tried different amounts of FB and all to no avail.

I'm not sure if this oscillation is caused by phase shift from adding output devices, or from something else.

The things I have tried make it better, but it still oscillates and causes terrible noise during medium to loud passages.

Can anyone help me on this at all?

[johnnyx]

It would be helpful to give us a schematic of your implementation.

I think that a cap between the opamp output and its inverting input will stop it. This will reduce the opamp gain to unity at some high frequency, and no oscillations can occur. If you have extra components within the feedback loop you will get extra phase shifts while the gain is still high, so reducing the op amp gain in this way will prevent the oscillations. I would try (say) 22pF at first, increasing it until the oscillations stop, the value depends on the circuit.
Hope this helps

[Duo]

Well, this is a terribly bad rendition of the schematic, since I don't have any software for drawing schematics properly, and my scanner isn't set up right now.

If you want any resistor values, just ask, I'll tell you.

[cunningham]

suggestion 1) try using a PNP current source for the positive side to get better symetry. minor improvement

suggestion 2) try placing a small pf capacitor from between the two diodes that bias the drivers, to the inverting input of the op-amp. This will reduce the high frequency bandwidth a little, but choose as small a value as possible to stop occilations. This approach has worked with descrete circuits, but I have not tried it with an op-amp. Don't get to discuraged if it doesn't work though.

[Eva]

Some facts to think about :

- You have forgot to place base discharge resistors between B-E junctions of output transistors. Suitable values are 100 ohms or less for each output device. A CFP whithout these resistors is almost useless since it will oscillate wildly by itself and may self-destroy due to cross-conduction

- This lousy op-amp aready shows poor phase margin [less than 45º] so the CFP should show very small phase shift at the unity gain frequency of the op-amp [1Mhz] to achieve stability

- CFPs usually show lots of current gain [up to 10.000] and lots of of phase shift at 1Mhz [probably 90º or more]

- To achieve stability, the open loop phase shift [from op-amp IN- to output] should be below 360º with some margin [ie 320º max.] up to the frequency where feedback ends [the frequency where open loop gain drops below excpected closed loop gain]

There is not much to do to reduce phase shifts in an inherently slow topology like CFP [the outer transistors are driven in an inefficient way imposing a current lag] so the 'end-of-feedback' frequency should be reduced by means of a capacitor from op-amp output to IN- [and also by increasing closed loop gain] until enough phase margin is achieved for stability [pase shift decreases for lower frequencies]

Capacitors from C to B of inner CFP transistors and from upper B to lower B of outer transistors may also help to reduce a bit CFP phase shift, and also to prevent excessive cross-conduction and wild CFP self-oscillation near clipping

....

Or you may just forget about CFP and go for a EF stage that may be stable even without compensation

[Duo]

Lol, this all turned out to be very interesting.

I made a mistake drawing the diagram, I did use discharge resistors in the amp, however, I missed them in the schematic. My bad.

Also, just a day before reading that last reply from Eva, I had already tried a standard EF output stage and was able to obtain much better results and a far higher inherent stability.

I have since removed the opamp in lieu of my own discrete version, which, undoubtedly, has far better characteristics. It employs a very standard implementation of a differential amp with CCS, mirror, and constant degeneration. This yielded a faster amplifier which needed only a small cap (48pF) across the BC junction of the VAS for stable operation in all of my test conditions.

Thanks for your help guys.

Also, next time, I'll be using a better opamp. haha, this is just what I pulled from the drawer at school and it was convenient so in it went.