Weird oscillation problem

[koff1]

Hi!
I have put together an amplifier based on a design by G. Randy Slone, with some modifications. The problem is I get some weird oscillations on the positive half period. They start at a peak value at about ~8 volts, and continue to about ~30 volts peak value (measuring on the output). After this the top of the curve is ok, but it is still oscillations below ~30 volts. Hard to explain put see the attached pics, they explain better than words.
Because of the NFB, this oscillation spreads in the entire amplifier so I cannot locate the source. Do anyone have any ideas.
Thanks

[jaycee]

This kind of split-pole compensation is tricky. I have been simulating this myself and it is easy to end up with overshoot and ringing.

What's the value of C5 ?

edit: IMO, Q8 does not need to be a 2SD669 - a 2N5551 here would be adequate.

[koff1]

C5 is 15pF.

Thanks for the reply

[koff1]

C5 is 15 pF.
One more important thing. The problem only exists with load on the output. The more load the bigger the problem becomes.
Thanks for the reply

[homemodder]

I agree with Jaycee, change Q8, a 2n5551 has less than a 1/3 of the cob of the 2sd part, speeding the vas up might help.

Next to try is to put a smallish value cap maybe 100 or 47 pf between base collector of Q15 and Q14 and see what happens.

I would remove the R11 C5 network at first or at least remove R and smaller C, creating too many poles here.

[jaycee]

I would not remove R11/C5, in my design, this is what stops the massive overshoot.

I noticed that your output stage is a CFP type. In this case, I would try 100-220pF Miller capacitor (collector to base) on Q14.

[koff1]

Thanks jaycee and homemodder for pushing me in the right direction.

I have experimented a little now.

1. I looked at the two-pole compensation c8, c9 and r13. The calculations where a bit “on the edge” so I increased c9 to 1000 pF. This changed the oscillation frequency a bit but did not really solve anything.

2. Next I tried to change Q8 to 2n5551. This did not change anything.

3. I then started to experiment with the NFB loop. C5, and R11 certainly had an effect on the problem. Disconnecting the C5, R11 nettwork made the problem disappair. But the network has a purpose. C10 acts as a phase advance capasitor. It is suppose to correct for high frequency phase leg (and thereby improving stability and lower distortion). So I started to experiment with the values, lower C5 helped an as expected lowering R11 increased the problem.

4.I tried to add a miller capacitor on Q14. This helped put the problem did not disappear.

I am now trying to find the middle road between a miller capacitor and changing the values to C5 and R11.

[timpert]

Do you have any info on the layout of your amp? Also, I see that you have made something like a Sziklai but with MOSFETs on the output. They can be quite challenging to get stable, and because of the different transconductance of the N and P channel device, the poles of the output stage move considerably when the output current crosses zero, making any compensation for this outside the OS a fruitless effort.

R11 and C5 are to compensate for the input capacitance of the input differential pair, effectively removing a pole from the transfer function of the feedback path. I'd expect the amplifier to be less stable without them, provided their values are OK.

[jaycee]

Not to hijack, but here's my take on this split pole thing:

[koff1]

Here is the PCB layout.
I have now put a 100pF capacitor between the base and cathode of Q14, and replaced C5 with a 4,7pF, this made the amplifier stable.

Thanks for the replies.