Peculiar oscillation

[dgta]

The amp below (power stage shown only) works fine. But on the bench, when I turn up the signal, I get an oscillation just as I approach clipping (see trace, taken at speaker output).

Here is the peculiar part: If I touch one 6L6 grid circuit (either side of the coupling cap) with a scope probe,VOM lead or just a plain piece of wire, the oscillation gets much worse and covers the entire wave, not just a small portion.

BUT if I touch the probe to the other 6L6's grid, the oscillation goes away.

Also, at lower signal levels there is no natural oscillation, but it does oscillate when I touch with a probe. Again, one 6L6 grid only, not the other. I did swap the tubes and the oscillation does NOT follow the tube.

Even more peculiar, this is a dual mono chassis with only the power switch in common. Both channels behave exactly the same. Signal fed is 1khz, oscillation roughly 300khz.

Any clue as to the cause and/or any ideas for investigation? There are no old or questionable components or layout issues.

[bigwill]

Is there global NFB and does this behaviour stop with the FB disconnected?

[dgta]

No there isn't any GNFB.

Also, the load is 8ohm 50W non-inductive resistor connected to 8ohm tap of secondary. Zero ohm tap of secondary is grounded.

[Hi-Q]

Worth checking any decouplers further down towards the front end. Your part schematic just shows one, a 47uf but I expect others will be present to decouple the front end. I know it is a stab in the dark but some of your output stage B+ variation may be getting seen by the input stage if the decoupling is not sufficient.

[dgta]

I did. The input and phase splitter are fed 350V with 100uF decoupling and the rail is perfectly clean. Anyway, there is no oscillation present in the preceding stages, i.e. the signal in the grid of the SN7 is clean and stable.

[Andrew Eckhardt]

You should be able to swap coupling caps and 6L6 grids and see if the chirp swaps sides, that would positively isolate the output stage from concern. Is it possible that capacitances are not the same for both halves of the sn7?

At least with two circuits doing the same thing you can be certain that the current design and/or parts wont work.

[kevinkr]

How about some snubber resistors on the screen grids, I don't see them. (Not necessarily common practice, but I've found them helpful.) I'd use something like 100 - 330 ohm 1W CF or similar.. Might consider flame proof MO resistors as well, since IMLE when a tube fails these resistors burn up.. YMMV and there may be other issues going on here relating to layout, etc.. Pix?

You haven't inadvertently swapped the UL taps have you? (plate and screen checked for phase relationship?)

I noticed that the trace seems rather thick at all points other than the peak where a burst of oscillation is clearly occurring.. Are you sure the output stage is not oscillating all the time or is this a mis-adjusted astigmatism control on your scope?

I think the design is otherwise sound..

[dgta]

Andrew: I did swap tubes and caps, as a matter of fact I tried a variety of coupling caps from 0.1 to 2.0uF, different film types, no change. The chirp does not follow the tubes or the caps. Different caps do affect the waveform of the chirp but to a minor degree.

Kevin: no I haven't swapped UL taps. As far as screen stoppers, I will try them but I can't imagine why that would by asymmetrical. The scope trace is not thick, that's an artifact of the camera. I turned down the lights in the room to avoid a reflection and the trace got overexposed. I should have turned down the trace intensity.

What puzzles me is this: oscillation is not that rare, stuff can happen. But I can't imagine 1) why one tube only and more importantly 2) why I can clear it by touching the grid of the opposite tube.

[Andrew Eckhardt]

I meant swap the grid connections at each cap, an attempt to pin down the asymmetry.

If the burst looks the same but switches polarities the difference is definitely a matter of something to the left of the coupling caps, although maybe handled well enough by changes to the power tube circuit as Kevin suggests.

[kevinkr]

Originally Posted by dgta

A<snip>

Kevin: no I haven't swapped UL taps. As far as screen stoppers, I will try them but I can't imagine why that would by asymmetrical. The scope trace is not thick, that's an artifact of the camera. I turned down the lights in the room to avoid a reflection and the trace got overexposed. I should have turned down the trace intensity.

What puzzles me is this: oscillation is not that rare, stuff can happen. But I can't imagine 1) why one tube only and more importantly 2) why I can clear it by touching the grid of the opposite tube.

Probably because both the leakage inductance and stray capacitances are different between the two UL taps on your output transformers. I have seen exactly the same issue, there may be more going on here than your scope is telling you. I had a VHF oscillation which I picked up with an FM radio in one instance, screen grid stoppers fixed the problem in that case..

It's just a thought, and more investigation and a different solution may ultimately be required, but it is a starting place and a new angle to investigate.