Sine wave scope question

[GainFreak]

Hi,

I'm using a cheap USB scope to check the output sine wave form of amps. At higher resolution I noticed those strange looking parts of the sine wave and was wondering what could be the reason for that?
It's a 2x6V6 power amp, voltages are around 430V, FB is 27k resistor from the 16 Ohm tap, OT is Hammond 1760H.

[Dave H]

A picture would be good

Here's mine. 4 x 6V6 amp, USB scope -

The wobble on the peaks is ripple on the power supply.

[R.G.]

A picture would be good

Actually, a picture would be mandatory.

[GainFreak]

I'm sorry, I forgot the most important thing of my question:



It looks to me like some high frequency oscillation but I'm not really an expert in reading scope waves.

[R.G.]

In solid state amps, that is called "bottom side fuzzies". It is, as you thought, a parasitic oscillation. The amplifier has a response peak at the frequency of those wiggles, but only under the conditions of voltage and current that you see them. The amp breaks into that oscillation - or ringing, if you like - when it crosses the edge into that section of oscillation.

In a tube amp, it's probably layout issues.

[Dave H]

It looks to me like some high frequency oscillation.

It does. Do the 6V6s have stoppers on the control and screen grids?

[GainFreak]

The amplifier has a response peak at the frequency of those wiggles, but only under the conditions of voltage and current that you see them.

You're right that this effect is more pronounced at certain output levels.

Do the 6V6s have stoppers on the control and screen grids?

Yes, they have.

I tried a quick fix - a 220pf cap on the 82k PI resistor and it eliminated the problem (before that I tried 100pf but it didn't fix the problem entirely). The same happens if I connect the cap between the PI anodes or after the PI coupling cap to ground on the 82k side.

[cyclone]

Thats very cool to know how you fixed your oscillation issues, targeted solution with small impact on overall performance. Thanks for posting that.

Here's a cool page to check out now that you have an O'scope, lots of fun!

http://www.r-type.org/articles/art-125.htm

[Chuck H]

IMHO a well considered top end bleeder is actually a good thing. As long as it doesn't affect the intended useful frequencies in any detrimental way. These circuits can help add stability under a wider range of variables than simply, say, fussing with lead dress until that specific oscillation is eliminated. The top end bleed circuit acts more like a dam to stop or regulate the problematic frequency under any conditions. That said...

Many modern designs over use them. Bumping top end with HP filters and then bleeding it off to ground or anti phase a dozen times in a single amp. I know the goal is usually to obtain individual response from individual circuits, but, again IMHO, it really isn't necessary to the extent that this practice is used and the end result is a million phase anomalies and just no sympathetic harmonic interaction resulting in bad, dry tone.

It's probably better to keep things symmetrical in the push pull circuit. So even though you seem to be having the problem on the PI input triode (assuming that's the one with the 84k plate load) you should probably use the cross line shunt for top end management. That's the one with the cap bridged across the PI outputs. I might be inclined to actually correct the problem if possible by manipulating lead dress and then put a 100p cap across the PI outputs anyway. JM2C, YMMV

[GainFreak]

I might be inclined to actually correct the problem if possible by manipulating lead dress and then put a 100p cap across the PI outputs anyway.

When I was testing the amp I was driving the PI directly at it's input capacitor so that the rest of the amp was "isolated".

After eliminating the problem I did some short A/B listening tests and I couldn't hear any difference between the cap on and off so I'll keep it this way or maybe move it across the the PI anodes as you suggested.

[Chuck H]

Were you feeding the PI input with the preamp nulled or disconnected somehow? If you just injected signal at the PI input there is still the potential for in phase interaction from outputs affecting the preamp and also getting injected. If nulling or isolating the preamp when injecting signal at the PI kills the oscillation without the bleeder cap installed this is what's happening. Another consideration would be NFB loop stability. If you have component networks within the loop or on the feedback line itself this could shift the phase and create loop instability. Usually at some resonant frequency as mentioned. This can even be tuned out of a simple NFB circuit by adding a cap and load that affect the resonant frequency in order to shift the phase alignment there. I think I'm making it sound more complicated than it is though.

[Dave H]

After eliminating the problem I did some short A/B listening tests and I couldn't hear any difference between the cap on and off so I'll keep it this way or maybe move it across the the PI anodes as you suggested.

220p across the anodes will have a -3dB point about an octave lower than 220p from anode to ground. Try 100p between the anodes if you hear HF roll off with 220p.

[GainFreak]

Were you feeding the PI input with the preamp nulled or disconnected somehow?

The amp has a Master volume so I disconnected it and injected the signal directly into the PI input capacitor.In this case I consider the preamp disconnected. Did you mean the the preamp itself could be injecting and/or capacitively coupling to the PI?

220p across the anodes will have a -3dB point about an octave lower than 220p from anode to ground. Try 100p between the anodes if you hear HF roll off with 220p.

I didn't hear HF roll off in neither of the cases.

[Chuck H]

Did you mean the the preamp itself could be injecting and/or capacitively coupling to the PI?

Consider that a tremolo oscillator does knock voltages around without adding any of it's own signal. This is what I was thinking. If there were an in phase induction condition affecting the preamp it could affect the linearity of the PI by modulating voltages at the input. But it sounds like that couldn't be your case since you literally disconnected the preamp.

In that case I would look to anomalies in the closed loop circuit. If you have anything like a variable feedback loop, presence control, resonance control or post PI master. Phase shifts and/or lead dress for these circuits may be responsible.

[GainFreak]

In that case I would look to anomalies in the closed loop circuit. If you have anything like a variable feedback loop, presence control, resonance control or post PI master. Phase shifts and/or lead dress for these circuits may be responsible.

I have both variable FB and presence control. Transformers are at the opposite sides of the chassis so I'm running a shielded cable from the 16Ohm tap to the variable FB (it's close to the PI) and then to the PI).
Maybe I forgot to mention that I'm experimenting with power scaling using the "screen voltage method" by adjusting the screen voltage, bias and FB simultaneously as in Marshall AFD/YJM100. FB varies from 29 to 7k (22k pot). This effect is more pronounced at lower power levels at their max non clipping output. I also noticed that when I dial the presence a bit those edges disappear but then the output goes into clipping.