AB763 Reverb Howl/Feedback

[stratomaster]

A common problem on the AB763 reverb circuit is a howl present at higher reverb settings. The accepted fix is to replace the shared cathode bypass cap on V4A/B.  The theory being that the cap is less than ideal and allowing some level of positive feedback between the two common cathode, but series stages.  Alternatively dedicated cathode networks work too.

Separately, we often experience howl and squeals when installing an OT on amps with global negative feedback.  The fix there is to swap the primary leads to eliminate the positive feedback condition.

I couldn't help but notice that the 3.3M/10pF  network and 1M driver grid leak looks a lot like a feedback network for the reverb circuit. 

Given that positive feedback is a possibility when applying a global positive feedback to an output section, could the same be happening in the reverb circuit? And would there be a benefit to reversing the primary leads?  Admittedly I'm ignorant with respect to the single ended nature of the circuit, the non-center tapped reverb transformer, and how the order of the primaries would affect anything--so I'm seeking guidance from the group.

[HotBluePlates]

I couldn't help but notice that the 3.3M/10pF  network and 1M driver grid leak looks a lot like a feedback network for the reverb circuit.

Make sure you see the "channel mixing" and "voltage divider" first that exists between the Dry and Reverb paths of, say, a Deluxe Reverb.




Pretend for a moment that the Reverb control is all the way down.  That means the left-side of the 470kΩ is grounded, and the 470kΩ is in-parallel with the 220kΩ.  The Dry signal has a voltage-divider of 150kΩ / (3.3MΩ + 150kΩ) = 1 / 23 ---> 0.0435x

Pretend the Reverb control is turned up:  it is on the left-side of the 470kΩ resistor, and sees a voltage-divider composed of 470kΩ and 220kΩ.  The 3.3MΩ is there (and there are paths to ground on its left-side), but it's >10x bigger than 220kΩ, so its impact as a parallel path is small.  The Reverb signal has a voltage-divider of 220kΩ / (470kΩ + 220kΩ) = ~1 / 3 ---> 0.319x


Can Reverb sneak-back through the 3.3MΩ?

The left-side of the 3.3MΩ has a 1MΩ to ground, in parallel with a 100kΩ plate load to "AC Ground" at a filter cap.  That's a total resistance of 90.9kΩ (even less if you choose to include the parallel resistance "looking into the plate" of the tube-stage there).  So this sneak-back path also has a voltage-divider present of ~91kΩ / (3.3MΩ + 91kΩ) = ~ 1/37 ---> 0.0268x.

One could argue the 10pF bypasses this divider some, but only above ~5kHz.



I'd worry about mechanical howl from the tank before worrying about feedback through the circuit (though your point about a failing shared bypass cap is well taken).

[stratomaster]

I absolutely see the dividers. Thanks for the illustration.  My question arises mostly out of studying the local negative feedback employed on triodes in Dumble designs and offshoots.  There the intent is to create a feedback loop, but the resistors employed are on the order of 10M-44M.  The actual feedback ratio is very small, but makes an audible difference. 

It also arises from an incomplete understanding of the role of a non center tapped output transformer in a single ended amp. If we pretend the 3.3M were instead 330k and that the secondaries were connected to a speaker instead of the recovery triode--and we were experiencing howl--would the usual tricks of reversing the primary leads flip positive feedback, or does the lack of a center tap ensure proper phasing regardless of primary orientation?

[pdf64]

...would the usual tricks of reversing the primary leads flip positive feedback, or does the lack of a center tap ensure proper phasing regardless of primary orientation?

Swapping the primary red and blue wires around inverts the signal - turns negative feedback positive, and visa versa. So does swapping the green and black secondary wires around.
Do both and it would be back where it started.

[HotBluePlates]

... If we pretend the 3.3M were instead 330k and that the secondaries were connected to a speaker instead of the recovery triode--and we were experiencing howl--would the usual tricks of reversing the primary leads flip positive feedback ...

In a past life I was around recording studios in Nashville with top-echelon gear.

If you'd ever seen a phase-meter on a Sony console (a pseudo-o'scope where the Left channel is applied to the X-axis and the Right channel is applied to the Y-axis), you would know that reverb takes a phase-coherent signal and smears it into "all phases" (and usually at a much lower level).  Therefore, 99% of the signal is "not quite in-phase" and only a tiny fraction is in-phase with the original signal.

If you're experiencing howling, I'd suspect mechanical coupling first.  Especially since no one has seemed to notice this "defect" in Fender reverb for 60 years or so.

[Bieworm]

I’ve had that howling a few times before. My fix was preventing the reverb tank to vibrate by adding some adhesive foam on the outside of the tank. Just thinking out loud

[mresistor]

I've also had the howling..  it was due to improper orientation of the reverb pan.

[Merlin]

Given that positive feedback is a possibility when applying a global positive feedback to an output section, could the same be happening in the reverb circuit?

Yes, positive feedback does occur through the passive mixing. You can make most any Fender reverb howl by increasing the 10pF cap enough.

And would there be a benefit to reversing the primary leads? 

Probably not, because the springs introduce a long time delay into the feedback loop that kinda throws the concept of its phase out the window. It *might* make a difference, at the very particular frequency where the loop wants to oscillate, but we're talking more luck than judgement. If changing the transformer phase does work initially, it may not work in the long term. The tank is mechanical and crudely made; almost anything could slightly change the total phase shift and throw it back into oscillation. Temperature changing the spring properties, a different nearby speaker imposing a different magetic field on the transducers, a new tube with fractionally different gain from the original... so many unknows.