Oscillation from Presence Control

[jewishjay]

I had a dead HRD so I gutted it and built this AB763 style circuit. It sounds great EXCEPT...

There's a high frequency oscillation if you dime it all out. It seems possibly related to the presence and NFB? If I turn presence all the way DOWN I can turn the master and the volume all the way up without problems. But in that (extremely loud) situation, turning the pres past 7 makes it flutter and squeal.

If I keep the volume and the master at halfway up, then I can crank the presence without incident.

My design goal was a loud clean pedal platform amp without the transistors, opamps, and PCB of the HRD. I'm almost there, but something funky is happening when you turn everything to 12. Which I admit I would never do....but what about future owners?

Did I mess up the design concept or the wiring? Anything you want to see closeups pics just lemme know. Appreciate your insight...

[sluckey]

Have you tried swapping the OT primary plate leads?

[pdf64]

AB673 reverb channel would have a 50k load on the 3rd stage. Whereas you’ve got a 1M load, which will double the gain.

Why is the series negative feedback resistor lower in value than the shunt resistor value, 3k3 and 4k7 respectively?
AB763 feedback ratio from an 8ohm tap eg DR use 17x higher, 820R and 47R.
The HRD power amp has been copied all but for that resistor, which has been reduced in value massively.

[stratomaster]

This is a cool project and it looks great.

Your feedback loop and presence control circuit needs work.  Revert to known/proven values and test from there.  Disconnecting the nfb altogether can help diagnose if the primaries are flipped as Sluckey suggested. 

I assume the clean amp design goal informed the decision to leave out the Ck on tone stack recovery?  Keep in mind you're also driving the reverb less with this choice coupled with the 3M vs 3.3M splitter/attenuator.

There's a series capacitor effect on the inlet of the reverb driver, and reducing the coupling cap as you have will affect the low end into the driver.  Shouldn't be much though, maybe be ready with a 10 to 20pf cap to parallel the 500pf you have (or if you're using a 470pf, look at a 47pf) currently if you find the reverb thin.

Also you're missing a jumper on the mid pot for the true AB763 tone stack. 

I personally like the sound of the lifted 50k load, but it does result in added gain and background noise.  Experiment with a parallel resistor across the outer lugs of your Master.  51k would be the obvious choice, but I'd mess with valves up to 120k because I like the effect. 

The split anode resistors on the upper half of your PI might be worth abandoning if you find it imbalances your PI too much.  This could have implications on you neg feedback loop too.

If all else fails, the trusty 120pf across the pi plates can help diagnose (or bandaid) an oscillation. 

[jewishjay]

Have you tried swapping the OT primary plate leads?

Yes. Thanks, but that's not it.

[jewishjay]

AB673 reverb channel would have a 50k load on the 3rd stage. Whereas you’ve got a 1M load, which will double the gain.

Sorry, I'm not understanding....you mean the 1M master volume pot? Or the 1M impedance to ground at the driver input?

Why is the series negative feedback resistor lower in value than the shunt resistor value, 3k3 and 4k7 respectively?

Ok, thanks...using that formula I changed the shunt to 200 ohms. All that did was defeat the presence control. I may try disconnecting the NFB next... and maybe increasing the NFB resistor?

[pdf64]

AB673 reverb channel would have a 50k load on the 3rd stage. Whereas you’ve got a 1M load, which will double the gain.

Sorry, I'm not understanding....you mean the 1M master volume pot? Or the 1M impedance to ground at the driver input?

The 1M track of your master volume control presents a 1M load to the 3rd stage. ie almost no loading effect.
Whereas in the typical AB763 reverb channel, there’s a 50k trem intensity control after the 3rd stage. The 50k track presents a 50k load on to the 3rd stage. That loads it down, reducing its gain and maximum available voltage swing.

Why is the series negative feedback resistor lower in value than the shunt resistor value, 3k3 and 4k7 respectively?

Ok, thanks...using that formula I changed the shunt to 200 ohms. All that did was defeat the presence control. I may try disconnecting the NFB next... and maybe increasing the NFB resistor?
[/quote]
Yes, changing the shunt feedback resistor value here to 200R will greatly affect the action of the presence control.

In any such loop, there’s 2 negative feedback resistors, creating a potential divider on the output voltage that being fed back. Lots of people only seem to notice the series resistor (R1 of a potential divider), 3k3 in your schematic, and don’t seem to regard the shunt resistor (R2 of a potential divider), 4k7 in your schematic, as having any bearing on the signal being fed back.

The HRDeluxe/Deville use a 47k series feedback resistor. You changed that to 3k3, I suspect that your rationale for doing that was ill conceived.

I suggest you revert the 200R shunt resistor back to 4k7, and change the 3k3 series resistor to 47k, as per the HRD
https://en.m.wikipedia.org/wiki/File:Resistive_divider2.svg

[stratomaster]

On the original HRD the negative feedback comes off the 4 ohm tap, were you attempting to compensate for this and missed a decimal point?  I haven't done the calculation but a 33k resistor makes more sense there than a 3.3k. 

[jewishjay]

Oh, thanks I wasn't perceiving the NFB and shunt as a voltage divider. Duh. My bad.

So... I changed it to 47k and 4.7k BUT I still have the issue somewhere in the presence OR reverb circuit. Now I can crank everything to 12, but if I turn the reverb past 3 or 4 it makes a motorboating sound. If I turn the presence down to zero, I can turn the reverb to 7 before the motorboating starts. Does that help narrow it down at all?

Thank you all for your help and patience. I'll work on an accurate schematic, and update this thread after the busy weekend....

[pdf64]

Yay, we’re making progress :)

The motorboating symptom is typically caused by a positive feedback loop forming between stages via inadequately decoupled HT nodes.

The reverb recovery stage is usually supplied from the D B4 HT node, whereas your’s is from B3.
So try swapping that.

[jewishjay]

Ok, here's a corrected schematic of what actually exists right now. Unfortuanatly, pdf64: the reverb recovery WAS already on B4 I just notated it wrong in the schematic. Also stratomaster: the nfb IS on the 4ohm so now the schematic reflects that also.

Next time I work on it I'm gonna triple check, make sure this really is accurate and chopstick some wires around just in case it's a lead dress issue.

What are some other possible causes of motorboating?

[pdf64]

Maybe a HT cap is dodgy or not actually connected?
Or the grounds have been mungled together, rather than being returned to the appropriate local star?

Or regular oscillation is somehow provoking motorboating?
The reverb return input is very sensitive, the ‘not super well shielded’ reverb return cable runs close to the speaker, but there’s no grid stopper etc.
So maybe try Merlin’s low noise 10k and 100pF input grid stopper filter at the grid of the reverb recovery stage.

I suppose the long tail pair LTP uses a 12AX7, same as the HRD?

[tubeswell]

^What pdf64 said^


Maybe check your B3 cap to see if it’s open

[jewishjay]

So maybe try Merlin’s low noise 10k and 100pF input grid stopper filter at the grid of the reverb recovery stage.

Winner winner chicken dinner! I played with all the ideas presented here, and this was the one that worked. 10k and 500pf mostly fixed it, but 820pf really got it.

I also lifted most of the bypass caps to reduce overall gain, and added a resistor to make a sort of fixed reverb dwell. I may still play with voicing.... but it's working for now.

So...did I fix the problem or just mask it? Would you suggest any other values? Other solutions?

Thanks so much for all your patience and help!

[sluckey]

Did you lose any brightness with the reverb?

[pdf64]

…
So...did I fix the problem or just mask it? Would you suggest any other values? Other solutions?
…

That seems a lot of capacitance. My guess is that your layout / lead dress might be causing excessive coupling between the reverb return grid circuit V3 pin 7 and the mixer stage anode circuit V3 pin 1.
Have you got a screened cable between the reverb return RCA and the V3 socket grid stopper etc?

[stratomaster]

The silver faces used 0.0022uF across the 220k grid leak.  I'm guessing the reason for this added cap is the relatively poor lead dress on those amps and a problem similar to what you're experiencing.  This filter is usually installed immediately at the RCA jack to eliminate the need for a shielded run.

I'd start there instead of increasing the capacitance of an added on low pass filter--mostly because it's a known good solution with historical usage in this style circuit. However the corner frequency you've created is ~20kHz.  You shouldn't be able to hear the effect of that through the speaker--and it could be a more transparent fix for your problem.  If you're leaving it in, use a shielded cable to your grid stopper since it is best to install it at the socket.

The "fixed dwell" 10k/1M voltage divider you've created at the reverb input is cutting the signal by less than 1% -- essentially nothing. 

Personally I'd get rid of the 10k and put the recovery bypass cap back in, and play with the values 1uF to 4.7uF to boost the useful reverb signal and minimize the mud, rumble, and motorboating.  Use a log pot on the Reverb control to get better control over the lower end of the range.  If it's still too much, increase the 470k resistor off the wiper to 820k or 1M.  This way you cut the noise proportional to the signal. 

Additionally, and similarly, put the bypass caps back on your triodes.  Increase the coupling cap to the Master to .047uF.  Put a 51k resistor across the Master volume pot.  This gives you essentially a stock AB763 in terms of gain when the Master is full up with the ability to essentially lock the amp along the tremolo circuit's swing.  A lot of the amp's inherent sweetness is in the compression that happens in the 3rd triode--Even when playing clean.  I'm afraid you might lose that with the 2nd and 3rd stage being unbypassed.  But these are voicing choices to be made once you're fully functional.

[AlNewman]

Would be interesting to see some voltages, especially in the PI area.  Looks like maybe you could increase the 6.8k resistor in the tail?