intro / AB763 PI design question

[electron obvious]

Hey all,
I mostly lurk around forums because I figure that I do not know enough to help anyone or my questions have been asked too many times (and if I searched enough I could find answers). When I stumbled onto one old thread and found posts by raiken and merlinb, I registered immediately. Merlin's book and the info on both of their sites are incredibly generous and helpful. I'm not sure I will get past being a mere modder compared to all the true designers on the forum but searching over a decade of posts can feel a bit futile.

I own a Fender DRRI circa '93-95. I've had some buzzing problems and the problem seems to be in the PI. I replaced the speaker a few years ago. I replaced all electrolytic caps. I did geek out a bit modding. (Once I had that PCB up I couldn't help myself.) Mid pots, tone defeat footswitch jack (norm only), switch to put the normal channel through effects (and last gain stage), slope slide pot (norm only, from Merlin's book), vibrato defeat switch, NFB defeat switch. I had some anxiety about going too far before fixing my problem, but after stuffing it all back together all mods seem to work with no noise from all the wiring. Problem is that the original buzzing is worse, maybe due to a tad extra gain (or leaks plugged) in the preamp.

The weirdness seems to become visible with a scope at the grid of the non-inverting side of the PI (12AT7). With a 1kHz sine input the waveform at the non-inv grid goes fractal-looking and subsequently, the plate and corresponding grid of the following 6V6. (I double-checked the connections of my NFB line, which seems to be cool.) Tonight I will be going back in to check/replace the coupling cap from that grid to ground hoping that some DC is messing it up.

1) Would that coupling cap cause this problem (resulting in an annoying buzz surrounding any note(s) strummed beyond a tickle)?

That's more troubleshooting than design and I still have work that I should do to answer that but I do have a couple questions about Leo's design after pouring over info from raiken and merlinb:

2) Why is 100K chosen as the load for a 12AT7 when for most preamp triodes it is chosen as roughly twice internal plate resistance?

3) In Merlin's book (preamp) he suggests a cool pot (I know, here I go again) to vary Rk (the PI bias resistor), suggesting a 47K or 100K pot. Would this value, or mod, fly with the Fender design and AT7 vs AX7? Too big for most of the pot's sweep? Is it hard for the tubes (or one of them) to run the bias cold and below the design Rk value of 470 ohms?

4) Can a grid stopper be inserted at the grid of the inverting side, all else remaining stock? 10K, less?

Power tube question related to 4):

5) Can the 6V6 grid stoppers (1.5K stock) be increased within reason (15K - 100K) without other changes?

I should probably stop my rambling.
Thanks

[pdf64]

1/ With regard to the LTP grids, I'm not sure which you are referring to as being inverting and non-inverting?
How about g1 as the main input grid, g2 as the feedback grid, as per Merlin's page The Valve Wizard -Long Tail Pair

If scoping g2, bear in mind that it's inside a negative feedback loop, and so what you are seeing is largely the error signal.
If scoping the power amp, it's much more helpful to open the feedback loop.

2/ I suspect that the LTP was designed for a 12AX7, and for some reason a 12AT7 was swapped in, with pleasing results. Perhaps the lower gain made overdrive less harsh.

3/ Not sure whether you are referring to the LTP bias (eg 470) or tail (eg 22k) resistor?

4/ Grid stoppers should be fine.

5/ Very high values may act with the tube's Miller capacitence and roll off HF at too low a frequency, perhaps making the tone dull. Also they add to the total grid-cathode resistance, which may then exceed the limiting value. Off the top of my head, I suggest 22k as a max.

[electron obvious]

Thanks for the quick pickup

"3/ Not sure whether you are referring to the LTP bias (eg 470) or tail (eg 22k) resistor."

Referring to the LTP bias 470 ohm.

Thanks

[Gingertube]

Grid Stops should be at least 8 divided by the triodes gm.

gm for a 12AT7 is typically 5.5mA/V reduicing to about 3.75 at low voltage and curernt. So for 12AT7 8 / 0.00375 = 2133 Ohms. Use 2K2, no point in going higher.

I have had a 12AT7 PI oscillate on me before and adding grid stops on BOTH triodes fixed it (no other changes). At the time I had'nt read the articles which talk about the 8/gm minimum value and I just put 22K in.

For 12AX7 gm is typically 1.6mA/Volt, use 1.2mA/V worst case to give 8 / 0.0012 = 6666 Ohms. 6K8 would do but just use 10K, again no point in going higher.

If you want to use either 12AX7 or 12AT7 then use the 10K value.

When I build these days every triode gets a 10K gridstop as "default".

Cheers,
Ian

[ThermionicScott]

Since Pete brought up that the Fender LTP originally used 12AX7s, it's worth noting that the original circuit also used a 47pF capacitor across the plates to quell HF oscillation. Although Fender apparently didn't see the need after switching to 12AT7s, it's something to try if you find your LTP oscillating.

[electron obvious]

Since Pete brought up that the Fender LTP originally used 12AX7s, it's worth noting that the original circuit also used a 47pF capacitor across the plates to quell HF oscillation. Although Fender apparently didn't see the need after switching to 12AT7s, it's something to try if you find your LTP oscillating.

Hey thanks,

What are the classic audible signs of PI oscillation? What is the downside of the cap tying the plates?

I solved my buzz problem thanks to forum replies to another thread/question i posed (My goofball mistake of putting 470K screen resistors instead of 470R), but I have been exploring /learning from Blencowe and Aiken and have changing up the LTP design in my DRRI a bit. It seems that (following their typical design process for a LTP) the component values for the AB763 fit only a 12AX7, though Leo must have liked something about the response of the 12AT7 (or had a lot of cheap ones to use). For a 12AT7, 12AY7 or 12AU7 the bias seems very warm and tail pretty small, and for the 12AT7 especially the loads seem large for such a small plate resistance. I made both loads 82K, increased the tail to 33K (and may up it to 47K), and intend to put in a pot to vary the bias (to have the choice to vary between tube types, listen to PI tube distortion in more isolation, and get the bonus for my family-- power dampening), along with the aforementioned grid stoppers.

Where is the source for the most accurate grid curves for design/analysis? I've searched a bit and find the curves differ between manufacturers/years and the scale of the anode current axis small, making the load line shallow on the graph (for all but the 12AX7).

Thanks all

[pdf64]

Since Pete brought up that the Fender LTP originally used 12AX7s, it's worth noting that the original circuit also used a 47pF capacitor across the plates to quell HF oscillation. Although Fender apparently didn't see the need after switching to 12AT7s, it's something to try if you find your LTP oscillating.

I think that the 47pF / 100pF cap across the LTP plates may have been necessary both because of the 12AX7 gain, and the lack of power tube grid stoppers.
BF amps didn't seem to have those caps, but tended to have 12AT7 LTPs / power tube grid stoppers, so there seems to have been several interlinked coincidental changes.

[ThermionicScott]

Hey thanks,

What are the classic audible signs of PI oscillation? What is the downside of the cap tying the plates?

If you can hear a high-frequency squeal, that's one, if there's a "mysterious" loss of volume but a scope shows activity well above the hearing range, that's another.

A cap between the plates (which are producing signals out-of-phase, as you know) allows high frequencies to cancel each other out. Use the right value cap, and it'll "waste" everything above what you need. If it's too big, it'll rob high-frequency audio as well. It sounds like you've fixed your problem, but it might be a good exercise to try it anyway and hear what happens.

I solved my buzz problem thanks to forum replies to another thread/question i posed (My goofball mistake of putting 470K screen resistors instead of 470R), but I have been exploring /learning from Blencowe and Aiken and have changing up the LTP design in my DRRI a bit. It seems that (following their typical design process for a LTP) the component values for the AB763 fit only a 12AX7, though Leo must have liked something about the response of the 12AT7 (or had a lot of cheap ones to use). For a 12AT7, 12AY7 or 12AU7 the bias seems very warm and tail pretty small, and for the 12AT7 especially the loads seem large for such a small plate resistance. I made both loads 82K, increased the tail to 33K (and may up it to 47K), and intend to put in a pot to vary the bias (to have the choice to vary between tube types, listen to PI tube distortion in more isolation, and get the bonus for my family-- power dampening), along with the aforementioned grid stoppers.

Where is the source for the most accurate grid curves for design/analysis? I've searched a bit and find the curves differ between manufacturers/years and the scale of the anode current axis small, making the load line shallow on the graph (for all but the 12AX7).

Thanks all

I would caution against too much rigidity in evaluating these tube circuits. Smart guys like Merlin and Randall have graciously posted helpful articles about designing gain stages and phase inverters from scratch, using classic hi-fidelity practices. However, there are many routes to a good-sounding circuit. Even though the plate resistors in an AB763 PI are higher than the textbook example of twice the plate resistance for a 12AT7, the circuit does a fine job of driving whatever output tubes follow, and Leo and several generations of guitarists think it sounds fine. Therefore, I submit that it's a perfectly valid circuit.

I would be careful about using a really big tail resistor, though. One limitation to be respected is the heater-to-cathode voltage rating was only 90 volts in the old tube manuals. In an AB763 Twin Reverb with a 22k tail resistor, the cathode is already up to 100V, and silverface versions pushed that up to 120V or more. Some new-production tubes don't handle that abuse as well as older ones. (I'm thinking specifically of newer 12AX7s that give up the ghost in cathode-follower duty.)

- Scott

[ThermionicScott]

I think that the 47pF / 100pF cap across the LTP plates may have been necessary both because of the 12AX7 gain, and the lack of power tube grid stoppers.
BF amps didn't seem to have those caps, but tended to have 12AT7 LTPs / power tube grid stoppers, so there seems to have been several interlinked coincidental changes.

Yeah, I think you're right.

It's always risky trying to pin down someone else's thought process, but my thinking has been that Fender saw enough instability with high-gain PIs like the 6SC7-driven paraphases and 12AX7/7025-driven LTPs that a cap across the plates was good insurance. When they switched to 12AT7s, they didn't need that insurance anymore. I took the cap off my 5B5 clone because I could tell I was losing high-end somewhere. Swapping tubes didn't make a difference and the rest of the circuit seemed fine. Voila -- the high-end opened up and the amp is still perfectly-behaved. It's weird, because I would have expected the cap (a 100pF) to be chosen so that it doesn't interfere with the audio response so obviously. I had used a NOS ceramic cap, but perhaps it was bad in some way. Since the amp is stable, I haven't bothered to try other caps, be they other unused 100pFs, or lower values.

The grid-stopper history is even more interesting to me. They started in the mid-1950s, but you can tell that Leo kept trying to get away with not using them, as they come and go from amps up and down the line. Champs and Princetons never got them to the best of my knowledge, but it seems that by the BF era, Fender decided to suck it up and install them on everything bigger to avoid issues.

[pdf64]

When I was tinkering around with 5F6A / plexi type builds, I found that the cap across the LTP plates seemed to make the margin of stability worse, ie when run open loop or presence off for a flat-ish response, and fed with a mid level square wave to the LTP input, a 47pF cap may add a bit of a peak to the leading edge.
It was sometimes necessary to increase to say 220pF to get rid of it.
Whereas (with a bit of attention to lead dress and 5k6 power tube grid stoppers) the stability seemed better of without one at all.
It may be different with an inductive load, but it's difficult to assess the wave shape then.

[ThermionicScott]

When I was tinkering around with 5F6A / plexi type builds, I found that the cap across the LTP plates seemed to make the margin of stability worse, ie when run open loop or presence off for a flat-ish response, and fed with a mid level square wave to the LTP input, a 47pF cap may add a bit of a peak to the leading edge.
It was sometimes necessary to increase to say 220pF to get rid of it.
Whereas (with a bit of attention to lead dress and 5k6 power tube grid stoppers) the stability seemed better of without one at all.
It may be different with an inductive load, but it's difficult to assess the wave shape then.

That's interesting! You've discovered another reason for Fender to abandon that approach in favor of grid-stoppers.

Now you've got me wondering if my cap was in fact causing parasitic oscillations at high audio frequencies thanks to positive feedback, not just a simple cancellation. There is that phase shift inherent in a paraphase PI after all...

[pdf64]

Grid Stops should be at least 8 divided by the triodes gm...

Can you remember where you found this guidance, and if there's anything similar for power pentodes / beam tetrodes? Thanks a million!

[Gingertube]

I don't recall exactly but given what I'd been reading about that time I think it was probably from Morgan Jones "Valve Amplifiers".

I haven't seen anything for pentodes / beam tetrodes but the inverse dependance on gm would still apply (so commonly see 1K5 for high gm EL34 and 5K6 for 6L6 which has half of the EL34 gm). I did'nt chase this down as it is my view that much higher gridstops should be used on output tubes for an entirely different reason. I want to "round" the grid current limiting waveform to limit higher order harmonic distortion products, so I rarely use less than 22K and often use as much as 47K. Higher value gridstops for output tubes are even more critical for amps with concertina splitters., to stop the concertina mis-behaving.

Cheers,
Ian

[mikepukmel]

Grid Stops should be at least 8 divided by the triodes gm.

gm for a 12AT7 is typically 5.5mA/V reduicing to about 3.75 at low voltage and curernt. So for 12AT7 8 / 0.00375 = 2133 Ohms. Use 2K2, no point in going higher.

I have had a 12AT7 PI oscillate on me before and adding grid stops on BOTH triodes fixed it (no other changes). At the time I had'nt read the articles which talk about the 8/gm minimum value and I just put 22K in.

For 12AX7 gm is typically 1.6mA/Volt, use 1.2mA/V worst case to give 8 / 0.0012 = 6666 Ohms. 6K8 would do but just use 10K, again no point in going higher.

If you want to use either 12AX7 or 12AT7 then use the 10K value.

When I build these days every triode gets a 10K gridstop as "default".

Cheers,
Ian

Hi Ian, picking up this post from last year, hope you don't mind an additional related question. I built a deluxe AB763 frankenstien. Before I started, I asked around about how to reduce blocking distortion since many players commented that this was a peeve about this amp. The suggestion was to add grid stops to the second side of the preamp tubes (after the tone stack), and up the value of grid stops on the 6V6 output tubes. Since grid stops help fix several problems, in your recommendations above, does the 8/gm value come from reducing oscillations only? If so, what calculations would I use to also help reduce blocking distortion?

Thanks!
Mike

[Gingertube]

Mike,
To limit blocking distortion you want output tube grid stops to be as high as possible without losing to much treble from the high frequency roll off that is introduced by the grid stop + miller capacitance at the output tube grid.

With 6V6 you should be able to use 47K without any worries, you might even try 100K. I would put in 100K and if you lose too much top end then just solder another 100K across them to reduce them to 50K.

That addresses blocking distortion in the power amp. Use similar in the preamp - as high as you can without losing too much top end. A high gain 12AX7 triode stage (preamp) actually has a much or more miller capacitance at the grid as an output tube in pentode mode does - somewhere around 100 to 200pF, so similar values should work.

In "nerd speak" blocking distortion occurs whenever the discharging time of the coupling caps is longer than the charge time - so bias shifts occur.
Increasing grid stops increases the charge time AND limits the charging current.

Cheers,
Ian

Some reading:
http://www.aikenamps.com/index.php/w...ing-distortion