Class AB2: $5, .5 hr

[Ray Ivers]

Watch out, my scanner's up again...

Here's the simplified version of the circuit from the February issue of audioXpress magazine:



This circuit or a similar one may already be in use over at the 18W Marshall page, I don't know - at any rate, here it is.

According to the article, the component values can be adjusted to allow positive grid shift during heavy drive conditions. This would allow the output stage to be idle-biased a bit cooler for a crisp and dynamic clean sound with the guitar turned down, with an automatic shift to a hotter bias point kicking in during overdrive for a more compressed/"woodier" dirty sound if desired; might sound pretty good!

I've got the complete amp schematic here too; I'll scan & post it if anyone's interested.

Ray

[Chris333]

I'd like to see the complete schematic. Probably won't understand why it does what it does, but I'd like to see it anyway...

[Satamax]

Hi vrybody!

Well, Ray, that's the zenner bias you were talking about? If yes, please explain how it works, cos i'm bit daft you see. There's neg grid bias, but how is it dropped by the back to back diode and zenner?????

Bye.

Max.

[Ray Ivers]

OK, here's the full circuit, resized to fit the site restrictions - it's really small, sorry:



and here's the first circuit again:



Operation is pretty simple, really - the Zener conducts when the negative signal swing exceeds the Zener voltage, dumping excess grid electrons to ground instead of allowing them to build up on the downstream cap plate and causing negative bias shift during grid conduction. The regular diode isolates the Zener from the grid during positive signal swings, and the resistor eases the transition into Zener conduction to prevent sharp transients within the NFB loop.

Ray

[BobW]

Ray, How does it sound? Have you performed any A/B tests to determine the ideal zener Vz?

[Ray Ivers]

Bob,

I'm gonna try it either tonight or tomorrow, and I'll post back with the results. From the article, it appears that a Zener voltage of about twice the idle bias voltage maintains a 'neutral' bias situation during hard drive, while reducing the Zener voltage from this point results in positive OD grid shift and increasing the Zener value allows increasing amounts of negative shift.

Ray

[R.G.]

The idea is that a grid usually sits at one bias voltage negative. To get grid conduction and bias shift, the signal has to be more than one bias voltage big on the positive excursion. If the zener is two bias voltages to breakover, then the next half cycle will break over the zener about the same amount as the previous one caused grid conduction, draining off about the same amount of charge, and keeping the bias on average steady.

[Matt T.]

OK, silly question: Could you use this (or something very similar to it) on a preamp triode? I'm thinking specifically about the input triode of a Long tail pair phase inverter. Sometimes, for example, a Marshall circuit will sound amazing and other times the exact same circuit, with the very similar components, at similar voltages will sound all fuzzy/fizzy and bad and it seems that whether the phase inverter is distorting 'well' or not is the issue.

[Ray Ivers]

Matt,

I still have yet to build & try this thing, but it should also work on preamp triodes provided enough current is available to avalanche the Zener. Unfortunately, one place where it may not work so well is at the input to an LTP PI, where very little grid conduction is going on due to the high tail resistance and input bootstrapping - but little or no grid conduction means little or no bias shift, so there may have been no real improvement to be had. This is just supposition as I haven't tried it, but if you do, be sure and take into account the fairly high grid voltage present at an LTP's input when setting up the Zener voltage.


PI idle point, tube type, component values, feedback situation if any, output tube loading - FWIW these are the things I feel affect the sound & performance of the PI stage the most. Different examples of the same tube type plugged into the same LTP PI circuit can often give widely varying idle bias points, with correspondingly wide variances in sound.

Ray

[R.G.]

Ray's right - there are a ton of variables there.

As to generalizing the technique for triodes:
- you could do this with any tube setup that is driven into grid conduction. What it really does is to compensate for grid conduction with driven-really-really-into cutoff conduction in the zener.
- how well it works depends on what it's fed. If the waveform is notably assymmetric, there may not be the balance of grid and zener conduction at all levels. There may be a certain critical level where one side of the waveform is clipped by the preceeding stage. If that clipped side is the positive going side, the bias will be driven off in the overconduction direction by zener conduction at signal levels where the zener coducts and the grid doesn't. If the clipped side is the negative going side, then you can have transient movement into the cutoff region as in normal grid blocking.
In a PI, you could put the zener network between grid and source instead of grid and ground. As Ray notes, the value of the zener will change for every different type of tube and bias point.

[Ray Ivers]

R.G.,

I actually managed to get the test amp ('73 Marshall Major) onto the bench tonight, and perhaps tomorrow sometime I'll actually manage to do something with it. I figured a high-power UL amp would be as good a test platform as any to try this thing out with - I'll be using only two output tubes, though, just in case things go horribly wrong.

It will be interesting to see how it handles asymmetric guitar signals you mentioned, as the article just used pure sine & (briefly) square waves for testing. I'm hoping it doesn't have a lot of idle-point 'bouncing', and closely mimics direct coupling instead. It would also be cool to get some real AB2 Ig and Ik flowing, but we'll see - this PA and PS will surely supply both if the drive is there.

I always wanted to try driving the PA with a signal consisting mainly of just positive PI swings; I guess this one achieves a similar effect, only on the far side of the coupling caps.

I wonder what a typical minimum Zener-breakdown current level is? I'll try throwing one across a preamp-tube grid and see what happens. In my 'symmetrical clipping circuit' testing (which I'll post about after I try it again using my own amp ), I saw a 12AX7 grid source over 2mA of current (!) which seems promising.

Ray

[Ray Ivers]

Preliminary results...

I've been screwing around with this thing the last couple of days, and while it works pretty well in some ways, it's not the drop-in replacement I thought it would be - and why I thought that, I do not know.

The Major's bias voltage is @ -76V, so I've got 100V and 51V Zeners in series with a 1N4007 on each driver output, as per the print in my original post. The stock 5.6K grid stoppers are still in place (mistake), and the stock 56K plate resistors are still on the 12AU7 driver stage (bigger mistake). Tubes are one new, unmatched pair of original Svetlana - now SED - KT88's. Bias point is 20 watts, or roughly 50% P.D.

As far as bias shift goes, the circuit works exactly as claimed - I can detect absolutely NO bias shift on the 'scope, at any drive level. The Zener-ed drive signal clips symmetrically around the bias voltage - the top half from grid conduction, the bottom half from Zener conduction. The NFB loop has been disconnected, and while the stock circuit shows ample evidence of bias shift/crossover distortion at full unclipped power, switching in the Zeners removes almost all of it. FWIW, I like what the Zeners do to the PA during heavy clipping - removing the chewing-tinfoil crossover stuff results in more of a smooth grinding distortion.

The trouble I have is with AB2 operation - so far I've seen little evidence of it, although as I mentioned above I can't believe I thought it would happen by just popping some diodes into a high-impedance AB1 driver circuit. I ran an 8 ohm load on the 16 ohm tap, and got pretty much the same power output with and without the Zeners switched in, although the Zeners did allow the grids to be driven a few volts more positive than the stock circuit before grid-circuit clipping.

The circuit in the article uses a 6BX7 driver - IMO the "heavy hitter" of small octal dual-triodes - along with 22K plate loads, gigantic 2.2uF coupling caps, and 1K grid stoppers; the Major uses none of these things. I think the 12AU7 can stay for now, but the "$5/.5 hr" thing is looking pretty unlikely at this point. I'm hoping to get it working well with minimal changes to most guitar amps, but it looks like a low(er)-Z 12AT7 driver circuit, bigger coupling caps, and smaller stoppers will be minimum requirements, in addition to the Zeners and associated components.

I'll keep at it, and see what I can come up with. I'm using an isolation box/12" Eminence Omega Pro/SM57/mic pre/headphones for listening tests - it's like putting the sound under a microscope, but if it sounds at all good through this setup, it always sounds good through a normal speaker cab.

Ray

[Ray Ivers]

Test Results (unbelievably long!)

First off, it's just come to my attention that this circuit was initially created by Paul Ruby, of the AX84/18 watt/surely many other BBS's. Thanks, Paul - it really works quite well!

OK, I've finished up testing of this circuit, and FWIW I'm quite happy with the results overall. I didn't get the AB2 operation I had (foolishly) hoped for, but what I got, I really liked. It should drop right into practically any SE or push-pull output stage w/no other circuit changes needed, and although I didn't try it on a preamp stage or self-split design, it should also work on those with a bit of tweaking. The test circuit was also fixed-bias; I won't cover cathode-bias here as this post will be WAY long enough as-is.





As an overview - this circuit consists of:

a) a conventional silicon diode (I used a 1N4007) w/cathode (stripe) to the 'upstream' side of the output tube grid stopper, if any - in series with

b) a half-watt or higher Zener diode, cathode to the output stage ground point; I used a 1N4757A (51V) and 1N4764A (100V) in series, as I had an unusually large -76V bias to contend with.

A small-value resistor (< 470 ohms) can be inserted between the Zener and ground to soften the Zener-conduction transition, but to my ears this just diluted the circuit's effect so I didn't use it.

The Zener value should ideally be about twice the idle bias voltage, at least for a starting point. Go higher than double the bias and the circuit has less and less effect; go lower and the bias voltage will actually shift POSITIVE during output-stage overdrive/grid conduction; way cool IMO. You can't go too far in this direction if you want to retain a clean sound, though, as the duty-cycle shift from Zener conduction on the negative signal peaks adds distortion 'shoulders' to the waveform slopes.

The test circuit was a '73 Marshall Major, 600V B+ in UL, using two National KT88-USA's (cool tubes, made in the U.S. w/the original GEC machines!). The driver is a 12AU7 differential amp, 56K plate loads, .47uF coupling caps, 68K output grid bias feed resistors. NFB was disconnected for testing, and levels were set such that only the driver and output stages distorted.

I first tried what I consider an extremely hot AB1 bias point - 50mA, or 70% plate dissipation. I didn't hear too much of what I call the 'springy mosquito' sound of crossover distortion with the Zeners switched out, but with them in it vanished completely. Chords sounded noticeably cleaner with the Zeners engaged, and with them out things got pretty muddy, as you'd expect using huge coupling-cap values like these, along with plenty of 'envelope effect' (swell/decay). I thought the clean sound wasn't too bad, but the night was young...

Next was what I considered the 'acid test' - I biased to 5mA, or 7% (that's SEVEN percent) plate dissipation, which even I consider ridiculously ice-cold. With Zeners out during overdrive, the mosquitos were a-springin' in swarms :>) - it was hard to keep from wincing on palm-muted 'chug' notes and chords - although no real muddiness was apparent. The clean sound was kind of thin, extremely percussive, and edgy/bright; cut-through-the-mix qualities, for sure. In comparison, the 70% clean sound was mostly mids and bottom end, and actually sounded like a compressor with high threshold and medium ratio was being used.

Zeners in: what a difference! The mosquitos vanished almost without trace, and apparent gain increased - it was more of an 'expansion' effect than compression, as the harder I picked, the more gain and overdrive I got. I had one 'scope channel on the grid, and sure enough, the entire signal shifted slightly positive during notes, with the bottom half clipped at the Zener voltage (obviously you won't hear this, as it occurs well below the tube cutoff bias point), and the top half smoothly round-topped as it went above the 0V grid-conduction threshold,up to @ +5V. Chords stayed just as defined as before - if not more so - and while the clean parts of notes stayed 'spanky' and crisp, the overdriven/sustained parts took on more of the higher-current characteristics (darker, more mids, compression, etc.) instead of becoming harsh and metallic as they did without the Zeners. I dug it!

I did more testing - ending up liking 25% P.D. for this setup the best - but this is more than enough out of me for now; I just wanted to get out a 'review' on this circuit before too much time passed, along with two thumbs up.

Ray

[Shea]

Sounds cool, Ray, thanks for the report.

I've been thinking that if you could generate a negative voltage rail that can supply at least double the expected bias voltage, then you could set up a bias circuit consisting of a pot, a 1:1 voltage divider, and maybe a couple of source followers, so that one bias pot would set both the actual bias voltage and another voltage of twice that amount. Then instead of using a zener, you'd just need one silicon diode per side to chop off the signal when it goes more negative than the higher (i.e., most negative) voltage. That way you wouldn't need to change zeners when you change tubes and rebias.

Shea

[Ray Ivers]

Shea,

Just when I thought everyone had lost interest in this thing... It was great to get back to the bench again, even if only for a few days.

Great idea - "Zener scaling"! Maybe you could even get away with just using a MOSFET "gain stage" set to a gain of 2, with the adjustable bias voltage fed to both the MOSFET gate and the grid resistors, the source thru an Rs to ground, the drain thru an Rd to the existing raw bias voltage (if it's not high enough, circuits tweaks or a doubler could be used), the threshold diode to the drain, and AC bypass caps? Maybe I'm missing something...

Ray