Designing for AB2

[Smitty]

Preamble/Rant

I would like to start a thread with a fairly narrow topic so that we may focus on depth of knowledge. I've started a few posts on this subject and they've quickly evolved into other topics. I'm normally not this pedantic, but I request that replies to this post respond to the topic only so that we may collect our knowledge on this topic in one area. Thank you.

Topic

Starting with a basic circuit of PI => CF => Output tube, what is the most elegant circuit to build for a blues guitar amplifier. The primary consideration should be reduction of blocking distortion as defined by Randall Aiken:

http://www.aikenamps.com/BlockingDistortion.html

The Ampeg SVT/V9 output section is a good starting place.

http://members.aol.com/portaflex/schems/svt6550a.gif

I'd like to focus on just designing for the cathode follower portion of that output section. Not the phase inverter or driver. Other forms of AB2 are not of interest. The assumption is that this will be transplanted into a Super Reverb type amplifier.

I think the way this circuit works is that the bias adjustment simultaneously changes the bias point of the CF as well as the output tube. Adjusting the 15k pot changes the bias point of the CF by changing the potential between the cathode and the grid of that tube. The potential between the cathode and ground is changed because the CF is drawing more current and there is an IR drop across the resistive path. This changes the bias point of the tube. Am I correct?

If so then designing the bias point of the CF is important so that it will work well when the output tubes are biased properly. Opinions on how the tone of the CF in this position is changed by bias should be an important part of this discussion.

My questions are:

1) Would a 12AU7 work well? What changes to the circuit would be needed?

2) I assume that the 150K current limiting resistor on the grid of the CF as well as the 47K on grids of the output tubes are not critical to the operating point of the tube. Is the critical design element here merely the trade-off between the limiting of grid current and the high frequency roll-off due to miller-effect capacitance?

3) What is the most elegant way to design the bias supply? The goal is to reduce bias shift during grid conduction. My preference would be to use only resistors in the grid circuit. Is the SVT circuit best? What is the optimum impedance for the grids of, let's say, 6550s in this circuit?

4) In the ampeg transformer is there a separate winding for bias or are these taps off the secondary? What are the issues here? When the output tube is conducting does the current through the bias winding effect the power supply voltage?

5) Assuming that I am removing the Vibrato circuit on a Super Reverb and using that tube position for the CF, what would be the best setup to prototype this? Would fabricating a separate bench negative supply with a CT transformer and adjusting it with a variac be the way to go? Where should I connect the CT of that supply if not to the CT of the transformer in the amp?

Thanks, I look forward to your responses.

[Bruce / Mission Amps]

1.) Not sure if it matters whether or not it is a 12AU7 but it is a 12AU7 type triode in the Ampeg so I'd go with what works now and tweak later.
2.) I'm not that savy about this circuit but I don't think the 150K resistors are not just current limiting and they are a signal load for the driver in front of it and subsequently the grid load of the CF triodes.... and it does form an AC voltage divider since one 150K is eventually grounded through it's respective bias balance pot.
3.) The Ampeg bias supply is super easy with the FW rectifier and probably better filtered too for very low AC ripple in the bias supply.
4.) They are totally separate windings on the power transformer. I have this same (salvaged) SVT PT in my stereo 4xKT88 amp.
5.) I don't know. You could probably steal low current high V to build a FW bias supply off the high voltage ends of the PT to experiment but I don't know what the ramifications of this would be either.

Sorry I can't be of much help with it but I can tell you we are working on designing a 4xKT88 amp with a similar CF driven audio stage design but it will use power MOSFETs instead of the 12AU7.

[Ray Ivers]

Smitty,

OK, I think I see what you're looking to do now - basically, improve the sound of your existing amp with an AB2-type driver stage (and if it's done right, IMO it will improve it, especially at medium to high volumes). I guess the SVT references were confusing me a bit, as the SVT circuit doesn't really lend itself to retrofitting easily into other designs, but it does work.

BTW, Enzo is right - Joe Piazza leaves out important stuff when he redraws schematics - so you might want to check out the original Ampeg print below if you haven't already (I'll be stunned if either of these links opens - I've read the FAQ, uploaded images as attachments, looked in vain for the "little triangle", used img tags - you name it ):

[img]http://www.freeinfosociety.com/elect...iew.php?id=830[/img]



1) I wouldn't recommend using a 12AU7 in the SVT circuit. It's a 330V tube, whereas the 12BH7A is a 450V (absolute max) tube - and the 6550 SVT circuit has almost 400V plate-to-cathode (it's less in the 6146 version).

2) The CF/output tube biasing works exactly as you described. The 150K and 47K resistors ensure that minimal grid current flows in either the CF or output tube grid circuits; obviously the SVT/V9 PA is not running in class AB2 (although it still works extremely well as the direct-coupled AB1 design it is).

3) 'Elegant' as in excellent performance using minimal components? 15mA Zener regulators on both positive and negative supplies, IMO (with the positive regulator being most important).

4) The Ampeg circuit uses a separate tapped winding for negative bias voltage and screen voltage. I'd say there are virtually no 'issues' with the SVT bias setup, as the CF's idle at 1mA and very little grid current flows, ever.

5) As this post is getting quite long, let me just give a quick overview of my idea of a "good time" in an AB2 driver:

a) +25V to +50V regulated supply
b) -100V to -150V regulated supply
c) low-rp hot-idling CF or MOSFET SF driver
d) direct-coupled PI (I know you mentioned not mentioning the PI but IMO the PI is critical to the AB2-driver design process).

And - last but not least - retrofitting an AB2 driver to an AB1 amp without changing the load line doesn't necessarily result in an amplifier that runs in class AB2; I realize you know this but I just thought I'd put it out there for those who may not.

If you want recommendations for what I think would be the best way to drop low-Z DC drive into a Super Reverb circuit, just let me know - all of the above is predicated on using the SVT circuit, which is what I assume you wanted to use if at all possible ("other forms of AB2 are not of interest") - correct?

Ray

[Satamax]

Hi Smitty!

Well, i'm not knowledgeable on AB2 amps, but i've always loved the look of this one. May be easier than using a CF!

Bye.

Max.

[Ray Ivers]

Here's the (partial) Ampeg SVT PA print - I finally figured out the picture thing (it's gotta be less than 48K and 640x480 or lower resolution).

[Carl / Zwengel Amps]

I'm pretty much in the same camp with Ray in that this little project isn't necessarily a clean retrofit. Also Im not sure if you're hell bent on constraining yourself explicitly to the SVT power amp circuit. If you're looking for an "elegant" ie simple yet effective design may I suggest this one as a starting point.

http://www.triodeel.com/heathw6.gif

Granted it's an ultralinear design but may lend itself a bit more easily to hacking into an existing Fender amp. You can use the power tube and tranny segment from the SVT and the driver section from the above for something of a hybridized design. You'll also want to ditch that voltage doubler in favor of a more standardized power supply as per the SVT. As an aside, with the heathkit circuit the AX7 inverter would be a much more comfortable fit with the the existing fender PI in terms of the overall layout.

[Smitty]

Thanks

Bruce,

Thanks for the info. That's one thing I like about the SVT is the FW. If I weren't stuck in the 60s I'd probably use a SF, too!

Smitty

[Smitty]

Thanks, I think were getting somewhere.

Ray and Carl,

Thanks for the help on this. Ray, you nailed it. A little better tone at high volume. Thanks for the SVT schematic.

1) I won't be running the 12AU7 off a screen supply like the SVT. I'm going to have to find a way to provide 250 - 275 volts at the plates. I think...

2) Maybe that's what I want: direct coupled AB1. Hmmm???

3) I'm not sure I'll be using another positive supply besides B+

4) I was thinking of idling the 12AU7 a little hotter than 1ma. Perhaps 3-4ma. Any reason I shouldn't?

5) Good stuff

a) I must be having a brain fart. What is the 25 - 50V supply for?
b) Yup
c) Um hmm
d) Ummm??? Could I direct couple the plates of a Schmitt to the grids of a CF and then direct couple the cathodes of the CF to the grids of the power tubes?? Wow! I betcha that would sound good if you could get it all sorted out. Can you tell me more?

Ray, I'd completely forgot about the W-6A. Nice call. I see it as very similar to the CF in the SVT. That 12BH7 is running way hotter, though. Only 4 volts of bias would mean not a lot of grid drive on the 6550s, right? 50 volts of bias on the 6550s and only 4 volts of swing seems wimpy. Looks like the SVT might be able to deliver 30 volts of swing. The right bias point on the CF and getting gain close to 1 is going to be mui importante.

[Satamax]

Hi everybody!

Smitty, remember i might be talking out of my ass, but, if the problem of driving 12au7 for the cathode followers at high voltage, wouldn't using cascode followers do the trick (well, i vaguely understand what cascode are but i thought they were able to cope with more voltage.) Sorry i'm going barmy.

Bye.

Max.

[Ray Ivers]

Smitty,

No problem! If you use 12AU7's, you could idle them up to 10mA if you wanted, maybe higher - depending on the B+ voltage they were running at, which would determine the idle plate dissipation. The higher the idle current, the lower the rp of the tube, and the easier it will handle sinking control-grid current.

The +25 to +50V supply was for MOSFET source followers; CF's would want more like a +100 to +150V supply, so they could idle hot as mentioned above. You could run them off the main B+, but then they'll have to idle pretty cold (like the SVT) and won't really handle the conduction transition well, which may or may not be a problem for you (it would for me, though).

Yes, you can DC a Schmitt to two CF's. The plates of a Schmitt splitter are inherently DC-imbalanced, so if you want both AC and DC balance (as I would) you would need to compensate one of the CF Rk values.

There will be > 50V of positive swing going into the W-6A CF's at full power; the 4V differential is the idle bias voltage. It's a cool circuit, although IMO the PS could use a little refining. I tried to find a Fisher 50-AZ schematic to post, with no luck. Thanks for the print, Carl! Tons of great stuff in the hi-fi world, just waiting for an open mind to come along and try it out.

Ray

[Ray Ivers]

Max,

Wherever you're talking out of, I understand you - a cascode or other 'stacked' circuit could work well in this application, especially when too much B+ and/or B- voltage is available. Check this one out:



Remove the "Cap" and use asymmetrical B+/B- supplies for a negative voltage at "out", et voila - bi-directional beef! You don't really need the low-Z negative swing for AB2 drive, but it would also make a cool direct-coupled headphone driver, low-power push-pull OT/PI-less amplifier, etc.

Ray

[Smitty]

Hmm, what about this?

Max,

Wow! Cascode Follower!! There's on for the annals of audio. Wouldn't that require two full twin triodes in between the PI and the output tubes to get the job done? Would there be a benefit over CF? Does a cascode follower have a gain greater than one? Is it lower impedance?

Ray,

Suppose I were to add just one tap off the CT winding that was designed to supply both the plates and the cathodes of the 12AU7 CF. One FW positive and the other FW negative. Then the trick would be to pick the exact right voltage winding that would provide the right voltages at the plate, grid and cathodes in an SVT/V9/W-6M type setup. Am I right?

The bias schema would be paramount in this type of situation, but it sure would be elegant. The bias circuit for the CF ends up being the grid circuit for the output tubes, so I've got an awful lot going on here.

I think you are right, the lower the impedance the better the tone, provided we can still drive the grids of the output tubes properly. I think that would also minimize bias shift under grid conductance, too. Right?

Remaining questions are:

1) Will a tap of the HT winding versus a separate winding work. Bruce has confirmed that the SVT uses separate windings. Can anyone speculate why that is necessary? On a 200ma winding the extra current from the CF (20ma max, right?) plus whatever goes on when the grids of the output tubes go past zero (how much is that?)

2) What is the right voltage for this winding/tap for a 12AU7? This will undoubtedly require some experimentation, which I will take responsibility for, but SWAGS are welcome.

3) Are there any refinements to the SVT/W-6M bias schema that would improve the tone or provide a wider range of operating points, or whatever...?

At this point it may be a good idea for everyone to take pot shots at the concept now.

Lastly, my tiny little brain cannot figure out how to do direct coupling on both the PI and the CF to the output. If the grid of the CF is somewhere between -55 and -75V, is that where the plates on the PI would be? What would Ebb and the reference point for the tail be? Would ground be where I connect the plate resistors of the PI and then shoot for an IR drop that equals the bias point for the CF grids? Would that mean the tail connects to some large negative voltage, like -400 to -500? Can I get this off of the HT winding? Would that mean that the CF bias schema would change such that the grid bias voltage came from the PI supply and the cathode voltage came from the 150V tap of the HT winding? Would I want positive feedback in this negative domain? Am I babbling? I know one thing, just thinking about a total DC output topology is giving me a woody.

My email just told me that someone just posted while I was writing this. I'm going to go read it.

Smitty Er...ah...or perhaps 'Woody' for now.

[Smitty]

Let's take a good look.

This is intriguing. Ray/Max, are you saying the I may be able to simply rectify a FW negative supply off of my existing HT and end up with the right bias voltage for the output tubes? Or by 'asymmetrical' do you mean another winding?

Are there other advantages to this circuit?

'Woody'

[Ray Ivers]

Smitty,

At this point, I've gotta warn you - IME first-time AB2 projects tend to have a certain 'trajectory', usually resulting in at least some measure of unhappiness. Often the first "improvement" to be made is to scrap the active regulation (who needs that SS crap, right? ), followed by the dedicated positive supply (B+ being just fine!), then the driver gets replaced by a 12AX7 (because they RAWK!), etc. - you get the idea. The end result can be a conventional PS supplying a conventional PI/driver/output stage to which CF's or MOSFET's have been grafted, with performance to match. I'm not saying that's what happening here, but I just wanted to throw this out in case things do start heading in that direction.

I've printed out your 16-question post (to be fair, four are for Max ) and I'll get to it shortly. As a preliminary reply - you're basically looking for a signal swing from slightly beyond output tube cutoff in the negative direction, to maybe +5V max in the positive direction (in true AB2 operation, especially triode AB2, the positive swing would be much greater than this). As you can imagine, a +450V CF plate supply is way beyond overkill - it's actually inappropriate, unless you're ready to dump 300V @ 20mA, or bias the CF's freezing cold and then bum out when they won't pull the grids positive.

After posting that a DC PI/CF setup will work (which it will), I then realized that the voltages required will seem kinda freaky to someone who's not used to direct-coupling tube circuits (to me direct-coupling seems normal, that's how wack I am ). You're correct - the PI plates would have to be slightly more negative than whatever resting bias voltage you wanted, with the Rp top well above that voltage and the Rk tail well below this - then there might well be heater-cathode overvoltage issues as well; things can spiral into a big-ass project pretty quickly, and we haven't even begun to deal with tube drift and DC bias stability, which can be a serious problem with cascaded DC tube stages.

FWIW, IMO the simplest drop-in AB2-drive circuit for a Super Reverb output stage is unquestionably a transformer; MV, 12DW7 pre-driver/driver w/12AU7 half driving the tranny primary single-ended (a 12AU7 would also work well if you didn't need the gain of the 12AX7 half of the DW7), Zener-regulated negative supply to a low-value bias pot & the tranny secondary center tap, secondary ends (w/loading R's to the center tap) thru low-value stoppers to the output tube grids, badda-bing, badda-boom. In a zero-NFB PA design I wouldn't be surprised if it was the best-sounding option either, and an excellent transformer (the Allied 6W68HF, about the size of a Fender choke) costs about $14 the last time I checked.

Ray

[Satamax]

Max,

Wow! Cascode Follower!! There's on for the annals of audio. Wouldn't that require two full twin triodes in between the PI and the output tubes to get the job done? Would there be a benefit over CF? Does a cascode follower have a gain greater than one? Is it lower impedance?

Smitty, huh you know i've just emited the idea, i know nothing about cascode followers Well, there's a class AB2 cascode "floating paraphase" amp here http://www.triodeel.com/sim1026.gif I found that expression in RDH4, and did a litle search! And there's a talk on audio asylum where they talk about a cascode long tail pair or hedge type! May this is the one http://www.diyparadiso.com/proj/sfm75.htm

I've dug this as well, where he talks about teh gain of cascode phase spliters! http://www.tubecad.com/december2000/page10.html
On page 14 there's something completely mad looking like a cascode.

nd on page 20 there's something looking somewhat like a cascode follower! http://www.tubecad.com/december2000/page20.html

Nother cascode follower. http://www.jensentransformers.com/as/as021.pdf

And that's a cascode phase inverter!

And the thing i had in mind talking about the cascode follower, is this one. Just to be able to cope with more voltage. That's all.



I never have done any calculations or read much about the real cascode follower on page 20 or the jensentransformers link, i know it has gain, only thing i don't know about is the output impedence.

Huh, Ray, please be easy on me! You know i'm a complete idiot when it comes to complicated tube circuits!
Bye.

Max.