Favorite easy 6V6 push-pull circuit?

[frank754]

So your opinion, Merlin, is that an interstage transformer would not provide as good
performance?

 

[refference]

Hi frank754 ,

I agree with Merlinb , in my opinion an interstage
transformer is not a good choice for hi-fi use .

Interstage transformer is GOOD for public-adress
amp . and is VERY GOOD for class AB2 amp .

We have a LOT of options of phase inverters cir-
cuits and I think that an interstage would be the
last one .

Regards ,

Carlos

 

[Garage Amps]

Wha?! On valve, socket and a few components compared to an interstage? The tranny would be at least three times more expensive- minimum!
Not to mention problems with hum susceptibilty, weight, lack of ability to tweak tone and drive levels, phase shift and associated feedback instability...

But the context there is electric guitar amplifiers. Hifi 20-20KHz bandwidth isn't necessary or desirable, 80-8KHz would be fine, and such transformers aren't very expensive or heavy. Fewer components mean lower assembly labor costs, too.

After all, various guitar amps with transformer phase splitters have been designed, manufactured, and sold over the years. Any drive level, hum, and instability problems were solved in the design; at least, players don't seem to complain about those things. But still none of them are regarded as being among the best guitar amps. IMO it would be somewhat interesting to do a controlled experiment on the topic.

--Paul

 

[Merlinb]

After all, various guitar amps with transformer phase splitters have been designed, manufactured, and sold over the years. Any drive level, hum, and instability problems were solved in the design; at least, players don't seem to complain about those things. But still none of them are regarded as being among the best guitar amps. IMO it would be somewhat interesting to do a controlled experiment on the topic.

I'm quite sure those older amps that did use interstage trannies were great, and I don't contest that using one will probably make a great amp. I have an interstage at home that I will one day put to good use in guitar amp.

However, I don't think those amps fell by the wayside because the bog-standard Fender LTP PI topology sounded better, but simply because Fender dominated the market and eventually became sinonymous with the 'best' amps, because they were what everyone was using!
And amps that came later are nearly all based on those Fenders. And I'd bet that Fender didn't use a tranny simply because it was cheaper not to, or possibly because the "new" LTP was in vogue at the time (after all, most of Leo's amps were just copies of datasheet circuits).

I have no doubts that you can build a greta sounding amp with an interstage. I was simply disagreeing with the statement that cost and practice between tranny and a regular valve PI is small. I think there is a BIG cost/practice difference, heavily in favour of the valve PI.

 

[Dave Cigna]

IHowever, I don't think those amps fell by the wayside because the bog-standard Fender LTP PI topology sounded better, but simply because Fender dominated the market and eventually became sinonymous with the 'best' amps, because they were what everyone was using!

True, but there was a good reason that everyone (namely professional musicians) were using them.

And amps that came later are nearly all based on those Fenders. And I'd bet that Fender didn't use a tranny simply because it was cheaper not to, or possibly because the "new" LTP was in vogue at the time ....

Again, I agree that Leo Fender probably did not choose to use something other than a transformer just because it was cheaper.

The part I really want to comment on is this:

after all, most of Leo's amps were just copies of datasheet circuits

That is completely untrue. The very first amps were similar to circuits commonly found in text books, but so what? The same is true of most 'hi-fi' amps at the time. In any case the early amps are not the ones that made Fender famous. If you look at the evolution of the designs you see a very consistent progression towards lower distortion and higher 'fidelity.' This is clear all the way up to the black-face amps of 1965 when Leo sold the company to CBS (and everything went to hell.)

The tone control circuit that first appeared in the larger amps in the late fifties - the one that has become the standard tone stack - is related to some that I have seen in text books. But as far as I can tell, it is an original Fender design.

The method of applying NFB to the bottom of the LTP has never shown up in any text book that I have seen (and I've read a lot) never mind a tube datasheet.

Look, let's get some perspective. Fender was in the business of making musical instruments for professional working musicians (and amateurs that could afford them.) A professional guitarist is not going to buy a guitar that is just a production line copy of something else. They will buy something that plays good, sounds good, and lasts. The same is true of an amp. That's why Fender amps became popular. They sounded great and were very well made.

All right, enough ranting. It's just that I have weak spot when it comes to people dissing good guitar amps as being poorly designed or uninspired - or as nothing but bad hi-fi designs that happen to work well enough for guitar. Merlinb, I know you didn't really say that, but my knee-jerk jerked. Thanks for listening.

-- Dave

 

[frank754]

Nice discussion, this provides some great opinion and info, I'll hopefully do this design at some point and see how it turns out.
For the one I'll be working on this coming week, the hi-fi design, it was based on the output transformer Stancor A-3311 10K center tapped primary.
If I use a Hammond T-1609 should I favor 8200 or 11600 as a primary setting, since there are several choices?

http://www.hammondmfg.com/125.htm

In the case of 8200 I would hook up the 8 ohm speaker to pins 2 & 5

 

[Merlinb]

The tone control circuit that first appeared in the larger amps in the late fifties - the one that has become the standard tone stack - is related to some that I have seen in text books. But as far as I can tell, it is an original Fender design.

The method of applying NFB to the bottom of the LTP has never shown up in any text book that I have seen (and I've read a lot) never mind a tube datasheet.

True the (particularly later) amps did include some different design features, but there were really just tweaks and add-ons to the tried-and-tested topology of older designs (a few of which were borrowed from the Hammon organs!). IMO they don't constitute an original and innovative design, just adaptations of existing designs. After all there're only so many ways to skin a cat. If the amp is a few gain stages, and LTP and a PP power stage, anything else you bolt on is hardly a revelation.
Applying NFB to the LTP may not appear in any text book, but why should it? You can apply NFB anywhere you like, and the LTP would be a pretty obvious choice. I know today Mesa would try and patent something like that, but it's hardly an "invention". It's just common electrical sense.

 

[Dave Cigna]

Applying NFB to the LTP may not appear in any text book, but why should it? You can apply NFB anywhere you like, and the LTP would be a pretty obvious choice. I know today Mesa would try and patent something like that, but it's hardly an "invention". It's just common electrical sense.

A) It is absolutely not "just copies of datasheet circuits." It is this comment and its implication that I was addressing.

B) The NFB is not applied to the LTP in any obvious way. In fact, the way it works is decidedly not obvious. It's not especially difficult to figure out when you look at it closely, but it's far from obvious.

I am not saying (and never have said) that these amps are revolutionary. But, they are also not just copies from a datasheet! The amps that made Fender famous were the result of careful and deliberate refinement. This is true of the late 50's tweed amps and even more true of those from the mid 60's. That is a very, very different thing from "just copies of datasheet circuits."

Concerning the NFB, take a look at this schematic - the power amp of a black face Deluxe. The NFB is applied to both the unused LTP input and the bottom of the tail resistor. This is not obvious. How they came up with it, I don't know, but clearly they felt that it was better (sounding) than the obvious method, which they no doubt tried as well. Not "just copies of datasheet circuits."

-- Dave

 

[Merlinb]

Well maybe it just seems like very basic theory to me then. NFB to the grid as normal, then a little bit to the tail to improve the CCS effect of the resistor to improve balance.
What later became evident though was that a lot of imbalance in a guitar splitter actually sounds better! Of course, no one was overdriving their amps in those days.

 

[frank754]

One thing that I've been curious about, is the tone controls of some of these guitar amps, this one for a more elaborate example.

On some of the hi fi amps, there are no tone controls, can something like the above be adapted to one of these, and it seems at the stage just before the phase inverter?
Is there any really good resource to learn more on this on the web.
It seems they are using a bunch of pots to navigate through various capacitances stacked up in a row, if I see correctly.
Do the varying interstage capacitances either clip/favor a certain range of the audio spectrum?
What determines if a coupling is more broadband (full spectrum favorable), rather than favoring a subset range (treble, bass), as far as the amount of power passed?

 

[Dave Cigna]

Well maybe it just seems like very basic theory to me then. NFB to the grid as normal, then a little bit to the tail to improve the CCS effect of the resistor to improve balance.

I have to admit, I don't see how applying the NFB to the tail improves balance.

In any case, the way I think of it is that the NFB is first applied to the tail. If everything were perfect that would do nothing at all. It's a common mode signal; it is effectively applied to both inputs of a differential amplifier. The 'fix' is that the unused grid is referenced to the bottom of the tail resistor and not to ground. The consequence is that the NFB is now applied to the upper grid, the one that also receives the signal.

My suspicion is that Fender wanted to use the feedback signal to gain a little headroom at the input to the LTP. Let's see ... at 20W the speaker terminal would be swinging about 12.5V rms. The 820/47 ohm voltage divider would bring that down to about 0.68V rms or almost a full volt peak. That means the LTP could handle almost a full volt more input before clipping.

Again, I admit that I don't see how applying NFB to the bottom of the tail resistor improves balance. Maybe you can help me out with that.

-- Dave

 

[Dave Cigna]

On some of the hi fi amps, there are no tone controls, can something like the above be adapted to one of these, and it seems at the stage just before the phase inverter?

These tone controls are not at all flat, so would (probably) not be well suited to hi-fi. You can download a very easy to use simulator from DuncanAmps.com that allows you to play with different component values and see graphically what the frequency response will be. It simulates this particular tone stack as well as a few others that might be better suited to playback (hi-fi.)

Is there any really good resource to learn more on this on the web.

There is a forum at the DuncamAmps site. Also, way back, a guy named Mark Garvin offered a very good analysis of this style circuit on one of the usenet newsgroups. The one devoted to guitar amps. I have no idea where to find that analysis today, but it was worth saving, so somebody must have. (I probably did, but finding it in my archives would probably be harder than finding it on the internet. )

-- Dave

 

[Garage Amps]

[ Also, way back, a guy named Mark Garvin offered a very good analysis of this style circuit on one of the usenet newsgroups. The one devoted to guitar amps. I have no idea where to find that analysis today, but it was worth saving, so somebody must have. (I probably did, but finding it in my archives would probably be harder than finding it on the internet. )]

Well, what is called Part 3 of Garvin's writeup is various places, e.g. http://ampage.org/htac/garvin.tone.html. Interestingly Part 1 and 2 seem to kind of have fallen off the face of the internet. (So Dave, dig em out!) But that 'Part 3' is by itself a great intro to the difference between Fender and 'hi-fi' Baxandall passive tone stacks.

 

[Garage Amps]

In any case, the way I think of it is that the NFB is first applied to the tail. If everything were perfect that would do nothing at all. It's a common mode signal; it is effectively applied to both inputs of a differential amplifier. The 'fix' is that the unused grid is referenced to the bottom of the tail resistor and not to ground. The consequence is that the NFB is now applied to the upper grid, the one that also receives the signal.

IMO that is a good description of what's going on. Aiken and others talk about the feedback being applied at the bottom grid, but that's not really right.

My suspicion is that Fender wanted to use the feedback signal to gain a little headroom at the input to the LTP.

OK but wouldn't applying the feedback directly to the upper grid (but taken from the other end of the output transformer secondary of course) have the same effect on 'headroom'? Intuitively the tube just sees the potential difference between its grid and cathode.

My feeling is that since you are deriving the feedback from such a low impedance source, you are going to have issues summing it to a high impedance circuit point (the phase splitter input grid). So, apply it to the cathode instead. But I'm guessing.

Just looking at schematics of what Fender called Bassman amps over the years, they tried various phase splitter topologies and various feedback schemes. Using a LTP with feedback at the tail is what got copied by Marshall and lots of others and used in most of Fender's later amps. But I still don't really understand why that should be.

Another question I don't understand is why if this scheme is original to Fender and is a big part of the secret of great electric guitar tone, why didn't Fender patent it?

 

[Wavebourn]

Another question I don't understand is why if this scheme is original to Fender and is a big part of the secret of great electric guitar tone, why didn't Fender patent it?

Patenting what others assume stupid idea means drawing attention to it. I can publish lot of my secrets on forum and in magazines so nobody can patent them anymore, but I don't believe somebody would use them because people believe that I am wrong.
I assume, Fender might have similar reasons.

 

[Merlinb]

Another question I don't understand is why if this scheme is original to Fender and is a big part of the secret of great electric guitar tone, why didn't Fender patent it?

Because it's not a secret of great tone, it's just one miniscule thing that happens to work well, but if you remove it you still get great tone. (I also suspect that it wasn't an original Fender design anyway, I dare say there's prior art out there somewhere.)

Anyway, first of all the feedback is to the 2nd grid of the LTP. That's where most of the action happens. It's still possible to do this without connecting it to the tail, and you probably wouldn't hear any difference if you did do it that way.

Here's a half decent explanation:
"...you are injecting feedback into the tail resistor, which is sent to both grids: the typical setup has all the grid grounds tied to "virtual" ground potential at the bottom of the tail resistor (remember: your input is bootstrapped, as well as the whole stage being bootstrapped by virtue of the tail resistor). The grid that the feedback is applied to is tied to ground potential by (typically) a .1uf cap that is where you are referencing the feedback loop. The NFB signal is fed partially through the tail resistor to help balance out the uneven voltage gains of the differential halves of the LTPI that is caused by the introduction of the negative signal. The tail is "split" to provide an easy point to add a voltage divider that will not upset the equalizing effect of injecting some of the negative signal through the bottom of the tail resistor.

So intranslation; If you had a better-balanced LTPI-- like say adding a CCS instead of a tail resistor-- then you can reference all of your grids to chassis ground, and you don't have to inject NFB anywhere but the second grid. Then you can make the grid reference resistor any value you like and adjust the feedback resistor according to your open loop gain"

 

[Dave Cigna]

OK but wouldn't applying the feedback directly to the upper grid (but taken from the other end of the output transformer secondary of course) have the same effect on 'headroom'? Intuitively the tube just sees the potential difference between its grid and cathode.

I had been thinking about it, and I think you're right. The reason for this has to do with the way the cathodes in a LTP 'follow' the grids partly. If you apply +1V DC to one of the grids it will cause the voltage at the cathodes to rise 0.5V. Instead, if you apply a 1V AC signal to one of the grids then you'll see a 0.5V AC signal at the cathodes. The result of this is that the headroom (before the grid goes positive) is twice the bias voltage. [Disclaimer: this is precisely true only if everything is balanced; perfectly matched tubes, identical plate resistors, infinite resistance in the tail, etc., but it's still close even in the circuits we're talking about here.] This explains why the LTP has only half the gain of an ordinary grounded cathode stage, but anyway ....

So what happens when NFB is applied to the other grid? If we make the simplifying assumption that the amp is producing little distortion to begin with then the feedback signal is essentially identical to the input signal. The 1V peak NFB signal that I mentioned earlier will result in 0.5V peak signal at the cathodes. Since this is in phase with the input signal it is something like providing 0.5V additional bias. As explained above, the extra 0.5V bias gives the input tube an extra 1V of headroom.

All of this ignores nonlinearities so is rough, but it gives us a good first approximation.

My feeling is that since you are deriving the feedback from such a low impedance source, you are going to have issues summing it to a high impedance circuit point (the phase splitter input grid). So, apply it to the cathode instead. But I'm guessing.

hmmm.... I don't see the problem.

Using a LTP with feedback at the tail is what got copied by Marshall and lots of others and used in most of Fender's later amps. But I still don't really understand why that should be.

Nor do I! I wish Leo were here to shed some light. I'd buy him a beer.

Another question I don't understand is why if this scheme is original to Fender and is a big part of the secret of great electric guitar tone, why didn't Fender patent it?

I don't know if it's a big part of the secret. Certainly it's evidence that they didn't just pick recipes from circuit cookbooks.

They did receive a couple of patents that I know of, there might have been more. One was in 1957 for a tremelo circuit that provided simple amplitude modulation by modulating the plate voltage of the phase inverter. It's interesting that they later came up with better ways of doing this, but didn't bother to patent them. That's US patent 2817708

The other was in 1961 for a much more interesting tremelo circuit that split the signal into low and high frequencies. The amplitude modulation was then applied out of phase; when the highs got loud the lows got soft and vice versa. The effect was intended to mimic the sound of a Leslie rotating speaker. They did use the circuit on the brown face Concert (and maybe another, I can't remember) but didn't offer it for more than a year or two. That was US patent 2973681

-- Dave

 

[Dave Cigna]

Anyway, first of all the feedback is to the 2nd grid of the LTP. That's where most of the action happens. It's still possible to do this without connecting it to the tail, and you probably wouldn't hear any difference if you did do it that way.

I disagree. The way I see it, if you don't apply it to the tail (by connecting the bottom of the tail resistor directly to ground instead of to the top of the 47 ohm feedback resistor) then you'll end up with positive feedback, not negative.

"...[snip]...The NFB signal is fed partially through the tail resistor to help balance out the uneven voltage gains of the differential halves of the LTPI that is caused by the introduction of the negative signal. The tail is "split" to provide an easy point to add a voltage divider that will not upset the equalizing effect of injecting some of the negative signal through the bottom of the tail resistor.

I don't see the tail as 'split.' Looking at the schematic I suppose it appears that there's a 22047 ohm tail resistor with a tap at 47 ohms, but applying the NFB signal to the tap does not mean that only 47/22047 of the signal is applied to the cathodes of the LTP.

I don't see the NFB being partially applied to the tail, I see it totally applied to the tail. The entire tail is sitting on top of the 47 ohm feedback resistor, at least 99.8% of it is. The cathodes of the LTP and one of the grids see exactly the same NFB signal.

Also, and this is important, I still don't see how applying NFB to the tail, partially or totally, has an equalizing effect. It seems to me that any feedback scheme used to balance the splitter would need to be local to the splitter. The feedback signal taken from the speaker terminal can't know which part of it is due to splitter imbalance. I'm fully prepared to be educated, but so far it isn't happening.

-- Dave

 

[Dave Cigna]

(remember: your input is bootstrapped, as well as the whole stage being bootstrapped by virtue of the tail resistor).

OK, after rereading a comment made by Brian Beck in another forum this is starting to make sense. If the diff amp were driven at both grids with identical out of phase signals there would be no signal appearing at the cathodes (ignoring HD and imbalances in the circuit.) But, applying the input to only one grid means that there will be a signal at the cathodes. That's the way the LTP works.

Applying a signal of the same phase to the bottom of the tail resistor bootstraps it (the tail resistor) making it appear larger. One way to understand this is to realize that the AC signal voltage across the tail resistor is smaller. Say we start with 1.5V AC at the cathodes resulting in 1.5V AC across the resistor. Ohm's Law tells us there will be i = v/R = 1.5/22k = 0.07mA of signal current through it. If we apply 0.5V AC to the other end with the same phase, then the voltage across the resistor will be only 1V. Now the current through it will be about 1/22k = 0.05mA.

OK, from the perspective of the cathodes, they're producing 1.5V and giving up 0.05mA. To them the tail resistor looks like R = v/i = 1.5/0.05 = 33k ohm. Now I get it.

Note that I pulled the 1.5V and 0.5V numbers out of my butt. The actual degree of bootstrapping depends on the actual amount of feedback applied and I'm not motivated enough to figure it out at this point.


-- Dave

 

[Dave Cigna]

Four posts in a row, sorry. I hope someone finds my thoughts interesting.

The cathodes of the LTP and one of the grids see exactly the same NFB signal.

Now I'm thinking that this might not be the case. If the NFB were applied directly to the cathodes from a low impedance source then the full signal voltage would be seen by the LTP. Applying it through the large tail resistance means that it is reduced by the voltage divider formed by the tail resistance and the input impedance of the cathodes.

Let's see, the resistance looking up into one cathode is something like (rp+Rp)/(u+1). With the 12AT7 we have rp=11k and u=60. The Rp's are 82k and 100k. So both cathodes in parallel present a resistance of around 830 ohms and the NFB signal there would seem to be divided down by a factor of 830/22k = 0.04 or 28dB. This is the basic function of a current source which is what the tail resistor is supposed to be in the first place.

Thanks, Merlinb, for forcing me to try to figure out what's really happening. I'm learning something, I hope someone else is too!

-- Dave