Doubling the power of hifi tube amplifiers

[DF96]

Using relatively low amounts of feedback to overcome gross distortion is not a good idea. Consider two amps. One is in its original state and, open loop, has 5% second-order distortion. 20dB of feedback pushes that down to 0.5%. There is also 0.25% of distortion generated by the feedback, but the feedback then pushes that down to 0.025%.

The second amp has been modded to overdrive the OPT. It now has 20% distortion open-loop, but 32dB feedback pushes this back down to 0.5%. The extra distortion is now at 4%, and the feedback pushes that down to 0.1%. Basic distortion is 0.5% as before, but the higher order terms introduced by feedback have grown four-fold. A THD measurement will hide this, but real ears might notice it.

If you want to have significant LF distortion, then straighten it out with feedback (which seems to be the basis of this design), then you really need solid-state levels of feedback. Not 30dB, but 60dB plus.

One of the criticisms of SS amps is that they suffer from rising distortion with frequency, because the open-loop gain has to fall with frequency in order to maintain stability. As a result the distortion residual (e.g. from crossover distortion) is, in effect, differentiated. This fits in with the general result (pointed out by Bruno Putzey) that high feedback means that the output distortion is proportional to the inverse of the open-loop forward gain function. Your modified amps are likely to sound more like SS than valves, at higher sound levels. This is because you have had to add extra compensation components. All this, for a mere 3dB increase in power which in most cases will hardly be noticeable.

 

[slafferty]

To Gilgy:

Never published before?

--- That is correct. I have repeatedly challenged posters on this thread to reference a published work which discloses the combination of all four elements of this invention. In spite of the prodigious levels of experience which have been professed, not a single reference has been produced. It is fair to conclude, that this invention has never been published before.

30dB NFB isn't unheard of, does anyone actually know the NFB level of a stock ST-70? I'd bet it's not far off 30dB.

--- I didn't say "unheard of," I said "extraordinary." You missed the comments which appeared on the previous page (7) of this thread. The original ST-70 had 18-20dB NFB.

Your other mods are utterly trivial, I'm amazed you even deem an extra heater transformer worthy of discussion.

--- "Utterly trivial"? I rest my case on the hyenas comment No one is claiming that ANY of the four elements of this invention is new. However, if you remove ANY of the four elements, you can't achieve nearly as great a result. So no. Not one of them is trivial.

Most inventions are a collection of otherwise unremarkable things, which combined, achieve a remarkable result. Most inventions seem obvious, AFTER they are disclosed. Picture this: You're a guy in Venezuela, without a lot of resources. You long to have a big-time tube amp. You can afford to add four EH 7591s and power them from a little heater transformer. You move a tap on the OPT; solder-in a couple of cheap silicon rectifiers; add a few cheap compensation parts. And PRESTO! You've just turned your little Eico ST-70 into a 124-watt amp, as powerful some rich dude's pair of MC60's! THAT is what the hoopla's about!

No one ever published an article before to tell Andres how to do this. He invented it himself.

All this thread has shown me is that 4 output tubes can produce more power than two. Which is sort of what I knew already.

--- If that is what you think this invention is, you have missed a lot more than the NFB figure for the stock ST-70.

 

[TheGimp]

But it looks good on paper.

 

[DF96]

Mullard 5-20 (first published in 1959)?) has 30dB of feedback. Not "extraordinary", just towards the upper end of 'normal' for a valve amp.

 

[Tubelab_com]

Remember this from networks analysis class somewhere in the past.

An amplifier must have gain in order to amplify. Assume it has 40 db of gain (a random number I pulled from a warm dark place). This is known as open loop gain (gain without feedback).

Applying feedback reduces that gain by the amount of feedback. So applying 30 db of feedback will reduce the total system gain (closed loop gain, or just loop gain) to 10 db.

For our amplifier system to have the original 40 db of closed loop gain the open loop gain must be 70 db since it will be reduced 30 db by applying feedback.

So to build an amplifier system with 30 db of feedback we need 70 db of open loop gain in the forward amplifier path.

The ideal feedback equation teaches us that as the open loop gain (A) of the amplifier increases the amplifier system takes on the characteristics of the (inverse of) feedback system and the amplifier itself has less and less of an effect. If the amplifier had infinite open loop gain the total system transfer equation is equal to 1 over the feedback system (Beta) and the amplifier is not part of the equation. This is why most solid state and high feedback tube audio amplifiers all sound the same. As we increase the feedback, the open loop gain must also increase. This will reduce the measured distortion but render all amps sounding the same.

Most tube amp lovers want some imperfection, the right kind and sound of imperfection of course. Application of large amounts of GNFB will eliminate this sound. 30 db has been stated as the limit where tube amps begin to loose their life. Some users prefer less, but several outstanding amplifiers have been built with 30 db of feedback. The Citation II is one of these as ar some McIntosh designs. All of these use multiple feedback loops instead of a single 30 db GNFB loop.

The ideal feedback equation, its application and reduction of terms can be found here. Ignore the complicated math. Just understand the 1/Beta part,

http://www.ti.com/lit/an/sloa017a/sloa017a.pdf

 

[DF96]

Paralleled output tubes, extraordinary NFB, OPT used at half-impedance, solid-state rectifiers and an add-on heater transformer.

Parallel valves are obvious, and already used in PA amps.
30dB NFB is not even remarkable, let alone "extraordinary". You need to get out more.
OPT does not have an "impedance", so it can't be used at half-impedance. However, I know what you mean. This will boost LF a bit, but increase distortion. Adding more NFB to overcome distortion is obvious, if perhaps misguided.
SS rectifiers are obvious.
Extra heater transformer is obvious.

An 'invention' is when other people say "Wow, that's clever. I would never have thought of that." Being patentable under US law is not a good criterion for invention. A new way to tie your shoelaces would probably be patentable under US law. In fact, if nobody has yet patented it, any way to tie your shoelaces would probably be patentable.

 

[TheGimp]

Here is another good one from TI (I'm reading it now as I'm quite rusty at this):

http://www.ti.com/lit/ml/sloa077/sloa077.pdf

Hmmm...

Actually it might have the same contents in a different form.

 

[slafferty]

To Tony:

that's what you think, tube builders have been doing that way before.....i read an article in the "Popular Electronics" Magazine in the 70's stating that using solid state rectifiers, your amp gains around 9 watts more power...

---Tony, you really should read a thread before you comment. If you had, you would see that the issue is NOT whether any of the four elements of this invention has been done before. It is whether the combination has been done. THAT is what it took to achieve this result.

To TheGimp:

The original ST70 used a 1K FB resistor from the 16 ohm tap to a 47R in the cathode of the input stage. Since the voltage at the 16 ohm tap is SQRT2 times the voltage at the 8 Ohm tap (which we are using as the output), the feedback factor (Beta) should be (47/1000)*SQRT2. Feedback it 20 Log ((47/1000)*1.414) = -23.55dB

--- Congratulations--you have just precisely calculated the wrong thing. When we talk about the NFB of an amplifier, we mean the gain around the negative feedback loop, including the open-loop gain of the amplifier. If you were able to do that, you would find that the NFB is 18-20dB.

As pointed out, one or more other amps should have this procedure applied to them to support the generalization that it can be applied to any amp (or similar wording).

--- I fully justified the claim in post#63. If you object to any of the statements in that post, you should address those directly.

To west_nm:

Name two commercial designs which use 30dB or more of global negative feedback. Name even one that uses 40dB or more.---I think I know of one. HK Citation II... 30dB feedback. I don't know if it's global.

--- That one was mentioned before. So far, no one has come up with another. So the claim that there was more than one isn't supported. In any case, my position that 30dB of NFB is extraordinary, is clearly justified.

To DF96:

If you want to have significant LF distortion, then straighten it out with feedback (which seems to be the basis of this design), then you really need solid-state levels of feedback. Not 30dB, but 60dB plus. One of the criticisms of SS amps is that they suffer from rising distortion with frequency...Your modified amps are likely to sound more like SS than valves,

--- The compensation and NFB of the Hotrodded ST-70 were built on my article about the ST-70A, which used 30dB of NFB. I looked carefully at the spectrum of the ST-70A distortion products and did not notice significant levels of high order harmonics. Your theoretical concerns about the fact that NFB is reduced at high frequency, are put to rest by the actual measurements of the ST-70A: Distortion is 0.13% at 10kHz and 0.33% at 20kHz. In any case, the primary distortion challenge for the NFB in the Hotrodded ST-70 is at very low frequencies, say 20-80Hz. All significant harmonics of those frequencies will fall well below the point where the loop starts to roll off. Thus, the primary distortion challenge is addressed by effectively flat NFB.

 

[SY]

I looked carefully at the spectrum of the ST-70A distortion products and did not notice significant levels of high order harmonics. Your theoretical concerns about the fact that NFB is reduced at high frequency, are put to rest by the actual measurements of the ST-70A: Distortion is 0.13% at 10kHz and 0.33% at 20kHz. In any case, the primary distortion challenge for the NFB in the Hotrodded ST-70 is at very low frequencies, say 20-80Hz. All significant harmonics of those frequencies will fall well below the point where the loop starts to roll off. Thus, the primary distortion challenge is addressed by effectively flat NFB.

What did the spectrum look like at 20Hz when pumping some power? Harmonic distribution? That's pretty easy to measure and post.

And could you answer the earlier question about what other amps this has been successfully tried with?

 

[ArtG]

I suspect the word "extraordinary" is being misinterpreted by some. Steve's correct use of the term is to point out that the amount of feedback used in the circuit is more than is normally applied in this situation, meaning "above the ordinary". The term is commonly used today to imply "tremendous" or "phenomenal", and that's not the intent, in this case.

 

[DF96]

I too would be interested in the LF distortion spectrum. Ideally, before and after mods.

Congratulations--you have just precisely calculated the wrong thing. When we talk about the NFB of an amplifier, we mean the gain around the negative feedback loop, including the open-loop gain of the amplifier. If you were able to do that, you would find that the NFB is 18-20dB.

Yes, he did calculate the wrong thing. However, you may be describing another wrong thing. You have described the loop gain. This is not the feedback factor.

Your theoretical concerns about the fact that NFB is reduced at high frequency, are put to rest by the actual measurements of the ST-70A: Distortion is 0.13% at 10kHz and 0.33% at 20kHz.

On the contrary, this confirms my concern. Closed-loop distortion is rising rapidly with frequency, even though likely distortion sources in this frequency region will probably be independent of frequency. This is not a sign of "effectively flat NFB" as you claim. In the HF region it is the shape of distortion, not the level of distortion, which I was referring to. If masking plays an important role in distortion perception then rising distortion with frequency is a bad thing.

Leak routinely used around 26dB of feedback in their valve amplifiers (according to Stephen Spicer's book), starting with the TL12 in 1949. A Leak advert claims 10dB of gain margin, so I'm sure that Harold Leak would find it quite funny that 60 years later someone is claiming 30dB feedback as "extraordinary".

 

[rongon]

Isn't there a problem with how the NFB loop is implemented? Looking at the schematic, the pot is wired so that the 8ohm tap from the OPT secondary can be shorted to the cathode of the 12AX7 voltage amp (first stage). Wouldn't that inject enough NFB to make the amp unstable? Basically, you can adjust the pot to have near-zero ohms from the 8 ohm tap to the 150 ohm resistor to ground.

Or am I misunderstanding something?

-=|=-

 

[slafferty]

To tubelab.com:

As we increase the feedback...This will reduce the measured distortion but render all amps sounding the same. Most tube amp lovers want some imperfection

--- By "sounding the same," you mean sounding perfect, as far as the ear can tell. I think it's fine if some people prefer distortion in their sound. However, my goal is audio equipment which reproduces sound as accurately as possible.

To DF96: (May I call you DF?)

Mullard 5-20 (first published in 1959)?) has 30dB of feedback. Not "extraordinary", just towards the upper end of 'normal' for a valve amp...30dB NFB is not even remarkable, let alone "extraordinary".You need to get out more.

--- 30dB of global negative feedback is extraordinary because few, if any, commercial tube amplifiers were ever produced with 30dB or more. Naming two such amps was the challenge presented to a poster who claimed he had seen more than one. The Mullard 5-20 you mentioned was not a commercial amp. It was a demo published by Mullard to sell tubes. Even they said in the article, "In spite of the high level of negative feedback...the circuit is completely stable." Yet they admit that it might oscillate with capacitive loads. (Something the ST-70A does not do.) The HK Citation II has been mentioned as a candidate but tubelab.com above mentions that it used multiple feedback loops instead of a single 30 db GNFB loop. That is my recollection, as well. So I would give you half a banana for citing the interesting Mullard amp. However, as long as not even one or two commercial amps with 30dB of global negative feedback can be found, I am on firm ground, calling the 30dB used in the invention, extraordinary.

To Sy:

What did the spectrum look like at 20Hz when pumping some power? Harmonic distribution? That's pretty easy to measure and post.

--- I'm sorry that I can't volunteer to pull it out of the system and do lab work for you. However, as I recall, it looked pretty much like you would expect from any vacuum tube amp: Mostly third order, some second order. Not much higher order. In any case, I think I clearly answered DF's concern that there might be reason to think that the harmonics could actually rise with frequency or be significant at high orders.

And could you answer the earlier question about what other amps this has been successfully tried with?

--- That poster was questioning the claim that the Supermods™ technique will work with almost any push-pull, fixed-bias, hifi, tube amplifier, which has room for more tubes. I did answer that, explaining in detail why the technique is so broadly applicable. (It's because the principles upon which it works are quite general--see post #63.) Since the article was only announced three days ago, it is too soon to expect that others have tried it, so there are none, to my knowledge. Go for it! You might learn something <wink>

 

[TheGimp]

So the feedback factor is simply the feedback network ratio expressed in dB? ie, in the case of a Dyanaco ST70 with a 1K feedback resistor and 47 ohm cathode resistor, 20 Log 47/1000= -26.55dB.

The feedback of the modified amp can not be calculated as no value is given for the 10 K pot.

Various schematics give different values for the cathode and feedback resistor.

47R / 1K - ST-70 (Dyanco Doctor copy).
47R / 680 - ST70 II (SCC distributor copy C to gnd, not 4 ohm to gnd as original)
100R / 6K - ST-70, Stephen Lafferty from ofiginal schematic
150 / 4.3K - ST-70 Modified, Stephen Lafferty "ST-70A Modification Package"
47R / 10K potentiometer - Stephen Lafferty "Hotrodding the EICO ST-70"
100R / 2.21K ST-70 MK III replacement board

 

[SY]

To Sy: --- I'm sorry that I can't volunteer to pull it out of the system and do lab work for you. However, as I recall, it looked pretty much like you would expect from any vacuum tube amp: Mostly third order, some second order. Not much higher order. In any case, I think I clearly answered DF's concern that there might be reason to think that the harmonics could actually rise with frequency or be significant at high orders.
--- That poster was questioning the claim that the Supermods™ technique will work with almost any push-pull, fixed-bias, hifi, tube amplifier, which has room for more tubes. I did answer that, explaining in detail why the technique is so broadly applicable. (It's because the principles upon which it works are quite general--see post #63.) Since the article was only announced three days ago, it is too soon to expect that others have tried it, so there are none, to my knowledge. Go for it! You might learn something <wink>

OK, then so:

- You haven't actually tried this with other amps, so your claim of universality is as yet unsupported by any actual evidence

- You don't have LF distortion spectra to see if the effects one would expect from running the core closer to saturation are present nor the effects of the dropping OPT primary inductance on the "extraordinary" feedback

- You don't have an actual harmonic breakdown of the rising distortion at higher frequencies

- The negative feedback is dropping sharply at higher frequencies (though I think that's not as big a problem as what will happen in the bass with transformers sized for lower power)

 

[DF96]

The Mullard 5-20 you mentioned was not a commercial amp. It was a demo published by Mullard to sell tubes.

It was available from several suppliers as a kit. The Leak amps had a similar circuit. So a 30dB fb amp was available as a kit. A whole series of 26dB fb amps were available ready-built, with sufficient spare margin that 30dB would have been reliably stable. Harold Leak did not pick 26dB because 30dB would be "extraordinary", but because 26dB provided him with good enough distortion and appropriate input sensitivity. The modern usage of 'extraordinary' means 'well beyond ordinary', not 'within the normal range'. 30 is not well beyond 26; component tolerances might get you half-way there. It is silly to persist with this extraordinary claim.

So the feedback factor is simply the feedback network ratio expressed in dB?

Yes, but this is not what people mean by, say, 30dB of feedback. Feedback factor is beta, in clg=A/(A*beta+1), A is open loop gain. A*beta is loop gain. A*beta+1 is the amount of feedback.

 

[slafferty]

To DF:

Yes, he did calculate the wrong thing. However, you may be describing another wrong thing. You have described the loop gain. This is not the feedback factor.

---I did not use the term "feedback factor." I have referred to the "NFB of an amplifier," in context, to mean the loop gain of the amp (as others here do).

On the contrary, this confirms my concern. Closed-loop distortion is rising rapidly with frequency...This is not a sign of "effectively flat NFB" as you claim. If masking plays an important role in distortion perception then rising distortion with frequency is a bad thing.

--- Rising rapidly? No, it's rising at roughly 6dB/octave, which is what you expect from loop rolloff at high frequency. What I actually said about "effectively flat" was that the primary distortion challenge [just mentioned as roughly the 20-80Hz region] is addressed by effectively flat NFB. You need to read more carefully. Moreover, the figures of 0.13% at 10kHz and 0.33% at 20kHz are excellent for this kind of amplifier and are in no need of "masking."

To rongon:

Isn't there a problem with how the NFB loop is implemented?...the pot is wired so that the 8ohm tap from the OPT secondary can be shorted to the cathode of the...first stage). Wouldn't that inject enough NFB to make the amp unstable?

--- Yes. Ideally, Andres would have put a resistor in series to protect people who don't know better. The article specifies an adjustment procedure which begins with the pot at the center of its range. It also mentions that the amp could go unstable if the pot is set too low.

To Sy:

You haven't actually tried this with other amps, so your claim of universality is as yet unsupported by any actual evidence

--- It is supported by understanding the principles of how and why this works. If you disagree with some part of my detailed explanation of that, you should address that directly.

You don't have LF distortion spectra...an actual harmonic breakdown

---Look Sy, it was made clear in the article that Andres does not have the test equipment to do that kind of thing. It was also made clear that this was about hotrodding, not space engineering. It is what it is--what-you-see-is-what-you-get, and it's, guess what--FREE! If you want that stuff, I suggest that you build it and measure it and report back to us, instead whining about someone else's effort to achieve something.

 

[ColinA]

I see so many posts quoting distortion figures from calculation.
Or computer simulations.
Or spectrum anylisers with ociscopes. (can't spell)
Does anyone on here actually build something and then listen to it?
I bet Leak, Quad & Radford used the Mk1 Earhole to design their clasic products.

 

[slafferty]

To ColinA:

Does anyone on here actually build something and then listen to it?

There are listening tests described near the end of the article, on the second page. We are trying to get more.

 

[Gilgy]

Regarding the claim that no one has done all these things in a single amplifier before (which I find difficult to believe but we'll ignore that for now) - to the best of my knowledge no one has applied extraordinary feedback, added a heater transformer, SS rectifiers, paralleled output tubes blah blah blah blah etc and then, wait for it, painted a smiley face on the volume knob.

If I do all these things, can I claim to have invented something new?