MAC Autoformers?

P.S. Regarding the distortion, it is in octaves and in multiples of the base frequency, thus the 7th harmonic of 100 Hz is 14,800 Hz and the 7th harmonic of 1000 Hz is 148,000 Hz... That is why the H-Fi standard measures only to 3rd harmonic only up to 7 kHz.

This statement is false. The 7th harmonic of 100Hz is 700Hz and the 7th harmonic of 1000Hz is 7KHz. See:https://en.wikipedia.org/wiki/Harmonic

What is bifurcation in feedback?

Put very simply, bifurcation is the process where feedback can cause a 2nd to become a 4th and so on. Norman Crowhurst wrote about this and I think later also Baxandall. If you look into Chaos Theory you will see quite a lot of papers on the topic. 

The Class A3 amp sounded good too.

The pot is not dual but single with a switch attached, thus the click into the high damping condition which cuts out all the Zmatic circuitry resistor losses. It is indeed a combination of voltage and current feedback as seen in this schematic

Yes- any way that the current and voltage feedback is balanced is how you arrive at constant power. Other amps I've seen use dual controls.

Whatever feedback they use is not of importance, its the resulting output impedance that matters.

I agree that output impedance is important, but if you wind up using too little feedback to get there, it can be detrimental to the sound thru the process of bifurcation. There's a range of sorts- 8-15 db or so is where this sort of thing can occur. Here's a nice article by Nelson Pass:
https://www.passlabs.com/press/audio-distortion-and-feedback

Speaker manufacturers are not so honest about their impedance range and a curve is the only way to know. If they didn’t do so many tricks in the crossover we would not have this problem. The drivers are not the problem, the crossover is.

In my experience most speaker designers do not know much about electronics or care about what the amplifier may have to do.

+1 on this, I’ve felt this for a long time.

I’ve heard this from ESL owners where the impedance rises in the bass to very high values. They believe that OTL amplifiers are better for their speakers because they can supply the extra voltage to drive the high impedance in the bass. They indeed get more bass if the damping factor is low, but not the bass the speaker was designed to give. They get more and perhaps like more but it is one note bass.

[snip]

This is where we get in trouble with the power paradigm which had led people to believe that the speaker wants constant power. It does not. I cannot think of or find a modern speaker that wants constant power, the varying impedance and flat response insure that the designer uses a constant voltage amplifier, ie one with high damping.

@ramtubes

The first paragraph is false. You don’t get a one note bass at all, even if the OTL has no feedback (and most do use feedback). Please keep in mind that one of my employees (who now works at Bel Canto) owned ESL57s, and don’t forget Bill Toberman (RIP) who had ESL63s. The one note bass phenomena only occurs when the panel speaker (it doesn’t have to be an ESL) is too close to the wall behind it, allowing the back wave to reinforce around one frequency. This happens a lot if the ESL user has a solid state amp because getting bass out of the speaker becomes a challenge- most solid state amps can’t make power into 30 ohms or more!

The last paragraph is a bit misleading. The Power Paradigm does not lead people to think speakers want constant power. If anything, most people think the idea is ridiculous. And as you state, most speakers are Voltage Paradigm. Its not until they hear what a properly set up Power Paradigm system can do that they might begin to think that its not bunk (ask clio09). And there are manufacturers that make speakers that are Power Paradigm and I’ve mentioned a few already- Audiokinesis, Classic Audio Loudspeakers, DeVore Fidelity, Lowther, PHY, Feasterex, Onken, Volti Audio, PureAudioProject, Evolution, Sonic Flare.... I can go on and on but I hope you get the point.
Most horns and panel speakers are Power Paradigm devices unless something unusual was done to change that (there’s a guy named Wayne Piquet who does just that with Quad ESL63s, by adding an extra panel to the speaker). For example, Magnaplanars are Power Paradigm but because of their resistive nature work fine with Voltage Paradigm (constant voltage) amplifiers.

Sometimes its the intent of the speaker designer that makes the difference, so the speaker might be acoustic suspension, like the AR-1, or it might be bass reflex and so on. Your comment at the top of this post has a lot to do with that- many speaker designers went to the school of ’by gosh and by golly’ and are simply trying to make the speaker work with a particular type of amplifier, like an SET (which is likely the most common application). If they get the latter right, they will have a Power Paradigm device.

In case you are thinking I made this stuff up, this is really all about history; take a look at this Google search, in this case one for a Fisher A-55:
https://www.google.com/search?client=ubuntu&channel=fs&q=fisher+A-55&ie=utf-8&oe=utf...

One of the first hits is an image from YouTube, showing the damping control on the amplifier, which is a dual-gang pot. If you look closely, it is labeled "constant voltage" with the control all the way down, "constant power" with the control at noon, and "constant current" with the control all the way up. In this case, voltage and current feedback are balanced against each other in the noon position.

Now current drive never really developed into a thing (meaning there are almost no speakers based on current drive), but power drive and voltage drive did. This amplifier was built about the time (mid-late 1950s; Fisher, ElectroVoice and a few others had such ’damping controls’ on their amps during this period) when the voltage rules were first being proposed and so amplifiers had to be adaptable to whatever was out there. Instead of balancing current and voltage feedback, you get similar results with no feedback at all, which is understandable because as you know, current feedback **raises** output impedance while voltage feedback decreases it.

@bifwynne
Roger has one of the best tube matching services in the country.

I suspect that one reason you have issues with the taps is that the ARC amp really needs a few more db of feedback; 15 db is slightly on the low side to get right.

If you have the tubes tested in a rigorous manner as Roger does, it makes sense that they would hold up better :)  It *also* makes a difference that they are mil spec- that might even be the bigger difference!

That does not seem like the sort of tube of which ample supplies exist. We stay away from the collector market as a general rule but if you mess with NOS tubes of course you can get better results.

We've had such poor luck with the garden variety of GA-style tubes that we don't mess with them. Usually what happens is someone buys a mess of them and installs them in the amps without any sort of testing. Then we get a call about a malfunctioning amplifier, but then it turns out to just be tubes. With the Russian tubes (which is what we've mostly been shipping recently) we have few enough problems that doing a warranty for a year is easy.

After the MkIIIs went into production about 14 years ago, we found we were shipping about 10% of replacement tubes that we were doing before that. Its good that you have a set of American tubes working for you- IMO and from all the feedback we have, they sound the best (and the amp makes more power too, an extra 20 watts).

The thing that concerns me about 6AS7s, it that they emit cathode sparks in my Tektronix 530 curver tracer at zero grid.That is consistant with your statement of cathode coating failure as it appears the sparks are bits of cathode coating. They actually look like sparks off of a 4th of July sparkler.

The 6AS7G begins to develop grid current at about -15V or so; 0 volts on the grid is pushing them pretty hard (a bank of 12 power tubes at 0v on the grid can pop a 15 amp fuse rather quickly BTW). Despite that our driver circuit can push them to +15 volts on the grid while maintaining linearity.
Here are some tips for reducing arcing:
Precondition the tubes by putting filament current on them for at least 72 hours if you have the Russian variant (6H13C). If you have the Chinese (6N13) or American (6AS7G) tubes you will want to precondition for 96 hours. If B+ is applied before preconditioning or at any time during the process, you may consider the process ended. We built a jig for just this purpose. Preconditioning reduces premature arcing and can double the life of any power tube.
The Russian and Chinese tubes hold up much better in this regard. The 6AS7GA and its 6080 variant to the GA should be avoided for use in our amps- the grid heatsink is much smaller and tends to warp with grid current.

I know of no tube amp designer who uses SPICE. can you name a few?

Nope. Maybe one? I know I don't! Victor Khomenko was trying to model our MA-1 in Spice many years ago before he became part of BAT. He called me up saying that the amp didn't run in Spice, but I had to point out to him that the amp actually worked despite what Spice came up with :)

As you can imagine, the ratings of the tubes tend to vary with load- and the amp will draw less power and run less heat if a higher impedance load is used. Also, as you've pointed out, the class of operation is affected. In our case the amp is biased to be class A2 on the proper load, but will be class AB2 if the load impedance gets low enough. We don't get crossover artifacts on that account though.

As far as high current spikes, at full power the tubes still have some dissipation left over, so they can handle spikes, however as the load impedance is reduced, eventually this will cause the cathode coating to fail.

If Ralph (Atmasphere) picks this post up, ... what is the tube life is in your amps. Btw Ralph, …. Roger mentioned a couple of times that negative feedback, *if used properly*, is not all that bad a design feature. Not taking a position, just passing along what RM said.

10,000 hours is typical- so we warrant the power tubes for a year on this basis, and always have.


I agree about the feedback 'used properly' (which many designers do not) comment. Proper application of feedback is tricky to say the least, and may not have been possible until the age of personal computing, due to the number of variables involved. Here is a nice primer on the topic:http://www.normankoren.com/Audio/FeedbackFidelity.html

He also had a set of our M-60s- really nice guy.

I did read your white paper before entering this discussion. Lets just agree to disagree. Peter Walker designed the 57 to be used with an amplifier with a damping of 20 and even specified the series inductance. To say that these speakers will play with a damping factor of 1 is not fair to the speaker, no matter that some people like a widely altered frequency response. How can we discuss little differences in distortion when the frequency response has been so altered to make the speaker unreconizable?

You may not remember one of your customers, Bill Toberman (RIP, lived a few miles from me), but he had a set of RM-9s which beat out any ARC introduced to his system. He had Quad ESL63s. At the time I didn't think our amps would work with ESLs, but when Bill convinced me to bring them to his house, I was proven wrong- the MA-1s and Quads proved to be a match made in heaven. No loss of highs either, and this in direct comparison to the RM-9. The ESL57 proved even easier to drive and more accessible, since one of my employees had a set. In a nutshell, the FR is not as unrecognizable as you suggest!

We've also gotten Best Sound at Show using our amps with ESLs... all I'm saying here is there is more to this than meets the eye at first blush.

As we all know, there is theory, but there is also practice and we have many happy Quad customers. The thing is, OTLs can provide far more current than most people realize; I've seen one of our 12-tube output sections take out a 15 amp fuse...

As an antique radio collector I have given a lot of thought to how early SE amps (the 45 in particular) got relatively flat response without feedback. In that case the driver was a single cone in an open baffle cabinet so the impedance rise was not so severe and relatively flat. These early radios sound pretty good. However that does not represent modern popular speakers. 

If you recall the old days, speakers like Altec, Klipsch, JBL and others had level controls for the midrange and tweeter. These controls were not for room adjustment, they were there because the amplifier to be used had an unknown voltage response. **That** was how frequency response was handled in the old days prior to the introduction of the Voltage Paradigm. This practice persisted into the early 1970s as the industry switched over; KLH had level controls for their tweeters as did the AR-1, the world's first acoustic suspension loudspeaker. Of course, now open baffle speakers are back- one of them is in that list I posted above and its not the only one by any means. Google:
https://www.google.com/search?q=open+baffle+speakers&client=ubuntu&hs=UEV&channel=fs&...


Nowadays there are entire generations of audiophiles that grew up thinking that the voltage rules are the only game in town. In mid fi, that's certainly true but in high end the power rules have been in play for decades. That is why SETs are around. We've been making zero feedback OTLs now for over 42 years. The reason all this is going on is due to objectionable distortions-  and what kinds the ear doesn't care about. That latter statement is why the tubes/transistor thing has been raging on and on, and is why tubes are still in production (horns too), 60 years on after being declared 'obsolete'...


Just to be clear, we've been working on our own take on a class D amp for the last 2 years; not using anyone's modules, since we realized we had something to bring to the table. That amplifier acts as a voltage source; please don't think that I don't understand what's going on here. The paper was written because many people ask why we don't use feedback, and an explanation was needed that went beyond 'we don't like it'.


The reason 'paradigm' is used is because someone inside the paradigm (which is a platform of thought) will regard that which is outside the paradigm as blasphemy or heresy- it can't possibly be right. But this is history: the Power Paradigm used to be how everyone did it. These days our understanding of physiology has greatly expanded so we can better see why there can be a benefit to this approach.  

@ramtubes
Roger, you might want to read this:
http://www.atma-sphere.com/Resources/Paradigms_in_Amplifier_Design.php
If the Quads were close to the wall behind them, they will often have one-note bass. This often happens when a solid state amp is employed; the customer will note that there is no bass on account of the higher impedance of the speaker into which the amp can't make power, and so will move the speaker around, eventually finding out that if they move it closer to the wall behind it, they can finally get some bass. But this is not how *any* planar is supposed to be set up.

ESLs don't respond so well to voltage rules, depending on the speaker- the Quad ESL57 and 63 being pretty good examples- also the Sound Lab ESLs, AudioStatic and even Martin Logan (if you use a set of ZEROs to allow for their low impedance).  The reason is, unlike box speakers, the impedance curve is not also a map of its efficiency, which is pretty much the same across its bandwidth (on account of its impedance being based on a capacitor rather than a driver in a box with resonance). This is typically a 9 or 10:1 change in impedance! If the amplifier output power varies with this curve, the result will be too much highs and not enough bass.

The solution we've found with our customers using Quads is to have them pull the speakers further out into the room, so there is at least 5 feet behind the speaker to the wall. In this way the bass normalizes and customers report that the Quad is playing bass quite well.

I know of no speaker maker who designs for high output impedance amplifiers.

I can name a few- Coincident, Classic Audio Loudspeakers, DeVore Fidelity, Audiokinesis, Lowther, PHY, PureToneAudio and many more. Essentially, any speaker that works with an SET is working with an amplifier of high output impedance.

We are quite a ways off-topic; if you would like to discuss this further it would be a good topic for another thread.

@bifwynne , Well, since you asked, the answer is ’distortion vs actual frequency response’.

All amps make distortion. Some distortions (IMD and higher ordered harmonic distortion) are more audible than others. Of the two mentioned, the latter is what makes solid state amps bright and harsh, and does that to tube amps as well, although they are not as harsh as solid state because there is simply less of this form of distortion.
In a zero feedback amp, there is respectively (solid state or tube) even less. That’s why they sound smoother. The problem is, solid state amps have non-linear capacitances built into the junctions of the output devices. This capacitance is responsible for higher ordered distortions (brightness and harshness) that still affects a zero feedback solid state amp.

Since the ear uses the higher ordered harmonics to sense sound pressure, it is more sensitive to these harmonics than most test equipment, since it has to cover a 120-140db range! This is an insidious, inconvenient truth that the audio industry does not like to face. This is further complicated by the Fletcher-Munson loudness curves, which place many of these upper harmonics in the most sensitive region of the ear’s response.

That makes things particularly tricky.


The reason zero feedback tube amplifiers exist is that a good number of designers have come to recognize the problem of higher ordered harmonic distortion. One fairly easy way to sort that out is to avoid using feedback, since it is known to add such distortions of its own in the process of otherwise suppressing distortion. So this means that the designer has to use other means to suppress distortion.

This results in an amplifier with a fairly high output impedance.

The thing is, if you chose the speaker carefully, the result is **far** more neutral than is possible with the conventional approach (its not subtle- you can hear it right away, whether an audiophile or not). The reason this is possible has to do with two salient facts: first, no speaker made anywhere at any price is flat. Second, the ear converts distortion into tonality, and has tipping points wherein that tonality *is favored* over actual frequency response.

The latter fact is why two amps on the bench can have identical frequency response but one will sound bright and the other will not!

I’ve found over the years and decades that if the speaker requires the amplifier to employ feedback to get flat frequency response out of the speaker, that the result has no chance whatsoever of sounding like real music! At best it will sound like a very good stereo (this bit is tricky; many people are happy with the status quo of a good stereo, because they think getting it to sound like real music isn’t possible!). Its this difference where I have drawn the line, and why over 42 years we’ve avoided feedback.

**That** is why bother with a zero feedback tube amp. Sorry for the derailment.

@bifwynne I run into that too- we make OTLs as you know, but there are speakers out there that on paper would seem to be too difficult a load, but in practice work out just fine with our amps. So your experience does not surprise me in the least.

Feedback is helping you out here, and my surmise is that most of the speaker impedance curve is actually higher than the dips in the bass (the bass region also has peaks BTW).  That's a pretty simple explanation for why this is working for you. 

Quite often the 4 ohm tap on an output transformer sacrifices some performance- in some cases you can lose an octave of bandwidth off the bottom just by going from the 8 ohm tap to the 4 ohm tap. So it can be a mixed bag in the real world (as opposed to the math world of theory).

So here is my question. What is happening electrically if there is an impedance mismatch? Is the amp producing distortion if there is an impedance deviation from 3000 ohms? If so, does having larger wattage and power supply capabilities ameliorate distortion effects?  

How about negative feedback?  In my case, I understand that the Ref 150 uses 14db of negative feedback and as a result, performs like a voltage source amp. Does that feature help to reduce distortion effects?  

I have not touched on phase angle or sonic coloration issues caused by impedance mismatches between the amp's output taps (nominally, the 4, 8 or 16 ohm taps) and the speaker's actual impedance at a particular frequency.  

As to sonic coloration, I recall that the output impedance of the ARC Ref 150 off the 8 ohm taps is around 1 ohm or less. This results in some sonic coloration, ... but not a lot. In a Stereophile review of the ARC Ref 150, John Atkinson measured the Ref 150's voltage output variation off the 8 ohm taps to be (+) or (-) .8 db when presented with a synthetic speaker load.  

Btw, are similar issues presented with OTL tube amps?

@bifwynne 

All amps always make distortion. If the amp's tubes are loaded below 3000 ohms, distortion will rise. If loaded above that, its power will decrease as will its distortion. 

The feedback allows the amp to behave as a voltage source. George has it wrong; the amp can make double the power into 4 ohms as it does into 8 because voltage sources do that (this is by definition). The deal is, the 4 ohm power is the maximum power for which the amp is rated and you have to be on the 4 ohm tap. If the impedance then rises to 8 ohms, the amp will make 1/2 of its output power. Its a bit confusing, because tube amps accomplish the task of being a voltage source in a way that is opposite of how solid state amps do it.

The issues are the same with an OTL, but of course no transformer is (usually) involved. Instead the tubes see the load directly.

But what about speakers with impedance swings all over the place, from 3 ohms to off the charts well at well over 20 ohms like 802D3s (nominally 8ohms). It’s swings like this that tax the best tube amps with their transformer outputs. How, exactly, do these impedance swings effect autoformers???

Are the electrical properties of autoformers similar enough to those of tube output transformers to cause similar issues when driving speakers with crazy impedance and phase curves?

See the thread below:

https://forum.audiogon.com/discussions/autoformer-vs-speaker-impedance-curve

I believe he meant that with an autoformer an amp can be designed with lax’ed parameters eg: that make it stable. And that an autoformer can then isolate it from the bad outside world speaker emf etc, that may make it go into oscillation or ring or whatever. This this autoformer makes this amp listenable and reliable. It's a Band-aid.

This statement is false. The reason is that in a Mac, the autoformer is included in the feedback loop (this fact is also ignored by @unsound). Further, any kind of output transformer does not provide 'isolation' as suggested above, instead it **transforms** impedance. So in this way, a variation of speaker impedance is transformed to a variation in load impedance to the output transistors (but at a different impedance as defined by the transformer or autoformer) and they respond in kind. Ringing and oscillation is a red herring - ringing is handled by the feedback, and if anything, the amp is less likely to oscillate when a transformer is employed!

For pure tube enthusiasts, the only solution is to find speakers that have flat and high'ish impedance functions (say 16 ohms) over their entire frequency ranges. I do not think there are a lot of beasts like that out there. Ralph, if you can make some suggestions, please do.  

Btw, another knotty subject that Ralph and Al have posted about some years ago is low damping factor with high output impedance tube amps. Ralph, I forgot what you posted. Care to re-educate us?

This is a bit of a digression on this thread. If others are interested in this topic, I recommend a new thread:

Actually tubes can handle speakers with variable speaker loads and do just fine- the Sound Labs ESLs are good example. Like all ESLs, their impedance varies by about 10:1 over their entire range, and tubes can usually manage them better than solid state.
With regards to output impedance, a high output impedance does not imply universal frequency response colorations. It does mean that you will have to be more careful about the match between amp and speaker, but IME this is an issue regardless of the amp and speaker anyway :) 

The elephant in the room is the fact that the ear converts distortion into tonality- and in this regard the ear has tipping points where the tonality of distortion is favored over actual frequency response. IOW, it can be more important to get rid of the colorations caused by distortion than it is to have perfectly flat frequency response; if you look at speaker response curves, the latter does not exist anyway!
The ear is far more sensitive to higher ordered harmonics than it is lower orders, by several orders of magnitude. This is why solid state amps can have the coloration of 'bright' and 'harsh' despite having very low overall THD. By not running feedback, it is possible to reduce these higher ordered harmonics if the circuit employs good design principles (in other words, has good open loop linearity). This is far easier to do with tubes than transistors! There are good solid state designs with zero feedback, but they are rare. 

An alternative to zero feedback is to use an autoformer that allows the output transistors to drive a higher impedance. In this way higher ordered harmonics are suppressed as all amplifiers make less distortion into higher impedances. The downside is that overall, the amp makes less power on account of that higher impedance. But in the world of high end, lower distortion is far more important then overall power.

Why hinder the performance of a well designed solid state amp with an autoformer, unless it was a bad designed one to start with.
A simple test is to put a well known Auto Former that’s used on OLT’s to make them "sort of work" into speakers they can't drive without them, on the rear end of good solid state amp (say a Pass Labs) and watch it transform into rubbish.

Autoformers are "band-aid fixes" for amps that are not right before them to be able to drive into loads they shouldn’t be on without them. Get the right amp to start with, don’t just put a band-aid on.

This statement is false. Most solid state amps will sound better when driving a higher impedance load as they make less distortion. This is easily seen in the specs.

The autoformers restricts the ability to vary power with impedance / sensitivity and ergo compromises frequency linearity with the vast majority of loudspeakers.
There’s a good reason so many other amps don’t use them.

This statement is false as well. The use of an autoformer or full on output transformer does not prevent an amplifier from acting as a voltage source.

Mac built tube amps in the 1950s that behaved as voltage sources (and just so we're clear, a voltage source will put out the same voltage into all impedances the speaker presents, which is what any good solid state amp will do) and many tube amps with output transformers act as proper voltage sources. In fact Mac lead the way in the 1950s and 1960s (along with ElectroVoice) in getting the idea going that a loudspeaker should be driven by a voltage source. To suggest that somehow 60 years later their ideas suddenly no longer work is ludicrous. 

Personally I don't think that having an amplifier that behaves as a voltage source is the most neutral way to go because the factor that is left out here is the function of loop negative feedback, which is used in the vast majority of amplifiers. But it is this design aspect that allows amps with output transformers to behave as a voltage source- add enough feedback and almost any amplifier will! I don't like feedback as it adds distortion of its own, but if you are going to take the position that an amplifier won't act like a voltage source (which is exactly what is posed in both quotes above), then you'll have to deal with the facts which are in contradiction to that position.

Now there are even CAR stereo amps that can run .5 ohms all day at negligible distortion.

The problem with not using transformers, even with solid state amps is that 'negligible distortion'.
It is the mark of a good engineer to know what is negligible and what is not.
The slight amount of distortion made by most solid state amps is not negligible. The reason is that the distortion is composed almost entirely of higher ordered harmonics, and the human ear is tuned to these harmonics in several ways, a sort of convergence.
First, there is Fletcher Munson- the loudness curve. If you take a look at it, you will see that the ear is most sensitive at birdsong frequencies- up to 7KHz. This is why alarms are higher frequency. There are a lot of instrument fundamentals that are a lot lower- in particular, instruments that are near 1KHz will have a 7th that high- but here's the tickler: the 7th is also one of the harmonics that all solid state amps have in common (which we've known since the 1930s imparts a metallic quality to the sound), and the ear is insanely sensitive to this (moreso than good quality test equipment), because it also uses that and other higher orders to sense sound pressure. This is why solid state sounds bright and harsh, its why tubes are still being made and why we argue about tubes and solid state endlessly on the internet.

So this is a problem, but actually an easy solution is to simply present a higher impedance load to the output section of the solid state amp. Right away it will make less of these higher orders and so will sound smoother and more detailed. That's what the autoformer is for!
One other point- its not to anyone's advantage to make **any** amplifier work hard! You can know right away that if you do so, it will have higher distortion. In this regard, if the most realistic audio reproduction is your goal, your amplifier investment dollar will always be better served by a speaker that is higher impedance- 8 ohms or more. There is little point to 4 ohms unless sound pressure is your goal rather than sound quality.