Andre Jute explains why an ultrafi tube amplifier has Zero Negative Feedback
KISS 123
Andre Jute explains why an ultrafi tube amplifier has Zero Negative Feedback
KISS 123
*****
The customer complained that his new suit didn't fit. 'The sleeves are too short,' he said to the tailor.
'Yes sir,' said the tailor, 'but if you hold your arm just so, at an angle as if you're drinking tea with your auntie, it will show just the right amount of cuff.'
The customer tried it. The tailor was right! 'But what about the other sleeve? It is definitely too short.'
'Just lower your shoulder, sir. Yes, yes, a little more. Put your foot out so you can lower your shoulder a little more still. Bend your knee. Yes, that's it. See how beautifully your suitcoat now fits?'
The customer had to admit the tailor was right. 'Wow. But now the leg of the pants is all twisted around.'
'That's easily fixed,' said the tailor. 'Just point your other toe westward, sir, and look over your shoulder to where I am holding up the hand mirror. See? Doesn't that fit beautifully?'
'Yes,' the customer said doubtfully, 'but-'
'Now would sir like to wear his brilliant new suit or shall we wrap it?'
The customer was too intimidated to argue. He walked out into the street in his new suit, his arm crooked as if he were drinking tea, his other shoulder well down over a bent knee with his foot out to the side, his other foot pointing westwards, his head twisted back between hunched shoulders as if complaining to God about a cruel fate.
Behind him he heard a boy say to his father. 'Oh, Daddy, look at the poor twisted cripple.'
'Hush,' the father said. 'Be grateful the poor man found such a brilliant tailor.'
****
The Ultrafidelista view of Negative Feedback
by Andre Jute
Negative feedback is the paradigm of modern electronic design. It is mother's milk to an electronics engineer. He learns to say '100dB of NFB,' in his sleep before he finishes his first week at the most humble polytechnic. At the great institutions the professor of feedback is the most honoured man in the department. In Massachusetts and Minnesota the feedback guru is the most honoured man on the entire campus, equal in stature to the football coach. When a guru of transistor high fidelity (and some in tubes) says, 'I studied under Ron,' one doesn't have to ask which Ron, one just knows it is the holy name of the prophet of feedback from the Midwest.
Before I even finished the design of the KISS 300B it was forcefully suggested by a wannabe guru that with only 50dB more gain (about seven times as much as is likely to be in the actual design) I can apply 50dB of negative feedback to linearize my amplifier.
Negative feedback, shorthanded as NFB, is the instant response of the audio engineering fraternity to all ills, real, perceived, non-existent. They don't even ask if there is a problem, they swing the club of NFB regardless. NFB has become a reflex axiom of mainstream audio design. An audio engineer with his negative feedback is like a policeman who runs out into the street with his stick and starts beating a confession out of the first housewife he sees. The difference is that the policeman is relieved of duty to await punishment and the audio engineer gets away with it. In the case of the policeman it is unacceptable behaviour, in the case of the audio engineer so much the expected norm that no one except the ultrafidelista notice. I guess that if one in ten million audio amplifiers does not have negative feedback added, it will be a lot...
No one asked if my KISS Amp requires linearization. The presumption by all except those already of the ultrafidelista persuasion was that I would welcome suggestions about A Good Thing.
In the face of such overwhelming acceptance by qualified engineers, why do we as ultrafidelista not take the same easy path of negative feedback? Especially considering that superficially NFB is easy to understand and apply.
How does negative feedback work?
Negative feedback is simply a negative voltage fed back from the output to the input amplifying device to offset part of the harmonic distortion which is present as a positive voltage. It costs nothing except a loss of gain and a few side effects such as phase shift and possible instability which are well known in the mathematical literature and more or less easily guarded against depending on the level of NFB.
'Wow!' those meeting NFB for the first time will now say, 'Something for free! I'll grab some of that for my amp.' Hey, I said it, and I am a professional intellectual, by definition an infinite skeptic. NFB is a thing of beauty that will draw you in. It is like an electronic Marxism which admits of no contrary arguments because it has subsumed them all into The Holy Measurements. To question The Measurements is to commit heresy. You need to be of strong mind to resist the blandishments of such a universal panacea and of strong stomach to withstand the hysterical assaults of the lesser engineers defending their holy grail. (And when you do get hold of a superior engineer to explain NFB to you, you need to be high-domed indeed because suddenly NFB can turn very intricate.)
Unfortunately NFB doesn't come without a price. It levies a cruel charge on the perceived quality of the sound. Negative feedback is what gives all those 'blameless' transistor and big PP tube amps their chillingly unnatural sound.
Then how did NFB come to be such a panacea in amplifier design?
Your guess is as good as mine. Hi-fi design is not prestige work for engineers, or highly paid. The most talented and best qualified engineers go into automobiles or military hardware or big construction projects or computer design. The left-overs design amplifiers in the time they have to spare from writing up specs for requesting a CE mark for a new electric kettle. Lemmings storming en masse over a cliff come to mind; such people don't see the necessity of original thought, or have the mental equipment for it. The exceptions to this rule are normally audio enthusiasts in charge of their own small audio manufactories with niche markets; those who grow larger from this base follow the mainstream mantra of "mo' NFB give lowa' THD" because the marketing channels demand it from them if they wish to grow. At this point they usually cease to offer anything different, only the exclusivity of a very high price. (I know, because a sub-board I designed for a supplier to the trade turns up in so many very expensive amps with so many different big names neatly silkscreened on it... it strikes me as the sort of detail a real designer, as distinct from a marketer, would take under his own control.) Those very few makers who will sell you an ultrafi amp without any NFB operate even tinier shops, usually one man and a cat, just hanging on.
The mechanism by which NFB wrecks your sound
Negative feedback at first acquaintance sounds good enough to take to bed and cuddle. It isn't. It isn't even as simple as a superficial acquaintance may suggest. Follow the steps with me, from the theory as she is received to what arrives at your brain as music:
1. In theory NFB reduces all harmonic distortion equally, without discrimination. Strictly in theory it does not reshape harmonic distortion by reducing the most objectionable third and higher order odd harmonic distortion to a greater extent than the relatively harmless 2nd harmonic. Thus NFB at its theoretically most benign is already useless in terms of psychoacoustics, as will become clear at point 4. If you disregard psychoacoustics, as many audio engineers do, NFB is brilliant in reducing total harmonic distortion to a number as tiny as you want. You just pile on more NFB.
2. In real life, as distinct from simplified theory, NFB adds artifacts of its own. Remember, it is a loop. The signal starts at the input and is amplified by devices until it reaches the output. From the output a part of the signal called the negative feedback is fed back to the input. Here a loop is completed and the combination, less distorted, reaches the output again, a part of the combination is fed back, endlessly. The artifacts we want to consider here are created by the fed-back residue of harmonic distortions adding to both the fundamental and the distortions already created by the amplifier, then some portion of the sum of the original and the feedback distortion is fed back again and added on, until the ooh-ah bird flies up its own fundament. It looks marginally less disgusting as a recursive mathematical formula with lots of nested parenthetical parcels of noise being loaded onto your music. But it is a monkey on the back of your sound, with a smaller monkey on the back of the first monkey, a still smaller monkey on the back of the second monkey, and so on ad infinitum. These additive artifacts are all higher harmonics and the more dominant ones are all odd. Suppose, for the sake of simplicity, a superbly designed ultrafidelista amp with some second harmonic and zero odd harmonics before NFB. Add NFB and the second harmonic will be lowered but the recombinant new loop now contains newly added intermodulation effects between the fundamental and the residual second harmonic, and that is third harmonic. In the next cycle a small but nasty dose of fifth harmonic that wasn't there before is added by interaction between the still residual second harmonic and reduced newly added third harmonic. In short, the artifacts NFB adds to the distortion mix are all of the most harmful kind. But, say the proponents of NFB, so what? Every time the loop cycles the added artifacts are smaller, even if there are more of them... The whole affair starts to smell of trying to argue with a Marxist who simply declares any inconvenient truth 'an anomaly'. (If this sounds like a mess from which you should run a mile, you have come to the right conclusion. Start running now. It gets worse.)
3. We thus arrive at a situation where distortion has been lowered by NFB but where the most disturbing odd harmonic distortions are still present to some measure, with the added disadvantage that new and extremely disturbing artifacts of higher harmonic distortions have been created by the very process of using negative feedback to lower distortion. Regardless of the absolute level of THD, or the volume setting, the mix of harmonics has been adversely affected and now includes a higher proportion of third and higher harmonics than before NFB. Let me say that again: after NFB, third and higher harmonics will make up a greater part of the distortion than before.
4. Low volume levels perforce accounts for 99 per cent of audiophile listening because we all have families or neighbours, and we would like to keep our ears. Unfortunately for the lowest common denominator of hi-fi designer, the one who specifies NFB as a conditioned response much like Pavlov's dogs slavered when the bell rang, human physiology and psycho-acoustic response is such that odd harmonics are disproportionately more disturbing at lower than at higher listening levels. This inescapable effect is independent of definition of 'listening level.' At the 110dB in-room SPL (only 14dB louder than an automatic riveter!) advocated by the already deaf Transient Overload Elite known on newsgroups as the Borg, this poisonous concoction of original distortions and NFB recombinant artifacts will be least disturbing (and soon not heard at all!). At any lower level perceived interference of this harmonics cocktail with the music will increase in inverse proportion to the volume level. At low volume levels the artifacts generated by NFB will by their nature as higher harmonic distortions be disproportionately far more disturbing. At these normal listening levels 0.75 per cent of second harmonic distortion may be below the threshhold of perception for sophisticated listeners, whereas tiny amounts of third and higher odd harmonic distortions grate.
And they still use Negative Feedback? Are they stupid?
No, they are not stupid. Most of them march to the drum of a cost accountant on whom we wouldn't spit if he were alight. NFB is as cheap in money terms as it is expensive in terms of perceived quality of music. We shall come to those who claim to be sympathetic to high-fidelity but insist on devices which do not work without NFB, who have another devious answer. Here, meanwhile, for you to keep in mind, is a single-sentence summary of a complicated interdisciplinary argument:
The case against NFB is that for 99 per cent of listening the NFB cure is worse than the disease.
But surely we don't have to do anything so stupid?
It follows from the argument above that ultrafidelista should choose an intrinsically linear topology and device which does not require added negative feedback to 'linearize' the output. The intrinsically linear device is the thermionic tube in either its triode form or as a pentode hogtied to work as a triode, which can be a most pleasing alternative both economically and sonically. The topology is often single-ended operation, chosen also for several other reasons described elsewhere in these articles, including KISS; if the chosen topology is push-pull operation, which is more difficult but far from impossible to arrange without NFB, operation should be specified as Class A1. Inside the argued case above lies too the overwhelming reason to accept the potential small disadvantage that may accompany the preferred topology in comparison to the discarded alternatives. The disadvantage is of course the potential for a residual second harmonic that measures high by transistor or NFB tube standards. (Note the word potential. With a conservatively designed DHT amp the potential problem should not arise.)
The ultrafidelista, who are as keen on silent amps as anyone else, accept this small potential difficulty because it is the lesser evil compared to NFB. Unbelievers (largely unwashed, according to reports) sneer that ultrafidelista like this approach because of the 'added euphonics', which is bow-wow techie talk for the warmth a big chunk of second harmonic lays on a zero negative feedback single-ended amplifier. But competent design can easily reduce the level of second harmonic to below the level of perception without the need for NFB and its deleterious after-effects. In any event, it is your amplifier. You paid for it. You have a right to tune it as you please. The key thing is to get rid of NFB and to understand why you did it.
Can we prove any of this scientifically?
We have already. All of this is the technical subtext to my longtime contention that what the ultrafidelista hear and love is not a directly heated triode sound as is claimed by many enthusiasts but a Class A1, ZNFB sound. (Admittedly, as we have seen above, the right sound is virtually guaranteed with a ZNFB DHT SE amp of conservative provenance but may have to be developed the hard way with more economical or higher-power contenders.) In comparative ABX tests conducted over a number of years, I found that professional musicians, certified golden ears, choose the triode-linked Class A1 PP ZNFB EL34 whenever it is present in the test over all other contenders including SE 300B and 'blameless' high-NFB silicon.
Science also proceeds by pure reason. Ultrafidelista have long doubted whether what engineers insist we measure (the absolute level of distortion, THD) predicts success in audio gear. This is the full circle, because I have just proven by logical, individually uncontested steps that what matters, once a certain modest level of silence is assured to an amplifier, is not the absolute level of disharmonics but their composition. The same proof demonstrates that a more beneficial distribution follows instantly from doing without NFB.
But transistor amps won't work at all without NFB!
That is not our problem. Those who choose inefficient speakers and consequently are forced to accept monstrous amps made possible only by gigadeciBels of NFB, will receive our sympathy - and the music they deserve.
Engineering hangers-on of transistor attempts at high fidelity, where the measure of success is vanishing THD rather than sonic hedonism, pretend to be enthusiasts for NFB. To make it work for them, they have attempted to change the rules so that we won't hear what their treasured NFB does to our sound. They sneer that low level listening, which 99 per cent of us prefer and where NFB does most to wreck the sound, is 'easy listening' and therefore not permissible. According to them we should all be forced to listen at the high volume level which suits NFB amps, which they call 'realistic'. This is a contemptible circular argument, only too characteristic of a fascist mentality in a part of the audiophile spectrum which wants to prescribe their arid vision without regard for our enjoyment.
We can recommend a good tailor to them. It hurts every time you wear his suit. No pain, no gain, fellers!
In summary
Almost everyone listens at low level most of the time. NFB wrecks everybody's sound at all levels but most wretchedly at normal listening levels. We started out with a contemptible circular argument and we have met another along the way. We can now put both in context:
An 'engineer' who designs an amplifier which does not work perfectly without negative feedback is like a tailor cutting the suit incompetently and then demanding that you walk like a cripple to make it fit, so that everyone can admire the brilliance of your tailor.
Negative feedback is a bodge. That is why it is despicable to the ultrafidelista.
Reality shock
Everything above relates to added feedback. The KISS amp has no added global or loop feedback from the output to the input. The KISS amp has no added local feedback at any individual stage. That is the generally accepted definition of a zero negative feedback amplifier. But that does not mean that there is no negative feedback whatsoever; an amp with absolutely no negative feedback whatsoever would be very nasty indeed. What ZNFB means is that The KISS Amp has its just, apt, natural share of negative feedback. You can't see it but it is there: invisible, built into every triode, God's own equilibrium, what I like to think of as "the good feedback".
KISS 104: ZFNB: Religious imperative or manner of speaking? discusses the matter further for those with a technical interest.
THE VOLTAGES IN THIS AMP WILL KILL YOU.
GET EXPERIENCED SUPERVISION IF IT IS YOUR FIRST TUBE AMP
All text and illustration is Copyright © Andre Jute 2001, 2004, 2006
and may not be reproduced except in the thread KISS xxx on rec.audio.tubes