FETs, and a question for PRR (and others)

[AP]

A very well known and respected audio designer says:

We have been using JFETs in the input stages of all our audio amplifiers for the last 15 years. However, the second stage always used bipolar transistor. With many years of listening and testing we have found that these bipolars are the limiting factors in terms of the finest resolution. We have therefore replaced all the bipolars with JFETs and MOSFETs in our phono preamp, lineamp and power amps.

So, what is happening here? Unless you do all of your own recording, the delivery chain for music goes through (probably) thousands of bipolar devices on the way to your ears. Is he suggesting that the (you would have supposed, if you buy his contention) irreparable damage already caused to the sound, is suddenly fixed by transiting a pcb that *only* has fets on it? Do we smell snake-oil?

What does the panel think?

Alan

Actually, this is not at all a 'tongue-in-cheek' question - I am truly puzzled by this.

Even the simple (and less contentious) suggestion that 'Fets are better than bipolars' for audio.... If this really were the case, would there not be a substantial degree of agreement, with documented evidence, such that it would be taught in schools (and BJTs would be totally disenfranchised).

At the moment, I am in the Doug Self camp on this one... But I very ready and willing to be convinced...

Over to you'all

(PS. I know about even -vs- odd harmonic distortion - I don't really think this explains everything though, as distortion can be made disappearingly small with either technology...)

 

[Gus]

IMO there is nothing wrong with BJTs used right. FETS to me are a pain you need to test them first to see if they will work. A good BJT design will allow the use of BJTs without matching or trimming most of the time.

I only use fets when I have to.

 

[tk@halmi]

Alan,

The decision to use one component over another or use one design over another often has nothing to do with detectable sound quality. Most of the time, design decisions and component choices are comunicated as leaps of improvement, but conceal a very simple marketing concept, The Differentiator. When a market is flooded with solutions like the Audiophile market the manufacturing or design firm needs to pick one or several aspects that differentiate their products from others. They proceed to put it on their advertisements to emphesize that they are doing things better than the competition. Most consumers are trained to take these differentiators as a matter of fact of improvement and will buy their products and fight to the teeth to defend their investment (regardless of the crap they have bought into.)

Also:
People keep making better mouse traps because they have control over them. They can't control the quality of the recordings they buy. There is a good reason for this. Should Audiophiles start recording music the price of doing business would rocket to the sky as gold wired consoles became absolutely necessary to record anything worth to listen. Artistic creativity would be swept aside and the focus from MUSIC would shift to the quest for the "best" recording of the banjo.

Finally:
Technical masturbation can be very satisfying. It is the act of deriving joy from examining technical details over and over without ever making quantitive conclusions to move on to something else. Sometime it is called analysis paralysis as well. Some people are addicted to it and are in a state of perpetual denial. There is little help, and one can hope that they snap out of it eventually.

Cheers,
Tamas

 

[FredForssell]

Should Audiophiles start recording music the price of doing business would rocket to the sky as gold wired consoles became absolutely necessary to record anything worth to listen. Artistic creativity would be swept aside and the focus from MUSIC would shift to the quest for the "best" recording of the banjo.

Uh Oh. I guess I better stop doing what I'm doing, since I consider myself an audiophile (and have for 30+ years), and I am trying to get the "best" recording of banjos.

However, one thing I have learned about recording banjo (which also applies to most other instruments as well) is that if you record really good players, playing really good instruments, in good sounding rooms, then all you have to do as an engineer is stay out of their way sonically (which take great equipment, IMO). I just recorded a great banjo and fiddle player named Ron Stewart and it wasn't difficult at all to get a great banjo sound. Same thing with Todd Philips on bass, Mike Marshall or Chris Thile on mandolin, etc. Just stay out of their way sonically and you are there.

And I'd rather listen to the music that I record on 90% of my audiophile friend's systems than on 99% of the recording studio monitor systems I've heard. FWIW.

 

[tk@halmi]

Uh Oh. I guess I better stop doing what I'm doing, since I consider myself an audiophile (and have for 30+ years), and I am trying to get the "best" recording of banjos.

Perhaps you count yourself as one, but you can be hardly put in the same bucket with the crowd that reveres the unidirectional RCA cable. :thumb:

 

[AP]

Tamas

... but conceal a very simple marketing concept, The Differentiator.

Hey - you cynic!
I must admit, it is my tendency to be the same, but when well-respected people make bold statements, I feel obliged to take note and try to understand whether there is really something to it. Maybe their view is a marketing stance, maybe just unfounded opinion, maybe it is just preference, but maybe they really do have something.

Uh Oh. I guess I better stop doing what I'm doing, since I consider myself an audiophile...

...well, I feel there are currently two different kinds of audiophile. Certainly it is a noble and worthwhile aim to deliver one's listening pleasure in the best possible manner. The other kind is who Doug Self refers to when he sugests they seem to believe that:

- Objective measurements of an amplifier's performance are unimportant compared with the subjective impressions received in informal listening tests. Should the two contradict the objective results may be dismissed out of hand.

- Degradation effects exist in amplifiers that are unknown to engineering science, and are not revealed by the usual measurements.

- Considerable latitude may be used in suggesting hypothetical mechanisms of audio impairment, such as mysterious capacitor shortcomings and subtle cable defects, without reference to the plausibility of the concept, or gathering any evidence to support it .

Fred, I don't believe you would choose to line up with the latter!

More on the point though, from a few of the things I have seen, I suspect you to be a FET fancier... If so, is there an objective case for FET superiority in audio applications? Mr Borbely seems to believe there is a case (his quote was in my first posting) - although he doesn't at all set out any objective evidence, and the subjective view he presents is set uot in a rather suspicous manner...

I take it as an article of faith (for most designers) that the *general* case is to minimise distortion and other impairments in the audio chain. (the less general case being when you are actually trying to either add something to, or take something away, from the sounds).

So I am intrigued, as it seems that it seems generally easier to get low distortion figures with BJTs than FETs, and as suggested earlier, they do have some other practical disadvantages in respect of needing selection more frequently than BJTs. I'm certainly not anti-FET (although I haven't used them much, and don't consider myself a FET expert), and I do have some sweet little configurations that have mixtures of BJTs and FETs.

...So, is there something inherent in FETs that makes them superior, even though even the best FET circuits tend to be eclipsed by BJT circuits in terms of measurable parameters? (Sweeping, probably unsafe, generalisations here...)

Alan

 

[Gus]

People

Look closer at that distortion myth. Go to Aron's stompbox page. the beginner project NPN BOOST is a circuit of mine. Its an easy build almost any NPN Si will drop in. The die size and Hfe will affect the tone.

If you are a guitar player build it you might be suprised at the tone clean and gain at max. works the best guitar to NPN boost to tube amp

With Si transistor the Hfe changes with Ic. With some transistors more than others so you openloop gain is doing something different at different Ic. So where you place the transistor(s) operating point and the curve of Hfe vs Ic you are changing harmonics. I believe this is why some people get "hung up" with certain transistor numbers. I tend to care more about the die size and mask type.

A great book is the GE (the company) 1964 transistor manual. Its one of the best books on transistors IMO.

 

[AP]

"objective" usually deals with math proofs.

Well, I would say that 'objective' here, refers to something that is

a) Repeatable. Can be reproduced at will.

b) Apparent. Some kind of physical evidence that any reasonable person skilled in the subject (but not possessed of uncanny powers) would accept.

Can you think of any?

Alan

 

[PRR]

> With many years of listening and testing we have found that these bipolars are the limiting factors in terms of the finest resolution.

Who is talking here? The Designer or the Marketer? (In the small world of snazzy-audio, it may be the same person in different hats.)

I doubt anybody can listen to music and say if the second stage is BJT or FET. Especially, as you say, with hundreds of assorted devices between the studio and the living room.

I also strongly suspect that if you put an emitter resistor under a BJT, equal to 1/Gm of the FET, the difference would vanish. (Unless that stage is starved for current and noticably non-linear... the answer for that is clear.)

I like Tamas' idea marketing concept, The Differentiator. I can't even fault a designer/marketer for touting some minor detail as a great discovery: the market is very broad and VERY stupid, you need to have a Differentiator to eat.

Get this designer alone in a room, where none of his customers will hear. Apply enough flattery or beer to make the Marketing hat vanish. Leave your firstborn son as hostage that you will not spill secrets. Then maybe you can find out what he really thinks.

> The other kind is who Doug Self refers to...

I like Doug Self's attitude, but I disagree when he sneers at the assertion that "Degradation effects exist in amplifiers that are unknown to engineering science, and are not revealed by the usual measurements". "Engineering science" does NOT know everything, or even a lot about, the complex things happening in real music through electronic amplifiers, electromechanical transducers, and the human hearing process. We aren't even close to having a standard IM test that looks like a musical signal, with output weighted anything like human hearing. There does not even seem to be any interest in it. The main IM signals are the old narrow-band 60/6KC test and these new ones that tease 18KHz and 19KHz through digital sampling systems. These are mechanically informative, but NOT music. Test signals are mostly steady-state, music is not.

 

[AP]

...but I disagree when he sneers at the assertion that "Degradation effects exist in amplifiers that are unknown to engineering science, and are not revealed by the usual measurements."

Well, I'm not certain he is sneering as such, and I am sure he is aware of the difficulties involved. I think he is pointing to use of this position to justify any old idea you might have, no matter how unlikely and crackpot, without feeling the need for even an attempt at substantiating it (his third point, actually).

A particular point he makes elsewhere is that in taking such a stance tends to cause you discount the value of objective measurement of any kind (ie. only the ears matter). In a way this is a kind of technical anarchy, where anything goes. The result of which is the emergence of some truly appalling products (products that are appalling from a technical standpoint), foisted upon the unsuspecting public at unbelievable prices in the guise of some kind of mystical 'audiophile' snobbery.

In his rant on the subject, he also says:

In the twenty or so years that have elapsed since the emergence of the Subjectivist Tendency, no hitherto unsuspected parameters of audio quality have emerged.

... I don't think we can deny him that one!

.......... anyway. Are FETs better in any sense? Even an objective *theory* (if there is such a thing) would do (or indeed, any theory that sounds like it might be true).

Or should I just be content with something to do with the makeup of harmonic distortion products? The latter is the best I've heard so far (although hardly new). The problem I have with it, is that since it was first touted, we have reduced distortion levels well below the threshold of hearing (even when the original signal has been removed), and then further orders of magnitude below that!

Is there really any audio differences between BJT and FET technologies, when competently used in a best in class design?

Alan

 

[PRR]

> taking such a stance tends to cause you discount the value of objective measurement of any kind

Seems to me it should provoke the study of better measurements.

But clearly it hasn't.

I don't think we've even caught up with some of the work in Radiotron 4th.

> reduced distortion levels well below the threshold of hearing

Disagree....

> something to do with the makeup of harmonic distortion products?

This is important. The ear distorts too. Large low-order distortions are masked: nothing much wrong with 5% 2nd Harmonic, or even 3% 3rd harmonic. But the ear's distortion mechanism falls very fast. Quite small amounts of high-order distortions are audible and "unnatural". Radiotron 4th proposes a weighting scheme that seems harsh (square of order) but does seem to be in the ballpark of ear-tests and decades of experience on different amplifiers.

> any audio differences between BJT and FET

Or tubes. They are all junction-effect devices. Tube cathodes are liquid-vacuum junctions but at that interface obey laws much like "solid state". FETs put the junction on the side, which increases input impedance while reducing transconductance. The BJT (for real-world currents) seems to be the best we can do for high Gm. If we add a padding resistor to reduce Gm to the level of an FET or tube, we get similar but lower distortion (lower because the resistor is more linear than junction-effects).

 

[AP]

Quite small amounts of high-order distortions are audible and "unnatural".

So your view is that we should still be very much targetting lower distortion figures (and perhaps better ways of measuring them), with an emphasis on reducing higher-order distortion products, if possible?

If we add a padding resistor (to a BJT) to reduce Gm to the level of an FET or tube, we get similar but lower distortion

You don't sound convinced that there is anything inherently 'better' about FETs or 'toobes' then? Except perhaps where you are deliberately trying to introduce colouration, in which case you will probably get more of what you want from a FET? (Or indeed, you might be trying to work up a plausible marketing story...)

In short, we (I) should resist the temptation to describe or characterise the 'sound' of FETs etc., as one would a ripe Camembert, or a fine wine (young, fruity, a hint of tobacco and chocolate, fat legs...etc.), and just get on with using the tools at my disposal to reduce the observable impairments... (and leave the marketing people to worry about the rest...)?

Thats a relief! (But I'm not certain the whole world sees it that way!)

Alan

 

[PRR]

> your view is that we should still be very much targetting lower distortion figures

No. We should be targetting buzz-words. It makes the user feel better, which is more important than actual sound. It helps put food in the mouths of designers and other audio professionals. If there are no new buzz-words, then sales are limited to new enthusuasts and replacement of failed gear. These markets are much smaller than churning the disposable income of the existing enthusiast base.

But it can't hurt to get the harmonic distortion to fall far below noise level a couple octaves away from the fundamental.

And while I have not really looked into it, that seems to mean either NO-feedback, or more than 40dB of feedback. Amounts like 20dB of feedback inject harmonics to be re-distorted and cause high-order hash. 40dB starts to fix it, but implies many MHz open-loop bandwidth.

In the no-feedback case, sure we can talk about "the 'sound' of... a ripe Camembert... (young, fruity, a hint of tobacco and chocolate, fat legs...etc.)". Tubes give a round mound of sound, and there are several vintages; BJTs become nearly unlistenable above a few dozen millivolts.

In the high-feedback case, we should only have to consider how feedback is failing, or the "sound" of the feedback resistors (all common resistors have voltage-sensitivity; in kilovolt amplifiers it becomes a large THD number) or those parts outside the feedback loop.

> You don't sound convinced that there is anything inherently 'better' about FETs or 'toobes' then?

No. The electrons don't know the difference.

> Except perhaps where you are deliberately trying to introduce colouration

That is "effects", not amplification and reproduction. This does not seem to be the thrust of your original question. IMHO, "effects" will slide off into the Computer: maintaining a Sexium 10GHz or Septium 100GHz will be cheaper than re-bottling old LA2a and BC3 boxes, and the sound will be JUST "the sound", not the stray buzz and lumpy response if that is not part of "the sound".

 

[AP]

No. We should be targetting buzz-words

Ha (LOL)...

And while I have not really looked into it, that seems to mean either NO-feedback, or more than 40dB of feedback.

It would seem to be a very fruitful field for research... Whoever sorts this one out (after so many years) will have.. um, errr... a good sized marketing department in no time flat....

In the high-feedback case, we should only have to consider how feedback is failing, or the "sound" of the feedback resistors (all common resistors have voltage-sensitivity; in kilovolt amplifiers it becomes a large THD number)

Was I dreaming it, or did you say only within the last day or so, that *you* couldn't hear resistors....?

No. The electrons don't know the difference..

Thank goodness - I'd been getting just a bit paranoid about the matter...

That is "effects", not amplification and reproduction. This does not seem to be the thrust of your original question.

You are exactly correct - I was just underling the contrast...

IMHO, "effects" will slide off into the Computer: maintaining a Sexium 10GHz or Septium 100GHz will be cheaper than re-bottling old LA2a and BC3 boxes, and the sound will be JUST "the sound", not the stray buzz and lumpy response if that is not part of "the sound".

Thats is exactly my own view. There will be Acoustic instruments, Microphones, Preamps, Midi Keyboards, ADCs and DACs, Power amps and Loudspeakers, and ...PCs (and not too much else).

Ive heard that UAD is nailing many of the much-loved 'vintage' sounds, very successfully (I don't have one at the moment, but I'll get one sometime). Plus there is the valiant attempt by the 'Liquid Channel'. And Convolution Reverbs seem to have most people convinced.

Its a brave new world out there (but I still like my 2N4403s).

Alan

 

[NewYorkDave]

Regarding the question of how much feedback is optimal: Peter Baxandall did an often-cited study of that matter in the '70s. Sadly, I don't have a copy of his original article, but his findings are summarized here in a page scanned from a book by John Linsley-Hood:
150kB GIF

Going by the crude graph, it seems that PRR's "40dB or more" is right on the money. But real-world tube circuits seldom operate with that much feedback; solid-state ones routinely do.

My own experiments with breadboarded tube circuits and a wave analyzer indicate that 20dB or so can be enough to put higher harmonics way down in the noise floor. Then again, triodes are inherently pretty linear devices, relatively speaking.

 

[tk@halmi]

Thats is exactly my own view. There will be Acoustic instruments, Microphones, Preamps, Midi Keyboards, ADCs and DACs, Power amps and Loudspeakers, and ...PCs (and not too much else).

What is the news with headroom and digital gear? We seem to be still better off with a compressor/limiter after the preamp before going into the converters. Especially when recording violent instruments like drums and electric banjo... Three years ago we got sick of tracking into a PC (endless Windoze problems) and started using a dedicated multitracking recorder most of the time.

 

[AP]

What is the news with headroom and digital gear?

Essentially, if you are recording at 16 bits, you are going to be right up against it. It will need to minimse your unused headroom as much as possible, to maximise s/n, and minimise quantisation effects.

However, if you record at 24 bits, you can afford to be fairly laid back. Currently, we don't seem to get the advantage of all 24 bits, but we get enough to have headroom to play with.

On the other hand, once you have captured it, anything you did before you recorded it, (typically) can't be undone. So, do as little as possible (or if you must put the signal through some pre-processing, record a dry track too - just in case...)

IMO.

Alan

 

[AP]

the best audiophile systems are maximized for listening to be an experience, not a requirement.

And that, really, is the nub of it. The aims of 'audiophile' setups and studio setups are different. The audiophile wants his equipment to deliver music that 'sounds' great, no matter what that means. "The ears are the only judge", and that is perfectly valid for that situation. The Amp may have 5% 2nd harmonic distortion, but if it lights your fire, then there is nothing at all wrong with it.

The studio monitoring environment is very different. You need to be able to hear everything very clearly - clinically in fact. You need to be certain that the tonal balance you get represents the actual tonal balance of the mix - so your eventual customer doesn't get a bass-heavy or bass-light mix, and the that the top isn't too hard or dull, etc.. If you like the sound too, that is great, but it really isn't too important (apart from fatigue issues etc.)

But don't mix up the situations there. If you do a critical mix on a highly coloured 'audiophile' system - how on earth do you know how everyone else will hear it. That lovely silvery mid/top (or whatever) you are hearing - is that a product of the Amp or speakers, or is it actually in the mix?

Ok, that is perhaps a bit extreme, there are certainly some very fine pieces of kit in the audiophile and domestic audio markets, and some of these are eminently suitable for studio use. However, I really do think that some aspects of audiophile development (measurements are unimportant - only the ears matter) have gone entirely contrary to the needs of the studio, and the original aims of 'HiFi' - High Fidelity - the naked truth - nothing added nor taken away - the 'straight wire with gain'.

Anything with distortion figures much more than can be very readily acheived by the current state of the art, have absolutely no case for being called HiFi (nor place in the studio, except as an audio 'effect), because High Fidelity it very plainly, and proveably, isnt.

Of course, it is entirely possible to have a piece of kit that has good distortion figures, but simply doesn't sound great. That does point to our current shortcomings in terms of ability to measure all of the qualities that the ears perceive. (As PRR said earlier - it should lead to better testing methods - but hasn't). But that is no reason to discard testing! It is actually a reason to work harder at testing!

And itt is certainly a reason to do comprehensive listening tests on any new design, but be careful what you are listening for. In respect of Hi Fidelity, you are not listening for equipment that sounds 'good' (in respect of adding something to the sound), you are listing for some that doesn't sound 'bad' (neither adds nor subtracts).

IMHO
Alan

 

[AP]

I don't think we've even caught up with some of the work in Radiotron 4th.

PRR, I've come across that concept before (weighting the harmonic products), but I cannot recall where.

Do you have any references to that, and indeed any similar or related subjects?

Thanks
Alan

 

[AP]

Separately, I am a bit disappointed that no-one has taken up the cudgel with respect to FETs particularly.

I really was hoping to hear a robust (basically technical) case put forward for FETs being particularly suitable for audio applications.

In particular, I had hoped that I had tempted Fred Forssell to 'have a go', as he clearly has some real depth of experience in designing with FETs. (Words to self - maybe I'll try again...)

Fred - is there any fundamental advantage in using FETS for audio work? I have a sneaking suspicion that there might be, but I really would appreciate a view from someone who has actually spent time and brain-cells on the issue...

TIA
Alan