The Audio Critic Seminar on the State of the Art: Part II (Vol.2, No.2: 1979)

This is the second and concluding installment of the edited transcript of our 8-man, 15-hour techno-talk-athon, already a minor legend in its own time. Of course, to get the most out of it, you ought to read Part I first.

The background and circumstances of our seminar, held early in 1979, were explained in detail in the preamble to Part I in the last issue (Vol. 2, No. 1). Mini-biographies of the eight participants were also given, so that none of this information needs to be repeated here, especially since we want all readers of Part II to have some previous familiarity with the contents of Part I.

Comment is due, however, on the absence of the eagerly awaited discussion of loudspeaker theory and design, which we decided after pro longed soul-searching to omit from the published transcript. Quite frankly, this final section of the seminar failed to come up to the level set by the preceding dialogues. A number of interesting ideas were presented, which we plan to draw upon editorially in future issues, but by and large the discussion was unexpectedly diffuse and in conclusive, with everyone bogged down in his own definitions, goals and values, and with none of the synergistic convergence toward a shared element of understanding that had characterized the seminar up to that point. Probably we were all tired, as it was late in the evening and midnight by the time we stopped.

At the current publishing costs per page, it doesn't seem to make much sense to print this material, especially on top of the large amount of solid information about loudspeakers in every is sue of The Audio Critic. Even so, we feel that the signal path from stylus tip to speaker terminals was explored in a uniquely illuminating way by the seminar, for which the less productive time spent on speakers was a very small price to pay.

As in Part I, asterisks (* * *) in the transcript indicate omitted sections, which are very brief and unimportant except for the loudspeaker discussion.

The opening of Part II picks up the continuity exactly where we left it off, without any omission.

EDITOR: Let's try to put together a recipe for state-of-the-art phono reproduction.

Let's see what we can agree on. We begin.

Do we all agree on a line-contact type stylus? Is there anyone here who would oppose in the light of these previous considerations a line-contact type of stylus?

COTTER: I would add I like it and I want to use it, but it marries us to a pair of constraints. If you go to a line-contact stylus, you have to realize that the cutting stylus is almost always precisely perpendicular; that if it deviates, and those of us who have experience with cutting know that if you deviate a little bit one way or the other, the stylus either digs in or it plows up and you get a noisy groove. So you're sort of naturally constrained, like when you cut with a lathe, to have that stylus pretty damn perpendicular.

HEGEMAN: You don't want a gray thread coming off.

COTTER: You don't want a gray thread. As a matter of fact, that means that the cutting face, the cutting facet and the cutting line of a cutting stylus--even though there is this vertical angle effect which moves the stylus in effect back and forth, that is down the groove as it moves-your aperture is lined up. If you have a line contact stylus, you're going to have an aperture azimuthal alignment problem not unlike the gap in a tape head, in which the need to maintain that exact alignment is going to be a function of the ratio of the length down the track, that is the gap length, to the gap width in the track width sense. If that line contact is of the order of an aspect ratio of say 10 to 1, then one sees a rather shockingly small angle as tolerable before there's a significant change in the aperture. For the same reason you have to do azimuth alignment on a tape.

EDITOR: A gentleman from the University of California, I believe, pointed this out in a letter to us just recently. He pointed out that even though we're very strong for correct VTA alignment, as he pointed out correct VTA alignment may be incompatible with maintaining the perpendicularity of the stylus itself.

COTTER: If the pickup you happen to have doesn't have its effective vertical angle in exactly the correct position for that particular record, when the stylus is aligned in the vertical direction, then to get the correct reproduction you're going to have to do something else other than move the pickup up and down, unless you will tolerate some significant loss in the effective high-frequency aperture. And that's a truth, a geometric truth; it has nothing to do with size.

So you're on the horns of a real dilemma, and one is forced to take another approach. | think there are some approaches, and we're pursuing that. But the fact is that the need for a line-contact stylus has been largely ignored, and it was invented for all the wrong reasons. It was invented for CD-4.

No one ever looked at the principle of playing in effect the wide-track recording with a narrow-track head.

EDITOR: From the practical point of view then, since we have to live with this trade off between correct VTA and correct aperture, how should we proceed? Should we still get a line-contact stylus?

COTTER: Yes.

EDITOR: And should we still try to align for VTA by ear?

COTTER: That depends on which records. I think, because there are records that are available whose vertical angles are extremely different. In that case . . .

EDITOR: All right, on modern records that are say in the 15 to 18 degree range, you would still do what we here have been doing for about a year now?

COTTER: I would say that's the best strategy at the moment.

RAPPAPORT: But it's certainly not the optimal strategy, though, because you can really deceive yourself. Because if you have a record that's recorded poorly, for in stance, and it sounds somewhat muffled, it's very, very tempting to lift the back of the tone arm just a little bit and maybe aggravate the front end of your phono preamp if you're speeding up the signals getting in there and just add a little bit of . . . You have to be very careful, because if you're not careful you can really deceive yourself.

COTTER: What we found was, though, that the correct angle produces a fairly less sensitive null in these effects than you would have otherwise considered had you not had a preamp that was not vulnerable to these very small mistracking effects in the sense that Matti has defined it, as the first, second, third, nth rates following . . .

ZAYDE: It sounds like an avalanche break down, in lieu of a better expression. You know when you're there, but if you exceed it--it isn't something that goes out the window.

RAPPAPORT: Right, well as you said, that's a function of the quality of the preamp.

COTTER: Not even quality. I think, you see, we're facing an interesting problem. We're now defining a set of requirements, a set of constraints for that preamplifier that had not been considered part of the design problem before.

RAPPAPORT: When I say quality, that's quality in the sense as it relates to the job that it has to perform.

COTTER: But fairly to the other designers that exist in the world, they pursued an objective that didn't contain these definitions. So their designs are in most cases, let's say today, probably optimal for the set of constraints. The problem is that those approaches are incorrect. We're talking now about a really substantial change in direction, in which certain topological and certain system requirements now appear as a requirement that didn't appear before. So we're saying-I don't like the use of the word quality, I think it's a question of attitude or, you used the word before, philosophy. We're talking now about a change in the philosophy of the requirement, and therefore designs could change and should change.

EDITOR: Let's follow through now. We have a line-contact stylus; obviously we want a stylus cantilever that's fairly lossy but stiff at the same time, which is a problem. But this can be accomplished, I sup pose. Now what about the generator mechanism?

COTTER: What about the VTA?

EDITOR: We have discussed the VTA.

COTTER: What about the vertical tracking force?

EDITOR: As large as possible, in the light of what we've said?

COTTER: I don't think one should run home and put a flatiron on the pickup, but I think

HEGEMAN: How about playing it with an old Western Electric 9D, for instance?

COTTER: It's interesting. As a matter of fact, talk about that for a minute. Those were rather astonishing, weren't they? They sounded awfully good; the grooves weren't that much bigger, and talk about the forces! That's an interesting . . . You used those.

HEGEMAN: We used to call it the 5-pound monster. It was on an 18-inch arm, so forth and soon...

COTTER: Vertical pickup.

HEGEMAN: It was used for either vertical or lateral on that early stuff. The interesting thing was that the stylus was vertical. One of the requirements they had to have for use in the studios and so forth was this back wards and forwards cueing for the early disc jockeys, where they had to be able to cue up to something. For that reason you had to rock the record back and forth. Those things were pretty heavy. You had a warped record on it, you had a good chance of getting the stylus locking into the groove, or just taking the groove out of there completely. But it was a very-to me, at least-very fantastic sound.

EDITOR: Was that a moving coil?

HEGEMAN: Oh yes. Vertical-lateral. It had the coils at the angles so you could series-parallel them, buck them either way, so you'd get a vertical or a lateral presentation and the other component cancelled out. But the thing that was interesting to me was that they did all their equalization right out of the cartridge. They had all kinds of-there were probably 8 or 10 positions on that equalizing switch to allow for playing 78's, 33's, lateral, vertical, all this kind of garbage.

'", . . nature made it simpler to get us the orthogonality we want with the moving coil, and to get more power from a given mechanical impedance."

EDITOR: They were able to throw away all that signal and still have some left?

HEGEMAN: It was a big cartridge. I wish we could work with those output levels now, Ill tell you. That was their full equalization;

all they needed was basically flat gain after that. So that a lot of the electronic problems . . .

COTTER: Microphone amplifier was used.

HEGEMAN: Yeah, they used a microphone mixer, one position . . .

RAPPAPORT: Very interesting, because I hadn't heard of that, and 1 have just de signed a preamp, in fact I've applied for a patent on a technique to equalize right at the cartridge output.

HEGEMAN: Well, I'm sorry fella, some body got there first, a long time ago.

RAPPAPORT: So much for my good idea.

HEGEMAN: But the interesting thing, of course, is that by doing the equalization at that point, you came out with essentially a flat signal to amplify, and you did not build in some of these crazy effects that we've been talking about.

COTTER: You escaped completely all of these problems, and you still had the scanning problems, but by virtue of the enormous mass and the kinds of geometry relationships, in effect, those pickups had much fewer of the problems we're describing, of ultrasonic behavior. What's interesting is, those damn things lasted pretty long, playing records.

FUTTERMAN: | remember a cartridge, I think it was made in England . . .

COTTER: At 30 grams, right? The 6A was 30 grams, I think there was . . .

HEGEMAN: The 9D, I think, was running somewhere around 45 grams weight, measured almost in ounces instead of grams.

COTTER: 3-mil stylus, but scaled down, this is: 3

EDITOR: Is there anyone here who feels that stationary-coil cartridges have advantages over moving-coil cartridges? Moving mag net, moving iron, variable reluctance types?

OTALA: One comment, of course. With the present embodiments of these principles, I don't think there is anybody who would prefer moving magnet. But the present embodiments are not necessarily the only possible ones.

COTTER: Except that I'd say if you were looking for principles where nature gave you grace, it would certainly be in the moving coil direction, because you've got that orthogonality inherent in the generator sys tem and you'd have to make an awfully . . .

HEGEMAN: I would say that they had one advantage, Pete. They're a damn sight higher output.

COTTER: Ha, I want to attack that.

EDITOR: We're going to come to an interesting point.

COTTER: Let me start by saying something that's interesting. Every pickup, whether it's moving field, moving magnet, moving iron, moving coil, moving zabbas, capacitance or . . .

EDITOR: How do you spell that?

COTTER: Oh, that's an Altarian technique for playing phonograph records. But they all have exactly the same noise, because the noise that they have is the thermodynamic noise. It's the basic Johnson noise. Pickups do not differ in their noise power. If they did, we could connect the better one to the poorer one at room temperature and neatly violate the second law of thermodynamics, by transferring energy from one body at the same temperature to another body at the same temperature. The fact is they all have exactly the same noise power. And it's noise power which is the thing that determines signal-to-noise ratio ultimately. The way in which they differ is in their ability to abstract energy from the movement. Now you're talking about a sort of electro magnetic generator efficacy. I submit that if you look at the existing designs and you also look at the fundamentals, it would appear very much easier to get more power from a given mechanical impedance at the stylus with a moving coil principle-moving coil, stationary field-than it would with a moving field approach. In fact, when you look at the results, present-day moving coils differ by anywhere from 10 dB in the poorest cases to as much as 35 or 38 dB greater power than the moving field versions, which are disguised because of the difference in impedance levels. It isn't voltage, it's voltage times current. All the things we've talked about before, including the newer under standing of what signal-to-noise ratio comes from in a pickup, suggest that it would be a worthwhile pursuit to get a better signal-to noise ratio inherent in the generator system of the pickup so as to be able to realize some of these improvements.

HEGEMAN: I didn't say impedance matching; I just said you get higher voltage output out of magnetics.

OTALA: Take a typical example, though, of unused possibilities. Nobody has, as far as | know, tried to use the transductor type of moving magnetic . . .

COTTER: Parametrics.

OTALA: Well, you could easily do that, and that would give you energy as much . . .

COTTER: It's just an alternative amplifier.

OTALA: That's an alternative amplifier, but probably with all the advantages of a moving coil.

COTTER: Yes. If you made a pickup amplifier that was in effect a parametric system, you would have a lower than room tempera ture noise system if you chose to do it in some particular way. There are certain kinds of conditions and restraints there with respect to the one-wayness of the amplifier.

EDITOR: That still wouldn't eliminate the Rabinow-Codier effect, would it?

COTTER: No, what Matti is saying is that one could build a lower noise preamplifier, or make a pickup perhaps that even contained the amplifier mechanism if you used a parametric approach. What we're both saying is that that wouldn't alter the basic transducer relationship, in which getting a lot of power still has corresponding ad vantages. So actually these principles apply even to the improved technique. I think that's the important idea. So what we're saying is that if you were starting with the easiest approach, it would seem as though nature made it simpler to get us the orthogonality we want with the moving coil, and to get more power from a given mechanical impedance.

EDITOR: Basically, then, are we all agreed that line-contact stylus, well-damped but stiff cantilever, and moving-coil generator are the way to go today?

OTALA: Who knows? Who knows?

HEGEMAN: By popular demand, I would agree to . . .

EDITOR: My readers want definite answers, none of this highfalutin academic theorizing.

WILCOX: Don't let the facts get in your way.

OTALA: Well, that is the present choice, and that is what you asked about the state of the art today. But we didn't answer the state of the art in the future.

RAPPAPORT: That's right. There's no point in discussing the state of the art today; it's a question of how we're going to get to the state of the art tomorrow.

HEGEMAN: I'm the state of the art tomorrow! You don't know me!

RAPPAPORT: I just hope it's different from today, that's all.

EDITOR: I suggest we track through the possibilities through the whole phono sys tem. Let's get the signal up to line level and then let's go back. We still have to cover tone arms and turntables, which are the passive components in this, but since we're discussing the electrical signal itself let's follow it through then. We're now at the crucial juncture of choosing a transformer, a head amp, right? Aren't we?

FUTTERMAN: Pre-preamp.

COTTER: Now we have the problem: how do we get the thing amplified? In what way is the problem different than in the flat amplifier?

EDITOR: Obviously if there's enough power coming out of a moving-coil cartridge, what you need is simply an impedance trans formation. The question is how do you accomplish that in the best possible way? I know that there are some differing opinions around this table on that very subject. So let's hear them.

COTTER: I have a bias. It says the simple thing to do is just transform the energy through a transformer.

OTALA: I have the same opinion, not be cause that would be an elegant solution necessarily or that it would be easy to make a transformer of that kind, especially for that amount of octaves. But having tried almost every preamplifier worth trying in the first place, I mean pre-preamplifier, and coming to the fact that the transformer type of pre preamplifier had the lowest psychoacoustic masking, then I tend to prefer the trans former as being state of the art. But suppose, somebody goes and invents a pre-pre amplifier which is capable of doing the job.

In the future we probably will see that kind of amplifier. How this is done, that's an other thing.

RAPPAPORT: Actually, there's no need at all for a pre-preamplifier. There ideally should exist no pre-preamplifier. If we accept the existence of the moving-coil cartridge as being the state of the art today, then the idea is to design the entire equalization and preamplification system around that cartridge, and there's no need for a trans former or an electronic flat amplifier to step up the voltage. A better approach might be to combine the equalization with the step-up and just deal with a preamplifier with a sensitivity of roughly 20 dB greater than our standard magnetic preamplifier.

HEGEMAN: Commercially, that's a very difficult one; Andy.

RAPPAPORT: I've been trying it for years and you're right; commercially it's very difficult.

COTTER: The thing is, you've got a problem which comes about in audio systems- maybe it bears a little notice at this point- that actually makes the problem for the de signer of the elements extremely complex, because of the lack of definition of inter faces-impedance, power, voltage levels and so on. We deal with the world of wild and variant kinds of interfaces, and that is certainly a non-optimum .

OTALA: The good old thing of having a high-sensitivity piezoelectric type of inter face-one volt, and that's it. Out of the pickup.

HEGEMAN: Isn't that what that strain gauge pickup produces approximately, or some thing like that?

OTALA: It has got an amplifier inside to do that.

EDITOR: The whole thing is an amplifier.

RAPPAPORT: That's right. It's just a vari able resistance which we put into an amplifier.

EDITOR: The thing is then, we have three alternatives. Put the moving-coil signal through a transformer that brings the voltage up to the level of a typical moving magnet or moving iron type of pickup-that's one possibility. The other is, put it through a flat amplifier that does the same; or, three, put it through an equalized preamplifier that brings it directly up to line level.

COTTER: But there's a consideration here of noise figure. Optimization, minimization of noise in the system. It would seem, for a variety of reasons, that it's easier to match a somewhat higher impedance level to get an optimum noise figure than it is at a very low impedance level. That's both a device and a circuit design problem. And there's the question of cost. In the ultimate, the noise injected by an amplifier will always be larger than the noise injected by a transformer.

The path I chose at the time I chose it was to go the transformer route because it looked like a better path.

OTALA: Well, you say the ultimate is al ways greater. | see no basic physical reason for that.

RAPPAPORT: We don't have devices yet which will allow us to . . . well, we could use hundreds of them in parallel. Theoretically it could be done.

OTALA: Yes, this is completely true, but it doesn't mean that we wouldn't have that kind of devices tomorrow.

RAPPAPORT: That's right. There's no reason why they can't be built.

COTTER: You do have a very practical problem in the amount of current that flows in your amplifying path; as long as it has that one-way electronic emission, you're going to be governed by the shot noise equation and the perviance equation, both of which are limited by materials and geometry. Even if you take the Richardson equation, which is the field emission situation, which is the most advantageous, that still gives you a very, very large current that you have to have in order to deal with the 3 or 4 or 5 or 6-ohm source impedance. It seems as though that is not only uneconomic but leads to some practical problems building semiconductors.

That is, the people who've tried to do this wind up with either very large or a very large number of devices which becomes very un economic. Or when you try to get into a small device, you start . . .

RAPPAPORT: When you speak about signal-to-noise ratio, though, the idea is that our limit is the signal-to-noise ratio in the groove, primarily. In other words, what is that signal-to-noise ratio, and then . . .

COTTER: Ah, but we just talked about that, and we don't know what that is.

RAPPAPORT: But the point is if that signal to-noise ratio is equivalent to say . . . with a moving-coil cartridge the signal-to-noise ratio is normally measured with a 1-millivolt signal . . .

COTTER: You're saying that's a 15 dB noise figure, why fight for a 2 dB noise figure.

RAPPAPORT: Exactly.

COTTER: I quite agree. The point that I think is important to make is that we don't know what the noise level of a disk really is yet. And we'd better start looking for it. But we're going to look for it with these line-contact styli with larger vertical forces, and we're going to do something about understanding what happens when we make a record to give us a noise level, and we may discover that we want another 20 dB.

OTALA: Coming back to another thing, I don't think for instance that using enough devices in parallel would be any problem.

We've got those devices-the National super-matched-pair transistor, for instance.

Just IC technology-how many would you want to connect in parallel? That's your choice.

COTTER: You keep running the collector current up.

OTALA: Right, that's true. If you don't do that you're having another horrendous problem, and that is if you decrease the unit collector current, the f; goes rather rapidly down. And it's a steep function of current at that level. So you're going to get a large variation of f. and consequently, since you are taking gain from that stage, you're then also causing phase modulation.

COTTER: Another view of the problem is that you're dealing basically with a current generator-we're talking about a low impedance generator-and amplification is equivalent to saying you're going to have many times larger current output which becomes these rather humongous currents.

And still at a fairly low impedance level. So this is a question of economics as well as a question of practicality. In the long run, you can build a transformer-if you assume you have theoretically no limit to the size, and to a certain extent you do have that virtue- you can make a transformer of vanishingly small resistance. To make an amplifier of vanishingly small equivalent noise resistance, like 0.5 ohm or something, gets to be a pretty large ensemble of conductors whose surface area becomes ...

OTALA: Making vanishingly small resistance transformers is also quite a problem because you're losing coupling then, in that case. And furthermore. the problem there is that. even if you would be able to do that, then you're still basically limited to let's say, four decades, five decades of frequency of proper operation due to various reactances.

COTTER: No. I think it's possible to make much more range than that. We've done it. But I find it very hard to build a preamp of comparable performance using what I know is available in the way of devices.

OTALA: You're exactly right: that's what I said too. Today the transformer seems to be the only alternative but not necessarily, I believe. tomorrow.

COTTER: Interestingly enough, when you go to the high currents, you wind up with a very significant current loop condition, which means you have to magnetically shield the system rather significantly because the current loop becomes a magnetic antenna. You wind up with shielding requirements.

We just opted for that because it seemed that in the long run, the opportunities pro vided by nature at the time-and perhaps even into the future, if you look at what semiconductors can do-we said with the available emissions didn't seem likely. So we went that road.

OTALA: The transformer has a very good property. though, and that is the filtering effect.

COTTER: Ah, you could make it anything you want.

HEGEMAN: Unfortunately, I'll have to vote a little in the other direction, because I have never yet heard a transformer in which I don't hear a change in the phase relationship of the upper harmonics of the violin, which acts as a constriction and a hardening element, as far as the string tone is concerned.

Now I admit I have not had any opportunity to work with Mitch's transformer, but this goes back to the early days, the early mono days with various cartridges.

OTALA: You know how many transformers you have in a signal path before it's cut into the record? HEGEMAN: You know what I listen to? I listen to recordings made with no transformers-for my strings-so I have perhaps a different perspective.

OTALA: Take almost any commercial mixing console; there's at least six transformers through which the signal . . .

COTTER: Most any microphone these days has a transformer.

OTALA: So that signal is already so contaminated.

HEGEMAN: I know, so why should we contaminate it any more?

OTALA: Yeah, well, there's something else coming. You know all that talk about Dolby records, and things like that.

---". . . almost invariably, when the distortion levels were small or reasonably small, the subjects picked the distorted channel as being the original." ---

HEGEMAN: To me that's unbearable contamination.

OTALA: That's true, but you see, the problem is that the more we get advanced in the technology, and the more we are aware of the contamination, the more we seem to start to contaminate. It was not long ago when they invented the noise reduction systems. It was not long ago when they started playing with . . .

HEGEMAN: 24-channel mix-downs.

OTALA: Okay. It's not long ago when they put all those LM-301's into the mixing con soles. It is not long ago when they started to use Kepexes and other units to spoil the rest of the recording. It was not long ago when they started to pre-distort the recording.

And heaven knows-it's not more than four, no five years ago when they invented the dynamic limiters to reduce the high frequency signal content, and they act exactly as TIM generators. Now for heaven's sake, this all has happened during this decade, and we are supposed to know something about sound quality!

COTTER: It's all proved as not important, because you can't hear the difference.

HEGEMAN: If you don't get a chance to hear the difference, you can't hear the difference. I agree with that.

RAPPAPORT: The problem is, we have two paralleling technologies. One is on one end of the stick, and they're going towards ease of production, and to a certain extent gimmickry, but it's allegedly with a purpose.

And then on the other side of the coin. you have us here, and we're trying to purify . . .

COTTER: Fidelitists.

RAPPAPORT: Yes, the fidelitists. Exactly. And they're two technologies which parallel but never intersect. And it's very un fortunate.

HEGEMAN: | don't consider them parallel, Andy; I think this is a divergent situation.

OTALA: But since this is a commercial world, and since we know that the percent age of unpolluted records, for instance, is decreasing rather than increasing, aren't we fighting a losing battle? I think we are.

WILCOX: I don't think that's really true, no.

HEGEMAN: I don't agree with Matti on that.

Let's keep on with the battle.

COTTER: Let me sound another note of optimism. Max is the one with the real note of optimism.

WILCOX: I think that it's changed. I don't know that this is the time you want to talk about it.

EDITOR: The very fact that this publication exists and is growing seems to indicate that there is some interest in the purist approach.

RAPPAPORT: The interesting thing is that every time we are allowed to listen to a superior component, we realize exactly how much better records are than we thought.

And they've been standing up, as Mitch says all the time, they've been standing up very well to the recent onslaught of improved components.

COTTER: I don't think we've ever played a record yet. I've gone around saying that, and I think there's probably more still lock ed up in the record than we are capable of extracting, that's good music.

EDITOR: Maybe this is not the moment to interject this, but this will soon stop when they switch to some kind of cockamamie digital process with a sampling rate of fifty thousand.

COTTER: I have an abiding faith it will not happen the way they claim.

RAPPAPORT & HEGEMAN: I hope not.

EDITOR: I would like to take this up later on; let's not forget about it.

COTTER: The fact is that there already exist more than a quarter of a million LP titles of incomparable majesty. When you look at the range and the character of music that is available to anyone today, no one in the past ever had access to that kind of wealth, be he king or prince or gold merchant of the universe. You can't possibly imagine the sort of resource that that made available. Some body at the age of 20 can have acquired more musical exposure than Mozart, Beethoven, Brahms and Telemann rolled into one. The number of hours of variegated music listening that this makes possible, however imperfect it is, is still an adequate stimulus to give you an exposure that you never had available.

OTALA: Which only goes to prove one thing. When we did our psychoacoustic experiments, we came to the staggering result that almost invariably, when the distortion levels were small or reasonably small, the subjects picked the distorted channel as being the original. That only means that we have become acquainted with distortion sources. Very few people go to concerts. They listen to radio and records.

COTTER: I was talking at lunch about the fact that the first 55 years of the phonograph history were spent listening to mechanically-acoustically recorded records, re produced acoustically-mechanically. And that we grew several generations of people whose identity with the record and the recording art was synonymous with distortion. Nobody doubted that it was distortion.

EDITOR: This is characteristic of many aspects of today's plastic culture.

COTTER: Orange juice.

EDITOR: Orange juice, coffee. Can you imagine--I always bring this up as an analogy-that hundreds of years ago, coffee beans were brought into Western Europe, and coffee drinking swept the Western world. Of course the original Arab way of doing it was to take the green coffee beans, roast them on an open fire then and there, grind them then and there, and then create a brew. And it was this magical brew that swept Europe, I believe in the 16th or 17th century. Can you imagine if it had been a can of instant coffee that was introduced at that point, that it would have swept Europe?

COTTER: On the floor, not in the cup.

EDITOR: The analogy goes for music.

FUTTERMAN: But | remember when, in the early days of radio, everything broadcast was live, at least 90% of it. So we listened to live music, even though the components

HEGEMAN: It was not live music, it was live performance.

EDITOR: Even in the early days of FM, Major Armstrong's station in Alpine, New Jersey, had lots of live broadcasts.

OTALA: But this is not what we are pointing out. It was live performance, which has nothing to do with live music itself. Live music is an acoustic sensation and if it is passed through a channel which has distortion, and if that's the only medium that people listen to, they get accustomed to the distortion, and they start thinking that that is the original. After we had these funny problems with people in the psychoacoustic experiments, we recommended them to go to a number of concerts. And they went. We later asked what did they like at the concert.

They said the tone quality was not very good, indeed, some kind of luster or brilliance missing-how can they ever sound that bad?

EDITOR: Of course this is a personal aesthetic that you can't really argue with. The plastic experience may have a sensory value that's incomparably better than the real experience.

HEGEMAN: I never come out of a live concert, as I walk out through the doors of the hall and down the steps--I say, that just set the cause of high fidelity back another ten years.

COTTER: Would you agree, Stew, though-that the general impression of recordings is that they are over-bright compared to live performances?

OTALA: There's some kind of glitter . . .

HEGEMAN: Yes. They're over-bright; they don't have the spatial, airy characteristic that a live performance has.

WILCOX: Yes, but there are various obvious reasons for that.

COTTER: Talk to me.

WILCOX: The obvious reason was the invention of the condenser microphone by-I don't know if it was Gerhard Neumann who invented it, but anyway ....

COTTER: It's a very, very old format.

WILCOX: I remember working with engineers who went from the days of the old RCA 44, the ribbon microphone, which was a flawed device but within its frequency range a pretty smooth-sounding device.

Then the condenser came along, and it came along with its pre-emphasis in the high end, and with the cardioid pickup, which also gave you more direct signal and less reflection. So you got two things. You got a brighter sounding microphone, and you also got a more direct sound. All of that added up to an edgy kind of sound.

COTTER: Why was it accepted?

WILCOX: Because it was supposed to be brilliant.

COTTER: Speakers I think, Max, were rather poor . . .

OTALA: There's another thing to this. When Armstrong experimented with FM radio here, he was particularly proud of the high frequency response, and that was boosted- and that created that sh-ch-ch type of sound.

HEGEMAN: I have recordings of a couple of his Army band concerts that came up from Washington, DC over his special 15-kHz line using Western Electric 640AA mikes.

They are very brilliant. It was very close miked, and it shouldn't have been. The 640 is not a mike you can close-mike with.

OTALA: This established some kind of standard synonymous of good sound.

WILCOX: I was an 18-year-old college kid, and I decided I was going to work for the summer and buy hi-fi equipment. So I went to Chicago to a place called Allied Radio, to their sound place. And the sound that was coming out of that place was so horrendous.

OTALA: Well, it is horrendous if you take a normal receiver and try to tune to New York stations.

EDITOR: Could we get back to preamps?

OTALA: We have stretched his patience.

COTTER: There are no easy answers. Do we all agree on that?

WILCOX: Peter, I think you just touched rather peripherally, and then zipped out of, something that's terribly important, that I would like to talk about for two minutes.

People say, why are things so brilliant-sounding in the name of high fidelity? I was telling a couple of them here. I re member deciding I wanted to buy a hi-fi system. I was 19, 18 years old, and I was a pianist; I was in college in music; I knew what music sounded like as most of my experience in music was live. I was playing violin sonatas with people, stuff like that.

So I went to Allied Radio. There was a Stevens, Stephens, I don't know how . . .

COTTER: Stevens.

WILCOX: Stevens Silver Coil, or something like that, loudspeaker, and there was the new GE variable reluctance pickup that didn't have any kind of preamp yet; it was just the naked output of the thing. I went into what was their audio salon. The relation of that to live music was nothing. It was the most screechy, horrendous . . .

EDITOR: It was all highs. If the variable reluctance cartridge wasn't equalized, all you got was highs.

WILCOX: That's right. And then the loud speakers were all tipped up. So from the Capeharts and Magnavoxes of the '30s and the early '40's, suddenly we came into the era of ''high fidelity."' And those of us who are old enough to remember that . . .

HEGEMAN: Those big old Capeharts weren't so bad, I'll tell you.

WILCOX: That was a very ugly period in the history of reproduced music. And I think only now are we starting to get away from that. Most recordings are still made with tipped-up microphones. They sound like very wide-range ''transparent'' recordings.

I would have to differ with you; there are some wonderful recordings; there are also some recordings which sound incredibly ugly when you play them on really wide range equipment because you see how screechy and awful-sounding they really are.

COTTER: But some of those older recordings in spite of their tipped-up responses have a cleanness due to their basic simplicity.

WILCOX: That's right. Because there are not so many microphones involved.

OTALA: Some of those recordings, how ever, also show another thing. With some of the cutters that are not really exactly the best, and some of the second-generation mixing consoles, we found a couple of re cords which by masking evaluation had about 35% TIM. 35% rms continuously is quite a number. You can't imagine!

COTTER: Coming back to this high fidelity phrase, I would like to introduce another idea. It seems to me that high fidelity came to mean something really very opposite from what high fidelity is supposed to be all about. Since we came together today to discuss state of the art in sound or whatever, but it seems to me that what we're talking about is largely something in the service of music. The idea was musical sound, music al ideas. It struck me that the phrase 'high fidelity' closely resembles the phrase ''painless dentistry.'" The minute the word ''painless'' appears, you know you have something to worry about. 'High fidelity' implies a certain struggle, and I think what we ought to make clear-because I think there are a lot of people who say 'Oh I know about high fidelity, I've heard high fidelity, it doesn't sound like music, it's too loud, and I really can't stand those high frequencies."" And that's what it means-it means almost too much pain for most people. I think what we're all talking about is something very different from that concept of high fidelity. What were talking about in fact is a kind of attainment, in which height no longer has any meaning because it's high enough. So we're really talking about music. And if we're talking about music, we're talking about un-musicality and musicality. It is possible, I think, to get a system that is less than perfectly accurate but which is more musical than other systems that blindly and foolishly pursue ac curacy without assessing the level of pain that is produced. I wonder if we can talk a moment about that aspect, because it comes to focus in records and in preamps. And with respect to such things as-not so much the 35% TIM, which is sort of unavoidable-but a question of a little excessive highs and brightness is something that can be in a sense mollified or modified to an extent where it becomes more musical by reducing the high content, the tone control idea. Is there any hope for improving the musicality of recorded sound at the expense of this accuracy?

OTALA: Don't talk about musicality or any trade-offs. They've probably contaminated the word '*high fidelity.' Lets just take the "high" off; let's talk about fidelity. You don't need to sacrifice anything, because if you have screeching highs there, then probably something is wrong. Either you've got distortion, or you've got excessive level.

Just take it off, that's it.

WILCOX: But unfortunately, a lot of it is built into the program source of the last 30 years.

RAPPAPORT: You can't do anything about that, though. If it's an equalization problem and it's simply a question of frequency response, then the tone controls come into play. But if you have 35% TIM or even 1% or whatever is above the audible threshold and it begins to sound hard and irritating, that's something you can't take away by turning down the highs or turning up the musicality control.

COTTER: Is that really true, that you can't take it away?

RAPPAPORT: You can't. Once a distortion is created, you can't take it away.

COTTER: Ah ...

OTALA: Nonlinear distortion by its very nature is such that it is a contamination of the original signal. You can't remove it un less you remove the original signal.

HEGEMAN: Again, modulation.

RAPPAPORT: You can compensate for a distortion that you can predict. If you can predict that there is a constant frequency response imbalance in a program, you can compensate for that. If there's a constant phase imbalance vs. frequency, you can compensate for that. But you can't compensate for a distortion element that is added to the signal. A nonlinear distortion.

COTTER: There are certain kinds of time distortion, one of the cleanest examples of which would be just the vertical angle pro cess, which puts 30% FM of the signal by itself on every stereo record ever made, approximately that for the last 15 years.

HEGEMAN: Is that why I like mono disks?

COTTER: Well, it's there in the vertical angle, is what I'm saying. It comes out when you play at the correct vertical angle and it'll vary in magnitude depending upon whether . . .

EDITOR: That's not a contamination then; it just washes out on the other side.

COTTER: No. It has a certain kind of removability. I think there is a whole family of distortions and disturbances that are separable. We ought to stop and think about that.

OTALA: As Andy says, if you can predict the distortion then you can remove it. But notably there's one form of effect that takes place which makes it impossible to predict, and that is the introduction of a frequency characteristic between the distortion generation and the distortion correction. Then you will start having a hard time to compensate 14 for it. Especially when you start talking about networks, you would in principle be able to say, all right, we've got an amplitude response and from that we can deduct the phase response. So, all right, everything is just fine-we can measure it, we can counterbalance it, then we can re-distort it, and it's okay. Unfortunately, this isn't the case, for two reasons. First of all, it is not the first pole which is important in the phase relationships-they are the second, third, and so on ad infinitum. And the second thing is, most of our transducers-right now, for instance, a few things, like cutter head, pickup, loudspeaker-are not mini mum-phase networks. They are partly non minimum-phase networks. Which means that the amplitude relationship and the phase relationship are not bound together with the Hurwitz transformation.

COTTER: They may not be Hurwitzian, and they may not even be Hilbert, but what I'm saying is, are they separable? I think that's a different question.

OTALA: Under these conditions, they are separable. If you can predict it, that means if you know the system characteristics be tween the distortion introduction and the distortion correction, in all dynamic do mains, then you can do it. Otherwise not.

COTTER: I happen to feel a lot of these are separable distortions. They aren't simple, but I think they are separable. The acoustic '". . . people who have been involved in these [preamplifier] tests have been using associated equipment of very, very limited resolving capability."' recording, for instance-as a matter of fact we were talking again at lunch about acoustic recordings having novel and different kinds of time domain disturbances, relatively simple.

RAPPAPORT: The point is that the problem with this program material that we have is not necessarily its amplitude characteristics or its phase characteristics, which are separable and are predictable if you know the system, and you can correct for that kind of thing-that's relatively simple. But the problem is the time dispersive distortions that we've been talking about.

COTTER: But I'm saying those too are analyzable and in some cases ...

RAPPAPORT: Not in all cases. Even simple distortions.

COTTER: Then that's a challenge for us.

HEGEMAN: You're gonna have a computer console, and you punch in an algorithm for every record that you have ....

RAPPAPORT: That's right. And for every mixing console, and every microphone . . .

HEGEMAN: You get a little complicated.

COTTER: I think we ought to look to Max on this score because it's a valid thing, I think, from a music point of view to say there's an awful lot of gorgeous musical performance, from the standpoint of the artistry and the concept and the execution, that is somehow or other entrained in media that make it less than the most pleasant thing to play back in the traditional way. The question I'm asking is, are those disturbances separable so that we can recover something more than we think we have there. We know of examples; we talked about the Stockham processing of the acoustical recordings of Caruso, that's a sort of example; it can be argued that they are or they aren't. Some years ago I heard one of the retired gentlemen from Bell Labs demonstrate a very interesting processor for injecting some detail into old acoustical recordings that cut off at 4, 5 or 6 kHz by the very simple expedient of processing them with a circuit that looked for transitions and injected a little white noise in proportion to the high frequency energy. It's amazing how much crispening of ...

HEGEMAN: Add a little noise, and the signal brightens up, I'll tell you, and you don't hear it as noise.

EDITOR: Is it perhaps for that reason that some people prefer head amps?

COTTER: Yes, possibly.

RAPPAPORT: Maybe. It's a very slight possibility.

COTTER: Anyway, the thing is that we have this tremendous wealth of recorded music.

Is it all completely lost? Is it, so to speak, destroyed irrevocably?

OTALA: No.

HEGEMAN: No, I don't think so.

WILCOX: I think that we're in the kind of plastic society that we're talking about. I would personally rather devote my energies to making good recordings of the great musicians who are now alive, and think it's lucky that we do have some kind of representation of those who are either retired, or those people who have been recording for the last 80 years.

OTALA: You can easily say, all right, go to the museum and look at the old clothes.

They were beautiful clothes, but you don't wear them, not them or not even their replicas. You wear your clothes.

COTTER: We can replicate them, though.

RAPPAPORT: We can replicate them but not with the same craftsmanship that they had, not with the same materials.

OTALA: It's a weak analogy.

WILCOX: I would like to deal with the fu ture, not the past.

EDITOR: Max, I want to ask you a question.

Of the great musicians that you've worked with-and you've worked with a number, you've worked with Rubinstein, you've worked with Solti, you've worked with some really great musicians-have any of them expressed any regret that in their early days they did not have the kind of recording techniques that they enjoy now? In other words, do they really care?

WILCOX: Some of them care. But they have again, what we were referring to at lunch as this wonderful suspension of disbelief. A musician will come in, he will be a little bit interested in the sound quality, maybe in the beginning. A few of them are very interested in it, not very many. Peter Serkin happens to be an exception; he's very interested in it. He's very interested in collaborating with me in getting a very accurate sound of the instruments. Most of them go in and immediately enter a world of fantasy, which is not an unpleasant world at all, where they’re listening to the tempo and the inflection, and are not really aware how accurately or with how much distortion their performance is being registered. I find them probably the poorest judges in some ways of recorded sound, because they don't really listen to it.

ZAYDE: But they’re not looking for replication of the live instrument per se, but rather the recorded event.

WILCOX: They already know that it's not going to sound like them, and it's not that dynamic range .

EDITOR: They already know that it can't be done. Isn't that what it comes down to? Whereas we here know that maybe it can be done one day. This is the difference.

OTALA: Well, not really. A musician is interested in how it is played, not how it sounds. He's interested in his fellow musician, the way he articulates, for instance.

EDITOR: That's true, but I have a purpose in pressing this point. The only way to make these major record companies pursue better technology, to go in the direction that we've been discussing here, would be for these people, for Sir Georg Solti to say, “I will not record for you if this is the garbage you give me.”

HEGEMAN: No, the only way to do that is to have the consumer say, “This is a lousy record'' and bounce it right back at the re cord store. If that happens enough, that's gonna hit where it hurts.

OTALA: 80% of American records sold are pop, and there the virtue is . . .

WILCOX: 95%.

OTALA: 95%, okay. There the virtue is to introduce as much distortion as you can.

EDITOR: Actually, among the pop musicians there are more hi-fi freaks than among the classical musicians. Cat Stevens, for example, is a hi-fi nut, and I am told that his place is full of panel after panel of Magneplanars.

ZAYDE: But is he trying to recreate a sound, or a reality? We have to look at that-what is he trying to recreate?

COTTER: Or create.

ZAYDE: Right, exactly, or create. There may not be a font for developing the original experience.

WILCOX: Most pop music now is not acoustical, anyway.

EDITOR: My point is that record companies can be approached only through their pockets. If the artist is fed up with the kind of sound he's been getting, that hits the pocket.

Nothing else does. The pressure comes from the technologists and the producers-I think the producer is in a somewhat better position to push than the technologist ...

WILCOX: In this country.

EDITOR: But if the pressure comes only from that side, it's not going to be as effective as if the pressure came from the artist. Wouldn't you agree with that?

WILCOX: I think the artist is the last person it's going to come from.

EDITOR: I am afraid so, too.

WILCOX: I think it's going to come from the people who know enough about what real music sounds like, and who buy records, and say, ''This is not really what it sounds like." I think there's been a refreshing re turn to simplicity in the last two or three years only, brought about by Doug Sax and Lincoln Mayorga, in the beginning almost

COTTER: Direct to disc.

WILCOX: Who are only copying techniques of people . . .

HEGEMAN: Another reinvention of the wheel.

WILCOX: That's right, exactly.

COTTER: You had not only the fidelity question resolved in many of those old 78 recordings, but you had a kind of artistic integrity, at least for the 6 or 7 or 8 minutes that took place, that is a hard thing to find today also.

WILCOX: That's a completely separate thing. But what I see happening is that in the small record companies that are now coming along, there is an interest in something like the Blumlein technique-which I am not so interested in-but in any case, there's a return to maybe making recordings with two or three microphones. I've heard some recent recordings made with all kinds of wonderful equipment-we were discussing it at lunch time-with two or three micro phones, which happen not to be good; all their 'state of the art,' all kinds of things.

So you have to have someone in the control room who can actually know what an orchestra sounds like. Just reducing the number of microphones to two or three isn't automatically going to give you a terribly accurate recording of the orchestra, any more than increasing them to 36 is going to give you more clarity. But I think there's a return to simplicity in the whole thing. Certainly in my work there is, and I think I'm not a pioneer in this at all.

OTALA: I think that we have not covered, almost at all, the preamplifier. We've discussed some of the problems, but . . .

EDITOR: I'm glad you said that, Senator.

OTALA: The problem is we have circum vented the problems. There's another thing I would like to add to the agenda, if that is of general interest, and that is measurement methods.

EDITOR: Absolutely. But couldn't we follow through? I would like to get the sense of this meeting as to what some of the basic considerations are. Our subscribers are interested in components. Whereas this is also an exchange of ideas among us, and it's for the benefit of each one of us, there has to be some kind of takeout for our subscribers when it comes to what they can expect of components. So maybe we could speed up the discussion.

COTTER: We got up to the preamp, then we went off in other directions.

EDITOR: We haven't really discussed the preamp. I know there are a number of issues on which all of you differ. For example, bandwidth limiting. Again, feedback vs. no feedback in preamps.

FUTTERMAN: Bandwidth limiting is a good subject, and we haven't even touched on it.

EDITOR: We haven't touched on that at all. I would like to talk about measurements, and not only measurements but also evaluation techniques, which is of course very, very close to my heart.

OTALA: One thing in particular I had in my mind when I mentioned measurement methods, for instance, is the new IHF Standard, which is just a catastrophe. It's the ultimate catastrophe. I mean we should do something about that, too.

HEGEMAN: Any time you standardize you reduce to absurdity, most of the time.

COTTER: Classes of fits and interfaces and so on, I think, no, I wouldn't agree. But I think what you're talking about is when you get a group of people to agree on a definitive method for the specification of products that they're all selling competitively, you're very likely to reach an absurdity.

HEGEMAN: Design a horse and end up with a camel.

COTTER: That's right. Can we get some opinion on what a phono preamplifier should do, in the light of all of the things we've said? Why do we have a situation where Al Foster and others feel the only difference is frequency response? Why do people take the attitude they do of specifying a group of preamplifiers which, upon measurement, verify that they all have these exquisite specifications? And yet they all sound different, and we're all striving to do better, even though the measurements are perfect. Are we dealing with time domain effects, again?

HEGEMAN: I believe so.

COTTER: Are we dealing with anything else?

RAPPAPORT: With the current evaluation of preamplifiers and the rather absurd opinion that seems to be very popular, that they all sound the same, I think there are two problems. One of the problems is the question of limits of resolution. Most, in fact all of the tests that I have heard about--I unfortunately haven't been involved in any--but that I've spoken to people about, people who have been involved in these tests, have been using associated equipment of very, very limited resolving capability.

COTTER: Which we feel. You're talking about the Shure M91.

RAPPAPORT: Well, going from the Shure cartridge to the-I'm going to lose friends-the AR speakers and this kind of thing, where the kinds of distortions that they will be able to hear are only the amplitude distortions, because these are the only things left. We've gotten rid of, in the cartridge and the speaker, we've gotten rid of all the ability to look at time problems.

EDITOR: This raises an interesting question about testing; even though I think we should pursue the discussion of preamplifiers, this comes in at this point or at almost any point.

The general question of whether garbage piled upon garbage sounds like plain garbage or a new kind of garbage. I have a feeling that it sounds just like plain garbage, which also puts a . . .

COTTER: I think what Andy's talking about is, there's a way of getting a question answered about those kinds of tests, that really reveals the nature of the problem.

That is to say, when people declare there is no difference, or they talk about the only difference being that of, say, frequency response errors and when they are correct that there really are no . . .

RAPPAPORT: That's really what they heard.

COTTER: That's really what they heard, and I'm inclined to believe them. There's one question I would ask and I think I know the answer. And that is I would say, 'What was what you were listening to in your comparison, something that was indeed very closely akin to live music?'' And I think the answer would have to be that it was not.

RAPPAPORT: And the point is that as we get closer to live music, the differences be come more and more apparent.

COTTER: In the department we're talking about here.

RAPPAPORT: In all departments. If there are x number of components in the chain, as x minus | of those become very, very good, the last one, you'll be able to show tremendous differences in the last one.

COTTER: I'm not sure you mean what you said, because we' ve had some discussion in the past about this. Do you think that differences in amplitude of .05 dB become important after you've cleaned up the time domain and these other effects?

HEGEMAN: No. They don't.

RAPPAPORT: I don't think they become important in and of themselves.

COTTER: Then what you're saying is not that these things, the other things, that is more of this frequency response, the kind of thing we were dismissing a while back in connection with amplifiers, that those be come important; they remain unimportant.

RAPPAPORT: They remain unimportant, but what I was saying was that as the associated components-say, we're discussing preamps-as the associated componentry becomes better in the sense of time distortions and that kind of thing, the ability of the reference system, so to speak, to resolve the differences in preamps that are time distortions or of that nature, increases. I believe the reason that these listening panels have come up with the conclusion that the only differences are frequency response differences, or cartridge loading differences-which is another very popular opinion-is that they did not have the ability to resolve the real differences in these components, which are the time dispersive types of distortions.

ZAYDE: Actually, it's even more coarse than that, if I may interject, Andy and Mitch. That is that they don't talk about frequency response, they talk about amplitude response, which is even a further coarsening of what's going on.

COTTER: But they're using pickups that introduce time dispersion, speakers that introduce time dispersion, and pre amplifiers themselves introduce also these

RAPPAPORT: There's a limit in the ability of the system as a whole to resolve the kinds of distortions that they really set out to listen for.

EDITOR: Does limited resolution in various parts of the system, in your opinion, invalidate straight-wire bypass tests as well?

RAPPAPORT: Yes, absolutely.

HEGEMAN: I think so.

EDITOR: In other words, if you're using, say, a Shure cartridge or some kind of low pass filter type of cartridge, and you're using a phono stage, say, full of TIM, and then you're testing line level preamp stages by means of a straight-wire bypass test, and listening to a speaker that again has limited resolution, you don't think you'll be able to hear valid differences?

RAPPAPORT: In most cases, no.

HEGEMAN: You may well hear differences, Pete, but I don't think you can really interpret what you're hearing in terms of the differences.

FUTTERMAN: Peter, I'm an Audio Critic subscriber, and I don't know what you mean by a straight-wire bypass. Explain it.

EDITOR: It's something that others have brought up. A straight-wire bypass is when you compare the sound of a component to the sound of a straight wire.

FUTTERMAN: Wait a minute. I didn't know a straight wire has a sound.

EDITOR: But it has a signal path.

RAPPAPORT: Sometimes a straight wire has a sound too, and in many of these straight wire bypasses that straight wire sounds very, very bad.

OTALA: The pot I talked about is a very good example.

FUTTERMAN: Give me an example of a straight-wire bypass speaker.

EDITOR: Of course you can't. What about a cartridge? You can't.

COTTER: Only with electronic things.

FUTTERMAN: Even with a preamp.

OTALA: Let me phrase what annoys me the most in all those tests that have been con ducted. In my opinion, basic scientific thinking is this: Everything is possible un less it has been proved to be impossible.

And even if it would be proved to be impossible, then there exists a distinct possibility that the proof was wrong. Therefore, if your conclusion from a listening test is that there cannot be any differences because we didn't hear any, this is just the opposite of every

'"We believe that the rise time or speed limit in the hearing perception is somewhere between 12 and 14 microseconds.' element of scientific thinking. It is the other way around. We say, in this test we did not hear any differences, therefore it is highly probable that there are differences, but we had a bad setup and we didn't find them.

You have to turn it around exactly, that attitude.

COTTER: I think it goes even further in the negating direction in that I can readily conceive-and I think you would concur, Matti- that if you were to compare a pickup that has, say, a low-pass 11-kHz cutoff and even with its mechanical resonance a rise-time property that's extremely limited, with some very much faster pickup which has very much faster rise time and much greater resolution, free of needle drag distortion as opposed to the other one which, say, had a lot, free of reactance modulation, time domain shifting effects which it had none, and you were to present this signal to the input of your pre amplifier under test, and compare the results obtained in the two cases, that you might well conclude that you got more junk out of the better, higher resolving system than you did with the lesser resolving system. You might infer that, therefore, the lower resolution system was ''better."' Now I could precisely reverse your opinion by getting rid of the problems that existed in the pre amplifier, and you'd get a reverse opinion. I think that we're talking about this pain pleasure ratio kind of thing, and the ability of the system to withstand the stress of the signals that are presented. The question be hind all that is what does it take to make it sound like music?

EDITOR: That's a very good point.

OTALA: You know, there's a very funny history to this to support your thinking. A certain Professor Bougaritz, who is Belgian, has conducted a number of tests where he injected distortion into Muzak. He's a psychologist. He has recorded the level of irritation in the workers given a daily dosage of distortion in Muzak.

HEGEMAN: You don't need to add any distortion to it.

OTALA: He was really jumping when he reported that he has found a strict, direct correlation, and he was very happy. And incidentally, he has been using TIM as one of his distortion mechanisms. He was even more excited about the fact that TIM seemed to be a very irritating pollutant.

COTTER: I think the problem of getting a poor result with a better stressing signal, which is what you would want to have if you wanted to get closer to the music, is not often addressed. That's where these comparisons can go astray. Andy was talking about this with respect to the preference for level being corrected, there's amplitude of--how did Bruce put it--just amplitude variations being removed, removed all of the remaining differences. None of it was what we would call good, clean sound. On the level of stress, one of the things that I found with respect to the TIM-like processes, is that the simple introduction of a 9-microsecond rise time limitation cleaned up remarkably most anything we could en counter. We later discovered that even if we had a very ultra-clean electronics, that mechanically, accelerations-the second, third, fourth, fifth moment, whatever you want to call it-that would be implied by allowing rates of change in the principal component-we're operating strictly in the time domain independent of amplitude here- that you wind up with accelerations that are up in the magnificent G's area, and you induce nonlinearities in the mechanical systems. This flies in the face of the DC-to light theorists, and is a subject that I think we ought to bring up at this point because the burden that is to be borne then by what ever the system is that we're considering, can't have infinite limits. Because obviously there is some stress level that will break down any system. Even if you have a gigahertz amplifier, there is some kind of signal that's going to produce an excess of garbage out with that sort of bandwidth exposure.

RAPPAPORT & OTALA: Not necessarily.

COTTER: There's an amusing little tautology, and that is you have infinite bandwidth, then the immediate reaction of connecting your loudspeaker to such a system is that it melts down because you have this infinite energy coming out of this infinite bandwidth system from the noise, just from the noise.

Anyway, aside from that, it seems to me that one of the problems that we ought to avoid in an audio chain is excessive stress, stresses that are beyond the range of that which is necessary to complement the human hearing apparatus. My simple expedient seemed to work, even though that value of rise time is clearly not discernible as different from something significantly faster or something significantly slower.

HEGEMAN: Mitch, why 9 microseconds? This is a number that I find I can't really translate into my hearing experience, or anything else. Why 9?

COTTER: We tried a lot of different values. I was looking for the longest time which would, as a final rise-time limiter in the system, not introduce something that encroached on the other limiting processes, and which still was sufficiently slower than leaving it wide open to much greater speed. I wanted something. in other words, which wouldn't be discernible as a loss in acuity; we established that that's certainly not. We believe that the rise time or speed limit in the hearing perception is somewhere between 12 and 14 microseconds. Certainly I think people would agree 12 microseconds seems to be, for an event, indiscernible from 2 microseconds. And 9 microseconds seemed to be a worthwhile limitation and yet not something that would encroach too heavily on the ...

OTALA: Single-pole filter?

COTTER: No, not a single-pole filter. A time-domain corrected design which minimizes the maximum transition.

OTALA: A Bessel.

COTTER: Well. it's a Bessel-type transfer, it has low ringing, and it has minimum maximum slope. A toe in other words.

OTALA: How many poles?

COTTER: Equivalent to 9 poles.

HEGEMAN: That's interesting because it's very different from my own experiments, and I admit that this is tape program, not record program.

EDITOR: That shouldn't make any difference.

COTTER: It's a time-domain synthesis, Matti, and it has less than a percent or 2 percent ripple.

HEGEMAN: We always found, and this was not only myself, the Citation II power amplifier had a 4-microsecond rise time.

COTTER: In the output, or as an put limiter?

HEGEMAN: In the output; that was what went out. At one point I designed a small amplifier for Lafayette which had a 2.5-microsecond rise time, and there was a difference in the sound: now true, that amplifier sounded slightly different. But what you could seem to hear was a better airiness around the top end of the presentation. It was more real in space. It was a spatial characteristic that you heard; not necessarily a sonic thing. I've been listening for years to amplifiers that have a 1-micro second rise time up to the full output of the thing, and frankly-I guess my ears are conditioned-I love them.

COTTER: I don't find anything inconsistent in that; it's just that I don't think the differences you're hearing come from the rise time phenomena as much as they come from other things.

OTALA: Let me put it this way. I just tried to calculate some experimental values that I've found. They boil down to some startling figures. I would say that for an amplifier having 26 dB of feedback, 0.8 microsecond is the maximum rise time that can be allowed for the amplifier itself; and about 4 micro seconds is just right when you put the input filter in. But under those circumstances, and assuming perfectly abrupt slewing, 0.8 microsecond rise time seems to be right.

There are some good rules of thumb.

EDITOR: But you're talking about two different things.

COTTER: But that depends very much on your definition of not only the amount of feedback but what the delay properties are in the amplifier.

OTALA: No. I'm not talking about that; that's a perfectly stable amplifier, and the delay therefore is relatively unimportant. But 0.8 microsecond comes from the simple feedback relationship. If you put a 100 kHz filter at the input, then you cannot allow the distortion to rise in the frequency range from 0 to 100 kHz. You may allow it to start rising after that, but that means open loop distortion design, to yield flat distortion spectrum.

COTTER: But this is outside of the feedback, Matti.

OTALA: I'm talking about inside the feed back.

COTTER: That's a different situation al together. I'm talking about a rise-time limiting device that is totally passive, that is not active, that has no feedback; it's simply a passive filter. And it's the input signal that is limited to this rate of change.

OTALA: I know that.

EDITOR: Mitch. to simplify matters, could I interject something here? Your contention, Mitch, is that if in this room, or in a music room where music was being played, we could somehow preserve all the information only to a speed of 9 microseconds, and somehow got rid of the . . .

COTTER: If we'd put a gas in the air that somehow or other had a 9-microsecond rise time limitation of the form of this quasi Gaussian, Bessel-type thing . . .

EDITOR: Then we couldn't hear any difference, you're saying.

COTTER: . . . that minimizes the maximum rate of change, at 9 microseconds slope, then I don't see that you would hear any difference whatsoever.

OTALA: I'm pointing to exactly the same thing. In my opinion, that is fully legitimate. The important difference, however, is that--I'm saying this following my own experience-say 4 microseconds would be all right as a rise time. So let's take the microphone, let's put a 4-microsecond or 9 microsecond filter after that. And then, after that I say that the amplifier following that filter must have at least 0.8 microsecond rise time.

RAPPAPORT: With 26 dB feedback. Exactly.

OTALA: With 26 dB feedback. In order to cope with the signal that is coming out from the microphone.

RAPPAPORT: So that the open-loop amplifier is not driven by the 4-microsecond rise time into nonlinearity.

COTTER: Unless it has no feedback at all, or has lesser feedback. I really don't care at all about that. I'm only saying there is some value of stress, in the sense of first, second, third, nth moment, which . . .

OTALA: Here is the big difference and the big confusion. The proponents of the so called DC-to-light frequency response are perfectly right in exactly the fact that the amplifiers following a rise time limitation, or a frequency bandwidth limitation, must have a bandwidth to light in order to cope with the bandwidth-limited signal.

RAPPAPORT: If it's a feedback amplifier.

And the bandwidth of the amplifier is deter mined by the amount of feedback around it.

So in a situation where you have a no-feed back amplifier, the amplifier should be simply as fast as the rise time limitation, maybe a little faster . . .

OTALA: To retrace this, an amplifier having 26 dB of feedback, in order to reproduce properly a 30-kHz signal, must have a band width of 1 MHz. Period.

COTTER: Yes, this is the basic Bode constraint. That's generally not adhered to, I might add, in the execution.

HEGEMAN: You noticed that, huh?

COTTER: Well, we do notice it in this interesting way. That in no system have we ever encountered less than a significant improvement in the overall sound when such a filter is used in the system between the power amplifier and the preceding stages.

RAPPAPORT: I've got a question for Matti, because you've obviously done work in this area. The question is, if you take your 0.8-microsecond rise time amplifier with 26 dB of feedback, and you put a variable rise time filter on the front end of it-and let's assume it's time-compensated and isn't going to create any problems in the audible range-do you notice a difference between 4 microseconds, 9 microseconds, or what I use, which is 14 microseconds?

OTALA: I recently conducted a series of experiments, where it was a 2-pole Bessel filter which was adjusted down from 1MHz to the lowest frequency that I was

"allowed" to try, 100 kHz. No audible effect was noted.

RAPPAPORT: And if you go from 100 kHz to say 30 kHz or 25 kHz?

OTALA: Well, I would say in a 2-pole Bessel, for instance, the problem is mostly that the amplitude characteristics are not sufficient. If you go to 50 kHz it already is 2 dB down from 20 kHz because it's so smoothly rolled off.

COTTER: You need a higher order of approximation.

OTALA: There are some other problems in higher order approximations, especially if you do it actively. So don't try that.

COTTER: Oh no, don't try it active. But passive, there are no problems . . .

EDITOR: There doesn't appear to be any disagreement among you. We all agree that the haphazard kind of bandwidth limiting will be audible for various reasons.

COTTER: There's an important amplification here. I said that I found that introducing this cleaned up a lot of the problems in the system by removing many of the TIM-like processes and some of these other time modulation effects which exist apart from feed back by itself. Matti is giving us some ideas about what a feedback amplifier would have to do that followed this situation. What we're all saying is that it's pointless to have stresses in the signal source beyond the range of that which is valuable from the standpoint of what the hearing experience is. Are we agreed on that? We don't know what that number is precisely.

OTALA: You're saying in fact that it is pointless having an audio frequency response, or frequency characteristic, beyond x kHz.

COTTER: I prefer to talk about it in the time domain.

OTALA: Well, all right, that's your handicap.

RAPPAPORT: It just requires a little extra math.

EDITOR: Mitch, what is the relationship there?

RAPPAPORT: 0.35 over the bandwidth.

COTTER: When you leave the minimum phase situation, you enter into some other

OTALA: It's 0.35 to 0.42, depending on the filter, on the type of cutoff.

COTTER: The thing is that it varies with the slope, but the essential idea is that there's some value here that we're saying doesn't remove information. But does it remove problems? I say it removes problems by preventing excessive stresses.

RAPPAPORT: It all depends on what these excessive stresses are going to do to the components. You can take an amplifier which has no problem at all with 100 kHz or 200 kHz or 1 MHz, and your filter isn't going to have an effect at all.

COTTER: What we found, Andy, that's interesting is that where you have a situation approximating that, at least insofar as we can tell, that it is the mechanical side of the system that gets into trouble. So it seems as though there's a rational basis for this, whether the electronics is vulnerable or not.

RAPPAPORT: This is a very common mis conception that I've heard in conjunction with your filter in particular. I am very confident that your filter does absolutely no thing for my amplifier. But I still recommend its use in conjunction with my amplifier because most of these tweeters that people are driving with my amplifier are really giving problems.

EDITOR: That's certainly true.

COTTER: I don't disagree at all. We did have another interesting experience when discussing filters with someone, and the problem was that they were distressed to find it made a difference in the sound. be cause the idea of such a filter had been presented to them as something that re moved a certain latent vulnerability to something, should it come along and cause a problem, rather than being an on-stream purification of the signal handling ability.

Am | making myself clear? In other words, the idea behind that person's impression was that this filter was a sort of protective device-it removed lightning bolt threats, and things of this sort. but it shouldn't affect the signal. They were distressed to find that it made a difference in the sound. I had to apprise them of the fact that that was the reason we made the filter.

EDITOR: Why would you make something that makes no difference at all?

COTTER: Yes, but the point is that their attitude was that that was just a protection from those threats that might come along.

My attitude was they're continuously pre sent, and you want to remove them. I think 18 it's important that we clarify what it is we're talking about in this connection.

OTALA: Let me read something from a paper. This is a paper published in 1970: that was the original paper on TIM. The conclusion section, if you will bear that reading, goes like this.

"Conclusions: (1) To minimize transient intermodulation distortion, it is advantageous to let the preamplifier limit the frequency response of the complete amplifier. (2) In the above case, it is the power amplifier frequency response without feed back that determines the desired pre amplifier frequency response. Therefore, the feedback in the power amplifier does not necessarily enhance the usable frequency response of the complete amplifier system. (3) If high-fidelity reproduction requires a 20-kHz upper cutoff frequency in the amplifier, the power amplifier should reach it without feedback. (4) In the above case, the upper cutoff frequency of the power amplifier with feedback must be at least this 20 kHz times the feedback loop gain. For in stance, in an amplifier with 40 dB feedback, 2 MHz."

COTTER: Those are the classic rules.

FUTTERMAN: Yes, and my amplifier follows those rules. But I have a filter in front of it.

COTTER: Yes, which makes it immune to the range in which it could conceivably get into some problems.

“Every amplifier should have an input filter which will restrict the input signal to just that with which the amplifier has no trouble at all in dealing.”

HEGEMAN: I still have a question on this 9 microsecond bit. Does that cutoff change the spatial characteristic of what you hear?

COTTER: Not so far as we can tell, Stew.

HEGEMAN: Because old Harry Olson once told me that he thought that time differences in channel-to-channel work on stereo could be down to several, like 1 or 2 micro seconds, or you find yourself . . .

COTTER: The differential delay in these systems is zilch. It isn't even nanoseconds. The point is that it wouldn't matter if there was a half-second delay in a filter as long as what came out came out together, and the differential delay between the two was zilch.

The differential delay even in the passband is zip because there is very minimal transient disturbance, and that's why you need a multi-order, higher order filter approximation to make the thing behave in a correct way in the time domain. The closer you move it in to the audio passband, the more significant that time domain correction be comes, because if you were to, say, have a cutoff that was really audibly inside the passband, then you might have to have your ripple component down 70 or 80 dB to be below the threshold of audibility, and that is indeed a heroic problem. So we bring it to within range; we have it outside the range of audibility; the ripple value we have certainly isn't going to cause amplifiers, or whatever, to zock from spiking or anything. And we allow the system to have some other limitations in some other portion of the system without encroaching too heavily, because these things roughly add by the square root of the sum of the squares of the rise times.

So that was the feeling about 9 micro seconds. Which is not a heavy feeling, not a strong feeling, so I don't mean that it's a real magic number.

OTALA: It's about 40 kHz.

COTTER: Typically around 40 kHz. That's essentially the idea.

RAPPAPORT: There's just one thing that Stew touched on, and that is that there's a very common misconception about limited bandwidth. There are those people, in fact, publishing audio reviews who maintain that unless a device has a bandwidth from DC to X-rays or whatever, or preferably below DC.

COTTER: That's going to collapse the space.

RAPPAPORT: That's right. The space and the air are completely gone. The idea is that this is because limited-bandwidth feedback amplifiers typically have been loaded with transient distortions, which collapse the space and take away the air and make things sound hard.

COTTER: The more feedback they have or the faster the rise time, the more time modulation effects and the more garbage there was, and the idea of going faster and faster was that you got a change in character but not a reduction in magnitude of these time modulations.

RAPPAPORT: Exactly.

COTTER: And I think that's a false goal.

We've talked about that in amplifiers. But I see no reason that an audio system should be required to go beyond that. Now one of the things about digital that is a very important concept is that digital allows you to perform a sort of arbitrary approximation. You merely have to tell it, so to speak, in an abstract way, you merely have to tell the system what it is you need. You can't have infinite bandwidth; that's infinite sampling density. So you have to make a decision about what's necessary. So in a sense what we're saying here is that a 9-microsecond rise time satisfies the audio requirement, certainly. We're agreeing. Maybe we're not absolutely correct, but we're not far off the mark. You say you use 14; Stew has used 4; we're all more or less somewhere in this ballpark.

RAPPAPORT: The idea is that there is absolutely no reason for the reproduction system to be able to reproduce anything faster than that which we can hear. Unless, of course, there is the idea of a safety margin.

EDITOR: Are any of you gentlemen aware of actual experiments without electronics, in actual air, to measure this phenomenon?

COTTER: Harry Olson constructed an acoustic filter to approximate that filtering condition with some limited properties.

EDITOR: What did that consist of?

COTTER: It was a slot structure.

HEGEMAN: It was a room with a variable basic door coming down on there.

RAPPAPORT: I'm embarrassed to say I was 10 years old when I read about that in Popular Electronics.

EDITOR: And what were those findings?

HEGEMAN: His cutoff was basically about 7 kHz.

COTTER: 7 to 10 kHz.

HEGEMAN: 7 kHz, 10 kHz, and full range kind of thing. Everybody-and he apparently pulled people off the streets for his listening panel-but he had a fairly large sample of people . . .

COTTER: In Princeton.

HEGEMAN: Was it in Princeton? Yeah, could have been Princeton, could have been. Everybody could hear the 7-kHz cut off. Not everybody could hear the 10-kHz cutoff, as he started, with live music, and so forth and so on. Now it's a very simple thing; you can do this in any concert hall in the world. You listen to somebody out on the stage playing a violin, and all you have to do is move forward one row from the edge of the balcony underneath it to the edge of the balcony out in the open. Your string characteristic opens up and changes absolutely, completely. But if you're under the balcony . . .

COTTER: But Stew, don't we wish we could get as good a sound as we can get just that one row back? HEGEMAN: No! I want to be that one row front.

EDITOR: Nobody has ever tried a 35-kHz acoustical filter.

COTTER: But I'm saying if we had a sound system that would be even one row back, we'd be doing pretty good.

HEGEMAN: But not when you have it as an A-B, Mitch. That gets totally unacceptable.

OTALA: But I would make one correction on what you said, and I don't think you said it on purpose, but it might convey some misunderstanding there. You said there is no need whatsoever for an amplifying sys tem to reproduce anything beyond that which is audible.

HEGEMAN: What is audible?

OTALA: That requires one correction. That is that in many systems, you cannot in the first place restrict yourself at the signal source to what is audible. Therefore, you cannot put the filter exactly on that point. And this goes for preamplifiers, for in stance.

RAPPAPORT: What I was referring to was an idealized case.

OTALA: Exactly. So you have to design the things, the amplifiers, to be capable of coping with what's fed into them. But we're allowed to limit that.

RAPPAPORT: That's right. Every amplifier should have an input filter which will restrict the input signal to just that with which the amplifier has no trouble at all in dealing.

And there's no reason for that input filter to be of any shorter rise time than that which we can hear.

OTALA: That's true, yes, fully agreed.

COTTER: What Matti is saying also, we might even have a criterion, a sort of standard for apportioning some other limitations elsewhere in the system and dividing it up say between the recording and the reproducing side, because certainly the disk cutter gets into some very serious limitations on its

OTALA: And remember that there's a summation of rise times, so that your 9 microseconds and your 14 might be okay just for that case. But don't propagate that as an acceptable value ...

COTTER: Well, the eleventh dub is going to be in trouble.

OTALA: ...because people are starting to put ten filters in the chain, and after that you're in dead trouble, because that's 25 microseconds.

RAPPAPORT: That's right.

COTTER: Let me say this. One of the things that Stew was poking a little fun at was the use of a lot of transformers, that often happened in the past. But one of the things that was accomplished in a system that had some transformers in it, especially systems with flat amplifiers like 300B's and no feedback, was that these transformers afforded a mea sure of protection from these excessively fast events. Since the systems were, basically. simpler, composed of fewer links and chains. we were often dealing with systems that had-though not described in detail and in a sense almost inadvertently-very significant speed limitations in light of the sort of times, speeds and rise times we're talking about. That kind of came for free with the case at hand. Undefined, but very much a part of it. Even today, the RIAA pre emphasis recording characteristic is defined as a straight line with a slope of 1 at the point at which it stops. at the top end of the audio spectrum. There's nothing to indicate that it doesn't go up infinitely, which is insane of course, because you have an infinite ...

HEGEMAN: It's impossible to measure.

COTTER: The fact is a matter of practical significance, very little known, is that the Neumann system turns over at about 35 kHz. and that the Ortofon people originally turned over higher but modified their characteristic to coincide, because of the processing and the electronic kind of music that was being cut, to make that agree with this undefined . . .

HEGEMAN: So that's basically about a 5-microsecond turnover?

COTTER: Something like that period. You recall when Jerry Minter and you and I and RCA tried to get a cut in that characteristic?

EDITOR: Have we got a clean signal path now, from stylus tip to power amp output? Have we neglected anything important?

HEGEMAN: I think one of the things we've neglected so far is the time adjustment of the preamp equalization network.

COTTER: Why don't you amplify on that?

HEGEMAN: As an old-timer, if you were going to measure the response of an equalizing amplifier, the simple way to do it is put in your inverse network. At which point then you can change your frequencies and go through there. and you always read the same, you return to zero on the meter, and if you have designed that inverse network properly. it's a very easy test procedure.

Now you take one of these, which I have, you start putting a square wave into it, and oh boy-you have discontinuities that you can't believe. It's been my practice to do a square wave adjustment, basically a trim.

Because I can't find RMA value resistors and close enough tolerance capacitors that I can afford, and I'm not even sure, if I finally came down to these esoteric values, it would always work.

FUTTERMAN: Excuse me, what's the frequency of the square wave you use?

HEGEMAN: 10 Hz to 100 kHz.

FUTTERMAN: Through an inverse network.

EDITOR: Stew, if all your stages of gain were perfectly linear, and your RIAA equalization is the exact mirror image of the RIAA preemphasis, then you should automatically get square waves.

HEGEMAN: From a practical standpoint, you either have to build a piece of measurement gear, or, if you have a production unit, you have to trim it.

COTTER: There's another problem here, and that is the RIAA boost characteristic that you use goes up how far? Where do you turn over?

HEGEMAN: I've turned it over with a 7.5”-- since we're working at time constants of 3180, 318, 75, I took 7.5.

COTTER: 7.5 microseconds on the top is where you stop boosting.

HEGEMAN: That's right. It's a turnover there. Working with my existing test equipment and everything else, I could get it down to about 2Y or something like that, but then I wouldn't have the drive to drive the network to make the test.

COTTER: To interpret what you're saying, you're saying that even using a 7 1/2-micro second turnover on the top, stopping the boost in other words at a 20 dB up value, sloping off at 20 dB up, that you still notice that preamps go a little bananas when you hit them with that sharp a signal. Is that correct? The rise time is still going to be the rise time of whatever your generator is. The turnover is 7.5 microseconds. What rise time speed do you use in a square wave?

HEGEMAN: What does my Hewlett Packard do?

EDITOR: Probably of the order of 30, 40 nanoseconds?

HEGEMAN: It's a 10 megacycle unit; it's probably, yeah, fairly fast.

COTTER: I wonder if that's a realistic input signal in the light of what we've been talking about. 30 nanoseconds rise time isn't likely to come out of a phono pickup, not any that I know of, anyway. What would happen if you slow it up?

HEGEMAN: I don't truly know.

EDITOR: I've heard this general concern about the phase characteristics of the RIAA equalization, but one should think that whatever the phase characteristic. it would wash out with a precise inverse curve.

HEGEMAN: Peter, it's essentially a multiple network. You drive out of a low impedance. But if you're doing a passive equalization job, you do it in sections. This section takes care of this, that section takes care of that, and so forth and so on. You end up with square waves that look very queer.

COTTER: I think that there are three different problems. The first problem is that you're applying a transition that is very, very fast. You said 30 nanoseconds. The second problem is that I think you're pointing to a different problem, which is that the topology of almost all the standard equalizers does not turn into a piece of algebra. If you state what the equations of that transfer function are, that resembles the transfer function stated in the RIAA or the IEC standard-simply because the components are not freed of their interaction. Even when corrections are made, as presented in a re cent paper, what happens is that you may

correct the pole values but the transition has a different form of algebra, and there are saddles-there are errors in the response that develop that inevitably lead to funny looking things in the square wave. The square wave doesn't come out looking like a square wave.

EDITOR: Yes, but he doesn't have that in his preamp.

COTTER: Well, you can do this correctly by partitioning the network so you don't have that problem. The third thing is, what do we see happening in a preamp where we don't hit it necessarily with this high a speed of transition? In other words, 1 would say that if you got down to the microseconds speed of transition, and you hit it with a pre emphasized signal that's more or less like the pickup you might use-we talked about a pickup that had a 70-kHz bandwidth-you look at some of the better moving-coil pick ups and you'll see a certain kind of rise time capability. What happens to a preamp if we hit it with a signal which has the 6-dB-per octave rising velocity characteristic, coming out of a pickup such as that? Neglect for a moment the errors in equalization and talk about these time domain effects that go on.

We find that most of the topologies are flawed in that they have the equalization in the feedback, and the system is capable of a lot of time modulation. The current levels at which they operate are very low fr's, and you've got a system that's very, very time modulable by the signal. Capable, in fact, of the triggering I alluded to before that can be controlled by marginal changes in vertical angle, which is the speed of the signal transients. I think that preamps suffer categorically from a lack of understanding of what their dynamic response characteristic is. With the equalization in the system, you're boosting the sensitivity, as I've said before, to all the difference tone change that can be generated.

RAPPAPORT: If I can add something to that . . . I think what Stew is getting at is the idea that if you have a given amplitude response in your preamp. and it follows exactly the RIAA standard, and you feed it a square wave which has exactly the inverse in amplitude and phase characteristic, you should get back a square wave. But the problem there, and the reason that you can't really trim in a preamp by using a square wave- unless you know the properties of the preamp and in fact I would contend unless the preamp is properly designed-is be cause, especially when the equalization is accomplished in the feedback loop, you're going to get overshoots. There's going to be an overshoot all the time in that case. A lot of people are now looking at equalization and using a step with a relatively fast transition time, and they're saying well, ideally, you should get a step at the output. And that's true, but the fact that you get an over shoot in many cases doesn't necessarily mean that there's an error in the equalization. It means there are other problems occurring in the . . .

COTTER: We're getting into these other things. I'm saying that I think 30 nano seconds is an excessive n moment, dE/dt or dV/dt or whatever . . .

HEGEMAN: What difference does it make?

OTALA: There's been some discussion about the DIM method, for instance, being invalid or too stringent in preamplifier testing. Of course, the original recommendation was that it should be inversely emphasized with an inverse RIAA curve . . .

COTTER: But you and John Curl did start to put some time limits of 3 microseconds-was it a | microsecond or 3 microsecond . . .

OTALA: No. no. The DIM 30 method is the ly method which has the 30-kHz roll off.

EDITOR: That's a single-pole filter, right?

OTALA: Single-pole filter. But then for pre amplifier testing we have now used the DIM 30-30 method, which has a two-pole roll-off in order to limit the rise times more or less after the inverse correction to, say, some thing like 30-kHz bandwidth.

EDITOR: Matti, how do you explain the fact, though, that every piece of junk that comes on the market now claims to do very well on your tests? I see this on the specs every time now. TIM, .004%; the sine square test . . .

COTTER: He covered that in his discussion in a recent paper on these thresholds. The problem is if you average it in the right way you get some very low percentage numbers, but when you look at what that represents in non-instantaneous spaces. it turns out it's 5% or something.

"I wish that were made very clear to everyone. Do you hear us out there? Just a good TIM measurement doesn't mean you have a good amplifier.''

EDITOR: All right, but they take the rms value of the . . .

OTALA: Let's put it this way. The DIM method detects primarily TIM and related phenomena. There's no problem making those vanishingly small for these band widths. We're still having 30-kHz band width, and for that bandwidth, with present technology, it is not a problem. It was a problem five years ago when the method was originally devised. It is not anymore.

There are other effects, and that's exactly why I've been here advocating a total pic ture, so to say. TIM is the past; it's history; it doesn't exist any more.

EDITOR: You're absolutely right. I have tested equipment here that invokes your spirit, obviously for endorsement. They say they've done your test and the device is perfect, or nearly perfect, in terms of your test-and the sound is terrible. Obviously there are other things that they haven't done right.

OTALA: That's true. There are 100 or 200 other possibilities, at least.

EDITOR: | wish that were made very clear to everyone. Do you hear us out there? Just a good TIM measurement doesn't mean you have a good amplifier! Do you all under stand that out there. please?

OTALA: You can quote my words by saying that, apart from the TIM psychoacoustics, we have not done any significant work on TIM during the last five years. Everything that has come out has been done previously and has just waited in the mills of publication routine. We finished that in 1974, something like that. We considered the case closed. There was no problem any more.

When it started, it was a problem, and a horrendous problem at that time. We had catastrophic amplifiers on the market in, say, 1969. Right now it's seemingly so that people have learned their lesson and they got rid of TIM. So, case closed. Let's concentrate on other factors.

COTTER: What effects do you think are still with us in preamplifiers that we should look at? Is the speed of change affecting the result? Is this rise time?

OTALA: Here we go to an area where I don't exactly know.

RAPPAPORT: You can't examine the problems of a preamp really independently of anything else, because a properly designed preamp . . . Basically you have a stage of gain, or two stages of gain, or three stages of gain, and an equalization network. The equalization network is reasonably cut-and dried; at least it should be. Then there is a stage of gain, there is a low-level gain stage, or two, or three, or four, or five, or what ever. The idea is that the considerations for those should be exactly what the considerations are for any active stage in terms of dynamic performance.

EDITOR: In other words, the criteria we laid down in connection with power amplifiers.

RAPPAPORT: Exactly. There are a few special cases. For instance, in a passively equalized system where you have a stage of gain, then equalization, then another stage of gain, the first stage of gain has to be able to deal with some pretty fast signals, unless of course you put a filter at the input. It also has to be able to drive an equalization net work which is a varying impedance, and a few other things. But it's really basically just a stage of gain.

EDITOR: How do you explain, though, the phenomenon we found here, that with a really good moving-coil cartridge aligned to the nth degree for lateral and vertical geometry, playing, say, massed sopranos singing high, nearly all preamps sound distressed-not all but nearly all. What phenomenon is that?

RAPPAPORT: It's probably the fact that there is a tremendous-because of the preemphasis and the speed of the cartridges which we're using-there is a tremendous amount of high-frequency energy. Most circuits are not designed in such a way that they will readily cope with this energy.

WILCOX: Do you know that the s. al is clean, Peter?

HEGEMAN: I was just going to ask Max, how do you handle that?

EDITOR: If it sounds clean and more de tailed on some preamps, then the cleanliness and the resolution must be there.

WILCOX: That's the kind of signal you hardly ever find clean on any record.

RAPPAPORT: You'll find that those are the preamps which have no trouble dealing with this high-frequency information, either be cause of the way . . .

EDITOR: How would you quantify that kind of distortion?

COTTER: Max, I want to stop at this point and say something about that. I think the problem that we're getting to is that we don't really know what's on the record. All we're talking about here is the playbacks that we've been making. We've seen many dimensions in which the playbacks that we' ve been making are non-representative of what you could call the latent signal in the record. We're talking about ways of optimizing that . . .

WILCOX: I'm just talking about hearing master tapes that have distortions in sopranos.

COTTER: That's another story.

OTALA: I would like to object to Peter that the problem is-the problem can be-just the converse. That is, those that you feel are clean and nice and beautiful might just be so terribly distorted that they take all the garbage away.

RAPPAPORT: This is of course a possibility.

COTTER: That's a good kind of distortion, if it makes it sound more musical.

OTALA: It certainly is, yes.

EDITOR: I think that's most unlikely.

OTALA: Most of the distortions, when there is just slight distortion, are pleasant. Most of the distortion that I've been working with adds musicality to the sound.

FUTTERMAN: In Japan, they like amplifiers with a lot of second harmonic distortion.

EDITOR: Do you think it's possible that we have preamplifier A and preamplifier B, and we play this particular type of sound that I find particularly stressful--either solo sopranos, or massed sopranos are even better-singing well above the staff, and we use a moving-coil pickup with a line-contact stylus very carefully aligned, correct vertical tracking force, correctly adjusted anti skating bias, everything-and on preamp A you hear a marrow-piercing scream, and on preamp B you hear a sweet, soaring and thoroughly transparent sound. Is it possible that A is accurate and B is inaccurate?

OTALA: Yes it is. | just said earlier that 90% of our subjects took the distorted sound as being the non-distorted.

EDITOR: But in the sense that I've just de scribed there's distorted and there's distorted. I mean a marrow-piercing, screaming edge. And that's what 9 out of 10 pre amps do under those circumstances. They can't take it.

OTALA: Okay, okay. If it's really that bad, then okay. But if it's slightly less, then there's a possibility.

COTTER: Matti, there is one way of working with a system where you think you may be on one side or the other of a threshold effect, and that is to iterate the system. We talked briefly about this before. If you think you are below threshold, and in fact you are actually above threshold, if you go through several such units, inverting and playing again and inverting and playing again, and you can't hear any difference between n such things strung in series and one or none, then isn't it a fair surmise that you are still sub-threshold?

OTALA: Well, possibly.

EDITOR: That was a pretty heavy sentence there. You're suggesting a whole system of testing there.

OTALA: | would suggest another system of testing, specifically for The Audio Critic.

COTTER: Without knowing what the details are, why-I'm not saying you'll find out why this way-you'll just find out whether or not you're above threshold.

OTALA: One way of testing would be to make a distortion box that could generate different kinds of distortions. and the no-distortion position would be absolutely clean. You would simply try. with the kind of situation you described. introducing more distortion. Now if one gets worse and the other one gets better, then you know which way things go. That. I think. is psychoacoustically the only possible way of detecting distortion.

FUTTERMAN: You might be cancelling distortions.

OTALA: Even so, yes.

EDITOR: You have to know what you start with. And this is one of the problems, be cause you never know exactly what you start with.

OTALA: Well, you can iterate there too.

COTTER: You're saying that you really ought to standardize it. I'm saying that this kind of approach allows you to make a de termination based on a belief that if you've got some process of this sort, then if you iterate you're going to increase the magnitude of what you've got. Now it's hard to imagine processes that are purely multiplicative that are not going to work something like that. It seems to me that it allows you to tell whether or not there's a change, and therefore whether or not you're near or be low the threshold. But it doesn't tell you what the mechanism is. As you aptly mentioned before, that you can synthetically introduce a certain value of distortion where you can get a good correlation of discrimination, but the jury, though consistent, is un able to tell you the character of what it is they're tracking in making the identification. When you're dealing with near threshold phenomena, this is possibly not an unlikely occurrence. I think what Peter is talking about is something that's a more gross difference, and I think it relates some what to this effect that the vertical angle changes induce sharp changes in some cases and much less sharp changes in others. We seem to be able to connect those with trigger-like-call it TIM or time shift-effects that a lot of people have described . . .

EDITOR: This is not TIM in Matti's classic sense.

COTTER: No, but this is a time area thing.

And it is not unusual to find that sort of thing described by observers who have some experience as mistracking. Because it has the sound of a fast, high-frequency mis tracking. We were surprised to find that it was not mistracking at all, but it was electronic in character. I think there remain problems in the speed time modulation processes within preamps that go beyond equalization.

OTALA: Let me also state one thing which is in parallel with your thinking. The DIM test, as it normally goes, only detects amplitude variations, more or less, of the 15-kHz signal.

COTTER: Unless you look at the sidebands we talked about.

OTALA: Yes. We are presently thinking in terms of-if that proves to be feasible- trying out a method where the same test signal setup is used but possible phase modulation effects on the 15 kHz carrier are detected. That should in principle provide a more sensitive measurement method.

COTTER: Do a direct frequency detection, then look at the output of the frequency detector?

OTALA: Yes, with a phase-locked loop.

That would probably be first of all a simpler test, and secondly would relate directly to the time domain effects in such a way that a fast-rising slope of, say, a 3-kHz signal would then phase modulate, in the time do main, the 15-kHz signal component.

COTTER: You have now got a heroic filter problem in your instrumentation. though, to eliminate your signal components from your detector.

OTALA: No. not necessarily. The phase locked loop technique is quite okay for that purpose. That is a heroic filter, by the way.

COTTER: I'm saying it's a Hilbert filter. but it's still a speed problem.

EDITOR: Where do we stand here? Are we more or less squared away on the signal path from stylus tip to power amplifier terminals? Is there anything significant to be added, anything this group would like to contribute to . . .?

COTTER: We said you should pay attention to all of these time domain effects; we didn't tell how. And we all, I think. disdain the classic THD and even some of the TIM measurements, because they're not sensitive to these things.

EDITOR: Let talk for just a little while about this multiple pass test that you suggested. Is this your recommendation in lieu of the straight-wire bypass test with its obvious shortcomings

COTTER: I think that if you think you have something that's sub-threshold. that you're in a much better position to believe that it is sub-threshold if you can put some significant number of those elements in series, and cannot track the difference between n elements and 1 element or no element. I've found very few pieces of equipment that will do that. I've never found, other than some of the experimental things we've built and the new thing we've made, two preamp stages that can be strung in series without distress, without an obvious . . .

EDITOR: Exactly how do you do this? You pad out the first one to reduce the . . .

COTTER: Pad it out, inverse equalize it back again. Very simple. Very straightforward.

We don't use 7.5 microseconds; we use a 35 kHz turnover. Close to the same.

EDITOR: Is it conceivable that inaccuracies introduced in the course of the first pass will mask all further deterioration?

COTTER: You ultimately become limited in your n, in your choice of n, by heroic accuracy problems. Even using the best possible components, and you start to get your net works turned out . . .

OTALA: With the far-out poles.

COTTER: Not only the poles, but the fact is that you've got the accumulating error. If you do five passes, and you're talking about a quarter of a dB being a JND, you suddenly require less than .05 dB in everything. And that's getting hairy.

EDITOR: Let's just take one device, one pass versus two passes.

COTTER: I think one versus two, if you can get it to that stage, you are outside the realm of most everything else.

EDITOR: But we're back to this question that [ asked quite a while ago. Does garbage overlaid with garbage sound different from plain garbage?

HEGEMAN: I think it sounds like more garbage.

OTALA: More garbage is more garbage, and that's it.

EDITOR: But does more garbage sound different from plain garbage?

HEGEMAN: If you have an A-B chance to listen …

EDITOR: I'm not arguing, I'm just asking.

COTTER: Not always. For instance, we know from the work on flutter that the mind reaches a saturation value of flutter at about 1.4% rms at a 3-Hz flutter rate, such that you cannot distinguish as greater or lesser 5%, 2%. 3%. 7%. They are indistinguishable in quantity. in quantitative value. So that there is for the time shift effect, at least in the low speed rate area, a kind of saturation effect.

There’s another set of experiments that deal with flutter rates in the area of 30, 35 Hz modulation rate that also show a kind of saturation effect. If you accept a place theory. and a place theory spread kind of critical band idea, for pitch and flutter side band recognition, then obviously there should be some kind of saturation value for the time shift, more of which is not going to be discerned as more.

EDITOR: In other words, two passes through a really bad device may not sound different from a single pass through that bad device.

COTTER: You've got to go back and ask that other question we asked. Is this sensibly like music, to start with? If Max would walk out of the room. then maybe the comparison isn't valid to start with. But if Max is intrigued by the musical sound-and we're taking you for argument's sake as a sort of paragon of musical judgment-if you would accept it as decently musical, then I think you ought to be concerned with whether it sounds different. If it's horrible to start with, then obviously I think that the com parison . . .

EDITOR: Mitch, why is this preferable to padding the pre-equalized stage out to unity gain and then comparing it against a straight wire?

COTTER: But how do you get a straight wire out of the pickup that you can listen to?

EDITOR: You can't get it straight out of the pickup.

COTTER: I submit that a pickup signal has all kinds of things in it that are very different from what you get out of, say. a tape recorder synthesis.

EDITOR: That's true. There's much more out-of-band energy. for one thing.

COTTER: That's right. I think you have to be able to handle that realistic signal. Having it work well on a tape recorder synthesis doesn't necessarily prove that it's going to be as good or even better than something else. when presented with . . .

RAPPAPORT: I was just going to say that Mitch's test, the iteration test, has one thing in common with the bypass test. That is the limit of resolution. As you say, if you start out with something that's decent. something that's good. and you're able to construct a padding network. or for that matter a straight-wire network that's valid. that has 22 no sound, that has no adverse loading characteristics or anything like that. then the test is valid. However. if you have a situation where you have a gross limit in resolution or you have some kind of distortion that's occurring. then there is the possibility that something's going to be masked. I think it's safe to say that if two trips through a device sound worse than one trip through a device, something is wrong. Which does not necessarily mean that if it doesn't sound different, something is right.

COTTER: I think. though. that with the experience we've had. you don't have to be quite as wary and look for quite as far-out things as you might suspect or you might want to consider. In the abstract it sounds that way but as a practical matter, when we've done these things, most of these systems emerge as grossly faulty where bypass 1 and bypass 2 are . . .

RAPPAPORT: There's a possibility that the gross fault may be due to your interface. It's not that cut-and-dried.

COTTER: We try to control those things, and I just think that in the case of phono preamplifiers. the difference is a lot bigger a difference than has been suspected. When we wind up with an improved system that's been designed this way and we listen to it, the differences are readily apparent differences. and they are in the direction of improved clarity, improved musicality. So I think what we're finding is that we've been looking with our traditional methods at aspects of the preamplification or low-level signal processing that don't consider these time shift effects.

RAPPAPORT: All I'm saying, though. in response to this test and also the straight wire test. because they are very similar, is that if the interface has a fault. if the inter face is audible. and that's taken as being a transparent given, then any test that utilizes that interface is going to be faulty. I think that most of the straight-wire bypass tests that are being performed right now are using a faulty interface. And sure-the straight wire is going to sound different than even a very good component. and a bad component is going to sound the same as the straight wire. It's the same thing with the iteration test to a certain degree. because through one trip. you don't hear the interface. Through two trips, you hear the interface. And if there's a problem in the interface, then two trips are going to sound different than one trip.

EDITOR: There is a difference, though.

COTTER: You have to contrive to make your interface relatively inert.

RAPPAPORT: You have to make sure that it is absolutely inert, not relatively inert.

COTTER: The problem is that nobody can say for certain that a certain amount of reactance change isn't a factor.

RAPPAPORT: That's the problem.

COTTER: I think again we're digging far deeper than you need, because there are certain kinds of things that come to light.

One of those areas is a very great difference in the time modulating properties of systems that pass even an imperfect interface system, as contrasted with systems that just are grossly faulty, where you get a one-two comparison that's just a black and white scene. You don't have to stop and think or do any extended switching.

RAPPAPORT: The point is that those cases are audible without an iteration test or with out a straight wire test.

COTTER: Then you're relying wholly on judgment. This is just a way of keeping you honest.

ZAYDE: What you're saying. you should buffer the straight wire to render interfaces as invisible as possible.

RAPPAPORT: The point is that once you begin to buffer you're assuming the transparency of the buffer.

ZAYDE: That's true. too. You're damned if you do and damned if you don't.

EDITOR: This is the superiority of the multiple pass test. that at least your first interface 1s typical of in-use interfacing.

RAPPAPORT: But that's only valid for one trip. Two trips. you begin to see an abnormal situation. and the iteration test is only of use if you begin with two. three. four. five, SIX trips.

COTTER: If a network that is a very high impedance. and whose reactance is a small part of the kind of load that a preamp is working into. is sufficiently disturbing to affect it. then perhaps there's a problem also. What I'm saying is that I think you're over-killing small problems when in fact if you do this. it turns out there's some grossly faulty . . .

RAPPAPORT: I have one fear. and that is that ultimately I think this discussion is going to appear in print. And your iteration test on the surface seems like a fantastic thing, and . . .

COTTER: Don't fly into it with the feeling that it's very easy to do. One can spend weeks building inverse equalizers that don't affect things.

RAPPAPORT: I've told you before that I've performed this test; in fact, we discussed this a few weeks ago. The idea is that 1 would hate to see a rash of tests like this that involve op-amp buffers and all kinds of things, and people are saying well, three trips sound the same as one trip.

COTTER: Horrors. No question, Andy, a very good point.

EDITOR: As a matter of fact, somebody once sent me a bypass box with an op-amp buffer in it. And, indeed, the device that it was supposed to prove to be transparent was indistinguishable from a straight wire.

COTTER: So was the box.

RAPPAPORT: I don't want you to think that I'm condemning Mitch's test, because in a very real sense I stole it from him. But you have to be very careful.

COTTER: But it's tough. And you work on components, because it carries us back exactly to the idea that Matti introduced, with just the pot. The pot can be a distorter in a very peculiar way. Capacitors are very, very big problems, especially . . .

EDITOR: At this point we're sort of picking at the remains of the subject on the plate. It might be time to pass on to loudspeakers.

Editor's Note: The rest is all about loud speakers, with only minor digressions. Our reasons for not publishing this portion of the transcript were explained above, in our brief introduction to Part II. Thus our two part series ends here, not with a bang but an outtake.

---------

[adapted from TAC]

---------

Also see:

The Audio Critic Seminar on the State of the Art: Part 1

Speaker Wires and Audio Cables: Separating the Sense from the Nonsense

Various audio and high-fidelity magazines