The Audio Critic Seminar on the State of the Art: Part 1, part b (Vol.2, No.1, Winter/Spring 1979)

(cont. from part a)

HEGEMAN: I know it does.

EDITOR: But an explanation for that was offered a while ago, of why it sounds good and why others that are.

OTALA: Let me straighten out one thing first. I've always been called an enemy of feedback-I am not. I've made it clear so far in this meeting already, three times, I believe, that there is an amount of feedback, overall feedback, which can be used in every amplifier and that the amount varies. In your case it might be that you are using just the right amount of feedback, or even less, heaven knows. But let's put it this way-in a given situation, the use of an infinite amount of feedback is as stupid as using no feedback at all.

EDITOR: This is pretty heavy now.

OTALA: Let me continue with another example. Let me explain the typical methodology of listening experiments.

Well, we say, this is a high-feedback design; it apparently sounds bad; so why does it sound bad? Well, I recently discovered a unit which did not produce TIM at all, although it was described as producing lots of audible TIM-like distortion. The effect was very simple. It was namely so, that since the poles of the transfer function just moved up and down with current excitation, so when used with a large amount of feedback, its phase margin was going up and down.

The frequency response varied, depending on the signal level. Therefore it created very much this kind of time effects, phase modulation or time modulation, whatever you wish. But here the important thing is, once again, that effects like TIM, or this phase margin shifting or whatever, are not related to the basic concept of the feedback itself, but a very poor application of the principles. So let me still say once again that your approach probably is okay if you have taken into account all those problems-you have done a piece of good engineering. There's nothing wrong with feedback itself; we can use tremendous amounts of feed back if it's applied correctly. Now correctly ...

EDITOR: Isn't it true that the situation in which you can apply it correctly and fearlessly is the very situation where you hardly need it?

OTALA: That's true, yes, I fully agree.

FUTTERMAN: You're all wet-because in my design, the loads, an 8-ohm speaker load, is so mismatched to the output tubes that it isn't even funny. The output tubes would like to see quite a few hundred ohms and here they're seeing only 8 ohms. So naturally I have a lot of open-loop distortion. So the more feed back I use, the lower the distortion. And as I pointed out, the feedback goes up with the load impedance because the gain of the last stage goes up with the higher impedance. And in fact it keeps going up and up.

RAPPAPORT: I want to say something.

The idea is, and I don't want to be-this is very difficult, because I don't want to insult you by this, I want you to under stand that an earlier version of your amplifier was the first amplifier I ever heard that I liked.

FUTTERMAN: Which one was that?

RAPPAPORT: This was an H-3a or something like that that a friend of mine had five years ago, and it was the first amplifier I ever heard that I liked. And I think you make a fantastic amplifier, but I think your amplifier may well be better than you think it is and better than we realize it is. Because there are various problems occurring from your use of feed-back. Now you get away with it, because your amp has very small delay, due to the fact that you're using vacuum tubes and their transit time is low--you get away with it. However, look at the modulation of feedback. The amount of feedback in your amplifier is greatly determined by the load impedance. Now as the load impedance changes, which it does with a reactive load, your feedback is changing. As the amount of feedback is changing, you're changing the parameters of the circuit as a whole.

FUTTERMAN: Exactly.

RAPPAPORT: And you're creating by that distortions that-I don't think we even realize what they are at this point.

FUTTERMAN: Now wait a minute.

We've come to the ultimate point.

HEGEMAN: Well, the pole at the top end compensation is not swinging around that way. If it's got itself out of that area, then that degree of feedback change is not as significant. As Matti has talked about, these things are kinda on the edge.

ZAYDE: It depends how you see the complex conjugate pair. Okay, so what does compensation do? You're gaining a handle on that, and you're clamping it.

You're permitting a specific aperture.

EDITOR: Bruce, you've done some calculations on this. Why don't you tell us about that? FUTTERMAN: Let's finish this.

RAPPAPORT: The idea is that, I think you're getting away with it-if your amplifier operated in exactly the same way, except built into the amplifier there were a delay of maybe a couple of hundred nanoseconds, something like that, instead of the few nanoseconds that it actually is, or a microsecond, or some thing like that, and you had the same kind of effect happening with feedback determined by load impedance, also ad ding of course the regenerative effects that I was discussing before, you would have a tremendous amount of trouble.

You get away with it and your amplifier sounds excellent because the delay you're beginning with is very small, so the modulation is unimportant. I wonder exactly how much better the amplifier could be, or if in fact it could be better, if it would work without feedback. As I said, it might be better than we think it is.

FUTTERMAN: It's impossible because of the mismatch I have in the output tubes.

RAPPAPORT: It's a practical impossibility.

COTTER: When you say mismatch, Julius, what you've got is a vacuum-tube type structure which is basically a cur rent source. And it's configured in such a way-in fact your whole patent is based upon maintaining its current-source qualities by having essentially screen grid rive. So that what you're dealing with is a situation that's not altogether unlike what we have in a collector output terminated power transistor amplifier. For instance, if you have complementary power transistors, and you take the output from the collector instead of the usual emitter-follower totem-pole type thing, then you have essentially a current source system. And again, let me bring this back to see what goes on in an amplifier by realizing that once you get inside the amplifier it doesn't know and it doesn't care whether it's got a feedback loop around it or not. It is unmindful of the fact that there's feedback. It's simply handling an error signal, which I think is Andy's point. It's handling an error signal, it's delivering the output. It's a current source. Now how is it going to deal with this reflected energy that comes back? That's really what the problem is.

FUTTERMAN: Well, isn't the proof in the listening, after all?

COTTER: Well, maybe, but we're trying to go a little further and understand it.

I'm saying when you have a current source and you're trying to make it damp out the energy that's coming back, you have an interesting problem. The better the current source is, the tougher it is to cope with that reflected power. Because it can't really, in a sense, absorb it. If you had an infinite current source, you'd be very hard put to deal with that reflected energy. In fact, what you would wind up doing is sending it back out toward the loudspeaker. Quite apart from the fact that there's feedback or no feedback.

OTALA: But remember, the thing that saves us here is that it doesn't matter how much energy you put backwards into your tube-it won't, unless it starts arcing, do very much harm there. So you can tolerate that.

“... I think there has been a relatively mindless pursuit of the bottom and the top of the meter scale . . .”

COTTER: But the fact is that you are dealing with this elliptical or circular load line where you've got current and voltage out of phase. What I'm saying is, that if you look at where the dissipation is taking place, if you really had an in finite impedance output stage, then this stage is actually incapable of absorbing power. And it will not, no matter how many zillion dB of feedback I have, it will not absorb or damp the load. Now one conceivable way to improve the situation would be to take your zillion dB-feedback infinite-current-source amplifier and hang a 750-ohm load resistor like the 300B across it, and it makes it into a magnificent amplifier.

FUTTERMAN: I'm still working.

HEGEMAN: Incidentally, just for the record-Mitch and I have been talking about the historic triode amplifiers that Western Electric used to make-the distortion measurements on that were about 40 dB on second harmonic; that's 1%, more or less, over the band. The third harmonic ran down around 37 dB, which is, what, 12% and so forth, and these things sounded so good.

FUTTERMAN: In Japan they're building them.

EDITOR: The subject of this seminar is, what is the State of the Art? So in amplifier design, what is the State of the Art? Can you agree on this?

HEGEMAN: We found that 25 years ago, Pete.

EDITOR: Could you define it as an amplifier in which these time modulation effects are reduced to an absolute minimum? Would that be a good way of defining it?

COTTER: Well, absolute minimum, we don't really know at this point how low you have to go to be inaudible.

EDITOR: Bruce, why don't you tell us about these calculations that you made?

ZAYDE: Basically, the similarity in the transfer function behavior or profile of the feedback-shall we say appearance-which is basically an expanded polynomial, to that of the filter theory suggests that there are methods in which we can get a handle on what the aperture is that one can expect to yield relatively low time-dispersal problems in the feedback approach. And it's closely related to such things we discussed, which is the transit times of the devices, etc. Because, to a large extent, when all this is collected together, it describes where the complex conjugate pair of poles are going to be located on a unit circle that describes this whole phenomenon. So by operating with compensation and the rest, we find that with given realistic devices surprisingly little feedback is tolerable using current solid state devices of typical transit time proportions; but in such cases as Julius's amplifier we're dealing with enormously short transit times . . .

COTTER: And very little time modulation.

ZAYDE: And very little time modulation, right, as a by-product, you can instill much, much larger amounts of feedback and still be within the correct aperture.

COTTER: It's sort of tragic in a way, isn't it, though, that-can we agree that vacuum tubes, which were so difficult to use-Julius got rid of the transformers and removed one of the great difficulties and so made a great deal of feedback possible-that in the solid state devices, which make feedback so very easy to use with this wanton abandon with which it is applied, they are the least tolerant of that kind of scheme. So one is driven to do things in pursuit of these traditional distortion and frequency response numbers that in effect carry you further and further into a state of problem. Because what we're evaluating isn't what we're hearing. And we're kind of agreed that time modulation effects are the trouble some and the sound-contributing aspects.

OTALA: I don't know whether we have agreed.

COTTER: Well, have we agreed that it certainly isn't the fact that one has 0.01% distortion and another one has 0.001% distortion?

OTALA: Yes. we have basically agreed that there are time modulation effects and that they are important. And TIM is one of that kind of effects in fact, so I support you wholeheartedly. However, I would not mix a distortion mechanism, which produces something, with the result of an engineering operation sequence called measurement method, which yields us a number. Because let's take any measurement method that we've got now-they are purely engineering methods and they yield numbers which may have no relevance whatsoever in this world, referred to either the distortion mechanisms or the psychoacoustic result. Therefore I don't subscribe to your saying, well, 1% is not audible in that and that respect and 0.00--something is not audible in the other one . . .

COTTER: I don't know where the borders lie, but they certainly . ..

OTALA: Yeah, but what you are talking about has nothing to do with amplifier design. It's just that you're applying a measurement method which has different sensitivities; and if it has nil sensitivity for a given phenomenon you can't still say that . ..

COTTER: But one of the problems I'm talking about is that I think there has been a relatively mindless pursuit of the bottom and the top of the meter scale, which are engineering techniques, in the belief, in the rather cultist or mythological belief, that somehow or other virtue lies in those particular di rections -0.01 dB of flatness and 0.0001% of distortion.

EDITOR: Wouldn't you even go further? If you are shown an amplifier and you are told that it has 0.0008% distortion, and that's all you're told about it, your red flag go up immediately?

OTALA: Not necessarily. I would ask who has designed this amplifier, and after that I would say, hey, it's probably okay. Or there's probably something wrong with it.

RAPPAPORT: You can have an amplifier, depending on what the amplifier's designed to do, with 0.0008% harmonic distortion with no feedback and no time modulation effects, either.

EDITOR: Has that been done?

RAPPAPORT: Yes, it has been done.

COTTER: I don't think most people are aware that it has been done or that it can be done.

EDITOR: Let's talk about it.

HEGEMAN: All you need is one of these hi-fi freaks coming in, “Look at this magazine! Isn't that the greatest thing you ever heard? Look at what that number is!” And it sounds like crap.

EDITOR: Let me structure the question a different way. I think Mitch is among those who has heard me phrase a question in this manner. If you were told by a tyrant that in six weeks you had to come up with a better amplifier than anyone has designed so far, or else be shot, what avenue of approach would you take, beginning tonight?

HEGEMAN: Get ready to get shot.

FUTTERMAN: I would make my amplifier, and say to the speaker manufacturers, give me at least 32-ohm voice coils or 32-ohm transformers in the case of electrostatics.

EDITOR: So you feel that the Futterman circuit with a high-impedance load is the ideal amplifier.

FUTTERMAN: Right.

EDITOR: Okay, that's one opinion. Let's have some more.

COTTER: Go around the table. Bruce?

ZAYDE: Basically I think, to recap and extend, that to get a retention away from static conditions and pay attention to such aspects which are very real as this convolution, which involves such things as h of tau convoluted with g of tau-which I don't know if I should go into, what that is...

EDITOR: Don't use dirty words around this table, please.

ZAYDE: Okay. But to pay attention to the dynamic condition almost exclusively puts us in the right direction.

COTTER: How would you do it, is what Peter's saying.

ZAYDE: How would I do it. Well, I would basically analyze, starting from the devices that I'm using, which I would optimize on their own without using any other techniques and extend from that point forward.

COTTER: Are you saying you'd take devices that had low inherent time delay?

ZAYDE: Exactly.

COTTER: Would you be inclined to use solid-state devices or vacuum tubes? Given that you had to run, you had to . . .

ZAYDE: That’s a leading question, well not leading, but dangerous. I'm familiar with solid-state devices; I would use solid-state.

COTTER: If you had to play it safe, I think is what Peter is saying. You'd use solid-state?

ZAYDE: If I had to play it safe and run really quick I'd use vacuum tubes.

EDITOR: What about MOSFET's? I tried to ask that before but nobody paid attention. Does anybody have any opinions here on MOSFETs?

FUTTERMAN: I worked with them; I worked with the siliconic MOSFETs.

Frankly, they went bad on me so quick and they cost so much money.

EDITOR: Don't they have very low transit time?

FUTTERMAN: I had to drop it. They're not high-impedance devices, no matter what people say-for DC maybe, not for AC.

RAPPAPORT: The point is, there is a way to get any device to work properly.

Bruce's equations and his work show that it doesn't matter if the device has very long or very short transit times. I use bipolars because I'm very comfortable with them and I found a way to get them to work satisfactorily. Julius uses tubes because he's very comfortable with those and he's found a way to make those work very satisfactorily. And I don't think that by digging into the use of MOSFET's or V-FET's or whatever you had, you're going to make, just by using different devices, 4 substantial improvement. Now they may allow you to build a circuit or a topology that has great advantages over what we currently have, but the devices aren't the answer.

FUTTERMAN: Listen, if I were to try to do it again-and you mentioned in a review that tubes are going to disappear-I have some AT&T shares and I got their quarterly newsletter and it says,

“Solid-State Breakthrough. Bell Lab oratories has devised a way to almost double the velocity at which electrons, subatomic particles, can race through solid-state circuitry, a development it calls a major advance in solid-state technology. A team of Bell scientists developed an impurity-free highway system for electron particles, a unique approach that could lead to even tinier microcircuits with higher speed and more capacity than any such circuitry available today.” I'd get to Bell Labs and get some circuits.

COTTER: I think what they're talking about is gallium arsenide Schottky barrier FET's. They exactly double the diffusion velocity, and that's a fancy way of saying something we know already.

But that's for shareholders and not for scientists and engineers. What Andy's saying, in a sense, is that it's a difference in perspective. You would look at the devices differently, look at their operating conditions differently. Are you saying, to put words in your mouth, you would optimize and minimize the time delay effects? The time delay modulation effects?

RAPPAPORT: Exactly. I said I use bipolars because I'm most comfortable with them-that doesn't mean I use them the way they've always been used. The idea is that in order to use them you have to look at exactly what they do, exactly how they work in a circuit and exactly how you can minimize the time modulation effects.

COTTER: Suppose it was possible to make an amplifier with a device that was some kind of real slow, electrolytic, slow poke kind of thing, and you put in a signal and it came out half a second later.

As long as it came out precisely, exactly half a second later . ..

HEGEMAN: Your internal total time delay makes no difference.

ZAYDE: Right. There [unintelligible] your independent variable, which is in a sense the device that you're using.

RAPPAPORT: That's right. A record or a tape is a time delay device.

COTTER: If we had a half a second delay, however, I think Matti and every body would agree that the use of feed back would be highly unlikely-it gives you nothing but trouble.

ZAYDE: Oh no, no, you don't want to do that.

RAPPAPORT: In terms of delay, a very gross example of the problems of time delay in a feedback network would be that, if Max is recording in Studio A at RCA, and we have this incredible feed back network that stretches all the way to this house, whereby any error that's created by the recording or playback process is miraculously picked up here and transmitted back to Studio A, where it is transformed into an error signal which Max takes into account when he is recording this-and ideally we have absolutely no error from microphone to loudspeaker-but Peter decides to play this record a month after Max has recorded it, and by the time the signal gets back to Studio A at RCA Max is recording something totally different.

We end up with much greater amounts of distortion than we originally bargained for-which is exactly what occurs in these feedback amplifiers.

FUTTERMAN: I think you're wrong.

ZAYDE: Regardless of whatever you were recording, Max, you wind up now with Charles Ives.

COTTER: I object to that on musical and logical grounds.

OTALA: I would like to object, too, because that was too gross an example.

RAPPAPORT: I said it was gross, but to a certain extent it will and does happen in these amplifiers.

EDITOR: Doesn't all this come down to the fact that it has to be quantified? We cannot discuss feedback qualitatively; we have to discuss it quantitatively. Doesn't it come down to that?

HEGEMAN: I don't think so, because I think the quantitative analysis depends exactly on the circuit's topology, the particular apparatus you're using, and there's no universal solution. There's just some guidelines as to how you can work and what you should be doing.

OTALA: Let me put it this way. Delay is a funny problem because it doesn't roll off the amplitude 6 dB per octave, which is supposed to be the case in a feed back amplifier. Therefore, you cannot go beyond the case of having a, say, mini mum 60-degree phase margin. You can apply feedback to that point. Taking that as an engineering rule, as the utmost limit where you can go, then in that case I'm not concerned so much about the delay itself, because when the delay gets larger, that automatically limits the amount of feedback you can use, unless you change compensation. The only problem is that the worst-case situation comes when you have a system which has initially a small delay, say a couple of hundred nanoseconds which is typical in audio amplifiers, and then has a large delay time variation with signal.

COTTER: Maybe another couple of hundred nanoseconds.

OTALA: Yes, wobbling up and down.

This is a dangerous situation, and that's easily verified. Have you seen, for in stance, an amplifier when you just juggle with it a little bit; put some kind of capacitor at the output, and you see it starting to oscillate at a high frequency during one portion of the waveform.

You feed in a low frequency signal and it starts oscillating just there somewhere. At that very point the phase mar gin went to zero.

COTTER: And recovered.

OTALA: And recovered. That is the most dangerous thing you ever can have.

HEGEMAN: Try that same amplifier that has that little squiggle on the sine wave, drive that to clipping and see what happens. It'll go absolutely wild.

OTALA: Oh yeah, that's another case.

What I wanted to say is-the delay. I think you said it already once before.

Perhaps you also, I don't recall who said it. It has nothing to do with the delay itself, and the only important thing as the end result is the variation. In both cases, the application of feedback cannot take any kind of extreme proportions because stability considerations will anyway limit the amount of feedback to an amount suitable in that particular application.

COTTER: You don't mean to say though, Matti, that just because an amplifier is stable the feedback that is applied is okay.

OTALA: No, I don't say that. You must remember my qualifications. First of all, the qualifications are that all the signals injected into the amplifier cause it to operate in the static domain. This means not necessarily only audio signals but al so ultrasonic signals that we get from records and pickups, you know, say typically up to 100 kHz or so. There will be no TIM or rate-of-change type of effects, any of those, in that region. Secondly, that the feedback applied is reasonable; and the reasonability is set by other criteria, and some of them are interface, some are stability criteria--I mean stability both as far as high-frequency stability is concerned as well as, say, gain stability or low-frequency stability, depending on the situation. Third, the important thing is to take into account also in feedback considerations-all the other things like interface inter-modulations and dynamic load variations, which we know. I've measured negative momentary impedances from loudspeakers: I can show you some graphs of that. So the amplifier, the whole concept, must cope with those loads, too. That again tends toward increasing the amount of feedback, not taking it completely off. Finally, slew rate is normally, in normal everyday amplifiers, caused by feedback. So if you get a little less feedback you would probably increase the slew rate. But again here, the importance would not be to consider slew rate as volts per micro second, because the important thing seems to be amperes per microsecond. Those are the important things.

" ... there are some kinds of benefit to be obtained from feedback, but I feel that with the sorts of systems that are involved, it's too dangerous to apply broadly.”

EDITOR: You're saying it's a balancing act in each circuit.

OTALA: It's a balancing act, yes.

FUTTERMAN: I wanted to say some thing in reply to Andy. I've lost the train of thought, I'll try to recollect it. He said if the delay was very, very long and it would be fed back to Max--was that it?--it wouldn't be good, or words to that effect. Well, just to be a little facetious, by that time it would be positive feed back, and Max would make a better record because of it.

RAPPAPORT: In one sense that's true, but you’re talking about what Peter would say about the record that Max had made and quite possibly it would then become positive feedback. That was just a gross example, and then you can shrink the dimensions to begin to understand exactly what's happening in the amplifier.

FUTTERMAN: I just want to continue.

Feedback is not only used of course in amplifiers; we have servo mechanisms, and all kinds of things. And our body has negative feedback in various ways, and it's a very useful tool if it's used properly.

HEGEMAN: I use it when I go to an audio show. I just turn my ears way the hell down, put a lot of negative feedback in, and survive,

RAPPAPORT: We all use feedback every day of the week. We have philosophical feedback, psychological feedback, physical feedback, in just everyday living. It has nothing to do with audio or electronics.

ZAYDE: We're not trying to match the ear, though. Like in airplanes, they use a balance control feedback for the control panels, but it's a very different type of sensory mechanism you're appealing to.

EDITOR: We have an interesting situation here, because I know for a fact that both Andy and Mitch are designing circuits without feedback; Julius is designing circuits with lots of feedback, and they sound good; Matti has designed some pretty good-sounding amplifiers and he says it's a balancing act; and I think Bruce is also, at least from his mathematician's point of view, saying it's a balancing act; and I think Stew, you're also saying essentially that it's a balancing act.

HEGEMAN: It is. It's a function of the hardware as to what you can do with the circuit.

EDITOR: So we have two representatives here of the theory of “Why use feedback if you can avoid it?'-or is that unfair?

ZAYDE: I believe in the balancing act, too.

COTTER: I would qualify that, I really believe in the balancing act and I think there are two dimensions of my concern that really don't relate to feedback. And I'm completely in accord with what Matti has said. The difference is that I think in certain situations where you do not have control over the entire system, the interface with the output, that your balancing act leads you to some very narrow possibilities, where the amount of gain of benefit to be obtained from the use of the feedback is marginally significant. And I think that in that case I would abandon it. I think my limitations come about because I choose, I opt to use solid-state devices where the delays are significantly large and the delay modulation effects are significantly small. In that sort of system, if you calculate it out and you find out that your optimal design is 82 dB of feedback, I don't think there's a whole hell of a lot of advantage.

If it gets up to 20, you then have to concern yourself with the interface problems that Matti has talked about, and that too becomes dangerous. So what I chose to do was to take a rather different approach, since I don't know at this point in time just how much of what kinds of time-delay modulation effects produce how much of what kinds of disturbances.

I tried to find something that was an “overwhelming” of the problem. And in doing that, I decided I was safer; without knowing these properties of the hearing mechanism, I was safer. I relied in order to prove my case on the iteration method to test whether or not I had a transparent channel. The iteration method is very simply to string up in series a number of the things that I believe to be transparent to see whether or not it's a verifiable or discernible difference between none and n strung in series, believing that with the best-resolving loudspeaker systems, and the best of demanding program material, that if I heard no difference--not an identifiable or consistent difference--then I had something that was at least noncontributory. I think that there are some kinds of benefits to be obtained from feedback, but I feel that with the sorts of systems that are involved, it's too dangerous to apply broadly. If I had complete control over a situation, I might be moved to use some, again as a balancing act. But I think the numbers that I see as viable within the constraints of present kinds of systems are very, very much less than what are in common use. I think Matti could agree that to him 26 dB of feedback is a healthy number, and not 60 . ..

OTALA: Something between, say, 15 and 26, or 16 and 26, something like that.

EDITOR: You're talking about power amplifiers?

OTALA: Power amplifiers, yes.

COTTER: Certainly numbers like 40, 50 and 60 dB of feedback on solid-state devices are dangerous to the extreme.

HEGEMAN: Get lost. You can't do it.

FUTTERMAN: Mitch, would you use local feedback, instead of overall feed back?

COTTER: Well, however you apportion it, I think Matti made the point that you've got a finite gain bandwidth product to deal with and you can apportion, but you still wind up with a very serious constraint. Now Julius, I think your advantage comes from being able to apply the feedback technique with considerably greater ease because your delay and your delay modulation process in the vacuum tube system are vastly lower, and you are therefore able to execute forms of solution, kinds of topology that I don't think work in a solid-state system simply because what you need in order to make them work is far more feedback than the delay modulation will permit. Now Stew has designed in both camps, and has a feeling for this too. Would you agree with that kind of a generalization?

HEGEMAN: Yeah, I do think that anything over about 26 dB feedback is really on the edge, and certainly, beyond the edge, or all the way on the edge, if you have an output transformer . ..

COTTER: You're saying solid-state or…

FUTTERMAN: Is not an emitter follower 100% feedback?

COTTER: No. Not at all.

FUTTERMAN: Not at all? Why?

HEGEMAN: It's not voltage feedback.

COTTER: An emitter follower and a cathode follower are somewhat similar and the fact is there's always a gain of less than 1, and the equations contain this (1-a) term throughout.

FUTTERMAN: Wait. The output of the cathode follower is subtracted from the input, and the difference . . .

COTTER: There are very few cathode followers where your feedback approaches 40 dB and it's a vacuum tube.

You have to have a mu greater than 100 incrementally. And there are very few emitter followers where your feedback is any more than that kind of a number. . .

FUTTERMAN: What number was that?

COTTER: 100.

FUTTERMAN: 100 dB?

COTTER: No, where you're talking about a 40 dB type ratio. As a matter of fact, one of the things that's a characteristic problem about emitter followers and cathode followers is that the minute you introduce any reactive loads, you get into some very serious problems because then the output following becomes-you expand and you modulate the difference term. Because in the negative stroke on a cathode follower, or in the negative stroke on an n-p-n emitter follower, the bass disconnects from the output when the rate gets greater than what you can pull down the output reactance with.

OTALA: My concern about emitter followers is somewhat more directed towards the frequency or rate-of-change effects. Firstly, remember that the emitter capacitance, internal emitter capacitance, in an emitter follower results as an inductance at the output.

Now this inductance being current dependent, we get usually the problem of having a phase modulation in the emitter follower. Having a phase modulation with the overall current. This doesn't of course apply to a single sinusoid but it applies to multiple signals, where the low frequency components will phase modulate the high frequency components. That is one thing. The second thing is that an emitter follower having an inductive base impedance and capacitive loading at the output, as is usually the case due to strays and loading effects, is a Colpitts oscillator, and therefore . . .

HEGEMAN: You get a very nice negative resistance on them without too much trouble.

OTALA: Yes, and that means it's oscillatory. Now, it is not that dangerous that it is oscillatory by itself, because anyway you're going to suppress the oscillation.

However, it exhibits a phase margin which is less than 90 degrees at high frequencies, and that phase margin is cur rent dependent. Therefore again you get a set of complex conjugate poles, and these poles are sweeping up and down like crazy rabbits when you . ..

COTTER: Signal dependent.

FUTTERMAN: And the emitter base capacitance varies with current. It's non linear.

COTTER: Yup. And what we're saying is even in a cathode follower case where the transit time is small, you can still get significant reactive effects. This has been known, was pointed out, admirably, in the MIT/Rad Lab series with respect to pulse amplifiers. You tend to get this asymmetrical charge and discharge characteristic with a cathode follower.

HEGEMAN: Besides which, they're non linear as hell.

COTTER: Maybe.

FUTTERMAN: It's a 100% feedback device. All the output is fed back to the input. In a cathode follower.

COTTER: Except for a very important difference, and that is that the transit time is very small; there is not a very large capacitance modulation . . .

FUTTERMAN: In the tube.

COTTER: In the tube. When you get into the higher frequency domain, transit time effects load the input with a real part dependent on the frequency squared, and you've got a whole order of magnitude, many orders of magnitude, difference in the range of time in which time dependent effects enter. So the realizability of that local 100% feedback -100% means all the available out put being fed back to the input-but the gain is still less than 1 intrinsically, which 1s what makes that thing stable. Matti points out that under certain conditions, like the Colpitts condition, if the associated strays exist you can get a gain greater than 1. And it does actually happen even with cathode followers that you get into instability situations. I would never use a cathode follower as an output circuit for an audio amplifier, where I didn't know what cable, what system I was dealing with or, if I did, I'd put a thousand-ohm resistor in the series with the output or something to protect it.

HEGEMAN: You noticed that, huh?

OTALA: But if you use an emitter follower in an output stage, there are ways to get rid of these things, or at least to decrease them to an acceptable value.

They will be there, but there are ways.

RAPPAPORT: That's one of the reasons that a class A output works, because the changes in current are diminished with a Class A output emitter follower, and also the heat developed by the output stage decreases the capacitance and decreases the reactive effects of the emitter follower circuit to a certain degree.

OTALA: That's one thing. Also with certain circuit tricks you can always say, let's have a situation so that the emitter follower sees a capacitive generator impedance. And that's the most important single thing to make it stable. Further more, of course, if you really want to be clever, then you try to stabilize the collector-base capacitance variation effect.

But all right, there's a lot of small minor things like that. I seem to be thinking nowadays that's the only output stage that you can really conveniently use be cause all the other circuits have even more problems. I don't know whether you agree.

COTTER: To summarize what we're saying, really, there's no basic disagreement between what Julius is saying and what we're saying. Simply that Julius did not choose to try his system with a solid-state output device.

FUTTERMAN: I did.

COTTER: You did and it didn't work, you said; it kept blowing up. And that therefore he uses the vacuum tube system and then is able to use a lot of feedback.

Whereas if we stick to solid-state devices then we are facing more time delay, and what we're all saying is it comes down to the same thing. I don't know whether Julius sat down and set out to make low time delay, low time variation as an objective.

FUTTERMAN: No, I didn't know any thing about it.

COTTER: So we're talking about a different understanding of how to assess what we've done. And we're coming to an interesting conclusion, which is that the standard specifications do not reflect this.

EDITOR: Yes, I was about to say that. We need a spec.

COTTER: We need a spec, we need an approach which assesses this time modulation process in a way that relates somehow or other to what we hear, and in fact it's sort of a strange and un defined realm.

RAPPAPORT: Just for the record, just so that it's clear that I don't agree with Mitch or with Matti, I want to express a different opinion towards feedback. And it's really very simplistic. And that is that in a solid-state design, which is really all that I've been concentrating on-and it's important that it be understood that my thinking comes primarily from solid state work--that feedback gives you nothing. Although you can set up a series of parameters by which it is a balancing act, as you say, and you can say--and I've talked to Bruce about this--that you can look at certain parameters of the amplifier and you can determine that amount of feedback which will not cause any kind of regeneration or distortion caused by regeneration that is above some audible threshold, I don't see that feedback gives you anything. I don't see that it gives you anything that you can't obtain without feedback, that I can linearize a circuit-I think that any good engineer can linearize a circuit to the point where feedback is made unnecessary.

EDITOR: Your statement assumes knowledge of the audible thresholds of ordinary harmonic and IM distortions.

RAPPAPORT: Yes.

EDITOR: You're saying you know those thresholds . . .

RAPPAPORT: No. I'm not saying I know those thresholds yet. Obviously I'm trying to reduce all distortion to as minimal a value as possible. However, I'm not uncomfortable if, when stressed, a circuit of mine exhibits half a percent harmonic distortion.

COTTER: I agree. And in fact I didn't differ with you.

EDITOR: I basically-in fact I've said so in print-that I basically agree. I just wish somebody could tell me what those thresholds are, because let us say, let us just assume that a quarter percent sounds a lot better than half a percent. Then let's have it.

COTTER: I think I mentioned a rather definitive but a rather old paper by the Feldtkeller/ Technische Hochschule bunch, Mr. Gassler, who did a series of studies that go a long way toward telling us that some very much larger numbers than we currently bandy about border on inaudibility. And I really agree with Andy because I said earlier that actually my feeling was that the kind of benefit that you could get, and the magnitude of the improvement-the magnitude of the amount of feedback that you could use when you defined all of this-was so trivial that I didn't think it was worth the candle.

RAPPAPORT: It doesn't give you any thing, plus the risk of using feedback is incredibly high.

. . . we've all come to the conclusion that it's in the time-domain department, where there don't exist criteria, that the problems arise.”

COTTER: Yeah, well, the risk part I emphasized heavily. And I feel that we do agree, and there's danger in feed back.

HEGEMAN: I would like to cast a vote for feedback, if I might. I think that any good engineer can take a given batch of hardware-transistors, resistors and so forth-and he can make a circuit work, and he can probably pinpoint that up to a maximum type of thing. Now the problem is, fellas, we're going to make hundreds of these things. I can assure you that the advantage of a small amount of feedback is gonna make these things come off the line and off the test bench and off the line, a hell of a lot easier than if you have to sit and adjust each one right on the nose before it goes out.

RAPPAPORT: That's really not the case, because a lot of the problem . . .

HEGEMAN: Andy, it is the case, I'm SORRY.

RAPPAPORT: No, because one of the things ... One of the things that I've found--and I've begun to produce products without feedback in production, although limited production, quantities--is that the feedback masks a lot of effects that I think otherwise would be

... If we use feedback to mask these effects I think we would not be getting anywhere, because I find that in my designs, which don't use feedback, I have to very carefully match transistors; I have to hand-bias everything; it's a very ticklish process.

HEGEMAN: That's what I'm saying.

RAPPAPORT: Now, the point is that in doing this, I'm coming up with a circuit that is far more linear than it would be if I took transistors out of a batch and threw them into a circuit. Now my question is, if we use feedback we're not noticing the effects that this transistor matching and this very, very careful biasing and control of various parameters have-we're not noticing these in a feed back circuit. Or we're not caring about them. Without the crutch of feedback we have to look at these. If we looked at these in a feedback circuit, would it make a difference? I think it would. Even though we've got 8, 10, 20, 40 dB of feed back. If you linearize the open loop further, if you match transistors, if you carefully bias-as I said, if you control all the parameters very carefully in production-you may end up with an even better amplifier. In which case, feedback still doesn't give you anything.

COTTER: Are you talking about power amplifiers?

RAPPAPORT: Power amps, preamps . . .

COTTER: Because most everything I've been talking about-I'm treating the power amp, the severity of the power amps .

EDITOR: We'll talk about preamps after lunch.

OTALA: Let me inject one thing. I don't see, really, such strong points; I said it earlier too. We can use local feedback, and that means feedback in one form or another-feedback it is, in fact, if you decrease the stage gain, that you're introducing, even if you would decrease the collector load-so why does the gain of a grounded emitter stage decrease? Simply because we've got the re which gives you local feedback. Feedback is the basic mechanism of gain adjustment in our cases. Now, what makes the difference between local feedback around one stage, a short feedback system which encompasses, say, two stages or three stages--what's then the difference between that and a big amplifier with overall feedback around ten, fifteen stages? What is the division between nested loops inside the amplifier versus overall or local?

COTTER: There is a difference.

OTALA: There is a difference, yes. I'm just advocating that there is a difference but it is not that dramatic. I would like to say that overall feedback of course has all the known pitfalls; but even that can be used to some extent, and anyway we're using local feedback. You say you don't use feedback at all; well, you use a lot of feedback in local stages. All right, this is a gliding scale.

COTTER: I would use feedback in increasing measure as I went down in power level to a certain point where then I would have to start decreasing the use of feedback. This is very much dependent on the properties of solid-state devices.

The maximum feedback . . .

RAPPAPORT: As you say, you can't control the gain of the stage without feedback. Even if there are no feedback components, the re as you say is a feed back component. And there's absolutely no way around that. However you're decreasing and diminishing delay in a local stage, and also you're eliminating feedback around an interface.

OTALA: Right, that's true.

COTTER: Your maximum error is a single quadrature, too.

OTALA: Right. I seem to prefer much local, some two-stage type of feedback, especially nested loops, and with just a slight overall portion . . .

COTTER: Which is what I think we're all saying.

OTALA: That's a balancing effect, isn't it?

EDITOR: We'll listen to that one very carefully, Matti.

FUTTERMAN: I'd like to point out that in my amplifier design there are only two stages. And then I apply overall feed back, so it's a very simple device. That's why I think I can get away with it.

EDITOR: I don't think anyone here is in total disagreement with anyone else.

COTTER: No, I think it's interesting, though, that we've all come to the conclusion that it's in the time domain department, where there don't exist criteria, that the problems arise. And we all fear feedback rather than respect it.

EDITOR: Do we all agree on the need for a standard, or a spec at least, in that area?

RAPPAPORT: We first have to figure out what we're trying to accomplish and then we can settle on specs.

EDITOR: We've almost concluded power amplifiers. Is there anyone here who feels that something important needs to be said on power amplifiers?

HEGEMAN: I think we beat it to death.

EDITOR: A lot of these things we've said are applicable-certainly the feedback considerations are applicable to pre amplifiers. Let's talk about pre amplifiers for a little while. Who wants to segue from amplifiers to preamps?

COTTER: I'd like to start in a different place altogether, because a preamplifier has to do a certain kind of job.

EDITOR: All right, but in that case we should first talk about loudspeakers.

COTTER: I don't agree at all Loudspeakers are the last thing in the system, obviously.

HEGEMAN: That's for 3 AM, Peter.

EDITOR: That's not the scheme that we picked, but I'm perfectly agreeable to take the course that's developing here, rather than anything preconceived.

COTTER: Let me bend the group out of shape a little bit then. Because I have a certain perspective. My perspective is that, in order to make a preamplifier, one must first find out what happens in the phonograph pickup, or at least what happens to start with, let's say at the stylus. If you understand that, then I think you're in a position to at least define better some of what has to be done with the preamplifier.

EDITOR: How does the rest of the group feel about that road map?

HEGEMAN: I thought maybe we could work up into that gently. Consider that the preamp, that we know what's going into the preamp, and handle it from there to get it to the power amp.

RAPPAPORT: The point is, I don't know if we do know what's going into a preamp.

COTTER: That's my point.

OTALA: Well, any way you wish, this is a round egg, you can view it from any angle.

EDITOR: That's true. Well, let's do it that way, whatever we're warmed up to.

HEGEMAN: Well, for instance I'm not so sure that the stylus is the only thing;

how about a tape head? Or something like that?

COTTER: The big medium that we all look at is a record.

HEGEMAN: I gave them up years ago.

FUTTERMAN: For smoking?

COTTER: As I recall, you had a lot to do with making some pretty slick records.

HEGEMAN: But they were done on tape, Mitch.

EDITOR: He doesn't listen to them in the vinyl form.

HEGEMAN: Well, that's not true. But I made them on tape.

EDITOR: I do know that he uses a lousy pickup.

COTTER: Most of the world has its music delivered in the form of a vinyl plate, and we ought to talk, I think, about that problem.

HEGEMAN: Are we talking about the State of the Art?

EDITOR: That's obviously what we're going to talk about.

COTTER: I'll talk about it later. Go talk about preamps. As a matter of fact, wait a minute, I've changed my mind. Let's talk about preamps. I'm interested in what comes out of the discussion as an approach to preamps, and then we'll talk about what goes on at the pickup and we'll see whether they track each other.

EDITOR: Let Stew launch that, since he…

HEGEMAN …Opened my big mouth, did I?

EDITOR: You broached the subject.

Preamps.

HEGEMAN: It's a very simple concept.

You have a signal coming in, you want to get a signal out, you put some gain in be tween that, and in our strange way, either off tape or off discs, we have to equalize that to compensate for prerecorded equalization that's in there.

OTALA: And then we all come up with one conclusion. In a miraculous way, it's all screwed up; I mean the sound.

HEGEMAN: Totally.

FUTTERMAN: I don't know; they sound pretty good to me.

EDITOR: You're designing one yourself, aren't you?

FUTTERMAN: Yeah. In my spare time.

HEGEMAN: He never has any spare time, if I know Julius.

EDITOR: Because he's always late on the delivery of the latest handmade power amps.

FUTTERMAN: And I'm improving it.

EDITOR: Yes. Well, do we all agree that not all preamps sound terrific?

HEGEMAN: Very few.

FUTTERMAN: I didn't know they had a sound.

EDITOR: Well, what might be some of the reasons?

COTTER: It was theorized, and loudly declared to be the case by some people, that the only difference between preamps were equalization errors. And that all the differences between preamps disappeared when they corrected these equalization errors. And I believe them.

FUTTERMAN: You do

OTTER: I believe the experiments that they performed probably got that kind of result. I happen to disagree with the conclusion and I think I know why.

EDITOR: You mean you don't dispute that they heard what they heard.

COTTER: No, I don't dispute that they heard what they heard, but I do think that I understand why they heard what they heard. It has to do with anomalies in the approach to the definition of the problem. Their conclusion was that frequency response differences was all that they heard that was left. The basic problem I think . ..

HEGEMAN: Possibly.

COTTER: Well no, this was reported, as you know, by in fact a fair number of people who all tended to corroborate.

EDITOR: We don't have to conceal their names. This was reported by a group up in Boston. I'm not aware of that kind of emphatic conclusion by any other group, are you?

COTTER: Other people took their tip from them and began doing some of the same things. I think it's important to see why maybe they got some of these results. For one thing, pickups are interesting creatures, and they can be represented in electrical analog by a pretty good low-pass filter. The characteristic impedances of pickups are shockingly significant low-pass filters.

OTALA: And you're pointing to the fact that they used the Shure V-15.

COTTER: Or others. And the cutoff frequency of these electrical generator systems is anywhere from 11 or 12 to 14 or so kHz, and their Q's at cutoff vary from highly undamped to moderately damped. They also have loss-eddy-resistance type effects that are somewhat signal-dependent, with the possibility of some time modulation. But the basic thing that they have is an inherently significant low-pass property.

It's always been amusing to me that one company in particular advertised the extreme frequency flatness of their particular pickup design, and they didn't lie; it was true. That was the data you would get if you did run a frequency response curve. They gave the parameters of the pickup which indicated something in the neighborhood of a 10 kHz low-pass condition with a Q of about 1%. So when you got this flat frequency response out of this pickup, the only conclusion left was that they had this enormous resonance in order to produce this flat response.

Which somehow or other nobody seems to have picked up on, maybe partly because of this whole emphasis on frequency response, where it's considered to be okay as long as what you come out with is flat.

EDITOR: Mitch, could I interrupt you for a second? Isn't there-you're refer ring to the Shure type of moving magnet pickup and others of that ilk--isn't it true that this kind of low-pass filter also has a mechanical pull elsewhere, conceivably, and the end result is a circuit that's a resultant of the electrical and the mechanical low-pass filter? Doesn't it work that way?

COTTER: Yes, you can have a mechanical low-pass filter, you can have all kinds of response, but no way, no matter how wide a band pickup you have, are you going to get significant speed or time effects out of a generator whose basic characteristic is that of a rather significantly low-pass filter condition.

It's also important to know, if we talk about pickups-and pickups have as a function the translation of the stylus/groove contact into an electrical signal-just what the scale of relationships is involved there, and how the stylus scanning process is inherently a bandwidth-expanding process. That is to say, if you have a low-pass filter condition in the recording-suppose you really recorded 15 kHz mechanically-the nature of the tracing process is such that one expands the spectrum, expands the speed of the events-let's shy away from spectrum-the speed of the events that are taking place is expanded in the time scale by virtue of the scanning process.

There are a couple of other things that are going on. When stereo records are cut, because of the vertical angle there is built into the movement an inherently self-time-modulated effect that's being speeded up and slowed down by the signal itself--and the magnitude is about 30% speed modulation by the signal, that's what the vertical angle condition is approximately in present, modern-day recordings. All of these processes amount to time-domain-operative effects, and they are applied right at the stylus. All other things being equal, they exist right at the stylus. So that one would expect that the output from a pickup that reproduces this is going to be itself a very much time-expanded, speeded-up set of phenomena, which-being a velocity sensor as most of the magnetic pick ups are-would further expand its time/speed effects; and that the whole equalization process is in effect a way of kidding yourself at the output about the relative magnitude of the speed effects at the input, and then enormously in creasing by a factor of 100 to 1000 your sensitivity to any of the lower frequency intermodulation phenomena that may be produced, since that's what equalization does-it expands your sensitivity to the lower frequencies and suppresses your awareness of the higher frequencies. This seems to me to be a condition that invites more serious problems than many if not most of the other portions of the electronics in the system. It seems to me an underrated problem in its consequences.

OTALA: I think that you forgot a couple of other effects that might also be important in that respect. First of all, if you take a pickup which is basically of the moving magnet type, then you have changes in the reluctance of the magnetic portion. Consequently, the inductance changes. Since the inductance changes and the capacitances remain the same, you have a sweeping resonant frequency.

COTTER: We have a time modulation.

OTALA: Yes, that's a time modulation effect.

COTTER: Especially since the low-pass condition is already well into the audio band.

OTALA: Yes. The second thing is that your compliance at the stylus suspension 1s normally highly nonlinear and therefore also the mechanical resonance, which also sits in the audio range, sweeps up and down depending on the actual point where the . ..

COTTER: There are two resonances in fact-one is the low-frequency arm resonance; the other is a midband resonance generally somewhere between 500 and 2000 or 3000 Hz, which is the resonance of the compliance of the pickup with the mass, or the effective inertia, of the system. And that midband resonance is the region of minimum impedance, where the motion imparted has the greatest ease.

". . . we're talking about preamps-and right away we're talking about pickups.”

OTALA: This means that there are a number of time-modulating effects pre sent in the pickup. This might be one of the reasons why moving coils are so much better because both effects are lower in moving coils.

COTTER: Yes. Let me introduce an idea since we're talking about preamps-and right away we're talking about pickups.

HEGEMAN: I noticed that.

EDITOR: I wonder how that came about.

COTTER: Maybe my original suggestion wasn't such a bad idea. I have a notion which I've described before to others, and it seems effective because what it seems to me occurs when we talk about pickups is a sort of perceptual mistake.

Because a pickup is so small an object, there's a tendency to collapse all of the dimensions into something that is in the zone of smallness. In the same way that if we think about the cosmos, and all but the most reasoning astronomer types don't really put any great weight into pools of galaxies, galaxies, interstellar distances, interplanetary distances--these things may differ in scale by enormous ratios but it's all vastness.

And it's either vastness or smallness that takes us out of the realm of our normal sense of proportions, sense of reason about the size and shape of things. It seemed to me there was a lot of mistake, a lot of mis-thinking going on, be cause the scale of relationships in the phonograph-stylus-groove-pickup system wasn't clearly in front of us. So I said, let's conceive that either we shrink our selves down so we can sit on the edge of a groove and look at it, or we can scale up the groove. And the convenient ratio occurs of 12,000 to 1-in English units, 12,-000 to 1 magnification. So that I thousandth of an inch becomes 1 foot.

And on that scale the present groove stylus system can be visualized. And of course we then have a groove that's typically about 3 feet wide and about 1% feet deep, and it's a V-shaped right-angle groove with a 2-foot-something sidewall on it. And our now nude, de minimus diamond, which is now about a 4-mil square shank, 13, 14, 15 to 20 mils long, sits in this groove. So if we scale this up and look at it, what we have is the stylus with a super-elliptical %-mil knob on the end.

Quarter mil means that there's two rounded surfaces sitting in this 3-foot wide groove that are roughly the curvature of a softball, in radius. And we have a 4-foot-square pole here that's 15 to 20 feet tall-that's our nude diamond.

And typically, as we get into it a little later, these two rounded surfaces sink into the surface of the record-these 2 foot sidewalls, 2-foot-plus wide walls-these two small spots sink in something in the neighborhood of 1/4 to 1/3 inch. And it's connected to a modern-day kind of cantilever, which is a smallish cantilever, maybe only 12 mils in diameter and 1.5-mil, 1-mil wall thick ness. So this is connected-this diamond, 4-foot-square, 20 feet tall-is pushed into this tube, this pipe, that's 12 feet or so in diameter and its wall about 1 foot thick, and a typical cantilever of about 250 feet long in this scale. The typical cantilever is 250 feet long, connected to a generator that's about 150 feet off the ground back up in the pickup somewhere. The pickup of course is 800, 900, 1000 feet away, the back end of the pickup. This begins to give you some sense of scale as to what's happening. We have this 0.25 or 1/3-inch depression in the groove wall on this “humongous” stylus connected to this great big giant cantilever, and there are copywriters and other true believers who evidently feel that somehow or other that is going to convey with great accuracy and delicacy the undulations of this groove wall that exist down here in this little contact. It's a highly suspect hypothesis when viewed from this scale.

Admittedly things like density don't scale, but there is some sense of the relationship here that one gets out of this. In approaching the problem of how to build a preamp, or what goes on with pickups, we also began to look-when I considered this kind of analogy at some of the dimensions and some of the effects and asked certain questions-when the pickup is playing a fairly quiet record, this scaled-up pickup is going to undulate for this fairly quiet record something in the neighborhood of a fraction of a milli meter. And that's a noise that we're going to hear as background noise. This thing is sitting in the groove wall, and one of the questions that we asked very early was, how far does this stylus push into this groove wall and how does that relationship vary with force? All the while analyses have been made of phonograph pickups and playback styli-all these analyses until very recent times have all been essentially geometric analyses-tracing distortion, all this kind of thing, are geometrical. And it's important to realize in geometrical analysis that there's no size particularly involved.

Geometry is geometry. It could be a giantific thing such as my scaled-up analogy or a little bitty thing, but it's geometry. And we learned about tracing distortion; and we learned about tracking error; and we learned about vertical angle and these things. And the elastic side of the equation, when it was considered that we had a hard, rounded thing pushing into a softer material-when that analysis took place, the analysis was based upon a classical Hertzian indenter equation which seemed to work. It was derived basically again from an idea of analyzing how a hard round thing pushes into a soft thing, again with no particular attention paid to size, because it was a purely classical physics analogy, scaleless. Just a hard round thing pushing into a soft material causes stresses and pushes and whatnot. And that even seemed to work when it came to measuring metals and looking at how certain things behave. So when the analyses occurred, rather recently-Miller, who was one of Hunt's students at Harvard about 20 years ago, did some analyses, a little over 20 years ago-and those analyses showed certain predictions which seemed to agree with some of the data except when you started looking into the details, and then it was found by a number of workers that they didn't agree. There were peculiar differences in the observed results of frequency response, for in stance. And it wasn't until the middle 60's that we picked this thing up and started to look at it. We found a very strange relationship. If you draw a curve of the Hertzian equation, which is force versus displacement-and it's convenient to represent a range from about a milligram to 10 grams since all the workers who have done any work have been within this range-and you cover the displacement in two orders of magnitude against these four orders of magnitude, you get a line which is a straight-line affair and looks like that.

And the observations that we made and several others have confirmed is a curve that looks like this. Practically vertical from the x axis; practically no change, and, importantly, a very large distance.

With a quarter-mil elliptical stylus, it is into the vinyl at 1 milligram something in the neighborhood of 3800 angstroms.

And it doesn't change a whole hell of a lot over this range. This theory is the one used by all the analysts, and these are the facts observed by myself and three other observers. Jim White, who was working as Hunt's last graduate student, did his doctoral thesis on it, and that was actually published and is available in the literature. That set of facts is so shockingly at odds with the theories that we are moved to stop and ask a number of very important questions. First of all, obviously, there is something very different than the classical physics situation; and the explanations for this involve the same kind of considerations of short range molecular forces that give rise to a very non-classical phenomenon in your teacup. There's a little meniscus at the side. Classical physics says that the cup should have a perfectly flat line of liquid that goes right out to the edge because that is the position of least energy in the cup. Only it climbs the edge. And as a matter of fact, interestingly enough, the meniscus at the edge will be the same irrespective of whether you have a wash tub, a teacup, or a test tube full of the same materials in contact with the same materials. That there is in effect something about the nature of small size that implies a range of forces and a kind of effect that has nothing to do with this classical geometrical analysis. In fact this does happen with a very tiny stylus on vinyl and even on other materials. As a matter of fact, if classical physics were to work, then the whole system of playing a record probably shouldn't work. Because we get into the situation of having these enormous tons of force per square inch, and materials should collapse, and all of these things that we have been blithely ignoring, which are obviously not at work to destroy the effort, seem to be responsible for making the thing work. And a lot of other things come out of a consideration of this sort, and in fact lead us to ask a lot of other questions about what's happening at the stylus. Not the least of which obvious thing from this is that certainly, the dynamical forces on the pickup, on the stylus, are moving the stylus a very much smaller dimension than we had surmised from the Hertzian type law, so the effect of tracing the groove is not quite so seriously disturbed by the dynamical forces. You're getting something much more like the groove.

However, what is the groove that we're tracing? These facts say that you are always immersed in the groove some considerable distance; there is no such thing as the surface. In fact the surface-what we know to think of as surface physics, in the sense of what we call surfaces in modern physics-is a range no greater than 10 to 100 angstroms, even 100 may be rather generous; 10, 20, 30 angstroms distance 1s considered a surface. Obviously a stylus is not playing a surface; it's playing a considerable region of sub surface. We were moved to ask what is the signal-to-noise ratio of a record and why? You're not playing the surface, you're playing something subsurface. An analogy was created to explain the signal-to-noise ratio of the playback process, and certain tests were made, and they verified this. If you flip for a moment to consider a tape track, a wide track 30-IPS magnetic tape, you have a playing head that's a wide-track head, and you have a large gap because at 30 inches per second you don't need a small gap. If you have a 1/2-mil or a 3/4-mil gap you will see most all of the magnetic oxide which is of the order of 3/4 of a mil thick, and with your wide-track head you pick up all the particles that exist to make the signal, and you get a very good signal-to-noise ratio. What happens if you play back that wide-track terrific 30 IPS master with a cassette head? A cassette head is a real narrow one, let's say it's a 10-mil-wide track; and of course it's also a very narrow gap, so it's only seeing the surface layer, and it's a little tiny percentage of the whole tape track.

HEGEMAN: Like a 50 micro-inch gap.

COTTER: A 50 micro-inch gap and you've got a 10-mil track. Quite obviously, you're going to get a lousy result.

Now let's suppose two observers are both nominally playing the 30-IPS tape track.

One says, “That's a terrific recording, sensational signal-to-noise ratio.” And the other one says, “That recording stinks.” Now it's not the recording that they're observing; it is the particular act of playback that they're observing. Now without going very much further in this, because there are a lot of other interesting things that come out of it, let's flip back to our analogy, our scaled-up analogy of this diamond playing the groove. We see immediately that we are with our elliptical stylus or our rounded stylus playing a very tiny percentage in width of this 2-foot-plus wide groove wall-like our cassette head. Also, one can imagine that in effect the gap width, which is somewhat like the aperture, is determined not only by the radius of curvature but also by the amount of depth that you are pushing down into this, or absorbed into, because a gram is not much of a push. The amount of distance that you are going through this sea of vinyl, this is like a volumetric scan. You are, in effect-you have a tape track; you have some thickness involved here. In fact, it is the number of particles that are averaging under the stylus that is very akin to the tape track system, and we're getting a signal-to-noise ratio that is the statistical averaging of both the number and the average distribution of sizes of these clumps. If we push harder, we're going to thicken the tape, in effect. If we push harder, we're going to see some what more material. Also, if we do the obvious thing that we do in the tape recording, if we go to a full-track head in stead of a cassette head, if we go to a very broad line-contact geometry, we will get the equivalent of a wide-track tape head.

So all other things being equal, we should be able to perform an experiment in which we use a very wide line contact head, and we push with a larger force,

and we obtain a very much better signal to-noise ratio. And it is a fact that if we do this we do precisely that-we obtain a very much larger signal-to-noise ratio.

We've been able to demonstrate 26 to, in some special cases, as much as 30 dB improvement in signal-to-noise ratio from the playback of the same record that you play with the smaller “cassette head” type of scanning. Now, this raises a lot of interesting other specters wherein people have surmised the nature of the recording process incorrectly. When a lacquer master is made, the lacquer master is examined by being played. We talk about how good the surface is-sheer non sense, because we're not playing the surface, we're playing the subsurface. In lacquer the penetration is even greater.

So when we assess the characteristics of the lacquer master, we are not assessing the thing we are going to determine when we metallically plate it, because it is then the surface that we're going to replicate.

When we then replicate the surface and examine it, it's strangely noisier. Well, it's not so strange because we never played the surface. The surface is in fact noisier, and it's also distorted. When we then replicate it in vinyl, and play it, we're not playing the surface either, fortunately, because it's terrible. We're playing this subsurface volume. And so the whole scheme of reality is quite inverted from what the mythology of playback has told us we should do. The mythology says light force-nonsense.

The mythology says narrow, rounded little bitty stylus-nonsense. The mythology says record wear will be reduced if we play at a light force-that, too, turns out to be nonsense, because from this scaled-up analogy it's quite apparent that if you want to know what a stylus is doing, you don't ask the pickup generator, you ask the stylus. When we contrived to measure the mechanical impedance of a stylus right at the stylus, we discovered a very interesting thing-that it bore very little resemblance to the output of the pickup, and that it had enormous energy storage capabilities-resonance in effect--large mechanical impedance and reactance at high frequencies. At very high frequencies, reaching out into the hundreds of kHz, where the device became very much like a squeaking chalk, a very beautiful ultrasonic abrader, and we're dealing further with a nonlinear compliance.

Quite obviously something that changes its value of compression by a factor of maybe 2 or 3 over a range of 10,000 to 1 in force is a highly nonlinear spring.

There is the physical reality, or at least some significant differences between the mythology of playback and the physical reality of what goes on at the stylus, which helps to explain not only why it works but some of the important differences that are going to occur when we look at what the signal is to be reproduced. Now realizing that this is going on, and that the basic geometry, the tracing effects, can be reproduced with great accuracy, we begin to have a picture of what the signal is that's going to be presented to the stylus. These higher frequency effects are going to be applied, as Matti has suggested, to a highly non linear reactant system in the moving-field kind of pickup and you've got also other kinds of elastic properties. So the problem of the pickup is that it presents to the preamplifier, if its behavior is correctly represented, a very, very much different kind of signal than had been surmised, and that it is able to propagate back to the generator through this *”250 foot long pipe” with any semblance of ac curacy is an act of sheer fortuitousness rather than any grand design. The scale of things is such that we're really only beginning to get an idea of what the stylus is doing. The bandwidth is greatly expanded over what the signal has and there are many things going on that don't belong in the realm of present definitions.

FUTTERMAN: I'm very much awed by all this. I'm wondering if Edison would have invented the phonograph if he knew.

COTTER: I think it's fair to say that if a pool of physicists got together and look ed at the problem recently, and it was proposed to invest millions of dollars in this scheme, it would be completely dismissed as utterly impractical. Because the reasons it works have nothing to do with the geometry and are based on effects which are very little understood and have only recently begun to yield to any kind of measurement. I think it's fair to say that the only reason we got it working is that people were too dumb to know that they couldn't do it, so they went ahead and they did it.

---”. . . 'm wondering if Edison would have invented the phonograph if he knew all this . . .” ---

HEGEMAN: Isn't that the way most things happen?

EDITOR: Mitch, does your analysis mean that these fantasies of say a laser beam tracking a groove-and I mean an analog groove, I'm not talking about the digital recording technique-a laser beam tracking an analog groove without any contact and therefore “no record wear” and all that, would result in a much noisier playback than what we get with a blunderbuss pressing down with 5 grams?

COTTER: We did some experiments.

Aside from the limitation of the laser as a scanning process-which has serious limitations as far as its resolution is concerned-there are schemes where playing back 'just the surface” have been accomplished. What's apparent from that is that the surface is a distorted wave form due to the dynamics of cutting.

When a lacquer is cut, that surface is not an exact replica of the signal because there are these elastic properties that distort it at the surface, and that the physical indentation of the playback stylus, again fortuitously, is just nearly exactly a compensation for this. That force-modulated scanning radius of the cutter is too, Matti, something of a time dispersive effect, even in the cutting. The playback partially compensates for some of this by having exactly a converting quality to it. So if you play the surface you get two things: you get a horrible noise and you get a distorted signal. So it is, again, a contrary-to-fact piece of mythology that the surface is both good and quiet. But it's neither.

HEGEMAN: Garbage in, garbage out.

EDITOR: Just to zero in on the practical end of it-tracking, attempting to track, at a quarter of a gram or half a gram is nonsense.

COTTER: For two reasons. You reduce the effective thickness, the effective size of the substrate that you're seeing, and each little grobble and bobble of the molecules that you're going to encounter is that much more effective in moving the position of the stylus.

EDITOR: Does that mean then that two thirds of the phono cartridge advertising that we see in the magazines today is basically nonsense?

COTTER: Just like I think we destroyed the advertising of amplifiers a little while earlier. This is a good place to begin to understand the problem of pickups.

FUTTERMAN: I found that out in practice. The heavier I usually made the cartridge, the better it sounds.

EDITOR: Most audiophiles have found that out in a purely pragmatic way. They heard distorted sound, they increased the stylus pressure . . .

FUTTERMAN: And it went away.

EDITOR: And it went away. Nevertheless, they're in an agony of apprehension about the damage that they're causing to their records.

ZAYDE: Absolutely. Most people believe the lighter the better. And they really need to let go of that false thinking. Because they really don't understand the quantum mechanical aspects that you just spoke about.

OTALA: There's one effect, however, which gets worse when you increase the pressure. That is with that viscous type of model of yours, when you increase the needle pressure, the apparent mass of the stylus tip becomes larger and therefore it also becomes force dependent.

COTTER: The mass of the stylus?

OTALA: The effective mass of the stylus.

Because part of the viscous material, the vinyl, belongs in the equation to the mass, the effective mass, of the stylus tip.

COTTER: While there's a partial truth to that, the fact is that what the data show is that that effective mass contribution at the interface is there; it isn't a very strong variable of force. It doesn't change greatly with the force. It's almost inescapable.

OTALA: What I'm saying is that at the high accelerations you are using, then at high pressure levels the added mass contribution becomes variable.

COTTER: No. It's not significant. The data show that there's a very, very small change-much less in fact than you would expect from a Hertzian law. Actually, the material behaves, for the volumes and the frequencies involved, except at very high ultrasonic frequencies, essentially as a stiffness component.

What does change with force is the real part, because this nonlinear capacitance 1s still nonlinear but it doesn't dissipate any energy. What does change is the mechanical resistance, which absorbs more and more energy as you go up in force.

OTALA: You're saying that the vinyl acts as a limited-slip oil.

COTTER: It's a highly non-Newtonian system. As a matter of fact, you improve the ultrasonic damping at higher forces in any given system.

OTALA: Yes, you improve the damping, but you cause variation, more variation, in the mass for lower frequency components.

COTTER: No, the increase in apparent mass is not a very large factor at all, because the change in distance-the in fluence of that zone is reduced. When you go up in force you're actually reducing the relative effect. Given geometry, given system may have a larger effect at a lower force. You get a larger percentage modulation. What's interesting is that when you look at the diamond itself at the stylus tip, there are many degrees of freedom in which it can vibrate, and there may be from 20 to 50 micrograms, but their moments of inertia are considerably less. And we found that there are modes of resonance, modes of vibration, that ex ist in example to example in a given design, that are quite different. Because they differ in their mounting, rather than in their frequency response. So the con trolled parameters are controlling something that is of no concern. They're not controlling the degree of tightness of the mounting, the exact position of the diamond, because it doesn't seem to in fluence these other effects. The consequence of which is that two different examples of the same pickup will wear quite differently. We find a very strong relationship between wear in both record and in diamond in the mechanical impedance at the diamond.

EDITOR: Are you aware of any pickup designers today who pursue their design efforts in the light of this information?

COTTER: No.

OTALA: Yes. Harbo Andersen at Ortofon. The MC-30 is partly the result of that kind of studies.

COTTER: He's aware of the non Hertzian law?

OTALA: Yes, and he's also quite well aware-for instance, he also designs cutter heads-he's quite well aware of the problem of various dampings, of the damping variation of the different resonances. For instance, the Ortofon cutter head there are five major resonances, and their damping behaves differently due to the elasticity of the lacquer master when he's cutting. He describes them as highly nonlinear; he's also discussing some of the penetration effects.

COTTER: He's not published anything, though.

OTALA: No, he never anything, he just designs.

EDITOR: What is the man's name again, Matti?

OTALA: Harbo Andersen.

EDITOR: Has he been there long, at Ortofon?

OTALA: Ten years at least, twelve.

EDITOR: That's interesting. The MC-30 I believe hasn't been marketed in this country yet, has it?

OTALA: I saw the first hand-made samples two months ago in Copenhagen.

Two weeks ago we got the first sample into the States. The quantity is set to be 2,000 for the whole world.

EDITOR: Per year? OTALA: No, total.

COTTER: What is the geometry of his line-contact stylus?

OTALA: Well, it's basically the same as the MC-20.

COTTER: It is not a Pramanik, even.

OTALA: No, it's not anything like that.

It's extremely light. I don't know the figures, I haven't got the latest data sheets-or any data sheets. I've seen the production. The frequency response, at least in the first samples, went to about 70 kHz flat, and all the major resonances are well above the usual, the 20 kHz normal range.

COTTER: Another part of the physics of this system that's extremely important and is responsible for some of the differences between moving coils and moving magnets is that there's a considerable drag force, which is the thing that gives us the skating force. This drag force is about a third of the magnitude of the vertical tracking force and it is a variable; it varies with the signal, but not exactly with the signal, but with an even ordered harmonic series of the signal.

This is the distortion named by Rabinow and Codier in a very old paper in the early 50s 'needle drag distortion.’ Now this force is of a very significant magnitude and is applied right to the stylus. If you constrain the system so that the stylus in effect doesn't move axially-be cause if it moves, you have a geometric error and the ball game is over-but if you constrain it so that the stylus doesn't move, you still have to deal with this force propagating up this ‘250-foot long tube’ to something or other back there that's going to respond to the energy.

Much the same as one can demonstrate in elementary physics lab when you have a billiard ball hooked on to a metal cylinder with another billiard ball hanging on the other end-if you rap this billiard ball, the tube may not move but the force is communicated back out there to the other end. So this force which propagates back to the generator will be inclined to induce some movement in the generator that is in the axial direction.

And the big difference between moving coils and moving fields is that the coil can move in two directions in which it will produce no output. Because of fringing publishes field, there is no way that a generator system of the moving field kind can be utterly immune to axial displacement; so one is getting in effect a very interesting kind of signal out of most pickups due to this axial modulated force that's a distorted signal. It's instructive to know that when the electronic music people want to fuzz a signal, they make a full-wave rectifier circuit to do it. That's very much akin to the kind of force variation applied to the end of the stylus from this needle drag process. So on every pickup there is applied a fuzz box at the stylus.

The question to be asked is, to what extent will the needle fuzz box force produce an output signal at the pickup? We've been able to simulate the sounds of some pickups by using a very low needle drag distortion pickup and injecting a full-wave rectified signal of the particular frequency characteristic of that pickup, and been able to simulate to the point where a panelist will identify the sound as the same as the sound of that pickup by injecting the needle drag component. We feel that the needle drag component is one of the strongest obvious differences between the sound of one pickup and another. Again, where they all show the same kind of frequency response, the same kind of normal IM and harmonic distortion. And this process is never measured. No one tells you what the axial force response . ..

EDITOR: This is time modulation, right?

COTTER: And ultimately this is a time modulation kind of thing, because it is a force moving the generator in and out that is essentially time modulation. Of course to this we have to add the effective vertical angle differences and the tracing distortion component, which are both time modulations.

EDITOR: Would you say this is the ultimate limitation of ordinary magnetic pickups? If this one thing were not pre sent then theoretically they could equal moving coils?

COTTER: Yes. But it's hard to conceive of a system in which your sensitivity in the axial direction isn't going to be a serious limitation, very high. In fact, there are pickups whose axial force response is more sensitive than their lateral deflection response.

OTALA: I've been waiting for you to mention the reaction type of vibration of the record surface itself.

COTTER: Again, going back to the mythology . ..

HEGEMAN: For us old-timers, are you talking about groove resonances?

OTALA: No, no. Simply due to the fact that you have a needle which you are trying to move. That's the action; that creates a reaction. Record masses are not that big, especially considering that the record material is elastic. Although it was classified as nonsense, Jean Hiraga's paper on the table mat material influencing the sound had some truth in it.

COTTER: We find that is valid, and what is more, the most important thing seems to be the degree of contact. And what seems utterly inane are these couple of little rubber bumper suspensions for the record which just convert it into a diaphragm so it becomes acoustically active and also mechanically active. Also realize that everything I've said about the nature of this interaction process would lead you in the direction of having a much larger vertical tracking force over a larger area and increasing the amount of energy that is being stored in that wall, and increasing this excitation. So yes is the answer. You would want to damp the record; you would want it to be in good intimate contact with a sound-absorbing medium.

OTALA: Here's a very good idea then for everybody who would like to use it.

Make a vacuum cleaner type of system that sucks the record down to the surface. That's the only way really to hold it in lace.

HEGEMAN: Well, how do you read instrumentation tape?

OTALA: Let's go on, because I think this is also one of the neglected factors. Due to the asymmetry of the record played, there's an interesting factor where the lateral reaction of the record being converted to vertical, so that you get a kind of cross talk. Vinyl elasticity type of cross talk from one channel to another.

COTTER: That problem exists significantly in the ultrasonic zone, where the pickup's resonance at very high frequencies, hundreds of kHz in fact, can be quite detached and very active. And twisting moment of inertia is one of the principal forms of abrasive resonance in the pickups.

OTALA: I was also talking of sheer record surface variation or vibration where lateral forces are converted to . . .or they are rotating in fact.

COTTER: What I'm saying is there's a circular kind of motion which has got at a lower energy state degree of freedom that is the vertical movement. This is very commonly the case; this is how styli twist and wiggle. Also, the magnitude of this axial force allows the stylus to do a lot of that kind of thing, skittering sort of action, so that there is applied to the signal an out-of-band ultrasonic spectrum of modulation which can by non linear process be demodulated and dumped back down into the audio band. And is quite influential on the apparent

EDITOR: And now we're back into preamps.

COTTER: And now we're back into preamps. Because all this junk is there, and it is necessarily always going to be with a phonograph pickup pre-emphasized by the velocity response if you have a magnetic pickup, very involved with ultrafast phenomena; and our sensitivity to the net intermodulation of this system greatly magnified by the inherent thousand-to-one difference between the gain at lower frequencies and the gain in the ultrasonic zone. Needless to say, the time, all these effects we talk ed about in amplifiers, are there in spades in preamplifiers, in consequence far pa I think, than anyone has realized.

EDITOR: I don't want to drop the subject of pickups and phono . ..

HEGEMAN: While we're on the subject, I have a couple of questions. Growing up in the phonograph industry, there were certain things we used to talk about when we had a new record. We talked about radius equalization, and we talked about the resonance of the stylus tip with the groove, or stylus and groove being a rather high-frequency rising characteristic. Now, I don't translate any of this into what we've been talking. You've been sure going through all of it, Mitch, but I don't quite bring it down to earth.

COTTER: What is interesting is that if you could get an ideal system in which you're biased at some point in this force curve, you're going to get a kind of compliance that will resonate with a singular mass point to give you a simple resonance. But the whole nature of this thing is such a nonlinear process that efforts to find the resonance are very involved with what the small-force modulations are. So that the exactitude of the resonance disappears.

HEGEMAN: So you're saying that it resonates over a band, instead of . . .

---“. . . there's a very different thing going on at the stylus. . . a lot more energy and a lot higher frequencies than anybody believed.” ---

COTTER: Right. A lot of what people have been identifying, Stew, as stylus resonances, were resonances of the stylus system, resonances of that “250-foot long” thing in torsional and in multi modal kinds of up and down propagation effects. Many, many pickups have been built whose axial Q in the audio range is in the high numbers range. A lot of energy can bounce back and forth down that tube, whose propagation time isn't exactly trivial, if you consider what's involved.

EDITOR: Is any of that dependent on the type of pickup we're dealing with? Is a moving magnet type of pickup more prone to that particular form of resonance than a moving coil? Or vice versa?

COTTER: No, I think that's the cantilever and system design . . .

HEGEMAN: I would think so.

COTTER: And in fact, there's another thing that merits saying, and that is that signal-to-noise ratio in the phonograph art is a very different thing from what people think it is. What I'm saying is, no one has ever played a record, actually played it completely in the sense of telling you what its signal-to-noise ratio is.

In fact, there is some doubt as to what it is; it's certainly a lot better than we know, because we've been using-to carry the metaphor all the way-cassette heads to play full-track recordings all this while. Furthermore, if you talk about these resonances and this propagation of energy back and forth, we are dealing with all the sorts of things that Matti was just talking about-that is, these non linear time-modulating processes. And if the beam can soak up a lot of energy and kick it back and forth, its propagation time is in the microseconds to many microseconds back and forth with significant Q in that range. There's an ample opportunity to have all sorts of events winding up from earlier time influencing the consequence of what's going on, be cause the energy keeps coming back and forth and is imparted to this nonlinear stylus contact and to other nonlinear processes in the generator and beam system.

OTALA: Which, by the way, only goes to remind me of a very early edition of Audio Handbook, which stated very, very bluntly that despite all efforts and all kinds of metal needles, the bamboo needle still is the best needle that you can buy.

FUTTERMAN: Cactus.

OTALA: Cactus? That was bamboo. . .

HEGEMAN: Cactus or bamboo, either one.

OTALA: Yes, that's only because it's a fibrous material and also very heavily damped.

EDITOR: It's a lossy material.

OTALA: Yes, so you're getting a very, very good needle.

FUTTERMAN: I have some cactus.

EDITOR: There were some problems there, though.

HEGEMAN: I don't know, do they still have a sharpener around?

ZAYDE: You're going to have several times the time constant that will persist two or three t as a result of inadequate termination or inappropriate termination, both at the generator and at the stylus tip.

COTTER: One could look at the cantilever as a transmission line. Then of course the differences are gross in the way in which the energy is terminated.

EDITOR: Let's zero in on some specific practical problems of audiophile or audio purist interest.

HEGEMAN: I see we just passed over my question of how about radius equalization. A very strong item back in at least old mono days.

COTTER: Radius equalization was derived from originally some observations, and then Kornei's paper in the SMPE Journal, which used the Hertzian indentor idea as a description of what this playback loss process was, and suffice it to say, that's in error. What turned out to be the case is that the short wave length effects caused changes, because the whole system was resonant within and around the audio band. And that if you could avoid that, then indeed there is an aperture process-there is a physical aperture process. The scanning stylus has a certain wave length. That aperture process is very much like the tape process-it tends to follow a sine x over x kind of relationship. Being that kind of function, as long as you stay far other enough away from it, there isn't much correction. If you're into it, then you're in a bad operating condition to start with.

If you're dealing with a quarter-mil kind of indentor, if you're in the line contact kind of condition particularly, then you're sufficiently outside the realm where radius equalization makes any sense whatsoever.

HEGEMAN: You're telling me that at a 6-inch radius on a 12-inch disc you get the same sound that you do on the entering and starting grooves?

COTTER: If you cut it appropriately, and if you play it back with a small enough radius of curvature, and you have little enough reactance in the stylus, the answer would be yes. You couldn't tell the difference.

EDITOR: But there is a limit to the information density you can achieve as you keep slowing down the linear speed of your groove.

COTTER: I'm saying you would want to do everything both in the design of . ..

See, it becomes possible to make a radius of curvature for the aperture that is very much smaller than you might otherwise have thought possible, and still have a fairly large area of contact, if you will, or still have a fairly good signal-to-noise ratio. If you keep all your ducts in order, then with a modern kind of stylus you can play without any significant loss from the playback process. But interestingly enough, a lot of the cutters, notably the SX-68-from which, by the way, the Ortofon is quite different in this respect-the SX-74, have significant re action effect at the cutting stylus from the smaller diameters that change what's cut onto the disc. If you correct for that, and the CD-4 people were faced with that problem, then you can make a record which inside to outside is not observably different in response characteristic. I think there's more cutting aperture loss than anyone believed. Because even if you have a 60, 50, 30-microinch radius cutter, the cutter is in effect exciting a much larger zone of influence. The fist stylus condition has more to do with easing the removal of the chip, it would seem, than it has to do with influencing the actual cutting, which isn't really a cutting, it's a tearing. You're influencing considerably more than just the surface layer of what you cut.

EDITOR: Gentlemen, where does all this leave us with respect to desirable pickup design and desirable preamplifier design? Which is really I think what our sub scribers would like to know.

COTTER: My purpose in talking about it was to show that there's a very different thing going on at the stylus.

EDITOR: We are very glad that you talked about it.

COTTER: It's relatively new information. It's a very different thing going on at the stylus. The consequence of it 'is threefold, of serious concern.

There's a lot more energy and a lot higher frequencies than anybody believed. This energy causes reactions in the pickup that hadn't been suspected, although Rabinow and Codier and others pointed to the existence of the potential 25 years ago. That the pickup, being a velocity sensor, presents to the pre amplifier, in consequence of this-particularly if you have, say, a moving coil structure that's not imbued with these low-pass filter properties-a very much more difficult signal to handle in face of the inherent equalization that you're applying than has been considered. For instance, in a power amplifier we may have high-frequency effects, time modulation processes, things of this sort, that will produce results in a flat amplifier that we have judged to be serious problems. Now what we're doing is, we're suppressing our awareness of it by in effect removing our sensitivity to it by a factor of 100 to 1000, and then boosting the response to the lower frequency consequences of its existence by a factor of 100 to 1000. So in effect the equalization demand and the preamp problem are vastly more serious than has been considered.

OTALA: There's a number of problems in that respect. The first one being, I believe, that we're exciting quite a lot of frequencies at the pickup mechanical resonances. But also at the preamplifier, which is normally a feedback type of system having a normal phase margin and a peaking response somewhere. They behave the same way; they're complex conjugate poles. Since they are complex conjugate poles, any type of transient ex citation simply gives us more or less a burst of “carrier” frequency there. It's ringing up there; it's creating a carrier type of signal; and this carrier type of signal occurs at frequencies where both the amplifier and the pickup are highly nonlinear. That mixes with all the ultrasonic signals present at that frequency region, and that's what you mentioned.

COTTER: Plus the base band, which is a Hilbert transform, which is then going to translate with incoherent distortion products back down into the audio band.

OTALA: Yes, it's going back to the audio range. This is one of the problems. The second problem is that most of the preamplifiers as far as RIAA correction is concerned, do not have a specified frequency response, say above 30 kHz or whatever.

COTTER: The missing pole.

OTALA: Yes. Therefore we have had some dramatic difficulties in the TIM psychoacoustic experiments, for in stance. We were forced to go to the STM-72 transformer with the Ortofon MC-20 pickup. And that was simply because of the fact that it filtered our ultrasonic response. There's a lot of this kind of effects, and although it is not strictly what we are discussing now, it might be good to tell you how we did select the records and all the components.

It was very simple in fact. We had the distortion generator there, which we could . . .

COTTER: This is a digital synthesizer.

OTALA: A digital synthesizer. And we could adjust the additional distortion that we created into the signal. We had a measurement system which showed us the percentage of distortion being generated. And the simple trick was-the only one that worked, in fact-was to replace or modify each part of the system so that the threshold where the added distortion just became audible was at the lowest. By this way, by adjusting the stylus pressures, changing pickups, changing preamps or transformers, and doing this and that we finally came to a very, very sensitive system. The same system applied for selecting the records.

And that was simply this-you play a record, you note what is the distortion level which just now becomes audible, take the next record and take again the percentage level-if it was lower, then this record has less masking. And that's the kind of effect we have to play here too.

COTTER: Let me tell you of an interesting consequence of this under standing that was revealed to us in some experiments we did with respect to vertical angle. And that's an observation that's continuing, and I think others have now formed similar impressions. It's always troubled me that a shockingly small variation in the vertical angle, usually in the positive or increasing vertical angle direction, would cause very marked changes in the sound character.

Usually a kind of hardness and bright ness, if you were to put metaphors on it.

EDITOR: To reinforce that, one way of tuning the VTA by ear is to get to that point where the sound hardens up, and then back off a little bit. That has been the point of greatest clarity.

COTTER: Now there are two things about that condition that have always troubled me. One is that when I measure the actual frequency modulation produced by a pickup, I would never come up with that angle, usually with a much higher angle, as the effective vertical angle of the pickup. In other words, the tendency by ear was to come out with an angle that was lower. Further, nothing in my repertoire of understanding could quite suggest a mechanism that was so abruptly changing in the region of this effect, because further increase did not cause any overwhelming collapse. We now think we understand that process as a trigger process; that in effect when one varies the vertical angle, what you are doing is varying the slope of the response of the stylus to the wave form in such a way as to modulate its slope, because changes in vertical angle are in effect changes in the speed with which you're reproducing.

EDITOR: The slope of which response?

COTTER: The slope of the needles motion. Its time derivative is being changed. Its rate of change is being changed very significantly by changes in vertical angle. In effect, what you're doing is getting a phenomenon within the electronics of the system wherein you are getting into some kind of marginal, incremental, slew-type phenomenon that is a trigger-like phenomenon.

OTALA: It is not really slewing, but it is that kind of change which seems to . . .

COTTER: It's that kind of change, for want of a better word. Anyway, we think we understand this, because if you dissect the topology of most of the existing systems that include equalization in the feedback loop-and it's important to notice that irrespective of which particular combination of artifice and technique you encounter, that the topology of most of these systems is essentially that kind of topology. That it resembles very strongly the basic Schmitt trigger circuit. You have a scheme within which, if you remove the feedback, and consider that the error signal can expand rapidly and dramatically, you are getting into an energy storage situation that on a short-term basis resembles a trigger snap. In effect, the character of the sound and the kind of marked influence for these marginal levels, affects it. There's one other experiment that you can per form to, inferentially at least, see whether there is something like this taking place. Since what we're talking about is a velocity-responsive sensor, and when you change the vertical angle you are in effect changing the amplitude of the pulse as well as its rate of rise, one should be in a position to suppress this critical moment, this critical angle, by simply reducing the magnitude of the signal.

Not afterwards, but as presented to the preamp. A moving coil pickup makes it very easy to do this; all you have to do is apply a loading resistor to trim the level.

When we tried this, remarkably, you could reduce the disturbance, you could go below the threshold, in effect. You didn't get the hardening. You could come up a little more before you got the hardening. We've been exploring this for a couple of years. I haven't been sure just how much of each contribution there was, and I've become convinced recently that this is an overwhelming effect in most systems. In fact, when you get a system that is incapable of triggering because you haven't got that topology, a feedback-less passive equalizer system, then you don't find any of this effect, and the null and the vertical angle correspond with the measured effect, and in fact it's much less critical than you otherwise find.

EDITOR: We found in our equipment evaluations here that with the preamps we've been leaning toward most recently, the VTA sensitivity is considerably reduced. It has become less critical. It was supercritical with less good preamplifiers. On the other hand, we haven't found that it's necessarily restricted to preamps that are equalized in the feedback loop. As a matter of fact I don't think there's anybody sitting around this table here today who equalizes in the feedback loop.

HEGEMAN: Strangely enough.

COTTER: Maybe, but there still are probably TIM-like time modulating processes which this could improve.

OTALA: Except in all the preamps tried by Alvin Foster. He just published a paper and said he didn't find any.

COTTER: Any what?

OTALA: Any TIM in any kind of pre amplifier he tested, starting from-well, he had 50 or so.

EDITOR: Who was this, Matti?

OTALA: Alvin Foster, the Boston Audio Society founder.

RAPPAPORT: Well, they're the ones who called you a charlatan anyway, right?

OTALA: I would like to come back to my statement in the beginning, that I believe strongly that it is important to reproduce the amplitude faithfully and the first derivative faithfully, and the second derivative faithfully-we just proved that. Let's put it this way: we have a series of transfer characteristics. The first transfer characteristic is the output voltage versus input voltage. This should be a straight line. The second one is the dynamic transfer characteristic, and that is output rate of change versus input rate of change, which should also be a straight line. The third is the second derivative--how would you like to name that, that's another thing.

COTTER: The rate of change, acceleration . .. The second moment, the third moment, the fourth moment . . .

OTALA: So all those plots should be linear and should be accurately reproduced. That is probably the consequence.

COTTER: Well all of these come down to non-time-dispersiveness, if you consider that rate puts its heavy emphasis on the time at which it occurs.

OTALA: If you take the nth derivative, and plot that nth derivative of the output signal versus the nth derivative of the in put signal, all derivatives with respect to time, and that nth plot is straight and linear, you don't have any time dispersion. That's the conclusion.

COTTER: It's important to go back again to the purely geometric description of what goes on in the phonographic system, and realize that with a rounded contact on a stylus compared to the sharper contact of the cutting stylus, that there is a time modulation that occurs such that as you go uphill, the point of contact moves in front, as you go down hill the point of contact moves around behind you. So one has a time-swinging effect, in which the rate of swing is proportional to the second moment, the second derivative; this acceleration component is modulating the time position, and that that is the tracing process. Its very nature is one of time modulation. That the vertical angle errors are a displacement-proportional time modulation, and that lateral tracking error is the same kind of thing, and that even some of these elastic effects which store energy can return essentially a time-modulated displacement force. And in fact, all of the phonographic disturbances have an intrinsic characteristic of being time modulated.

EDITOR: And not all synchronous, right?

COTTER: And not all synchronous; obviously displacement in one case and acceleration in the other case, and an energy storage thing which is a complex thing because it varies with frequency, if we may reintroduce that specter.

OTALA: It also varies with the place on the groove where you are because you have different geometry, whether you're in an outer or inner groove, or outer or inner curvature.

COTTER: Because it's dependent as well.

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 . . .wave-length

To be concluded in the next issue.

---------

[adapted from TAC]

---------

Also see:

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

Various audio and high-fidelity magazines