ACOUSTICAL MANUFACTURING's Mike Albinson, co-designer of the revolutionary Quad 405 power amplifier, and outspoken critic of many current fashionable amplifier theories, is our subject this month.
Practical Hi-Fi: The Quad 405 represents something fairly unique in modern Hi-Fi amplifier design. What led you to the concept of feed-forward instead of the more normal feedback?
Mike Albinson: Difficult question, but I have a stock answer to this which isn't quite adequate. If you look in my notebook for the year 1969, you will see the rather naive idea that if we could build a kind of giant emitter follower we could get it extremely linear. The only trouble was that to do this on any kind of realistic Class A conditions involved heat sinks from here to infinity and the general efficiency was lousy. Added to that was the simple idea that all we had to do was monitor the current that this poor emitter follower was trying to dump and off load the hard work on to pair of current dumpers. That was the first part of the idea. Then there were, on odd occasions, attempts to interpret this into a sensible amplifier, most of them, in retrospect, really uninspired. Several years later, we rumbled the fact that what the statement that we were trying to make in hardware was expressible in our famous equation
Z1/Z2 = Z3/Z4 where Z4 was a sensing resistor. P.J. Walker's unique contribution to this thought was that this equation held very usefully if you made Z3 and Z4 reactances instead of resistors. This had the happy effect of solving the stability problems which the original current dumpers suffered from. From then on we were able to develop the 405, though we went through all the usual few blind alleys and hiccups in the process. Nevertheless, that gave us a framework where we could see a means of getting ourselves out of our, up to then, impossible stability problems.
PHF: Many engineers we have met have complained that their design has been cut short before they have really finished it. Commercial pressures often dictate that it should be moved quickly into production, so, in retrospect, was there anything about the 405 design which you found worthwhile altering in subsequent production versions?
MA: Very little which is concerned with the basic concept of the amplifier. Detail changes there certainly have been, but not anything connected with the current dumping problem. As with all amplifiers, time goes by, and better output devices are made available. In the case of the 405, better output devices were offered to us, this improved the amplifier's high frequency performance, but produced some slight stability problems which led to modifications.
When you have had twelve months mature consideration, if you couldn't say anything better, then I reckon you are a bloody awful engineer!
PHF: But then surely you don't regard the current dumping concept as the first and last word in feed-forward design? Surely, there are other advances that you can make?
MA: Well, there most certainly are, but as of today, I couldn't tell you what they are. Basically, the feed-forward concept is a very nice way of overcoming the difficulties which existing output transistors present. If a new generation of devices were to appear, we would have to re-evaluate whether or not there wasn't an even more elegant solution to the problem. If I were looking at power amplifiers at the present time, we would have to examine things like the big power FETs which are now available.
PHF: To what extent do you think that the idea of the 405 came out of a group of engineers sparking ideas off each other, or did it come from a piece of solitary thinking?
MA: There is no such thing as solitary thinking! The 405 arose because we had the idea, that's quite clear. People say that the 303 is a bit old in the tooth, but in terms of doing what an amplifier has to do, it is still a commercially and electrically viable design. The time had come to produce a bigger amplifier, mainly because we were aware of the fact that loudspeaker efficiencies were declining over the years.
PHF: You talk about loudspeakers getting less efficient, but as amplifier loads, they are not well specified and certainly some loudspeakers seem to be quite awkward loads for amplifiers.
MA: well of course, this is the amplifier designer's nightmare. We would like to think that a 100 watts amplifier would deliver 100 Watts come what may. Some people think that we should deliver 100VA come what may. I would not have thought that was unreasonable but alas, there are a lot of loudspeaker designers who seem to think that a 100-watt amplifier should be able to deliver 400VA. One has to draw a line somewhere. One also has to give the customer a good result for this money and if you assume that loudspeaker designers have total freedom to demand 40 Volts peak across anything down to a fraction of an ohm and up to a farad, the only possible amplifier designs left are the BGW-type of approach which does nothing but turn the whole house off when you overdo it. It is a valid approach, but it is extremely expensive and quite clearly not really called for in the average domestic environment.
PHF: Did costs play much of a part in the design of the 405?
MA: Not as design criteria. It so happens, though, that for any product which you sell to the public, one would like, as far as possible, to obtain the best return for the money you spent on their behalf. It is not in our interests to rook the public and if we go and buy as nice big pair of output devices, we ought to give the customer the best possible benefit of them. If there are people around who label two-ohm speakers, eight-ohms, it's not in the customer's interests to give him four of these output devices just in case he happens to buy that specific loudspeaker.
PHF: We remember the 405 brochure very clearly because it stated don't try and stuff the front end of the amplifier with signals outside a defined spectral range, and don't connect the output to the output of the amplifier, loudspeakers which fall outside a particular specification. This was a rather unusual step for a hi-fi amplifier manufacturer to be taking. Was this because you were aware that signals outside the normal audio range would produce unusual 'nasties'?
MA: The reason is much simpler that that. The limits which we specified are quite generous. There are very few programme sources available to the public which come anywhere near them. The written specifications were partly inspired by the Finnish amplifier designer, Dr. Matti Otala proposing that audio amplifiers should, at the time, be tested with fast-rise square waves at 30 kHz. He has since wisely retracted that idea!
Now, no way can a thing like the 405 digest a 30 kHz square wave and no way is the public really looking for amplifiers with a megahertz-wide bandwidth, either. We simply wanted to specify the signal limits the amplifier could digest without audible degradation of performance. They are far outside anything which can be achieved by a tuner, or even a fairly fast tape, and certainly not a gramophone record.
PHF: This suggests that the preamplifier needs to have a fairly well defined band pass?
MA: On of the most common amplifier killers of all time are the stereo decoders which leave in the output, the whole of the 38 kHz square wave switched wave form. there are even designs around, where the manufacturer has even managed to avoid de-empasising this! Such signals do present a severe problem, not just to amplifiers, they would also remove the tweeter from any loudspeaker on the market.
Our intention was purely and simply to draw up a specification in the context of a lot of strange statements about transient intermodulation distortion. We wanted to specify what the amplifier would tolerate before transient intermodulation distortion or any of its relatives, would set in.
PHF: Why didn't you avoid the problem simply by using an input filter to shape the signal?
MA: Mainly, because with any sensible programme source which is available to the public, including electronic music, this sort of troublesome signal does not arise. I would personally like to think that if you put in a 20 kHz low-pass filter at all the accessible points in an audio chain and did a good sensible job of it, it would make damn all difference to the sound.
If it does make a difference, it is because something has got in there which is certainly not an audio signal.
PHF: But isn't it the case that power amplifiers are being subjected to stray signals outside the audio band, simply because cartridges and preamplifiers are wide open to them?
MA: It is very difficult to conceive of a cartridge which, played through anything resembling an RIAA network, would produce signals which exceed our limitations.
PHF: What about harmonic distortion generated in, say, the preamplifier?
MA: No problem at all because we have specified the maximum slew rate which in fact corresponds to a 6dB roll-off characteristic typical of normal harmonic generation. That means that via an RIAA preamplifier the pickup has to produce 200 per cent distortion to obtain a critical input level at 20kHz. You should remember that this was purely a paper limitation aimed at curbing people who test audio amplifiers as if they are medium-wave transmitters and then complaining that they don't work!
I firmly believe that the extended bandwidth of some amplifier designs is the result of taking a relatively small problem and then overreacting on a scale that is quite unique, it's engineering overkill. Designers might discover a small amount of TID or slew rate limitation and therefore produce an amplifier that will go to 1000 V/ms. It certainly eliminates the problems, but produces a new one in the process because now Radio Three can come through the amplifier quite happily, and, without being demodulated, can pass straight down the speaker cable and melt voice coils and tweeters!
Were I to write the specification for the 405 again today, I think we would probably not include any specifications on test signal limits.
PHF: That seems to be a reflection of the power of the press, in that a company like Quad should have written a specification for the minority of readers - those who were likely to sit down with an oscillator and test the amplifier.
MA: Yes, and it was a mistake, but unfortunately, such people tend to be vociferous and can do a lot of damage in the magazines by raising needless doubts and worries. A lot of customers worry about problems which they haven't really got. There are enough of those in the business of reproducing music, which are very real and worthy of a great deal of effort. It seems a pity that so many energies are diverted down paths which are not very productive, when there is so much to be done which could improve the overall system response enormously.
PHF: In the concept of feed-forward design, were you concerned at any time about whether it was going to sound right, or were you convinced from calculations alone that there would be no difficulties in reproduction?
MA: I firmly believe that if the voltage that appears at the output of the amplifier is a fair facsimile of the input, over a frequency range that encompasses all that the human ear can hear, then the amplifier has done its job adequately. I don't honestly think that an amplifier with a given transfer characteristic sounds any different whether it is executed in valves, transistors, feed-backwards or feed-forwards.
The problem is simply that if you put half a volt in at this end and obtain thirty volts out at the other, as long as those thirty volts are related to the input in the proper manner, it is bound to sound the same no matter what is in the box.
PHF: Effectively you are saying that theory is sufficient to predict the success of an amplifier design provided the theory is grounded on good knowledge of the input signal?
MA: Well, quite clearly, if a man is making an amplifier to reproduce music and the user injects a one Megahertz signal and judges it upon its performance at those frequencies, you may well come to a different conclusion to mine. there is a tendency in this world to measure an audio amplifier at 50 Hz, 150 Hz or 10 kHz and find it is beautiful, but this isn't very newsworthy. So 100 kHz is injected and out comes a rather poor looking signal and the reviewer says, 'Now I've got something!'. Is that relevant? Can you hear 100 kHz? Have you any way of detecting the problem other than with instrumentation? The answer must be no!
PHF: Can I come back to the subject of knowledge of the input signal - economic design considerations surely dictate that you will have to tolerate a certain amount of distortion and do you need to know what is tolerable to the listener. How do you cope with that situation?
MA: There has been a lot of fundamental work done on the threshold of detection of harmonic distortion. If harmonic distortion could be present without intermodulation, which is not theoretically quite possible, it looks as though several percent of harmonic distortion, especially, the lower orders such as the second and probably a percent or two of third, would not be frightfully detectable. What we write in specification sheets is determined to a certain extent by general usage. Harmonic distortion figures are largely irrelevant.
PHF: Does that mean when you are designing an amplifier, harmonic distortion takes a back seat and other things are taken to be much more important?
MA: The problem with both harmonic and intermodulation distortion is that they tend to be closely related and if a linear amplifier is built, neither will be a serious problem. I think that the reason why people like ourselves build amplifiers with very small amounts of THD is because it is very easy to do. It then gets a large problem out of the way easily. It must also be admitted that if we were to market an amplifier with one or two per cent distortion, ti would be a hard grind to sell it.
PHF: So, in a way, Quad has got into the commercial game of 'specmanship'?
MA: Of course we have, but bear in mind that if we lowered our standards, nowadays, it wouldn't save on brass farthing!
At one time, the discovery was made that the published distortion figures and the degree of crossover distortion did not correlate and considering that crossover distortion gets worse with decreasing level, it's not surprising that amplifiers with 0.1% or 0.2% crossover distortion at full output (where distortion is often measured) were producing several per cent at low outputs. There is a very good argument for claiming that distortion in amplifiers should never get worse as the output is reduced, but the standard specifications for amplifiers that you see never bring this out clearly.
PHF: Why don't Quad do this?
MA: We do attempt to produce, in certain of our spec. sheets, at least a plot of distortion content against output power. The fact that, on most good amplifiers, this vanishes towards the origin as you come down to about 80% of the full output power is just one of those things that makes the picture a little meaningless. We are also 'governed' by standards bodies who determine to some extent what we write in our specifications. For example, the American IHF standard insists that we shall define our output into an eight-ohm load at a certain fixed distortion figure. We are not allowed to say that an amplifier produces 100 watts unless it will give 100 watts into eight ohms. We could make amplifiers which, under IHF conditions, will have superb performance but wouldn't drive a loudspeaker at all. This situation is now being changed, but certainly it has been the case until quite recently.
It is our job to produce amplifiers which don't produce any audible degradation of the input signal. There are not really well agreed methods of producing figures to support this claim. The customer has to take this on trust to some extent, which is unfortunate. You have seen us do our bridge test; it pleases us that our amplifiers pass it.
PHF: Having mentioned that, in the light of all the arguments and criticisms there have been of Quad's ideas in that particular direction, do you still believe that the bridge test is a valid on? For example, in the comparative side of the bridge, matching the phase characteristics of the amplifier under test is quite a problem.
MA: Yes, I do think it is valid. The difficulties you mention are sufficiently considerable to make the test unsuitable for application as a general purpose comparative machine, but for a given, specific amplifier, it is possible to sort out the non-linear functions from the linear ones. You may well argue that we have to do a little bit of time and phase correction which might be detectable. But it shouldn't be beyond the wit of man to devise a subsidiary experiment to prove the point - it's no good us doing it though.
We can quote you many very able authorities from the BBC to the Bell Telephone Laboratories who suggest the form of phase shaping used by us in the bridge test is not perceptible, and, in fact, is several orders apart from being perceptible. This isn't as simple a statement as it sounds because if you take a square wave and phase shape it, you rapidly come to a point where the peak-to-mean amplitude has changed so much that we perceive it. In any practical programme material, though, it's swings and roundabouts. On the specific proposition of the square wave, phase shifts make the relative amplitude increase, or rather, if you introduce phase shifts without attenuation, the relative amplitude is increased. With any signal which is not so highly organised, something is both gained and lost by moving the phase of the various harmonic components. There aren't many square waves that occur naturally in music, even when engineers use soft limiters to cram the signal on the record.
PHF: I get the impression that you believe that power amplifiers have reached the stage where there is no requirement to improve their reproducing qualities.
MA: True, you are dead right! You'll find Peter Baxandall was recently shot to pieces for saying the same thing. If I can't hear the difference between the input and the output of the amplifier, what can I do with that amplifier which would make the difference still less?
PHF: Would you agree that most of the amplifiers in magazine tests invariably come down to interface problems?
MA: Yes, though oddly this situation is improving, since, after many years of absolute madness, there is now an attempt to draw up a specification of what an eight-ohm speaker should look like when examined via its terminals. For example, we went through a period some years ago when the way to make an eight-ohm loudspeaker sound louder than another, was reduce the impedance ot three ohms. As long as the amplifier managed to hang on to its breath, the speaker sounded 3 dB louder. It was a beautiful arrangement, but unfortunately it is one that sooner or later gets flogged to death and the point comes where amplifier designers can no longer tolerate it.
If I have to design an amplifier which will produce about 40 Volts peak across one ohm, this means that I need to produce about 1.5 kW! If I now sell a 100 watt amplifier designed to produce this sort of output under unusual conditions, either I am going to be benevolent, and give this extra power to him, or the customer has to be very rich to afford the actual cost of the safety margin. This is clearly an impractical situation.
There are one or two loudspeakers around which still cause problems. I'll mention one by name and that is the IMF. Its main problem is not that its impedance gets a bit out of control, but that it is so damned insensitive. Owing to that fact that it is a large loudspeaker, and is bought by people who want to make very loud noises, a 100 watt amplifier is just not enough to satisfy. In fact, the interface problems there are not the ones that you were talking about - it is simply one of providing enough power.
PHF: If you are of the opinion that power amplifiers should not be called upon to be too flexible, are you of the same opinion in the case of preamplifiers? For example, until quire recently, the Shure V15 Series cartridges required a capacitive loading which was quite different to the normal run of cartridges.
MA: It is certainly true that at the present time there is a demand for connecting capacitance across pickups. There is also a generally agreed standard that pickup loads should be somewhere near 47 kOhms. I don't suppose that Mr Shure is much concerned whether his cartridges have the correct value of capacitance connected or not, because he has the alternative choice of dropping the load impedance. Both will produce similar results.
Probably, in time, this difficulty will disappear, already there are some cartridges around where the inductance of the coils has been reduced to the point that added capacitance is not needed because the electrical resonance occurs outside the audio band.
PHF: How do you feel about the moving-coil cartridge?
MA: Moving-coil cartridges have one great virtue and that is that most of them are completely indifferent to load impedance. Perhaps this is part of their attraction, I don't know, but there are certainly some very good ones around. There are also a lot of theories circulating that moving-coil is good and moving-iron is bad. This has very little technical back-up, and if it is true, then I would imagine that there must be some physical argument to demonstrate the point.
PHF: Have you any firm views on the valve-versus-transistor argument?
MA: I seriously believe that if you regard amplifiers as a black box which has a defined transfer characteristic, it is possible to execute this equally well in either valves or transistors. The fact that the compromises used by valve and transistor amplifier designers are different, may have some bearing upon the apparent differences, but it certainly has nothing to do with the fact that one glows red hot and the other comes in a TO3 metal can.
PHF: If you accept that there are some differences possible, what do you think they are due to?
MA: I think we must concede that there is a possibility that, under overload conditions, they will differ and that a lot of people have got accustomed to the sound of certain valve amplifier overloads. Below the overload point, I don't think the two are very different. There is a tendency for valve amplifiers to have inductive output impedances which, into most loudspeakers, produces a high end roll-off. People may even like that sort of sound, but it doesn't make it right, we could provide a filter to do the same thing and you could be sure that they wouldn't want to use it!