News

We were saddened to hear of the passing, on December 10,of Audio Research founder William "Bill" Zane Johnson. Bill, who founded Audio Research in 1970 and became its Chairman Emeritus in 2008, is survived by his wife Nancy (left in photo) and family. We are preparing a tribute to Bill, to be published in the March 2012 issue of Stereophile, but meanwhile, we are reprinting here an interview Paul Messenger and I conducted with Bill that was originally published in the June 1983 issue of Hi-Fi News. (My thanks to HFN editor Paul Miller for permission. Stereophile's 1994 interview with Bill can be found here.)—John Atkinson

"Hot & Glowing" from Hi-Fi News & Record Review, June 1983

One would have thought that the introduction of transistor circuitry in the mid-1960s sounded the death knell for the valve/vacuum tube as far as its use in hi-fi was concerned. The high power consumption and inefficient waste of energy as heat, coupled with its high voltage/low current output nature tendency to develop mechanical problems such as microphony, caused many to greet solid-state devices with open—and perhaps uncritical—arms.

With hindsight, one can now admit that early transistors, particularly those made from germanium, were far from ideal for audio amplification, but the almost exponential progress since then has given electronics designers devices that would seem beyond reproach. Relatively inexpensive op-amps such as the NE5534, which pops up in such diverse products as the Meridian preamp, the Studer A80 professional recorder, and the output stage of the Philips and Marantz CD players, have specifications that would seem to be many orders of magnitude better than those required for audio work, while a power amp designer can choose from as wide am variety of MOSFETs and high-speed bipolar as his imagination can cope with.

Many engineers and audiophiles have refused to be weaned for their glowing tubes, however, despite the apparent limitations; the Audio Research company, from Minnesota, USA, has encouraged would-be anachrophiles for the past 13 years by producing a virtually unbroken series of valve amplifiers whose reputation for high-quality sound reproduction would only appear to be rivaled by their pieces. Their "budget" SP8 preamplifier, reviewed by Martin Colloms for HFN/RR in February 1983, still costs a UK purchaser more than £1400, while if you need to ask the price of the classic D79C power amplifier (£3950 for the curious or foolhardy) you most definitely will not be able to afford it! Audio Research founder and designer William Z. Johnson paid a brief visit to the UK in March, following the Festival du Son in Paris, so we took the opportunity to ask him why he sticks with what most people would regard as an outmoded technology?

William Z. Johnson: Well, we do make both solid-state and vacuum-tube products, and all our vacuum-tube models make extensive use of semiconductors and IC chips in their power supplies. The simple reason we use valves for the amplifications circuits in our best products is that intrinsically, valves are the best device for amplifying analog waveforms. The point of building equipment to reproduce music is just that, to reproduce music, and we shouldn't be hung up on the active devices, that's very narrow thinking. If at a given moment in time the best device happens to be a vacuum tube, then why not use it? But obviously, if we were building computers or portable radios and propounded the virtues of tubes, you would have us hauled away.

John Atkinson: But don't you feel that transistors make it easy to design audio circuits with excellent extension at low and high frequency extremes?

Johnson: Of course it is easier to design wide-bandwidth products with semiconductors, but frankly a wide bandwidth is not required for audio. The range for most of us is, say, somewhere between 15Hz and 15–20kHz and while it is true that it is necessary to have some extended bandwidth above and below to minimize phase shifts and give wide stability margins and so on, there is no need for megahertz-type responses and super-fast slew rates. Rather, they're examples of a red herring, a buzzword sort of thing that is supposed to be a reason why an amplifier is "good." I mean, 25–30 years ago it was "How many pounds of magnet does that speaker have?" We've been through frequency response, harmonic distortion and damping factor; two years ago it was Transient Intermodulation Distortion, and now it's slew rate that is supposed to be the important factor.

But every one of these things, if I might say so, is a red herring. For instance, take slew rate: in all probability there has never been an amplifier made that had an inadequate slew rate for the normal energy distribution of music. You might think that a very strong statement, but if we inject a modicum of common sense into the situation, let's agree that if a power amplifier can reproduce, let's say, 40kHz, it has a sufficiently high slew rate. Frankly, it's more than can possibly be needed; to reproduce 5000 cycles at full power is probably all that is required, but if you build a nominal 100-watt amplifier to have a power bandwidth of 40kHz, it only takes a few moments with a calculator to determine that 8V/µs is adequate.

If you then do something with the amplifier and come up with 16V/µs and it sounds better, the automatic conclusion is that slew rate must be the magic ingredient. But in all probability, it's nothing more than a byproduct of whatever you did to improve the sound.

Paul Messenger: The approach in the UK is to make sure that you have to make the preceding stage of an amplifier can never pass a signal through that is capable of driving the next stage into slew limiting. Even if the preamp gets an unlikely high-frequency signal from a record scratch, it won't overdrive the power amp.

Johnson: I can accept that, even agree with it, but I still take the position that it isn't necessary for an audio amplifier ever to receive information that will drive it into slew limiting. Where is it coming form, if it isn't off disc or tape?

Messenger: Well a moving-coil cartridge is certainly a wide-bandwidth source, particularly with the possibilities of groove chatter and new stylus shapes. I've seen measurements that showed output from moving-coil cartridges up to half a megaHertz. The amplitude might be lowish, but there might be problems with the sort of distortion signal that you could get from a moving-coil cartridge that doesn't have the internal bandwidth limitation of a moving magnet.

Johnson: Might be, but I've used moving coils exclusively for at least a decade and I've never any such problems.

To return to amplifiers, just yesterday at a UK dealer's, I had a demonstration of what we've been discussing. We had two amplifiers, one with a relatively low and the other with a very high slew rate. We listened first to one then to the other, and it was very clear that the amplifier with the low slew rate was much faster than the one with the high slew rate, much faster. When you hear that one amplifier is faster than another, the attack transient is coming out in the proper time alignment, this is nothing to do with slew rate. Slew limiting would occur at very high frequencies and while it might give us a weird form of hard "glassy" HF distortion, this wouldn't be down in the range where we would recognize this subjective area of "fastness." What I believe we are actually listening to here is the speed of the amplifier's power supply under dynamic conditions.

Atkinson: You're talking about the effects of current starvation?

Johnson: Exactly, or worse than that, interaction between supply and amplifier.

Messenger: It has been suggested that a reasonable rule of thumb would be for the power supply to have a bandwidth an order of magnitude greater than that of the actual amplification circuitry.

Johnson: That might at least be a good starting point. The power supply is certainly a neglected area in amplifier design/

Messenger: It is very difficult to find valid measurement techniques for assessing amplifiers.]

Johnson: Again take bandwidth. Our vacuum-tube preamps, the SP8 and SP10, have –3dB points around 0.15Hz and 250kHz, which is wide enough for audio but the phase shift [within the audioband] is relatively low. We recently looked at our SP6 preamplifier, which is not yet discontinued but I sort of obsolete. (The '8 and '10 are our later models but are comparable.) AT 10Hz, the phase shift is about+3° and at 20kHz it's about –2°, which is low for a tube product. However, I don't think the absolute phase, that is precisely 0° or not, is as important as its stability under dynamic conditions. In other words, if the circuit design is such that that the phase angle shifts under dynamic conditions, this is very bad. It's far better to have a few degrees of constant phase shift than to have a phase shift which is not fixed—too many designs are that way.

Atkinson: But how do you assess phase aberrations. . .

Messenger: . . . under dynamic conditions?

Johnson: It's just as difficult to measure that as it is to make any dynamic measurement. Hi-fi is one of the few industries where virtually all the measurements are done one way but the product is used in another!

If the power supply parameters were to shift under dynamic conditions, then you can be assured that all of the things we're talking about, phase and other kinds of distortion, will occur. The power series are in series with the active circuit elements and the more complex the circuitry, the more trouble the power supply can be. Obviously, if the power supply can be made effectively inert, if you can get down somewhere near to zero ohms at al frequencies under all dynamic conditions, then you have it out of the way, so to speak, and you're then free to optimize the rest of the circuitry.

You might find it interesting to know that the impedances in the SP10 power supplies—there are individual capacitance multipliers for each active stage—are in the high micro-ohms in the midrange, around 1000Hz. That's lower than the power supply impedances found in most solid-state power amps, yet this is for a valve preamp. But it contributes to quality sound.

Atkinson: So effectively driving the power supply rails for the preamp is an active DC power amp.

Johnson: We have a very high-speed active amplifications system providing the power supply, typically a series valve driven by another valve with an IC, a modern BiFET chip, used as an active error sensor and amplification device. The loop gain may be as high as 10 million but as it is degenerated to just 1x, to unity gain, you end up with incredibly low source impedances.

Messenger: Didn't your totally solid-state products get a somewhat different critical reaction on their introduction in 1976 to the tube designs?

Johnson: They got a reaction all right, but it wasn't the same!

Messenger: Any thoughts, with hindsight, on what went wrong?

Johnson: I'm not sure that we did anything wrong, nor that there was anything particularly wrong with the market. I think that the problem was that Audio Research's place in the market was created with valve products and when we abandoned valves, our market niche changed and we simply didn't enjoy those customers any more, we had to create a new market. Actually, I think we have some very good solid-state products but we are certainly more successful in selling valve equipment.