Those Durable Bottles

I believe it was 1958 when I first heard a transistorized audio product. The Fisher TR-1 was a small battery-powered box that provided microphone preamplification and inputs for three magnetic phono sources.

I remember being completely bowled over by the sound. It did some things far better than any tubed component I had heard, and I was so impressed by those things that I did not at first detect the things the TR-1 was doing worse than tubes. I am confident that I would run screaming from the room were I to listen to it today, though a lot of its shortcomings were doubtless being masked by complementary shortcomings in the loudspeakers of that day.

Soon the TR-1 was joined in a proliferation of more sophisticated and better sounding solid-state components, some of which did many things well enough that I, and the audio community at large, was confidently predicting that Lee De Forest's little glass bottle, the clumsy old vacuum tube with its finite life and its wasteful generation of heat, would be dead and gone within 10 years. Twenty five years later, the vacuum tube is not only still with us, it is enjoying a renaissance among audio perfectionists.

What is the continuing attraction of this technological dinosaur? Certainly not convenience, reliability, or cost. A tube is very inconvenient to use, requiring as it does plenty of space around it for ventilation and a separate, high-current power supply for its heater (fig.1). Its microphonic tendencies are evident in high-gain circuits, and a tube needs special flexible mounting to keep vibration away from it.

Fig.1 An amplifying vacuum tube consists of at least four basic components: the heater (1) which agitates the electrons on the surrounding cathode (2) so that they more easily escape, weaving their way through the grid (3), which modulates the current flow—on their meny way to the anode (4).

It's not very reliable either; output tubes tend to break down at the drop of a hat, and even a tube that doesn't suffer heater burnout before its estimated lifetime will, as inevitably as a living organism, come to its demise. Even worse, the sound of many tubes deteriorates over time, a known cause of tremens audiophilius.

Given enough time, every tube will wear out and have to be replaced. Which brings us to cost, another thing definitely not in the tube's favor. Premium brands tend to be quite expensive, ranging from $10 for a low-current, dual-triode 12AX7 to $30 or more for a high-current output tube like the EL-34.

What, then, does the vacuum tube have going for it? Nothing at all, except its sound. And that sound is so unlike solid-state sound that a tube enthusiast can tell one from the other immediately. Solid-state components tend to be crisp and sharply-etched at the high end, somewhat laid-back through the presence range (as though the instruments reproduced were slightly backed off from the recording mikes), and very taut and deep at the low end. Tubed components tend to be a bit soft at the extreme top—though this is less true for the very best modern components—rather forward and alive through the upper middle range, some what warm and ill-defined through the mid bass, and weak at the extreme bottom—again, less true of the best designs like the Berning 2100 and the EAR 509. They also seem to reproduce depth and front-to-back perspectives better than do transistorized components, sometimes with an uncanny feeling of three-dimensionality around the sound sources.

The tube sound has been described as "musical" by those who like it, and as "euphonically colored" by those who don't. It is in fact a little of both. Livemusic highs do tend to be more soft than crisp, and the overall sound is somewhat warmer and richer in spectral balance than is usually heard from solid-state electronics. While many tubed components seem to overdo these qualities, it can be argued just as convincingly that many solid-state components underdo them, making the sound too lean, dry, and crisp. (This sounds like a discussion of fried chicken.)

Many observers have suggested through the years that the ultimate truth—true accuracy—may lie somewhere between the sounds of tubes and transistors (footnote 1). And indeed, there is increasing evidence that this may in fact be the case, as some of the best solid-state products come to sound increasingly like tubes, and tubed products have picked up the virtues of transistors: extended bandwith at top and bottom. Yet the characteristic differences persist, and to those who hear those differences, there is often little choice. Either they prefer the tube sound, or they prefer solid-state sound. Yet few are able to explain their preference, except in such simplistic terms as "more musical" or "more accurate" or, lamely, "I prefer it."

The continuing consumer interest in tubes still baffles most of the so-called scientific community—aka the test-bench crowd. Seemingly unable to hear what all the fuss is about, or to cite measurements which correlate with the differences audiophiles claim to hear, they cast about for "rational" explanations of the observation that relatively high distortion measurements (tubes) should sound better to some listeners than products with vanishingly low measured distortion.

The most commonly heard explanation is that tubes "overload more gracefully. " This is often actually true, because the transistor's superior (or exaggerated?) detail makes the tiniest amount of waveform clipping immediately and obnoxiously evident. Not only that, the nature of tube clipping is more like a compression (fig.2) where the transistor-clipped waveform is an irritating-to-the-ear truncation (fig.3). But that argument can't hold all the water, because the things that distinguish tube sound from transistor sound are audible at all output levels, including those far below the component's overload point. (Nonetheless, many rock musicians prefer tubed amplifiers for use with their gee-tars and Fender basses because they do sound better when overdriven to the excessive levels beloved of all rock musicians.)

Fig.2 "Soft" clipping typical of tube power amplifier.

Fig.3 "Hard" clipping typical of solid-state power amplifier.

In truth, the reasons for a tube component's unique sound are not clearly understood, although there have been some canny theories presented. David Hafler (president of the Hafler Company, which makes only solid-state electronics) has hypothesized that a lot of the difference has to do with the relative distribution and strengths of the spurious harmonics generated by tubes and transistors. For example, a tubed amplifier which measures 0.05% harmonic distortion might produce two thirds of that as 3rd harmonic content and one third as 4th, whereas a transistor amp with the same measured THD may produce half of that as 3rd-order distortion and half as 4th-order.

And while the tubes may yield diminishing amounts of distortion with increasing order (5th, 6th, and so on), the transistors may yield the same content at each harmonic all the way out to the l0th. This strikes me as being one of the most convincing explanations we've heard to date, but it hardly explains all those other differences.

Bill Conrad (of Conrad Johnson Design) thinks that tubes simply obscure less of the very-low-level information that give us subtle aural clues as to the size of individual instruments and their location in space. Others have speculated that tubes are "nicer" to electrons that the electrons "like" travelling across a vacuum more than they like travelling through a semiconductor substrate (footnote 2).

It is probable that tubed amplifiers' typical weakness at the extreme low end is related to the limitations of their output transformers, whose ability to deliver LF current is directly related to the size and weight of the iron core and is thus constrained by economic realities (as well as by its movability by anyone but a sumo wrestler). But the "aliveness," midbass warmth (footnote 3), and seemingly enhanced depth all defy any explanation in terms comprehensible to design engineers.

Tubed component design remains rather much a black art, where measurements must at some point in the design process yield to the designer's subjective judgments as to how the damned thing sounds and to heck with the measurements. This is a singularly unenviable position for any designer to find himself in, for the whole area of subjective evaluation is fraught with more pitfalls than the video game. (We who skirt these pitfalls on a regular basis can sympathize more than most!) First, he must divest his listening from personal biases, and listen for what he believes consumers will listen for—easier said than done. Then he must attempt to second-guess his target market in terms of what they will like and won't. (I hear that, but will buyers consider it a strength or a weak ness?) He must convince himself that the signal sources he uses for evaluation. are, if not unquestionably accurate, at least representative of what his potential customers will use. And then there's the question of his loudspeakers, and that's a loaded question.

All loudspeakers are colored. He knows that, intellectually at least. So are all amplifiers, more or less, including his prototype tubed amp. What speakers should be used in evaluating the sound of his amplifier? And best in what respects? Electrostatics, generally, tend to have scads of detail and, often, rather thin low end. For electrostatics, he should go for a soft, "sweet" high end and a full low end. But most dynamic speakers require just the opposite qualities from an amplifier: crisp high end and tight, controlled bass. Should the sound be tailored to one kind of speaker and slight the other? Or would he do better to compromise and end up with something that sounds passably good, but not superb on either? Decisions, decisions.

As a result of all this, good tubed electronics inevitably differ in sound from one to another, much more than do good solid-state units. This raises legitimate questions regarding the "accuracy" of any of them—they can't all be accurate when they sound so different. But it also greatly enhances a buyer's opportunity of finding an amplifier that exactly meshes with his favorite speaker to produce the best sound that speaker is capable of delivering. Thus, in a perverse sort bf way, the variety of tube sounds can be an avenue toward incredibly listenable, satisfying, and accurate sound.

Tubes are not the only such avenue, of course, though I won't hesitate to admit that some of the best sound I've heard has been from systems using tubed electronics throughout. All of Sheffield Lab's recent recordings, as an example, are brought to us through the courtesy of all-tubed electronics. On the other hand, some of the best sound I have heard (in other respects) has been from solid-state systems. I could happily live with either, as long as they were hooked up to speakers which complemented their respective characteristics.

Perhaps, then, the main reason why the vacuum tube persists in audio (when the transistor has replaced it in every other area of electronics) is that it continues to provide the kind of listening pleasure that every audiophile envisioned in his mind's ear when he got into audio in the first place. I can't think of a better reason.

Footnote 1: Pontius Pilate would have queried, "What is truth— tube or transistor?"—J. Gordon Holt

Footnote 2: And no one wants to listen to an irritated electron.—J. Gordon Holt

Footnote 3: The midbass warmth of tubes is actually not so mystical. Introducing a 30Hz rolloff point in a solid state amplifier would bring about midbass warmth also; in fact almost all rolloffs have audible results prior to the frequency where the rolloff commences on a frequency response graph.—Larry Archibald