Landmarks in Power Amplifier and Preamp Design: Tubes vs. Transistors vs. Both: Audio Research M300; Audio Research SP11; Boulder 500 ; MESA/Baron M180 (modified)

--------

We take a revisionist (or is it just sober?) look at Audio Research's hybrid flagships and then find greater comfort in unmixed circuitry.

It is our pigheaded conviction that a high-fidelity amplifier's output should strongly resemble its input. Let us be more specific. A power amplifier's output should be an exact replica of its input, at an amplitude determined by its gain, into any resistive or reactive load down to a certain minimum impedance, over a certain dynamic range deter mined by its power rating. A preamplifier's output should similarly replicate its input at a higher amplitude, but modified by whatever intended equalization, filtering and/or other response-shaping characteristics are incorporated in the circuit. All that may amount to no more than coming out in favor of motherhood, the flag and apple pie, but it is surprising how often designers and reviewers ignore, or weasel out of, such truisms. There are even those who believe that totally accurate amplification does not sound good. The fact is that totally accurate amplification reveals the input that does not sound good.

A thoroughly knowledgeable designer is able to make the output bear a high degree of resemblance to the input by means of either vacuum tubes or solid-state devices. Tube sound vs. transistor sound may be a legitimate issue where routine (or shall we call it classic?) circuitry is used, but any previously quantified transfer function can be predict ably duplicated either way, as we point out at some length in another article, and thus the same sound achieved. The only exceptions to that generalization we can think of are due to tube aging, which can modify the transfer function, and to output transformers, which impose certain limitations at the lowest and highest frequencies. (Output transformerless vacuum-tube power amplifiers, on the other hand, cannot drive very low-impedance loads, e.g., New York Audio Laboratories' otherwise remarkable Futterman OTL amplifiers.) Since all of the drawbacks, minor as they may be, are on the tube side in a rational comparison, we would always opt for the unlimited flexibility of solid-state circuitry when planning a new design from scratch. That does not mean that a designer who is more comfortable with tubes cannot come up with a superior amplifier.

The widely divergent designs reviewed below represent some of today's strongest statements, credible or not, on how to deal with the above considerations.

Audio Research M300

Audio Research Corporation, 6801 Shingle Creek Parkway, Minneapolis, MN 55430. Model M300 Hybrid Monaural Power Amplifier, $4900.00 ($9800.00 the pair). Tested samples on loan from owner.

A pair of these not particularly handsome invitations to hernia, weighing 110 pounds each, will set you back about the same amount as a 1988 Honda Accord LX sedan.

So, when we first set eyes on them, we assumed we would find extraordinary parts under the cover, worthy of a piece of NASA gear. Not so. Everything inside is of commonly good quality; the transformers are huge, of course, as the rated power is 300 watts into 8 ohms, and there is a number of audiophile-brand film capacitors on the board, but we could see nothing in the way of hardware or workmanship that would account for more than half the price of the M300, including all markups. The other half you pay for Bill Johnson's high opinion of his amplifier.

The circuit is called hybrid because it uses FETs only up to the output stage, where eight beam-power tubes (6550s) and an output transformer take over. "Most hybrid power amplifier designs use small signal-input vacuum tubes, then rely on FET's for the output stage," explains Audio Research's blurb on the M300. "Audio Research has chosen instead to eliminate these low-level input tubes- tubes which can be prone to problems with hum, noise, microphonics, drift and, of course, require periodic replacement. The new FET 'front end' is combined with Audio Research's long-famous, patented cross-coupled circuit... [and] with yet another Audio Research patent: an output stage utilizing vacuum tubes and partially cathode-coupled ...but with the screen (or accelerating) grids operating with a signal voltage precisely in phase with the cathode volt age," etc. We buy the criticism of front-end tubes but fail to see, despite further rationales offered by Audio Research, why a much simpler transformerless output stage, using power MOS FET's or other solid-state devices, could not have been designed to achieve the same performance goals.

We suspect that the reason for the use of tubes in the M300 is political; Audio Research decided that the image of their flagship power amplifier must somehow be associated with tubes, and never mind the FET's, because the company's prestige is built on a tradition vacuum-tube components.

Our measurements resulted in a mixed bag of goodies and not-so-goodies. The claimed power output of 300 watts minimum CW at 8 ohms from 16 Hz to 25 kHz with less than 0.5% THD was confirmed only up to 12 kHz; further up, 15 kHz clipping was observed at 288 watts with 0.68% THD and 20 kHz clipping at 264 watts with 1.2% THD. It is an outside possibility that our line voltage was a little lower than the specified 120 volts, although in the region of 100 Hz to 500 Hz the clipping point was barely below 350 watts. More disturbing than that small discrepancy was the bizarre way the amplifier went into clipping at the higher frequencies, starting at approximately 4 kHz. Instead of symmetrically flattening the sinusoidal waveform, clipping superimposed a large sawtooth pattern on it just before (but not after) each zero crossing. This may be a feedback-related anomaly; we have no way of determining its effect on the sound we heard.

Another cause of raised eyebrows was the frequency response peak of 1.1 dB at 96 kHz. Undoubtedly due to an output-transformer resonance, it was sufficiently low in Q to spill down into the topmost octave of the audio range, effecting a minuscule treble boost, of the order of 0.2 dB at 20 kHz, equivalent to a tiny RIAA equalization error in a preamplifier. Audible? If so in a preamp, then so in a power amp-right? Output impedance at the 8-ohm tap was measured to be 0.281 ohm, yielding a damping factor of just over 28 and confirming the specification of approximately 30; the amplifier acts as a current source to an ever so slight degree. Hum and noise were negligible.

After a good many hours of warm-up, we did some serious listening. We cannot say we were disappointed, as our expectations were somewhat short of sky-high by then.

Yes, the M300 sounded "good" in a very general sense; it is, after all, a reasonably clean and very powerful amplifier.

To the critical ear, however, the sound was definitely on the rough and sibilant side, slightly spitty one might even say, and not really pleasant overall. Was it a case of a less than perfect input accurately reproduced? If so, why did the all tube MESA/Baron M180, reviewed below, sound so natural and so right in the same system? Or maybe the M300 needs a five-day warm-up-the kind of tweako ritual that should not be necessary with a good design. That, dear readers, we shall never know. Our listening tests ended much sooner and with considerable finality.

What happened was that, without any provocation, one of the pair of M300's suddenly quit on us. A short from control grid to screen grid in one of the 6550's created some kind of avalanche effect that traveled upstream and took out all the FET's-a catastrophic failure mode apparently endemic to this model. A well-known Audio Research dealer, who must of necessity remain nameless, confessed that all of his M300 customers, without exception, had run into this and similar problems, requiring massive warranty repairs.

The moral? There is no amount of unhappiness in this materialistic world that $9800.00 cannot buy.

Audio Research SP11

Audio Research Corporation, 6801 Shingle Creek Parkway, Minneapolis, MN 55430. Model SP11 Hybrid Stereo Preamplifier, $4900.00. Tested sample on loan from owner.

This is Audio Research's all-time top-of-the-line pre amplifier, already canonized by the high-end pontiffs and, one can assume, the input of preference into a pair of M300's. The price is the same as that of a single M300, and again we must point out that we found nothing inside the two chassis (the power supply is housed separately) so obviously costly or exquisitely wrought as to justify more than half that amount, if that much, typical markups included. Bill Johnson certainly knows how to charge; the Light Brigade could have taken lessons from him.

The hybrid circuitry is even harder to rationalize than in the case of the M300; the vacuum tubes are embedded be tween the FET's within the same gain stages of the SP11, and in such a redundant way that one could bypass the tubes and still have a working circuit. Audio Research seems to have some sort of cookbook philosophy to the effect that tubes in a preamp give you a musical and natural flavor, whereas solid-state devices are for quickness and dynamics;

put the two together and voila-the sauce is magnifique! Here again, our best guess is that Audio Research's flagship preamplifier simply had to have some tubes in it, regardless of function, for purely political reasons. (We are reminded of a story from the late 1960's about the tyrannical Avery Fisher, who objected to a blank space on the escutcheon plate of a prototype FM tuner and told the engineers to put a button there in the final version. When it was tactfully explained to him that there would be no function for such a button, he snapped, "Find a function!" and stormed out of the lab. They found one.) The SP11 has a plethora of controls, including a few slightly idiosyncratic ones; their operation has been amply covered in other reviews and need not concern us here. The same goes for all the inputs and outputs. On the laboratory bench the SP11 acquitted itself with distinction in nearly all respects. The phono stage has 46 dB of gain, which is a little scant for very low-output moving-coil cartridges, but with the line-stage gain of 26 dB, resulting in 72 dB (i.e., 4000 times) total available amplification, there will be relatively few phono sources that cannot be accommodated. We were impressed by the almost unmeasurable THD of the phono stage, as well as the excellent overload characteristics of both stages. Signal-to-noise measurements were also entirely satisfactory, and for once the RIAA equalization is dead accurate, as close to the standard as we have ever seen.

Interestingly, the high-frequency response of the line stage rolls off to -0.2 dB at 20 kHz and -2.0 dB at 100 kHz, thus compensating fairly exactly for the minutely rising response of the M300. We refuse to speculate whether or not this is deliberate or even marginally significant. Of greater concern is the fact that the input impedance of the phono stage cannot be modeled simply by a resistor in parallel with a capacitor; the complexities of the phono circuit create an impedance requiring a much more elaborate model, and that is not good design practice to the best of our knowledge. (This has nothing to do, by the way, with the switchable resistor values for cartridge loading at the phono input of the SP11.) Of course, the proof of the preamp is in the listening, and the SP11 sounds very good indeed. After warming up the unit for several hours, our initial listening reaction was much, much more favorable than in the case of the M300, although we cannot say we were transported to a realm of unearthly delights like certain reviewers. We merely found the sound to be open, clean, focused, highly defined, well controlled and quite neutral. Something we could live with.

The mystical masochists of the high end, to whom Audio Research has often meant the promise of redemption through suffering, are unlikely to get their five grand's worth of delicious misery out of the SP11, as it seems to be a square-shooting, not particularly temperamental piece of equipment with excellent mainstream performance. Its special piece of bad luck in the course of our listening tests was that, through sheer coincidence, we happened to have temporarily available for side-by-side comparison a reconditioned Citation I, the late Stew Hegeman's more than a quarter-century old all-tube kit preamp.

The SP11, so nice by itself, somehow began to sound a little nasal, rough and strained next to the utterly smooth, unflappable, musically un-contradictable Citation I.

The latter is one of the very few designs known to us that does not take the RIAA playback equalization standard at face value but takes cognizance of the fact that, in the record cutting process, the preemphasis characteristic cannot logically rise at the rate of 6 dB per octave to infinity, as the playback standard seems to imply. It has to flatten out at some point, and the ideal playback curve would mirror that flattening out. Although there are as many ways to trim the top end of the preemphasis curve as there are cutter heads and mastering engineers, Stew Hegeman had a very good gut feeling for the typical deviation and adjusted his design accordingly. Is that what accounts for the difference in sound? Possibly, but we also suspect that the elegant all tube circuit of the Citation I may in some very basic way be superior to the hybrid complexities of the SP11. We are not going to spend sleepless nights trying to resolve the matter because (a) the Audio Research SP11 is not a good buy for the money and (b) the Citation I is extinct except for a few second-hand specimens still floating around.

In all fairness, we should point out to new readers that The Audio Critic has never found an Audio Research component to be the absolute best of its kind, not even once. Sorry about that; we just call them as we see them.

Boulder 500

Boulder Amplifiers, a division of Silver Lake Research, 4850 Sterling Drive, Boulder, CO 80301. Boulder 500 Power Amplifier, $2875.00. Tested sample owned by The Audio Critic.

This unique power amplifier has been around for about two years, but not many audiophiles and even fewer reviewers have had much experience with it. It is quite a bit better known in the professional field. We have been using it as one of our reference amplifiers since shortly after its debut, so we are obviously in favor of it and would like to call the attention of our readers to the Boulder story.

Jeff Nelson and Randy Gill, the two refreshingly un pretentious but extremely competent engineers behind the Boulder line of audio components, come from the world of professional studio electronics and are totally devoid of the usual hi-fi phoney-baloney. Their amplifier technology is based on the work of Deane Jensen, formidable author of some of the most powerful computer programs for electronic design and developer of the JE-990 discrete operational amplifier, the circuit concept at the heart of each Boulder product. The JE-990 op amp is not an integrated circuit but an amplifier module made up of discrete transistors, diodes and other components. It has been in the public domain for seven or eight years and has an almost legendary reputation among professionals, but to our knowledge only Boulder has used it in consumer audio equipment.

The 990 circuit enables the amplifier designer to eat his cake and have it, too. He can eat up harmonic distortion until it is reduced to near-zero levels. That, as we all know, requires large amounts of negative feedback, which in turn will create transient instability, right? Well, with the 990 he can have it, too-mucho negative feedback, that is- because the open-loop compensation is so sophisticated that there is sufficient phase margin to prevent all overshoot and ringing. The Boulder 500 stereo amplifier circuit uses two sequential 990 op-amps in each of its channels, the first for most of the voltage gain and a second, highly beefed-up one for the power stage. That may sound simple, but the in nards of the amplifier look immensely complicated, not only because of the inherent complexity of the 990's but also on account of a whole arsenal of high-tech goodies that would take several pages to do justice to. Balanced and un balanced inputs, mono bridging, protection circuits without audio intrusion, visual indicators for everything, high quality input level controls-and the list has hardly begun.

With all that stuff in there, we cannot even get too upset about the price; we wish they could have done it for less, say $1995, but the difference is not enough to raise the suspicion of an early retirement fund for Jeff and Randy.

As far as measurements are concerned, the Boulder 500 is very close to a distortionless voltage source. The power supply is not quite the Krell-like beast capable of pumping increasing current into decreasing impedances down to a virtual short circuit, but it is still awfully good; the "500" designation is earned with 250 watts per channel of totally clean, continuous power into 4 ohms across the entire audio band, 20 Hz to 20 kHz. When we say totally clean, we mean that the THD is difficult to measure with a Sound Technology 1701A distortion meter at that power between those frequencies, the residual distortion of the instrument being 0.0009 % at 1 kHz and a little higher at the frequency extremes. Transient waveforms are also improbably clean; bass response goes below the range of our signal generators because of DC coupling (but there is a very good servo to prevent DC at the output); the output impedance is also hard to measure because it is so close to zero, resulting in an almost perfect voltage source. In effect, over a band width considerably wider than the audio range and within the current capabilities of the power supply, the transfer function of the Boulder 500 approximates very closely that of a straight wire with gain. The question is-how does that kind of transfer function sound? Well, it sounds undeniably different. Those who are used to the warmth and lushness obtainable with a soupgon of second and third harmonic distortion will call the Boulder sound a little dry, or perhaps on the cold side, or even a bit hard or bright. We call it accurate. With the right program material and speakers designed to be driven from a voltage source, the sound is a new experience in transparency and detail, beautifully balanced and controlled from bottom to top, and smooth as silk. If the recording engineers were pushing to punch through anticipated veiling, or if the speaker designers were pre-compensating for the possibility of slightly current-sourcey amplifiers, the Boulder 500 un forgivingly calls the listener's attention to those practices.

The basic statement it makes is, "If you don't like my out put, you don't like my input." That is why we like it as a reference. As a bland mediator of varying degrees of audio quality in a record collection it does not quite make it.

You may want to read Jeff Nelson's brief monograph,

"Too Many Notes: Harmonic Distortion Analyzed," which is generally included with the Boulder literature. It explains that "an amplifier's harmonic distortion... can easily mis lead those with even the best of ears into thinking that they are hearing something closer to the original when, in fact, they are not. The original recording is actually clearer and more distinct than that which a high-THD amplifier is capable of reproducing." We basically agree, but you have got to understand that to Jeff 0.02% THD is high.

MESA/Baron M180 (modified)

MESA Engineering, Inc., MESA/Boogie, Ltd., 1317 Ross Street, Petaluma, CA 94952.

Baron M180 Differential Feed back Amplifier (mono), $650.00 without mod (1300.00 the pair). Tested samples on loan from owner.

Ordinarily, we would probably have decided against publishing a full-length review on a discontinued amplifier, especially with a modification that was made only to a few pieces. This is no ordinary case, however; there is a genuine hope of some eventual benefit to our readers and perhaps other audiophiles, so we want to explain the situation as clearly as possible.

MESA/Boogie is a well-established, medium-sized manufacturing company in the highly specialized field of vacuum-tube amplifiers for electric guitars and basses. They have about 60 employees on the payroll, and their annual volume is in the neighborhood of $6 million, including sales in a large number of foreign countries. Thus, even though you may not have heard of them, this is not a case of tweaky amplifiers built by some tube freak in his basement. Randall Smith, the owner, is a musician by training and a tube circuit designer by sheer necessity, he claims, as there are no other designs out there that he finds musically satisfying. When asked why he designs only vacuum-tube amplifiers, he replies that he does what he knows best and what his musician customers like best. No harangues on the overwhelming superiority of tubes over transistors.

Equally refreshing is his assertion that no musician in his right mind would be willing to pay much more than $650 per channel for a guitar or bass amplifier; that represents the high end of the market. MESA/Boogie sells directly to the end user at list price, as well as to musical instrument stores at trade discounts that are much smaller than in the audio retail business. So much for background.

The M180 was designed as a monophonic, single chassis power amplifier for guitar or bass, but suitable also for general hi-fi applications. MESA/Boogie saw in it a possible entrée into the quality home-audio market and re named it MESA/Baron to take away the jive flavor. Not many were sold and it was discontinued some time ago, but not before an audiophile-oriented circuit modification was made on just a few samples. It was a pair of these that we stumbled into by the sheerest coincidence and tested upon the insistence of the owner, before even finding out about MESA/Boogie and Randall Smith. Currently, and some what confusingly, the company is marketing with greater success a stereo power amplifier called MESA Strategy 400 ($1250.00), which is identical to a pair of unmodified M180's on a single chassis sharing a common power trans former. More about that below.

The modified M180 we tested looked like a solidly made piece of professional equipment with parts of good quality (Randall Smith goes to great lengths to find the right vendors), giving a totally different impression from domestic-type audio components in the same price range.

The circuit uses a differential amplifier as the input stage, feeding a cascode driver stage (which is the mod, not used in the stock version), followed by an output stage of six beam-power tubes (6L6's) in push-pull, going into a very unusual output transformer, plain-looking but amazingly efficacious and clearly the result of prolonged R and D.

If tested as a "black box," without any knowledge of its innards, the amplifier could almost be mistaken for a solid-state design on the basis of some of its measurements.

Small-signal frequency response is ruler flat from 10 Hz to 100 kHz, without even a hint of an output transformer. Clipping, abrupt but very clean in the solid-state manner, occurs under most load conditions in the neighborhood of 220 watts at any audible frequency from 30 Hz up; power bandwidth referred to 220 watts is 20 Hz to 24 kHz. Thus the M180 (viz. 180-watt) designation is quite conservative.

At 1 watt into 8 ohms, THD is between 0.02% and 0.06% throughout most of the audio range, rising only at the frequency extremes: 0.2% at 20 Hz, 0.11% at 16 kHz, 0.42% at 20 kHz, all of it strictly second harmonic. At 180 watts into 8 ohms, distortion becomes a mixture of second and third harmonic, typically under 0.15%, but kicking up to 0.46% at 30 Hz and 0.85% at 15 kHz. Square waves look very good, even at 20 kHz. Input impedance is 470K, drop ping to 100K at 20 kHz. Output impedance is far from zero; we measured 0.873 ohm at the 8-ohm tap, yielding a damping factor of only 9 (i.e., not a perfect voltage source).

These test results, highly respectable as they are, did not prepare us for the sound of the modified M180. It is simply the best-sounding vacuum-tube power amplifier in our experience, with the possible exception of the utterly impractical NY AL Futterman OTL-1 at close to ten times the price. We had only a few other power amps available for comparison at the same time (see the Audio Research M300 review above), and only a limited number of speakers for testing the audible consequences of the slightly current sourcey interface, but the results were consistent; the M180 came out on top in every respect, even to our solid-state conditioned ears. Extremely transparent, totally neutral, beautifully smooth yet sharply defined, never on the verge of coarseness under stress, not really "tubey," either-the sound simply raised no problems and gave eminent satisfaction. We wish we had been able to compare the almost straight-wire-like Boulder 500 or one of the unshakable Krells with the M180; the contrast might have helped to bring out the all-tube Gestalt more clearly. The high output impedance remains puzzling; could that be what we really liked? Regardless of anything else, the quality-to-price ratio of the amplifier is truly astonishing and an embarrassment to the high-end tube scene in the audiophile market.

There remains the $64 question (or perhaps the $650 question) of whether or not the matter is academic, since the modified MESA/Baron M180 is not in production. Well, we bring you glad tidings. Randall Smith has assured us that he is willing to resurrect the design, even repackage it in any form that audiophiles might prefer, if he perceives some kind of demand for the product out there. He has the parts, the tooling and the production facilities; it would take him only a month or two, he claims. Our recommendation is that you write to him at the above address if you are seriously interested. As we said, MESA/Boogie is a real, grown-up company with worldwide distribution, so you will not be merely fueling the fantasies of some amateur genius looking for tweaky custom business. You could also look into the alternative of the MESA Strategy 400, which is immediately available, but without that far from trivial cascode modification we cannot see how it could possibly sound identical to the pair of amplifiers we tested.

--------

[adapted from TAC, Issue No. 10]

---------

Also see:

The Fourier Loudspeaker Story: Engineering, Finance, Politics

Various audio and high-fidelity magazines