Dear La Folia 11: This Time ’Round, Amplifiers

Russell Lichter

[August 2003.]

Dear La Folia,

Picture a summer evening, a little house on a hill in the country, cool evening air flowing in through the screen door, a couple of middle-aged audiophiles (who obviously are neither successful stockbrokers nor consultants to the Reelect Bush campaign) squinting and scratching their heads. Perceived acoustic certainties appear like Venus on the half shell, only to crumble in a penumbra of doubt: Do these two amplifiers, a Bel Canto eVo2 and a Spectron Digital One, really sound different? And, God save the mark, if they do sound different, which one sounds better?

Both of these “digital” amplifiers are Class D, or Pulse Width Modulated, which, in terms we lay folk can understand, means that a low-level analog input signal (from a preamp, or DAC / CD player with volume control) is converted to square wave pulses at a fixed or variable sample rate, where the pulse width varies according to the voltage of the analog input. This PWM signal triggers a pair of transistors per channel, one switching a positive rail voltage and the other a negative rail voltage. The output of these devices is the same pulse width modulated square wave as the input, but at a higher voltage and with lots of available current. Sending these square waves through a second-order low pass filter (an inductor in series, a capacitor in parallel) removes the “carrier” (the sampling rate pulses, usually in the neighborhood of half a megahertz), leaving the original analog waveform. Far as I’ve been able to determine, these types of amplifiers have two characteristics in common: very high efficiency (90+%) and very fast and accurate correction, either in the form of feedback or digital signal processing.

I don’t want to worry the “Class T” propaganda from Tripath, faithfully parroted as it is by reviewers and manufacturers around the globe. The Tripath chip, used in the Bel Canto, is very sophisticated, employing a variable sample rate based on predictive algorithms, capable of “learning” the characteristics of the switching transistors and so forth, but PWM it is and PWM it remains. I once got into a fairly tepid email exchange with a well-known reviewer as to whether these so-called “digital” amplifiers — all of which have PWM output stages, even if they sport S/PDIF input — are digital or analog in nature. I took the latter view, since pulse width variation / polarity is directly analogous to the amplitude / polarity of the input signal.

So there we were, rolling around in Earth’s diurnal course, with green pens, cones and cables at the ready. Is there any among us so level-headed, acute, unbiased, and of such healthy cochlea, that their perceptions are unencumbered? The truth is, I don’t have golden ears. Or, probably, silver ones. (This is the one thing an audiophile must never admit.) Also, from an objective viewpoint, there were things wrong with the test procedure. The eVo2 is Bel Canto’s latest, whereas the D1 is long out of production. Spectron would not regard it as representative of the best they have to offer (the Musician II using proprietary loudspeaker cables that provide feedback from the speaker terminals). Worse, though the test was ostensibly single-blind, switching between the amplifiers took 20 to 30 seconds, plenty of time for acoustic memory to fade. Finally, and perhaps fatally, due to stepped volume controls on the preamp, level matching was at best only within 0.75db. [Russell makes an excellent point. It’s an old showroom trick. If the salesman plays the item the salesman’s pushing louder than that with which it’s being compared, it generally sounds better. Ed.]

Nonetheless, we were at a loss to find a difference between the two. We liked both amplifiers, thought them wonderfully transparent, articulate, dynamic and powerful. We would have liked to find differences to prove we’re real men, like the reviewers who have extravagantly praised one over the other. Maybe differences were there and would have been obvious if the levels were matched, or if it didn’t take so long to switch between amplifiers, or if we’d had acupuncture on the soles of our feet. But I will say, if you’re looking for an amplifier, these two ought to be included in your auditions.

Going back a few years, I remember when there were sporadic shootouts on Usenet over whether the $2,100 Spectron Digital One amplifier was even comparable to the $10,000 TacT Millennium. As an owner of the former I was gratified to find a small preponderance of opinion deeming the Spectron “more musical” than the Millennium, despite the latter’s ball-bearing volume control and chassis machined from a solid ingot of kryptonite. (A bit of the old apples and oranges: The Millennium had only digital input, whereas D1 owners could swap DACs to their heart’s delight.) Then there was the guy who insisted that while my Digital One was a mere Class D (Pulse Width Modulated) amplifier, his Millennium was really and truly and actually digital. Bull. The output stage of the Millennium is PWM and employs a second-order low pass filter, just as the Spectron and the eVo do.

A lot of water has passed under the bridge and other locations since then. As you may know, the audio groups on Usenet are superb illustrations of that French adage, Plus ça change, plus c’est la même chose. For a period of almost two years I remained unsullied by rec dot audio dot anything, but when I revisited a couple of weeks ago, lo, I found the same boring players endlessly arguing over the same boring things: The same spiteful angel peppering the marketplace with claims of rip-off, quoting imaginary blue book prices, the same frustrated engineers pointing out the difference between subjective and objective, talking about Nyquist, RIAA equalization, and on and on. Is vinyl better than the silver disc? Do cables make a difference? Are SACDs better? Are cones really mechanical diodes? I must admit, no one seems to be talking about the Millennium any more. But the TacT 2150, at a paltry $5,000, is apparently a contender. The 2150 boasts no negative feedback, presumably relying completely on the capacity of its proprietary digital processing algorithms to predict and control distortion. Negative feedback, as you probably know, compares output with input, takes the result (which is distortion), inverts it and applies it to the input, thus (theoretically) canceling distortion. Even though a finite time is needed to effect this feedback, the system does work, and is employed in audio amplifiers as well as guided rocketry. One factor that may account for the transparency and resolution of digital amplifiers is the speed of correction. The Spectron’s feedback loop is 200nS, compared to a typical 2000nS in Class A and A/B designs. Predictive corrections as employed in the eVo and TacT should be almost instantaneous. The accuracy of these predictions is another issue.

I remember one guy on Usenet who told us how he accomplished a really vast improvement in his system for an investment of just a few dollars. Was it a pair of Dynarski Room Pebbles strategically placed due north on his loudspeaker cabinets? Rare-earth magnets taped to the power cord of his preamplifier? A new set of machined titanium cones placed under his CD transport? No. He had a doctor clean out his ear canals. As he said, this produced greater improvements in imaging and frequency response than upgrading all his electronics would have done. I don’t doubt it. What surprises me is that years have passed since then, and still there are no audiophile-grade Q-tips made of six-nines cotton wool and non-resonant, green MDF stems.

Take it easy.