Halcro dm38 power amplifier
The dm58 is physically large; when I tried the review samples out in my system prior to measuring them, they dominated my room. So when Halcro's affable US distributor, Philip O'Hanlon, asked me if I would be interested in reviewing the company's first two-channel design, the $18,790 dm38, I didn't need much persuading.
Two channels, not one
The dm38 is basically the circuit of the dm58 monoblock, duplicated for stereo but with reduced output power because the same heatsink area now has to be shared by two channels instead of being dedicated to one. The aluminum chassis has an attractive brushed-silver finish, and keeps the same H-shape of the '58. The uprights, sitting on wooden plinths, contain the output circuitry and heatsinks; the lower of the two crossbars is the power-factor-corrected switch-mode power supply, the upper the input circuitry and the I/O connectors.
With two channels to handle, as well as balanced and unbalanced inputs, the latter offering both voltage-mode and current-mode operation, the rear panel of the upper module is crowded. But there is still room to connect fairly heavy-duty cables, and the output binding posts are usefully insulated with rubber shrouds.
AC power is supplied via an IEC connector on the base of the lower module, where the main power switch resides. The amplifier is switched in and out of standby with a button on the underside of the upper module.
Vanishingly low distortion
When Halcro burst onto the US high-end scene in 2001, they made much of their amplifiers' astonishingly low levels of distortion, and of the fact that the excellent linearity had been achieved without the compromises introduced by traditional circuits that use excessive amounts of negative feedback, such as the shifting of energy from the relatively benign low-order harmonics to the much more objectionable (because they are not musically consonant) high-order harmonics. Such amplifiers would have low distortion only as long as you weren't using them to amplify a signal! And there would be other problems, such as slew-rate limiting and marginal stability.
To find out more about the Halcro amplifiers, I met with their designer, Bruce Candy, in London last November. ("Halcro," a Scandinavian name of Orkney Islands origin, is Bruce's middle name.) He was passing through en route from Helsinki and Australia, and I had just finished the sessions for Antony Michaelson's Mozart Clarinet Concerto project (see "Letters" in this issue). The most important part of the design is the output stage, Dr. Candy told me. He uses complementary vertical FETs, of a type normally used for switching, because they can turn on and off fast enough to eliminate crossover distortion. (Candy hates crossover distortion.) But the problem with these so-called switching FETs, he explained, is that it is "hard to make them stable with temperature so that all the devices draw the same current"—which is perhaps one reason they are not in widespread use in audio circuits.
Candy was reluctant to go into detail about how his patented Halcro output stage works, but he did tell me that the key to the design is that the distortion-canceling circuitry "float" up and down with the output stage. So even though the output-stage voltage rails are conventionally high, this important circuitry "sees" only a small voltage, of around 10V. This allows him to use very wide-bandwidth small-signal devices in the error-canceling circuitry. "This is all classical microwave stuff," Candy explained, which left me none the wiser, other than to assume that circuits that handle microwave frequencies must, by definition, be fast.
Continuing our discussion, we touched on: why tube circuits tend to sound good; protection circuitry (Candy is for it, when it is done right); how hard it is to design a switching-mode power supply that won't overheat while operating at high speed with high efficiency; the optimal loading for Shure moving-magnet cartridges, which he feels still sound great on classical orchestral music (though he admires how a moving-coil design will reproduce opera); why the ubiquitous use of simulation programs such as Spice is not necessarily a good thing for electronics engineers; and even jet engines—"you must think of it in terms of gain and gas flow, the relative impedances," Candy elaborated, "if you want to understand how the engine works." The conversation veered back to audio in the shape of class-D or switching amplifiers—Halcro plans to introduce multichannel amplifiers for home-theater systems that use some innovative thinking on Dr. Candy's part.
But eventually we returned to the subject of Halcro's amplifiers and how hard it was for me to comprehend how an amplifier with as little distortion as one of Candy's designs could sound so radically different from a conventional amplifier when the distortion artifacts of the latter are already well below the threshold of human hearing. "That's a good question!" he laughed, venturing no answer.
When I got back to New York, I did some Web work and uncovered Bruce Candy's patents on the Halcro amplifier designs (footnote 1). Rather than use conventional negative feedback to reduce the level of errors, he uses a complex, active error-correction circuit. Correction of errors in amplifier design goes back, in my memory, to Quad's "current-dumping," feed-forward design of the mid-1970s. As Dr. Candy points out in his patents, Malcolm Omar Hawksford, Masayuki Iwamatsu, and Robert Cordell all published papers or patents on this subject in the 1980s, and I believe Nelson Pass also attempted something along these lines with the "Stasis" circuitry used in 1980s Threshold designs.
The circuit diagrams in the patents are way beyond my ability to grasp. In Candy's words when I spoke with him, "It definitely would take the average amplifier designer quite some time to recognise the circuit as that of an amplifier if presented with no direct clue as to its purpose"—and I am not an amplifier designer at all, average or otherwise! However, it appears that the essential innovation of the Halcro amplifier is indeed the use of a floating power supply for the distortion-canceling circuitry, to allow the use of small-signal devices, including a differential-input op-amp with an extremely high (100MHz) gain-bandwidth product, coupled with multiple nested correction loops and enormous care taken with isolating the input circuitry from the magnetic coupling effects of the high-current output stage.
Candy stresses that there are significant differences between his circuit and earlier distortion-cancellation designs. Unusually, the Halcro amplifier doesn't appear to use the usual fully complementary topology. Candy's feeling is that the complete cancellation of even-order distortion this produces works against his design goal of having less odd- than even-order distortion. He also explains that, wherever possible, the circuit is arranged to keep the voltage across and current through semiconductors under "very near constant conditions," to eliminate the variation of the devices' operating parameters with frequency, voltage, and current. (I conjectured 21 years ago, in the August 1983 issue of Hi-Fi News, that a possible reason amplifiers with class-A output stages sounded good was not so much the elimination of crossover distortion, but the fact that the output transistors were carrying a constant current and were therefore immune to having their temperature, hence gain, modulated by the music signal (footnote 2).)
"To this day, I have yet to hear any amplifier that equals the dm58's combination of complete neutrality, harmonic generosity, lightning reflexes, and a sense of boundless power that is difficult to describe," was how Paul Bolin summed up his experience with Halcro's dm58 monoblock. It also nicely describes my reaction to the dm58 when the review pair briefly spent some time in my listening room.
Footnote 1: Two key patents are US 5,892,398, "Amplifier Having Ultra-Low Distortion," April 6, 1999; and patent application US 2003/0058044 A1, dated March 27, 2003.
Footnote 2: My thanks to Dutch reader Hans Polak for sending me an informative analysis of this subject, and for drawing my attention to the writings of French engineer M. Perrot on this subject, which Perrot calls "Thermal Distortion" and claims is a reason tube amplifiers are preferred to solid-state.