Bel Canto Design Aida D/A processor

John Stronczer, Bel Canto Design's technical spark plug, meets my definition of an electronics renaissance man, ranging as he does from designing single-ended amps that glow in the dark (the Orfeo) to digital processors (the Aida). Actually, digital circuitry is one of John's specialties, dating back to his days at Honeywell.

The Aida is a prime member of Bel Canto's "operatic" series, which includes the Tosca and the Fidelio. John told me that the Aida was developed to provide a digital audio source that could compare with and be enjoyed alongside a high-quality analog system. In developing the Aida, the goal was to identify and correct the primary error mechanisms which afflict DACs.

A case of the jitters
As described in Stronczer's White Paper on the Aida, his first priority was to reduce the noise on the critical word clock used by the DAC. For theoretically correct conversion, this clock must be virtually free of noise or jitter—a task made exceedingly difficult by the digital interface standards (S/PDIF and AES/EBU), which specify that the critical clock be referenced to the incoming data stream. Because the data stream is susceptible to many types of corruption from the digital source through the interface electronics and interconnect cables, jitter creeps into the conversion process.

Like many commercial DACs, the Aida uses the Crystal 8412 input receiver chip, which can generate a clock with under 500 picoseconds of jitter. Unfortunately, this level of performance is inadequate for high-end performance. To improve jitter reduction, a clock regeneration method must not only deal with external jitter caused by modulation at the digital source or in the interface, but also must cope with Logic Induced Modulation (LIM)—a major, recently identified source of jitter produced by variation of the digital data representing the audio signal. These data variations modulate the reference clock through the phase-locked loop (PLL) typically used by the clock regeneration circuitry. The simple expedient of providing a low-pass filter in the PLL is inadequate to address this type of jitter. By its nature, LIM is correlated to the audio signal and can produce discrete tones at or below the system noise-floor. This is plain and simple sonic garbage that obscures low-level detail.

So how low an induced or LIM jitter level is required for high-end performance? Steve Harris at Crystal Semiconductor has shown that jitter levels as low as 200ps generate extraneous tones at or above the noise floor of a 16-bit DAC. Therefore, to push jitter artifacts a full 20dB below the noise floor would require jitter levels on the order of 20ps. Stronczer feels that LIM jitter is a potential source for the lack of "soul" in digital audio, and a major impediment to long-term listening enjoyment.

The Aida is outfitted with a proprietary dejitter circuit to provide the cleanest high-frequency clock to the Crystal 4328 DAC chip. The circuit is claimed to maintain digital signal–induced jitter to below 10ps. Separate power supplies are used to prevent modulation of the clock, and to isolate the critical clock from the rest of the digital circuitry of the processor.

The 4328 single-bit Delta-Sigma DAC was selected because of its inherent low-level linearity and circuit simplicity. Stronczer points out that this latest offering from Crystal Semiconductor has several advantages over earlier single-bit converters: the use of a fifth-order modulator to reduce quantization noise artifacts to lower than –120dB at frequencies where the ear is most sensitive; use of switched capacitor techniques instead of continuous-time op-amps for the initial filtering of the one-bit DAC output; and the inclusion of two MOSFET ICs in a single multi-chip module to optimize both the analog and digital circuitry, and to reduce radiated high-frequency noise.

The Aida's analog outputs are buffered by a class-A open-loop follower circuit using separate power supplies for each channel. This unity-gain output circuit was chosen over a stage offering gain to keep the signal path as pure as possible, and because it was felt to be unnecessary when using the processor in a typical system with an active line-level preamp. As a consequence, the nominal analog output is only 1.4V—3dB lower than the 2V CD standard.

The analog ground plane and power supplies are separated from the digital circuitry to prevent digital noise from coupling to the analog output. A broadband RF filter is also used to prevent RF noise from leaking into other components in the AC circuit through the power-supply cable.

After several months of flawless performance, my sample of the Aida developed a hiccup: it sounded as if the input receiver IC was malfunctioning. I suspected that electrostatic discharge (ESD)—perniciously common in New Mexico's dry climate—had done the chip in. This was confirmed by Bel Canto Design. They found that their on-chip ESD protection circuitry is inadequate under extreme ESD conditions. Supplemental protection is now being added in the form of high-speed diodes in parallel with the input IC. The unit performed fine thereafter, but as cheap insurance against static shock, I now run a small humidifier continuously in the listening room.

At first it was difficult for me to accept the inevitable conclusion, but after several days, there was no escaping the truth: a sonic coup d'état was being consummated in my listening room. My long-term reference, the Theta Generation III DS Pre, was being unceremoniously booted out of office—the Aida had established itself as a more capable resolver of low-level detail. It was unearthing more information on recordings I was intimately familiar with, and the soundstage was more transparent, easier to step into, more convincingly fleshed-out in 3-D relief.

Hooked on detail
This business of detail resolution needs some clarification. Back in 1982, JGH was the first kid on the Santa Fe block with a CD player—a Sony CDP-101, the embodiment of Sony's prophetic promise of "perfect sound forever." We were struck by some positive attributes of God's new digital gift to humankind: bass lines sounded awesome; detail simply sparkled to the surface of the soundstage.

In hindsight, however, I can tell you that this dastardly machine transformed JGH's listening room into a sonic horror chamber: we cycled through one CD after another, and almost nothing sounded right. Bright and edgy was how musical textures were typically being translated by the Sony. Because we were intimidated by the technology, we tended to blame the software rather than the hardware; it seemed in those days that no recording engineer could get it right.

That was my first lengthy exposure to digital sound, and even now, many years after the fact, my feelings of disgust remain clear in my memory. In two words, the Sony's sound was musical vomit. Morsels of partially digested detail, especially treble transients, were expelled from its bowels with an edgy disposition, and with such a tinge of sizzle and brightness that they screamed and splattered at me. There were also textural artifacts—synthetic detail added to the natural fabric of the music.

This is a far cry from a live performance, where detail is perceived as an integral, organic ingredient of the musical tapestry. It's all in there, but it doesn't assault the senses. The sort of detail I'm after has to do with such musical nuances as the discerning of the various micro-resonances that musical instruments possess; or the ability to enjoy the delicacy of a musical chord as its harmonic envelope blooms to full glory, then decays into the noise floor of the hall; or the ability to discern the ambient signature of a recording venue; or even the various reverb settings used in a multi-track recording.

Another barometer of detail resolution is the ability of a system to delineate the individual phrasing of various instruments in an ensemble. Fuzzing over a complex passage so that individual voices are lost in the blend is the mark of a low-resolution device—the layers of an ensemble ought to be preserved in natural fashion by the reproduction chain. This is not to say that I'm always looking for that sort of detail. I may choose to focus on the whole of the music the first time around, only later shifting my attention to the substrate of the music. With the Sony, however, detail was relentlessly forced in my face—a constant barrage of exaggerated, synthetic musical minutiae.

You've come a long way, baby.
The Aida did its detail thing naturally—I was able to sift through music at my leisure, picking up detail naturally as I might at a concert hall. The treble registers were reproduced with an airy refinement, although the flavor through the top octaves was a bit on the solid-state side of reality—namely, a shade hard and grainy. Processors with a tube-based analog buffer/gain stage tend to do a better job of smoothing out and slightly softening harmonic textures. I noticed, but rarely objected to, the Aida's somewhat transistory treble presentation—probably because I consistently used tubed line stages with it.

The upper mids were gloriously pristine. My litmus test in this regard is soprano voice and violin. The sweetness and gorgeous bloom of Arturo Delmoni's violin on Music for Violin & Guitar (Sonora SACC 102) shone through unabated. And to cite one of many examples, Kathleen Battle's timbre and upper-register purity on Baroque Duet (Sony SK 46672) were undiluted.

The lower mids and upper bass projected a convincing rhythmic expressiveness, and were nicely balanced with the rest of the midrange. Whether coping with Bach, Mozart, Brubeck, or Delta blues, the Aida propelled the music forward with consummate persuasion and exquisite dynamic bloom. Bass lines were tightly defined—as long as I was careful in my choice of power amp and matching speaker cable. For some reason, the Aida did far better with Acrotec's 6N-S1040 than it did with TARA Labs' RSC Master. Check out Rob Wasserman's backing of Jennifer Warnes on Leonard Cohen's "Ballad of the Runaway Horse" (Duets, MCA MCAD-42131): if there's such a thing as a passionate double-bass, this is it.

The soundstage was projected with convincing depth and width. Transparency? The soundstage was transparent to the point of readily illuminating every inner recess. Image outlines were tightly focused in space, and stayed that way through the full dynamic roller coaster of the music. Massed voices were resolvable into their individual constituents. Commendable indeed; so many inexpensive players don't get the job done here, and smear the living daylights out of a chorus.

Final report card
In addition to engaging in hand-to-hand combat with my Theta Gen.III, the Aida did battle with CAL's Alpha processor (review forthcoming), the original PS Audio UltraLink, the Sonic Frontiers SFD-2, and, just for the hell of it, the Accuphase DP-90/DC-91 combination that I recently purchased.

Against the PS Audio UltraLink, the Aida highlighted just how far digital processors have advanced in the last few years. The UltraLink's upper registers sounded significantly grainier, and I was constantly aware of a reduction in soundstage depth. There was a forced, almost mechanical seasoning to its delivery—very transistory in its treatment of musical textures. Bass lines were exceptionally strong and forcefully delineated.

The Aida also fared well against the Sonic Frontiers SFD-2—the unit to which RH recently gave high marks, especially when used balanced. I've tried it both single-ended and balanced, and have thought it to be the finest processor I've heard (even when operated single-ended)—at least until the Accuphase came along. The Sonic Frontiers' reproduction of the treble octaves was smoother than that of the Aida, its midrange textures were more liquid, and image outlines were portrayed with greater palpability within the soundstage. However, if the SFD-2 were a perfect 10, then I'd rate the Aida a 7—not bad for a unit retailing at less than 40% the price of the SFD-2.

Then came the Accuphase DP-90 transport and DC-91 processor—Holy cow! A total of 16 hand-selected, 20-bit Burr-Brown DAC chips operated in parallel!—and my life changed forever. Well, at least my pocketbook got a lot lighter; this combination was, quite simply, head and shoulders above anything else I'd heard. The degree of detail retrieval was stunning; the multiple multi-bit–based processor was unearthing more information than I'd heard before, and doing it with elegance and grace. What clinched it for me was the following sonic episode:

I was comparing a digital master of Anyone in Love—my wife, Lesley's, latest album (Lesley & the Santa Fe Sound Machine, Vital Music VF003)—with a standard-production CD. This involved listening to the DAT master through a TASCAM DA-30—a pro DAT machine—then switching to CD playback through the Accuphase DP-90/DC-91. Imagine my shock when I realized that the production CD actually sounded better than the master itself. By better, I mean that the CD came oh, so close to capturing the essence of the live mike feed. Now, that's magic!

If I were then to rate the Accuphase DC-91 a perfect 10, the Sonic Frontiers SFD-2 would rate a 7, while the Aida would earn a 4. Still not shabby for a unit costing not much more than a tenth of the DC-91's asking price.

I suspected that the Aida's toughest competition in this price range would come from CAL's Alpha processor. Not surprisingly, the Alpha's sonic character was more tube-like, with slightly softer, more liquid harmonic textures.

Final thoughts
The Aida represents a new breed of digital processors, with access not only to the latest in DAC technology, but also to the latest technical thinking in the field. Using the latest in digital circuitry and paying close attention to clock-jitter performance, the Aida redefines what can be achieved for moderate cost. The unit served me well for many months, and I can confidently recommend it—it's a must-audition at the $1700 price point.

Bel Canto Design, Ltd.
221 N. First Street
Minneapolis, MN 55401
(612) 317-4550