My first "audiophile" amplifier was a JBL SA600, early solid-state with a sophisticated and uncluttered circuit design and quite decent sound. I used it for years ‘til I rather smugly challenged a local audio dealer to a comparison test against the latest Hafler in his showroom. I think that's when I discovered that one of my core criteria in evaluating audio equipment was detail. The Hafler revealed instrumental sounds that were simply missing from the JBL; nothing subtle about it. Of course an amplifier has to do much more than provide detail. It should accurately reproduce harmonics so that an oboe sounds like an oboe, and that oboe should not get buried when the strings and brass crescendo. Instead, it should be able to play loud without fraying one's nerves and allow extended listening without fatigue. A number of amplifiers have come and gone since the JBL. The Hafler (built from a kit), Parasound, Bel Canto, and Spectron (four versions) to be precise. I now have a Spectron Musician III, which I playfully refer to as my $6,000 night light because of its rather bright, blue illuminated logo. I have a high regard for this unfazable, nuanced and accurate reproducer of music; all fifty-two pounds of it. The Spectron is emblematic of the pioneering Pulse Width Modulation designs of John Ulrick over thirty-some years. A while ago I received a review pair of deceptively small monaural amplifiers from Bel Canto, the e.One M300s. They weighed in at a mere nine pounds each (and most of that weight contributed by the casework). Some weeks later a pair of Bel Canto e.One REF1000s also arrived, plucked directly from John Stronczer's home system. The REF1000s look exactly the same as the M300s but weigh 13 pounds each. Both models’ specs indicate a voltage gain of 27dB, although my measurements showed the REF1000s have about 0.5dB more gain with a 100Hz test signal. In comparing the two, I compensated for this difference.

The Design. Inside both the REF1000 and M300 are circuit board modules from Bang and Olufsen's ICEpower division, a company created especially to develop, build, and license the designs of Dr. Karsten Nielsen who holds two United States patents on his amplifier designs. Bang & Olufsen thinks so highly of these designs, that they granted Dr. Nielsen joint ownership, something they've never done before. Now, I was not particularly surprised to find several reviews of ICEpower-based amplifiers talk about sonic differences influenced by enclosure material, mounting methods, and – in one case – circuit board modifications. (Note that modification of the circuit board voids the Bang & Olufsen warranty, so not many manufactures indulge.) One reviewer, for example, found that replacing the stock feet on the REF1000 with three cones (costing, naturally, several hundred dollars) significantly improved the bass. Bel Canto, of course, also made considered choices about how to package the modules. John Stronczer detailed some of these in an email to another reviewer: "...there is...a 3M damping pad on the top cover to reduce any chassis ringing. The chassis material is coated steel to provide optimum EMI and RFI isolation...We also take great care in the type and construction of power and signal cabling [employing] braided power wiring with a large RF bead, pro-grade copper-shielded twisted-pair input wires, and single-crystal copper speaker wiring...we also use crimped connections internally as these sound better.” The input connectors are gold-plated copper RCAs, and silver Neutrik XLRs, and there is “a very high grade power switch.” Silicon caulk is applied to critical components for isolation and stability from shock or vibrations.

The specific modules chosen by Bel Canto, the 1000ASP (REF1000) and 200ASC (M300), feature switching power supplies integral to the circuit board. Bel Canto claims that a good measure of the amplifiers' performance derives from these power supplies. In an early experiment, they modified a Tripath-based Evo amplifier to run from a switching power supply and found the sound was remarkably improved. Now, an amplifier power supply must do two things: supply a constant voltage under rapidly changing current loads, and supply that voltage without artifacts of the power conversion process or the AC power line. Traditional (linear) power supplies do this at the power line frequency – 60Hz – using an appropriate transformer to step-down the voltage from 120VAC. This transformer is necessarily large, heavy and expensive, particularly in a high power amplifier, even one that's better than 80% efficient overall. (Class-A/B amplifiers are a maximum 50%, class-A a maximum 25%, and therefore require much larger power transformers for a given power output.) The stepped-down AC voltage on the secondary winding of the transformer is converted to DC by a bridge rectifier. If you look at the output of this rectifier with an oscilloscope, you’ll find a series of half-sine waves. Using high value filter capacitors – and sometimes a series inductor called a choke – which store energy during the ascending voltage curve and release it during the descending voltage curve, smoothes out the pulsing DC, making it look more like a steady voltage. Such a supply is typically unregulated, relying on the robustness of the components to maintain a constant voltage under varying loads. There is always an attenuated artifact of the AC line frequency, called ripple. Sixty hertz ripple and its near harmonics are audible, and if you place your ear near a loudspeaker woofer, you'll usually hear a subdued hum. Of course a high-current linear power supply can be regulated, but a regulator would add considerable expense and is necessarily inefficient, requiring a higher input voltage than the desired output voltage, which is to say it generates heat. And heat is the enemy of electrical components, being the critical factor in MTBF (Mean Time Between Failure). Most audio amplifiers use unregulated linear power supplies. Most of the cost of a traditional audio amplifier is for power supply components.

Now, switching power supplies (also called switch mode power supplies) are electrically noisier, more complex and more difficult to design than linear power supplies, but for audio amplifier applications they offer several advantages. Decreased size and weight have already been noted. Switching supplies work by increasing the frequency of the line voltage from 60Hz to 50,000Hz or more, which exponentially increases the efficiency of the step-down transformer, and also places residual conversion artifacts (like ripple) well outside the audible range. Because of the increase in efficiency at these frequencies, the power transformers in the REF1000 and M300 are only a couple of inches square, and filter components for the bridge rectifier can have lower values and much smaller physical size. It is almost counter-intuitive to imagine that such a small transformer can supply so much power. Moreover, these power supplies are inherently regulated, using a closed loop feedback topology to control the duty cycle of the square waves that power the primary of the step-down transformer.

If this description rings a bell, it should: it is essentially Pulse Width Modulation (PWM), the same method used in the ICEpower audio output circuit, and every other switching, sometimes called “digital,” amplifier, be it Spectron, Tact, Linn, Yamaha, Sony, Theta, Behold, Rowland, et al. (The proper nomenclature for this type of amplifier is analog switching amplifier or analog pulse modulated amplifier; the term “digital” is a creation of marketing departments, not audio engineers.) In the case of an audio amplification stage, the pulse width over a segment of time – as determined by the carrier frequency – is analogous of a specific voltage at a specific point on a complex musical sine wave. These square wave pulses pass through a series inductor [L] and a parallel capacitor [C] which integrates them, in effect “converting” them into a sine wave. In the case of a switching voltage regulator, the pulse widths are controlled by feedback from the output and cumulatively maintain a stable voltage on the secondary of the step-down transformer. (There are transformerless ways of accomplishing the same thing – lowering and regulating voltage – called buck regulators. But using a transformer offers greater isolation from the AC line and a greater safety factor.) And the level of regulation attainable with a switching mode power supply is greater than is attainable with a linear power supply, even one that is regulated.

[The term “carrier” is used in the same sense as it is in AM and FM radio transmission, where amplitude or frequency – rather than pulse width – are modulated by the data being transmitted.]

The part of a switching amplifier that converts the analog input to an analogous series of pulses of varying widths is called a comparator. (Amplifiers accepting digital input from a CD transport must contain an internal DAC.) In its simplest form, this circuit compares a complex musical sine wave against a sawtooth wave of a fixed frequency many times higher than the highest audio frequency. The frequency of this sawtooth is effectively a sampling rate. A Maxim IC technical paper on Class D amplifiers states, “During a switching period, the comparator output is low when the sawtooth exceeds the input signal and high otherwise...For a given input level, the comparator output is a duty-cycle modulated square wave with period determined by the sawtooth frequency...the amplifier's dynamic range is determined by the noise floor and the sawtooth magnitude.” If this succinct description sounds very technical, the basic principle is understandable by plotting the two wave forms together. When the amplitude of the sawtooth is greater than that of the sine wave input, the comparator output goes low; when the amplitude of the sawtooth is less than that of the sine wave input, the comparator output goes high. The result is a series of square waves of varying widths. It's very ingenious. However, using a separate sawtooth oscillator is not the only nor the best approach to designing a comparator. For one thing, the stability of the oscillator is absolutely critical. A self-oscillating comparator eliminates this potential source of nonlinearity.

The ICEpower amplifiers are designed around a self-oscillating, variable carrier frequency comparator that ICEpower refers to as a Controlled Oscillation Modulator (COM). This stable, closed loop circuit uses feedback from the output of the switching MOSFETs (Metal-Oxide Semiconductor Field-Effect Transistor) to generate a sine wave modulating carrier, rather than a traditional sawtooth wave. “The COM modulator”, the ICEpower white paper goes on to state, “is the heart of the technology, as it operates as a highly linear modulator and at the same time a wideband ultra fast feedback control system...The effective feedback bandwidth is equal to the switching frequency, i.e. the COM modulator is in effect an extremely wideband, fast, cycle-to-cycle error correction system, effectively compensating for the artifacts of the switching output stage.” In addition, this closed loop topology provides a high power supply rejection ratio, meaning the amplifier is especially immune to any noise that may be present on the power supply rail.

As can be seen with an oscilloscope, even a very complex audio signal can be represented by a series of positive and negative voltages (which coincidentally is precisely what digital encoding does). The ICEpower COM first generates a series of square wave pulses of amplitudes identical to the audio sine wave (Pulse Amplitude Modulation, PAM). It then logically reverses the X and Y axes, in effect converting amplitude to time (Pulse Width Modulation, PWM). The effective carrier frequency varies from 200,000Hz at high amplitudes to 1.3MHz at low amplitudes. And by varying the carrier frequency, the overall energy of these non-audio frequencies at the output terminals of the amplifier is reduced. A certain amount of this carrier energy inevitably bleeds past the stop band of the output LC filter, but lowering the overall energy results in lowered electromagnetic interference (EMI) being transmitted down the loudspeaker cables, and lowered thermal load on the loudspeaker voice coils. (It is fair to say that no ordinary loudspeaker can convert frequencies of 200,000Hz up into mechanical energy, so it dissipates as heat in the voice coil.) EMI is not just an abstract issue: I read a review of a quite expensive, and highly praised, analog switching amplifier that generated enough EMI to make listening to FM radio impossible.

The ICEpower modules use double feedback loops to maintain a remarkable level of linearity. (The newest generation of ICEpower designs are measurably superior in this regard, 0.003% THD+N versus 0.007 THD+N. Either figure, however, is vanishingly low.) ICEpower refer to this system as Multivariable Enhanced Cascade Control (MECC), which uses feedback from the output of the switching MOSFETs and from the output of the LC filter (ie., both square wave data and sine wave data). Basically, negative feedback reduces distortion by comparing the input signal to an inverted version of the output signal, adding the difference to the input, thus (theoretically) correcting any non-linearities. The feedback scheme used by ICEpower helps to solve two problems inherent to class D topology: load dependent frequency response and distortion caused by timing errors in the switching MOSFETs. Dr Nielsen's patent on MECC (6,297,692) states that it provides “...very low sensitivity to errors in the switching power stage...[and] to load variations and filter errors.” But feedback takes a finite transit time to effect correction, so there is a phase shift, however slight, between input and feedback. As John Ulrick notes on the Spectron web site, global feedback in a typical linear – class A, A/B – amplifier has a 2000 – 3000 nanosecond transit time. A properly implemented analog switching amplifier – because it uses digital logic instead of linear circuits – has a transit time that is ten to fifteen times faster. This also minimizes group delay, so different frequencies remain in phase and, given phase coherent loudspeakers, arrive at your ear at the same time. The better the group delay, the better the timbral accuracy. In my experience, these characteristics of switching amplifiers are immediately audible as transparency, the retrieval of subtle detail, and as nuanced harmonics.

The output stage of the M300 and REF100 consists of four bridged N-channel MOSFETs operating as high speed switches on the power supply rail. They have an internal resistance of <0.030 Ohm, which translates to high efficiency and low heat dissipation; and a switching speed of 0.00000003 second, which minimizes timing errors and results in the accurate reproduction of the comparator-generated PWM data. The use of a full bridge output stage is less efficient than half bridge (two MOSFETs), but it eliminates certain distortion mechanisms, simplifies power supply design by allowing a single voltage that in turn eliminates load dumping (voltage fluctuations) and prevents power supply pumping. Both the 1000ASP and 200ASC incorporate a soft-clipping circuit that, according to the Bel Canto white paper, “...mimics the soft tube overload characteristic, preventing the harshness that occurs with many solid state amplifiers... [but] does not cause any compromise in the performance at levels below clipping.”

One final note on the designs of the 1000ASP and 200ASC. The former is designated a “professional” module. It is designed to be, as they say, bullet proof. It is larger and heavier than the 200ASC, its output devices are mounted on a heat sink, and it is both physically and electrically more rugged, able to withstand repeated shocks up to 70G. These are not considerations likely to be of real importance to living-room audiophiles, but they are noteworthy. Also, the 200ASC has an internal fuse to protect the amplifier from fault conditions, the 1000ASP accomplishes this without a fuse. So if you're in the habit of swapping cables without turning off power, maybe occasionally shorting or grounding the output terminals, or you're shy about opening casework to replace a fuse (and Bel Canto do recommend having this professionally done), this design difference may be important to you.

The Sound.

These two amplifiers have been extensively reviewed, seemingly more so than other implementations of ICEpower technology (of which there are a growing number). This past spring,  the REF1000 was given “Publisher's Choice; Most Wanted Component” status by Clement Perry. Both the M300 and REF1000 are capable of driving my 91dB efficient loudspeakers to ear-damaging levels; both amps run very cool; both amps are small and unobtrusive; and both amps offer a level of sound quality I've rarely encountered, equaling or exceeding the performance of my Musician III. I'm not sure I can add much to the universal praise lavished upon these amplifiers. But I was curious to find out if there are sonic as well as design and component differences.

Beethoven: String Quartet Op 74, Borodin String Quartet (Chandos 10191). One of several submissions to a long-standing fantasy to write an appreciation of what I, and many others, consider the greatest classical (European) music ever written. This is music that talks to you in an utterly unique voice, with vitality, clarity and sureness, but with absolute humility and honesty. It is profoundly spiritual music, but it’s a spirituality firmly rooted in the human heart. I chose this recording because the Borodin Quartet does an inspired job on the last two movements of Opus 74, and the sound engineering is quite good. For eights months I've listened to versions of the Beethoven quartets, and what I want most from my stereo for this particular music are those qualities that facilitate a connection with the “spiritual goings on,” a sense of connection with the hearts and minds of the musicians, a sense of connection with the composer. I'm not sure what particular sonic attributes tend to do this, but the REF1000 may, I think, do it just a tad better than the M300. It is to be expected that these two amps will sound very much the same, as indeed they do, slightly better bass being a generally distinguished trait of the REF1000. Subtler differences are elusive. So when I suggest that the M300 is a sweeter and airier amplifier, and that the REF1000 does a better job of pinning down the positions of the four musicians sawing away at cat gut, that the REF1000 sets the musicians further back in the venue and that the M300 brings them forward in a more intimate perspective, I justify it not by the certainty of my perceptions, but by the fact those perceptions have recurred over several months. Other than these qualifications, both amps do a splendid job preserving the dynamics, sonority and transcendent energy of this music, both amps facilitate that all-important connection. At the risk of paraphrasing St. Anselm, what goes on in the Late, and some of the Middle, quartets is that which is beyond conception, beyond words, and these amplifiers reveal this without intrusiveness.

Nojima Plays Ravel (Reference Recordings RR35CD). This is supremely difficult piano music that calls for the utmost in technical mastery and musicianship, qualities for which Minoru Nojima is legendary. A quarter of a century ago I was privileged to hear him at an auditorium near LAX and the experience was as breathtaking as, and more rewarding than, hearing Vladamir Horowitz in downtown LA from my $75 balcony seat. This recording is a particularly good choice to test piano reproduction, in part because it stretches the pianoforte (a Hamburg Steinway) to its limits, and in part because I was able to correspond with Keith O. Johnson about the sort of imaging one can, ideally, expect. The imaging I do get is quite impressive, though it occasionally leaves me dreaming of a dedicated music room constructed of cedar and redwood, with lofty ceilings and faded Persian carpets. (Many recordings occasionally do that.) The piano was not close miked and should sit well back of the loudspeakers. In this regard, the REF1000 does a better job than the M300, placing the piano further back and clearly delineated; whereas the M300 creates what might be termed a “more intimate” sort of image. Dr Johnson also used a pair of omni-directional microphones to capture some of the room ambience. The opening notes of Oiseaux tristes are revealing: while both amplifiers reproduce room ambience, the REF1000 distinctly captures a sound – perhaps an initial reflection – that the M300 does not seem to. Both amplifiers do a wonderful job of reproducing the elaborate dynamics and nuances of this music, and the incredible speed and delicate touch of this pianist.

Beethoven: Complete Piano Sonatas, Stuart & Sons Piano played by Gerard Willems, Vol 1 (ABC 465 077-2). It is unusual to include two solo piano recordings in a equipment evaluation, but the differences revealed by this recording demanded it. There were two of us auditioning and when I switched amplifiers we stared at each other in disbelief: the ringing overtones of the piano strings were that much more evident with the REF1000. Why would one excellent piano recording reveal relatively subtle differences, and another, very unsubtle differences? Because the two pianos are very different animals. I think at least part of the explanation may lie in the design of the bridge in Wayne Stuart's piano (www.stuartandsons.com). Unlike traditional designs in which the string is terminated by two laterally opposing pins embedded in the bridge, the patented Stuart bridge terminates the string with a sort of clamp in the vertical plane, thus applying pressure in the same plane as the striking hammers. With a traditional bridge the string begins vibrating in the vertical plane, but lateral pressure from the bridge pins creates elliptical polarization; the vibration changes to an ellipse and ends up horizontal. These changes in motion have a profound effect on sound and sustain. The Stuart bridge enables the string to remain vibrating in a vertical plane, so it sustains longer and produces unique overtones, during attack, sustain and damping. Overtones, as you probably know, are precisely what enable one to distinguish, for example, a pipe organ from a violin, and the fact is the Stuart & Sons piano sounds quite unlike any other. Not only is the overall sound different, but ranges of tones on the keyboard – bass, upper-bass, treble – have a character of their own, rather than being homogeneous. In no other recording I tested was a difference between the REF1000 and the M300 as obvious. Both amplifiers are strikingly musical, rather than impressing with acoustical pyrotechnics (remember in the dark old days when audiophiles showed off their equipment with recordings of trains?), they impress with sheer beauty and ease of listening.

Bartok: Concerto for Orchestra, Fritz Reiner, conducting Chicago Symphony Orchestra (RCA 09026-61504-2). This classic performance from the mid-1950s has the further virtue of being very simply miked using 2- and 3-track stereo tape decks. While modern recordings, using the latest encoding schemes, the latest microphones and the latest electronics, can be superb – revealing an oboist turning a page of music or the creaking of a seat in the first violins – some of these older recordings, including Mercury Living Presence, have a quality that is closer to live music. Less than a minute into the Andante, the first violins begin a vibrato accompaniment, followed by the second violins. As might be expected from what has been noted previously, the REF1000 clearly delineates the entry and locus of the second violins, whereas the M300 is noticeably less precise. This is true of front-to-back positioning as well. If you haven't painted yourself into a corner from which you're compulsively checking image and sound stage, these are not absolutely essential qualities: imaging is not everything. The M300 has a quality of musicality (an adjective I am fortunately not required to define) that is extremely appealing; it is a quite wonderful amplifier, and quite a bargain. Its particular shortcomings in this comparison are soon forgotten in the sheer enjoyment of the music.


Kendra Shank: Afterglow (Mapleshade 02132). As my Mapleshade tee shirt states: no mixing board, no overdubs, no noise reduction, no compression, no multitracks, no EQ, no reverb. The result of Pierre Sprey's devotion to purity and simplicity is some of the most realistic sound ever put on CDs. I depart from my standard – Live At Ethell's – in order to test reproduction of a solo female voice. This one was bound to be a close call, and I'm afraid it was just that: the most difficult CD I used in this comparison test. I suppose better ears than mine, or for that matter better equipment (and a better listening room) than mine, might have discovered obvious distinctions, but I could not. But if I were to choose between these amplifiers on the basis of this CD alone, the M300 would be my choice. It gets back to that “sweeter and airier” quality of the M300, more suitable for a beautiful female voice. Somehow just a bit more present, making it just a bit easier to slip into that act of pure imagination. Can I account for an amplifier that's “better” reproducing a piano, being not as “good” reproducing a singer? Only by suggesting that the recording engineer alone could tell us which sound it the one he intended.

                            __________________




Bel Canto M300/REF1000 Specifications:
Analog Inputs
Unbalanced RCA, Balanced XLR, switchable on rear of chassis
Output
WBT 5-way binding posts
Specifications
Power Output: 150W into 8 ohms, 300W into 4 ohms (M300);
>500W into 8 ohms, >1000W into 4 ohms (REF1000)
Frequency Response: +/- 0.5db, 20Hz – 20,000Hz all loads
THD+N: 0.01%, 1W, 1KHz, 4 ohms (M300); 0.007%, 1W, 1KHz, 4 ohms (REF1000)
IMD: 0.002%, 1W, 14:15KHz, 4 ohms(M300); 0.0007%, 1W, 14:15KHz, 4 ohm (REF1000)
Output Noise: 90uVRMS A-weighted 10Hz-20KH
Damping Factor: >1000
Output Impedance: <0.008 ohms, 100Hz
Dynamic Range: 111db (M300); 120db (REF1000)
Standby Power: 5W (M300); 15W (REF1000)
Weight & Measurements
Dimensions: 8.5” W x 12.5” D x 3” H
Net weight: 9 lbs (M300); 13 lbs (REF1000)
Cost: $1990/pair (M300); $3990/pair (REF1000)

Bel Canto Design, Ltd.
221 North 1st Street
Minneapolis MN 55401
USA

Tel: 612-317-4550 (9AM to 5 PM CST M-F)
Toll-free (866) 200-7342
Fax: 612-359-9358
www.belcantodesign.com