Bel Canto Design Black amplification system Measurements

Sidebar 3: Measurements

As Bel Canto Design's Black is a complex product—both the ASC1 control unit and MPS1 monoblocks have both digital and analog inputs—I performed most of my measurements with the control unit and amplifiers connected with the ST-optical datalinks, with checks of the analog inputs of the amplifiers. I measured the Black using the digital and analog outputs of my Audio Precision SYS2722 system (see, and the January 2008 "As We See It"), and my 2012 MacBook Pro running on battery power to feed the USB input. Because the MPS1 is a class-D design, I used an Audio Precision AUX-0025 passive low-pass filter ahead of the analyzer, which eliminates noise above 200kHz. (At the Black's speaker terminals, without the filter and with no signal, there were 300mV of ultrasonic noise with a center frequency around 453kHz in the left channel, and 478mV of noise centered on 460kHz in the right channel.) After an hour of running at moderate power levels into 8 ohms, the chassis of the two MPS1s were warm to the touch, at 94.6°F (34.8°C).

I didn't test the Ethernet input but all the ASC1's other digital inputs locked to data with sample rates ranging from 44.1 to 192kHz. Apple's USB Prober utility identified the Black as "Bel Canto uLinkUSB 2.0 Audio Out," and confirmed that the USB input operated in the optimal isochronous asynchronous mode. Measured at the MPS1s' speaker terminals, the ASC1's digital inputs preserved absolute polarity; and with the volume control set to "100," data with a level of –20dBFS gave rise to an analog signal of 12.08V RMS into 8 ohms. This suggests that the volume control shouldn't be used above "94" to avoid clipping the amplifiers. The control itself operated in accurate 0.5dB steps.

The Black's reconstruction filter offers three different rolloffs, as well as a setting called Filter Off. Fig.1 shows the impulse response at a sample rate of 44.1kHz with Filter Off; this is a conventional, linear-phase, half-band impulse response, with the ringing symmetrically placed around the single sample at 0dBFS. By contrast, Filters 1–3 offer minimum-phase impulse responses, with (almost) all the ringing following the high sample and the amount of ringing decreasing as the number of the filter increases. Fig.2, for example, shows the impulse response of Filter 2.


Fig.1 Bel Canto Black, digital input, Filter Off, impulse response at 44.1kHz (4ms time window).


Fig.2 Bel Canto Black, digital input, Filter 2, impulse response at 44.1kHz (4ms time window).

White noise sampled at 44.1kHz reveals that the Filter Off filter has the usual fast rolloff above the audioband (fig.3, red and magenta traces), with a small rise in the ultrasonic noise floor centered on 65kHz (footnote 1). Bel Canto claims that Filters 1–3 are "apodizing" types, meaning that they eliminate ringing at half the sample rate. The red and magenta traces in fig.4 show the Black's output with Filter 1 selected. You can see that it is indeed an apodizing type by the sharply defined null at exactly half the sample rate (vertical green line). Filters 2 and 3 are identical other than the fact that the rolloff starts a little earlier and the height of the lobe in the stopband above the green line is lower in amplitude. The blue and cyan traces in these graphs are the output with a full-scale tone at 19.1kHz; all four filters completely suppress the ultrasonic image at 25kHz, and the distortion harmonics are all at or below –76dB (0.015%).


Fig.3 Bel Canto Black, digital input, Filter Off, wideband spectrum of white noise at –4dBFS (left channel blue, right cyan) and 19.1kHz tone at 0dBFS (left red, right magenta), with data sampled at 44.1kHz (20dB/vertical div.).


Fig.4 Bel Canto Black, digital input, Filter 1, wideband spectrum of white noise at –4dBFS (left channel blue, right cyan) and 19.1kHz tone at 0dBFS (left red, right magenta), with data sampled at 44.1kHz (20dB/vertical div.).

Fig.5 is a more conventional way of showing digital frequency response, with Filter 1 and sample rates of 44.1, 96, and 192kHz. The overall shape of the response is the same at all three rates, with, as expected, a sharp rolloff at 20kHz at 44.1kHz. But the ultrasonic rolloff occurs a little earlier with the higher rate. With 192kHz data, for example, the output is down by 9dB at 52kHz rather than at the expected 90kHz or so. This measurement was taken at the speaker terminals, so it includes the contribution of the class-D modules and, of course, the Audio Precision low-pass filter. However, repeating the measurements without the AP filter or with an analog input to the MPS1 gave the same result. The class-D stage has, of necessity, a curtailed ultrasonic response, due to the fact that it cannot have an infinitely high switching frequency; so while the full bandwidth extension offered by high sample rates can't be realized with the Black, it looks as if Bel Canto's design team has achieved a careful balance of what is possible. Filters 2 and 3 offer earlier rolloffs with 44.1kHz data, trading off the slight loss of top-octave response against the better time-domain behavior of these filters.


Fig.5 Bel Canto Black, digital input, frequency response at –12dBFS into 100k ohms with data sampled at: 44.1kHz (left channel green, right gray), 96kHz (left cyan, right magenta), 192kHz (left blue, right red) (1dB/vertical div.).

To test the resolution of the Black's digital inputs, I set the volume control to "90," equivalent to an output power of 165W into 8 ohms. Under this condition, increasing the bit depth from 16 to 24 with dithered data representing a 1kHz tone at –90dBFS dropped the noise floor by 23dB (fig.6), which is equivalent to a resolution of almost 20 bits: superb performance. Referenced to the Black's specified clipping power, this would increase to 21 bits, though I didn't test that due to my concern that I might inadvertently break the amplifier before I finished the measurements. This graph was taken with AES/EBU data; repeating the test with USB data gave the same result, indicating that the USB input correctly handles 24-bit data. With its superb low-level linearity and low noise, the Bel Canto had no problem dealing with an undithered 16-bit tone at exactly –90.31dBFS (fig.7). The data describe three DC voltage levels; these were clearly evident, and the waveform was nicely symmetrical about the time axis. With undithered 24-bit data, the result was a clean sinewave relatively free of noise (fig.8).


Fig.6 Bel Canto Black, digital input, spectrum with noise and spuriae of dithered 1kHz tone at –90dBFS with: 16-bit data (left channel cyan, right magenta), 24-bit data (left blue, right red) (20dB/vertical div.).


Fig.7 Bel Canto Black, digital input, waveform of undithered 1kHz sinewave at –90.31dBFS, 16-bit data (left channel blue, right red).


Fig.8 Bel Canto Black, digital input, waveform of undithered 1kHz sinewave at –90.31dBFS, 24-bit data (left channel blue, right red).

Tested for its rejection of word-clock jitter with 16-bit J-Test data (fig.9), the Black produced no visible jitter-related sidebands, and the odd-order harmonics of the LSB-level, low-frequency squarewave were all close to the correct levels (indicated by the green line). This graph was taken with AES/EBU data; USB data gave an identical result, and with 24-bit data, the noise floor was clean and free from spuriae (fig.10).


Fig.9 Bel Canto Black, digital input, high-resolution jitter spectrum of analog output signal, 11.025kHz at –6dBFS, sampled at 44.1kHz with LSB toggled at 229Hz: 16-bit data from SYS2722 via AES/EBU (left channel blue, right red). Center frequency of trace, 11.025kHz; frequency range, ±3.5kHz.


Fig.10 Bel Canto Black, digital input, high-resolution jitter spectrum of analog output signal, 11.025kHz at –6dBFS, sampled at 44.1kHz with LSB toggled at 229Hz: 24-bit data from SYS2722 via AES/EBU (left channel blue, right red). Center frequency of trace, 11.025kHz; frequency range, ±3.5kHz.

Turning to the Bel Canto Black's performance with analog signals, these are digitized by the ASC1. As the ASC1 has AES/EBU outputs as well as the ST-optical outputs to feed the MPS1s, I could examine the efficacy of its digitization by looking at the digital data directly. The ASC1's A/D converter operates at a sample rate of 192kHz and a 24-bit word length. The maximum volume control setting was "110," with settings above "100" applying gain in the digital domain. At this setting, it took an 830mV signal at 1kHz to reach –0.1dBFS. Reducing the volume appears to increase the digital headroom. Fig.11 shows the ASC1's analog-input frequency response, measured in the digital domain. The output is down by 3dB at 53kHz, the fairly slow rolloff suggesting good time-domain behavior.


Fig.11 Bel Canto ASC1, analog input, A/D frequency response (left channel blue, right red) (2dB/vertical div.).

The ASC1's single-ended analog input impedance was 9k ohms at all audio frequencies, this a little on the low side for use with tubed source components. The MPS1's balanced analog input impedance was 18k ohms at low and middle frequencies, dropping to 2300 ohms at 20kHz. Both components' analog inputs preserved absolute polarity (ie, were non-inverting) at the factory default setting. The MPS1's audioband output impedance was very low, at 0.09 ohm (including 6' of speaker cable). As a result, the modification of the amplifier's frequency response, due to the interaction between this impedance and that of our standard simulated loudspeaker, was minuscule (fig.12, black trace). The data plotted in this graph indicate that the MPS1's response is down by 1dB at 20kHz, –3dB at 31kHz, and –9dB at 50kHz. This ultrasonic rolloff was the same driving the ASC1's analog input and measuring at the MPS1's speaker terminal, nor did it change with different settings of the ASC1's volume control.


Fig.12 Bel Canto MPS1, frequency response at 2.83V into: simulated loudspeaker load (gray), 8 ohms (left channel blue, right red), 4 ohms (left cyan, right magenta), 2 ohms (green) (0.5dB/vertical div.).

Fig.13 shows the Black's reproduction of a small-signal, 1kHz squarewave into 8 ohms. The tops and bottoms of the waveform are obscured by the ultrasonic switching noise generated by the class-D output stage. Repeating the measurement with the Audio Precision low-pass filter ahead of the analyzer gave the well-shaped waveform in fig.14. I suspect that the small amount of overshoot and ultrasonic ringing is due to the ASC1's A/D converter. A 10kHz squarewave (fig.15) was reproduced with just one cycle of ringing, and lengthened risetimes associated with the ultrasonic rolloff seen in fig.12. Even with the Audio Precision filter's rolloff, there is enough leakage of the output stage's switching noise to interfere with both the unweighted wideband and audioband signal/noise ratios. The A-weighted S/N ratio of the MPS1 alone, taken with the analog input shorted to ground and ref. 2.83V into 8 ohms, was 79.2dB; that of the complete Black with its analog input shorted but its volume control set to "110" was 60.2dB. Channel separation, measured at the ASC1's analog inputs, was >100dB below 2kHz, and still 70dB in both directions at 20kHz.


Fig.13 Bel Canto Black, small-signal, 1kHz squarewave into 8 ohms.


Fig.14 Bel Canto Black, small-signal, 1kHz squarewave into 8 ohms with Audio Precision low-pass filter.


Fig.15 Bel Canto Black, small-signal, 10kHz squarewave into 8 ohms with Audio Precision low-pass filter.

I tested the MPS1's clipping power by driving it directly, to avoid the possibility of overdriving the ASC1's analog input A/D converter and confusing the result. The Black is specified as delivering 300W into 8 ohms, 600W into 4 ohms, or 1200W into 2 ohms, all equivalent to 24.8dBW. However, in this continuously driven condition, the amplifier's protection circuit operated at a level 3dB below the specified maximum power. This can be seen in figs. 16 and 17, which respectively plot the THD+noise percentage against output power into 8 and 4 ohms. The traces stop at 149.5W into 8 ohms (21.75dBW) at 0.00223% THD+N, and at 293W into 4 ohms (21.67dBW) at 0.002%, which in both cases was where the protection circuit went into action. Into 2 ohms (not shown), the protection cut in at 545W (21.5dBW), at 0.015%.


Fig.16 Bel Canto Black, distortion (%) vs 1kHz continuous output power into 8 ohms.


Fig.17 Bel Canto Black, distortion (%) vs 1kHz continuous output power into 4 ohms.

But below the continuous powers where the protection operated the MPS1 offered very low distortion. The lower trace in fig.18 shows the distortion+noise waveform at 59W into 4 ohms. I had to average 64 captures to reduce the noise to the point where the distortion waveform emerged from the higher-level noise; it is primarily third-harmonic in nature. This was with the review unit designated as the left channel. As can be seen in fig.19, the other sample (red trace) had just a trace of second-harmonic distortion. Intermodulation distortion at a fairly high power was also very low (fig.20), the 1kHz difference component resulting from an equal mix of 19 and 20kHz tones lying at –96dB (0.0015%).


Fig.18 Bel Canto Black, 1kHz waveform at 59W into 4 ohms, 0.005% THD+N (top); distortion and noise waveform with fundamental notched out (bottom, not to scale).


Fig.19 Bel Canto Black, spectrum of 50Hz sinewave, DC–1kHz, at 86W into 4 ohms (linear frequency scale).


Fig.20 Bel Canto Black, HF intermodulation spectrum, DC–24kHz, 19+20kHz at 86W peak into 4 ohms (linear frequency scale).

Overall, I was very impressed by the Bel Canto Black, especially with its behavior as a D/A converter that is effectively capable of driving loudspeakers. Once I have finished the review of another amplifier I have been working on, I am going to set up the Black in my own system for a listen.—John Atkinson

Footnote 1: This test was suggested to me by Jürgen Reis of MBL.
Bel Canto Design, Ltd.
221 First Street N., Suite 300
Minneapolis, MN 55401
(612) 317-4550

JRT's picture

Bruno Putzeys' NCore variant of switch mode (class D) amplifiers, in combination with switch mode power supply, represents significant progress.

If well implemented, Putzey's NCore can be fully adequate for high quality audio, but that is not why it represents significant progress, because a well designed conventional class AB amplifier with conventional power supply can also produce high quality audio at similar power output.

What represents significant progress is that Bruno Putzeys' NCore variant of switch mode amplifiers, in combination with switch mode power supply, can also produce high quality audio at substantially reduced cost, reduced size, and with reduced power consumption, as compared to a conventional class AB amplifier capable of similarly high quality audio at similar power output.

I would suggest comparison of this Bel Canto MPS1 amplifier to a Parasound Halo JC1.

Here is a link to Stereophile's review of the JC1:

The Parasound Halo JC1 sells for less than $5K.

Bel Canto chose a $15K MSRP price point for this MPS1, and by doing that they have failed to deliver the real progress that Putzeys' NCore could and should provide.

Bel Canto's small dealer network in the US will sell a few of these, probably very few, and I suspect this Bel Canto product will soon fade into oblivion. I hope Putzeys' NCore does not fade as quickly.

JRT's picture

I'd like to see Bruno Putzeys' NCore implemented in active loudspeakers, with DSP crossovers, attached to a wired gigabit Ethernet network using Audinate Dante multicast. Dante uses conventional gigabit Ethernet switch and wiring.

The front end of the system could be as simple as a netbook running JRiver and Audinate Dante virtual soundcard. Dante virtual soundcard uses the wired Ethernet port to attach to a dedicated audio Ethernet network. The downstream end would be active loudspeakers, subwoofer(s), etc., also attached to the dedicated wired Ethernet network. In between would be nothing more than a conventional gigabit Ethernet switch and inexpensive cat5e wiring. That would eliminate a lot of the conventional resource hogs, and the majority of the system budget would reside in active loudspeakers. Setup and configuration could be extremely easy.

JRT's picture

A system using Dante could begin as a pair of loudspeakers (or even just one) attached to the network, and could be expanded by attaching new speakers to the network and adjusting settings on processing.

Audinate Dante over gigabit Ethernet can provide 16x16 24bit/192kHz (16 bi-directional channels), 32x32 at 24bit/96kHz.

edit: I should mention that I have no affiliation with any of these products or vendors.

JJSIII's picture

In response to JRT's comments, yes indeed nCore represents significant progress, the premium, customized 1200Watt capable module used in the MPS1 is NOT an inexpensive device. Also note that the MPS1 is MUCH MORE than a simple amplifier. Bel Canto has released a new e.One nCore based >500W mono power amp that retails for less than $5K a pair, the REF600M. It does not use the same nCore module as the MPS1 units but it does provide remarkable audio performance and value.