Bel Canto USB Link 24/96 USB-S/PDIF converter Measurements
I examined the measured behavior of the Bel Canto USB Link 24/96 using the Audio Precision SYS2722 system (see www.ap.com and "As We See It" in the January 2008 issue) and the Miller Audio Research Jitter Analyzer. I played back test tones at various sample rates and bit depths from Audacity on a dual-core Pentium PC running Windows XP, and from Audacity and BIAS Peak Pro 6 on a Mac mini running OS10.4.11.
To check whether the USB Link was bit-transparent, I fed its S/PDIF output to an RME soundcard fitted to a second PC and recorded the data using Adobe Audition. Playing back that recording simultaneously with an inverted, bit-synchronized version of the original audio file would reveal any difference. Fortunately, the files nulled perfectly at both 16- and 24-bit word lengths with playback from both PC and Mac platforms, proving that the transmission chain was indeed bit-transparent.
Using RME's DIGICheck utility, I looked at how many bits were active in the Bel Canto's output: 16 with 16-bit files on the Mac; 24 with 24-bit files on the Mac, and with 16- and 24-bit files on the PC. With both Mac and PC, DIGICheck confirmed that the Bel Canto USB Link 24/96 passed on the original file's sample rate, and operated at all four specified rates: 44.1, 48, 88.2, and 96kHz.
The only meaningful measurement of the Bel Canto's output quality is the amount of timing uncertainty or jitter in the datastream. Assessing this, however, is not straightforward, as you need to have a measurement system with a clock accuracy at least an order of magnitude better than that offered by the device under test. Unfortunately, while the Audio Precision SYS2722 has a sophisticated digital-data interface analyzer, its own timing uncertainty appears to limit measurements to data sources whose jitter levels are greater than a nanosecond or so.
Nevertheless, the AP can operate as a digital oscilloscope, overlaying successive snapshots of the datastream to show what is called the "eye pattern." With an ideal transmission system, all the pulse transitions in the datastream will overlay one another to give a wide-open "eye," with apparently just one trace visible. Fig.1, for example, shows the eye pattern of the Boulder 1021 disc player's data output, plotted over one "unit cycle" as, for 60 seconds, it fed the Audio Precision a 16-bit AES/EBU datastream comprising the Miller/Dunn J-Test Signal. The calculated jitter level was the lowest I have so far measured with the SYS2722, at 1.75ns, and you can see that, other than a slight thickening of the horizontal sections of the traces, the eye is indeed wide open.
Fig.1 Boulder 1021, eye pattern of AES/EBU data output carrying 16-bit J-Test signal (±3V vertical scale, 175ns horizontal scale).
For comparison, fig.2 shows the eye pattern of the Bel Canto USB Link's output feeding the same data to the Audio Precision, sourced from a WAV file on the Mac mini. You can see that while the eye is still wide open, there is greater timing uncertainty at the start and finish of the unit cycle than there was with the Boulder. The measured jitter level was 5.85ns; ie, three times greater than the Boulder. To put this in perspective, the jitter from the E-Mu 0404 fed the 16-bit Miller/Dunn data was 7.9ns, and that of the M-Audio Transit USB was 2ns.
Fig.2 Bel Canto USB Link 24/96, eye pattern of S/PDIF data output carrying 16-bit J-Test signal (±500mV vertical scale, 175ns horizontal scale).
It would seem essential, therefore, that to get the best sound quality from the USB Link, it should be used with a D/A processor that offers excellent rejection of timing uncertainty in the incoming S/PDIF datastream.1 I looked at the effects of datastream jitter in the reconstructed analog signal with four D/A processors: the Benchmark DAC1, Bel Canto e.One DAC3, Musical Fidelity X-24K, and Assemblage DAC1. I fed each first the 16-bit Miller/Dunn J-Test signal over TosLink S/PDIF from the RME soundcard on my test PC, then the same signal via my Mac Mini's USB output fed to the USB Link.
The Benchmark uses a sample-rate converter to upsample the incoming data to 110kHz before feeding them to the DAC chip. It thus offers superb immunity to datastream jitter. Fig.3 shows a high-resolution plot of its analog output while it decoded the J-Test data transmitted from my PC via TosLink. The central spike, which represents a high-level tone at exactly one-quarter the same rate, is narrow and well defined, and the noise floor is at the 140dBFS level, about 6dB lower than the Compact Disc's theoretical limit. While low-frequency sidebands can be seen to either side of the tone, these lie at the residual level of the test signal and have not been exaggerated by the bandwidth limitations of the optical datalink.
When I analysed the Benchmark's analog output while it decoded USB data converted to electrical S/PDIF by the Bel Canto USB Link, to all intents and purposes, the spectrum was identical to that in fig.3. It is shown in fig.4. Regardless of any jitter in the Bel Canto's output, it will not affect the Benchmark's sound quality.
Fig.3 Benchmark DAC1, high-resolution jitter spectrum of analog output signal, 11.025kHz at 6dBFS, sampled at 44.1kHz with LSB toggled at 229Hz, TosLink S/PDIF data. Center frequency of trace, 11.025kHz; frequency range, ±3.5kHz (left channel blue, right red).
Fig.4 Benchmark DAC1, high-resolution jitter spectrum of analog output signal, 11.025kHz at 6dBFS, sampled at 44.1kHz with LSB toggled at 229Hz, USB data. Center frequency of trace, 11.025kHz; frequency range, ±3.5kHz (left channel blue, right red).
Bel Canto's own e.One DAC3 was almost as good as the Benchmark DAC1 at rejecting jitter, so it would be a good choice to use with the USB Link. The older D/A processors fared less well. Fig.5 shows the spectrum of the Assemblage DAC1's analog output, which was the worst of the bunch. This DAC is basically wide open to incoming jitter. Not only are the data-related sidebands very much higher in level than they were with the Benchmark, the central peak has developed wide "skirts" due to the presence of random low-frequency jitter components, and sidebands can be seen at the power-supplyrelated frequencies of ±120Hz. Dreadful.John Atkinson
Fig.5 Assemblage DAC1, high-resolution jitter spectrum of analog output signal, 11.025kHz at 6dBFS, sampled at 44.1kHz with LSB toggled at 229Hz, USB data converted to S/PDIF with Bel Canto USB Link 24/96. Center frequency of trace, 11.025kHz; frequency range, ±3.5kHz (left channel blue, right red).
Footnote 1: There are many articles about jitter in Stereophile's free online archives. A good starting point is "A Case of the Jitters," which also includes many useful links.