Columns Retired Columns & Blogs |
Channel Islands Audio Transient Mk.II & VDC5 Mk.II USB D/A processor & power supply Measurements
Sidebar 3: Measurements
I used Stereophile's loan sample of the top-of-the-line Audio Precision SYS2722 system (see www.ap.com and the January 2008 "As We See It") to examine the Channel Islands Audio Transient Mk.II's measured behavior. I used my 2012 Apple MacBook Pro running on battery power to examine the processor's performance via its USB port. For all tests, the Transient was powered from its VDC5 Mk.II supply.
Apple's USB prober utility identified the Transient as the "CIAudio USB Audio 2.0" and confirmed that it operated in the optimal isochronous asynchronous mode with 24-bit resolution. The Transient worked with data sampled at 44.1kHz up to 192kHz, and made all stops in between. The maximum output level with the DAC's volume control set to its maximum (top blue LED lit) was 2.07V and the output preserved absolute polarity (ie, was non-inverting). The output impedance was a moderately low 551 ohms at low and middle frequencies, dropping slightly to 536 ohms at the top of the audioband.
Fig.1 shows the Transient's ultrasonic behavior with two different signals: white noise at 4dBFS, and a 19.1kHz tone at 0dBFS, both sampled at 44.1kHz. (This test was suggested to me by Jürgen Reis of MBL.) The cyan and blue traces show the spectrum of the DAC's analog output with the 19.1kHz tone: the aliasing product at 25kHz (44,10019,100) can be seen to be suppressed by 60dB, and while all the distortion harmonics lie at 92dB and lower (<0.003%), there is a tone present at 69kHz. However, this does lie at 70dB, and so will be inconsequential.
The magenta and red traces in fig.1 show the spectrum with white noise. The rolloff above 21kHz is a little slower than is usually found, and there is a striking series of low-level peaks in the reconstruction filter's ultrasonic stopband. This behavior is very similar to "Filter 1" in the data sheet for the Wolfson WM8742, which other reviews have suggested is the DAC chip used in the Transient. However, it also resembles the behavior in the data sheet for the filter in Wolfson's WM8523 DAC/line-driver chip. When I inspected the circuit, this chip appeared to be connected directly to the Transient's analog output jacksI couldn't see a WM8742, but some of the circuit board was inaccessibleand, like the WM8742, the WM8523 offers a digital volume control. On the other hand, the Transient's impulse response (fig.2) reveals the reconstruction filter to be an asymmetrical, non-halfband type, which is one of the options offered by the WM8742 but not, according to its data sheet, by the WM8523. Perhaps CIAudio's Dusty Vawter can clear up this matter in his "Manufacturer's Comment."
Fig.3 shows the frequency response of the Transient plotted with sample rates of 44.1, 96, and 192kHz. The rolloff is very similar at the two higher rates, the only difference being the sharp drop-off above 47kHz with 96kHz-sampled data (cyan and magenta traces). The DAC behaves a little differently with 44.1kHz data (green and gray), there being a small ripple in the top audio octave. Channel separation (not shown) was good, at 74dB in both directions from 20Hz to 20kHz.
With both 16- and 24-bit dithered data representing a 1kHz tone at 90dBFS (fig.4), the second harmonic at 2kHz was visible 20dB below the level of the primary component. Even-order distortion correlates with waveform asymmetry; if you look at figs. 5 and 6, which show the waveforms of undithered data at 90.31dBFS, you can see some asymmetry: in each graph, the top half of the waveform is smaller than the bottom half. There is also more analog noise in these graphs than I usually find with high-performance D/A processors; looking again at fig.4, you can see that the increase in bit depth has dropped the noise floor by only 8dB or so. This implies that the Transient offers resolution of just over 17 bits, even with 24-bit data.
At high levels, the Transient Mk.II offers low levels of harmonic distortion, with the subjectively benign second harmonic the highest in level, at 90dB (fig.7). The third and fifth harmonics lie around 6dB lower in level than the second, but, commendably, none of the distortion harmonics increased in level when I reduced the load impedance to a punishing 600 ohms. Intermodulation distortion was also very low (fig.8), though the "leaky" reconstruction filter allows the ultrasonic images of the primary tones to be seen in this graph. Again, this spectrum didn't change into a 600 ohm load.
When I examined the Transient's rejection of jitter, I got anomalous results. Though the Miller-Dunn J-Test signal, which comprises a high-level tone at exactly ¼ the sample rate and an LSB-level squarewave at 1?192 the sample rate, is not formally diagnostic for the streaming of audio data via USB, I have nevertheless found it a useful test signal, as it can reveal problems other than word-clock jitter. Fig.9 shows a narrowband spectrum of the Transient's output while it decoded 16-bit J-Test data sampled at 44.1kHz. The central spike that represents the 11.025kHz tone is well defined, with no "skirts" that might result from random low-frequency variation in the clock frequency. And while a pair of low-frequency sidebands can be seen, these are at 125dB and presumably innocuous. However, what is missing in this graph are almost all of the odd-order harmonics of the low-frequency squarewave, which should slowly decrease from 124dB at the left of the graph to 129dB at the right.
These harmonics are apparent in a digital-domain spectral analysis of the same J-Test data output from the Transient's S/PDIF port (fig.10), so these low-level data are being lost somewhere in the Transient's D/A circuitry. (The DAC's volume control was set to its maximum for these tests.) And by comparing figs. 9 and 10, you can see that, via the analog outputs, the pair of sidebands at ±229.6875Hz is being accentuated by almost 10dB.
The Channel Islands Audio Transient Mk.II's measured performance was a little disappointing, though there is no reason why it shouldn't sound acceptably good with CD-resolution recordings. As Jon Iverson noted, it "holds its own comfortably" in its price class.John Atkinson
- Log in or register to post comments