Digital Processor Reviews

I measured the CanEver ZeroUno with my Audio Precision SYS2722 system (see the January 2008 "As We See It"), using both the Audio Precision's optical and electrical digital outputs and USB data sourced from my MacBook Pro running on battery power with Pure Music 3.0 playing WAV and AIFF test-tone files. Apple's USB Prober utility identified the CanEver as "xCORE USB Audio 2.0" from "XMOS," and confirmed that its USB port operated in the optimal isochronous asynchronous mode. Apple's AudioMIDI utility revealed that, via USB, the ZeroUno accepted 24-bit integer data sampled at all rates from 44.1 to 384kHz. The optical and electrical S/PDIF inputs locked to datastreams with sample rates up to 192kHz. I started the measurements with the CanEver DAC set to its factory defaults: Jitter filter on, Oversampling filter on, and resolution at "9 bits pseudo." All the measurements were taken from the single-ended RCA jacks; the XLR jacks don't offer a balanced output but duplicate the single-ended outputs.

With the volume control set to its maximum of "0dB," the maximum output level at 1kHz was 1.7V, which is 1.4dB lower than the CD standard's 2V. The outputs preserved absolute polarity (ie, were non-inverting), and the output impedance at middle and high frequencies was a moderate 850 ohms. This rose to 7800 ohms at the bottom of the audioband, presumably due to the presence of an output coupling capacitor following the single-ended tube stage. To say the least, the specified impedance of –1 ohm is erroneous; the ZeroUno needs to be used with a preamp or power amp having an input impedance of at least 40k ohms if the low frequencies are not to sound a little lean.

The Sharp and Smooth FIR reconstruction filters are available only with USB data, which the ZeroUno's display identifies as "i2s." Similarly, the option of turning off the oversampling filter is not possible with S/PDIF data—you can turn it off, but the DAC then loses lock with the incoming datastream. Figs. 1 and 2 respectively show the DAC's impulse response with the Sharp and Smooth filters—both are linear-phase types with time-symmetrical ringing, but Sharp is a conventional long filter, Smooth a very short filter. The impulse response with the oversampling filter turned off is shown in fig.3—it looks like the impulse response of an analog low-pass filter.

With 44.1kHz-sampled white noise, the Sharp filter offered a rapid rolloff above 21kHz (fig.4, red and magenta traces), with almost complete elimination of the aliased image at 25kHz of a full-scale 19.1kHz tone (blue, cyan). The Smooth filter offered a much slower ultrasonic rolloff (fig.5), and I had to reduce the level of the 19.1kHz tone by 3dB in this graph to avoid the spectrum filling up with aliasing products. (As AD's auditioning was performed primarily with the Smooth filter, I suspect that this behavior correlates with his finding the "strings had texture, but also a bit of fuzz: too much texture, if you can imagine such a thing." Without the oversampling filter (fig.6), the high-frequency rolloff slope was shallow, with aliased images abundant both within and above the audioband.

With S/PDIF data, the default reconstruction filter appears to be identical to the Sharp filter with USB data. Fig.7 shows the frequency response with that filter and with data sampled at 44.1, 96, 192, and 384kHz. Other than a sharp rolloff below each Nyquist frequency (half the sample rate), the response at the lower rates follows the same shape: a slight boost in the low bass and an ultrasonic peak (+1.8dB at 80kHz). The response with 384kHz data (blue and gray traces) extends to above 100kHz, but then rolls off sharply before reaching the data's 192kHz Nyquist frequency. Fig.8 shows the frequency response with 44.1kHz data with the oversampling filter on (blue and red traces) and off (green, gray). To my surprise, the DAC's output with the filter turned off drops off sharply in the mid-treble, reaching –3dB at 7kHz. The HF cutoff increases with the sample rate—eg, with 96kHz data, the response is now –3dB at a more acceptable 15kHz—but providing the ability to turn off the oversampling filter seems perverse.

Channel separation was excellent, at >110dB below 10kHz. With its use of a tube stage, I expected to see power-supply–related spuriae in the ZeroUno's output. However, these spuriae—almost all odd-order harmonics of the 60Hz power-line frequency—are all close to or below –110dB ref. full scale (fig.9), and so should not be audible. However, the presence of very low-frequency or "flicker" noise moves the waveform of an undithered tone at –90.31dBFS away from the time axis (fig.10). The three DC voltage levels are well defined, however, and the level of higher-frequency random noise is low enough to allow the Sharp filter's ringing to be seen. The analog noise floor was low enough that when I changed the bit depth of the incoming data from 16 to 24 with a dithered tone at –90dBFS, the floor dropped by 18dB (fig.11), implying resolution of at least 19 bits.

Harmonic distortion was relatively low, as long as the load impedance was high. Fig.12 shows the spectrum with the ZeroUno reproducing a full-scale 50Hz tone into 100k ohms. The second and third harmonics are almost equal in level, but at –80dB (0.01%) are negligible. The levels of the distortion harmonics increased, however, with loads below 10k ohms. When I tested the CanEver for intermodulation distortion with an equal mix of 19 and 20kHz tones, the combined waveform peaking at 0dBFS, and with the Sharp filter on (fig.13), the second-order or difference product at 1kHz lay at a respectable –74dB (0.02%).

Tested for its rejection of word-clock jitter with S/PDIF J-Test data and its Jitter filter turned on (fig.14), the CanEver ZeroUno's output spectrum revealed that all of the odd-order harmonics of the low-frequency, LSB-level squarewave were at the correct level (green line), but the spike that represents the high-level tone at one-quarter the sample rate was surrounded by sidebands spaced at 60Hz. These most likely stem from supply noise on the DAC chip's voltage-reference pin. Without the jitter filter (fig.15), the data-rated sidebands were all still at the correct level, but now there was a 9.55kHz idle tone present at –100dB.

The CanEver ZeroUno offers respectable measured performance, with only its low-level noise signature and its intolerance of low load impedances hinting at the presence of vacuum tubes. But if you're interested in this DAC, keep its jitter and oversampling filters switched on.—John Atkinson