Digital Processor Reviews

I measured the Ayre Acoustics QX-5 Twenty 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, USB data sourced from my MacBook Pro running on battery power with Pure Music 3.0 playing WAV and AIFF test-tone files, and via Ethernet with Roon 1.3 playing the same files. Apple's USB Prober utility identified the Ayre as "Ayre USB Interface" from "Ayre Acoustics," and the serial number as "Streamlength(tm)." The QX-5's USB port operated in the optimal isochronous asynchronous mode. Apple's AudioMIDI utility revealed that, via USB, the Ayre accepted 24-bit integer data. The optical inputs locked to datastreams with sample rates up to 96kHz, the AES/EBU and S/PDIF inputs to streams of up to 192kHz.

In DAC mode, or in variable mode with its volume control set to "100," the Ayre's maximum output level at 1kHz was 4.47V from the balanced output jacks, 2.23V from the unbalanced jacks, 8.96V from the balanced headphone jacks, and 4.5V from the single-ended headphone jacks. All outputs preserved absolute polarity; ie, were non-inverting. The output impedance was a low 153 ohms from 20Hz to 20kHz from the unbalanced line-output jacks, this doubling as expected from the balanced jacks, to 305 ohms. From the single-ended headphone jacks the output impedance was a super-low 3 ohms, this doubling to 6 ohms from the balanced headphone jacks.

The QX-5's impulse responses with 44.1kHz data with its Music (fig.1) and Measure (fig.2) reconstruction filters reveal that both filters are minimum-phase types, but that Music has very little post-sample ringing. The Music filter appears to be identical to the filter used in Ayre's Codex D/A headphone amplifier and QB-9^DSD D/A processor. The red and magenta traces in fig.3 show that the Music filter has a slow rolloff above the audioband (footnote 1), with the result that the image at 25kHz of a full-scale 19.1kHz tone (cyan and blue traces) is suppressed by just 10dB, and that higher-order images are folded down into the audioband, though the latter lie at or below –70dB. Again, the behavior of this filter is identical to that of the Codex.

With the filter set to Measure, the ultrasonic rolloff (fig.4, red and magenta traces) is more conventional, in that the output plunges above 20kHz, though it hasn't quite reached full attenuation by half the sample rate (vertical green line). This time, the image at 25kHz is suppressed by almost 100dB, and no audioband spuriae are visible. The second, third, fourth, and fifth harmonics of the 19.1kHz tone can be seen, though the highest-level of these, the third at 57.3kHz, is still 66dB down from full level (0.05%). This is impressive circuit linearity, given the QX-5's lack of loop negative feedback.

Fig.5 shows the QX-5's frequency response with the Music filter, taken with spot tones at sample rates of 44.1, 96, 192, and 384kHz. The slow rolloff with 44.1kHz data seen in fig.3 results in an output down by 1dB at 13kHz and by 3dB at 20kHz, which is the same as I found with Ayre's Codex. The slow rolloff is again seen at the higher sample rates in this graph, but the audioband response is basically flat. Channel separation (not shown) was superb, at >120dB below 2kHz, while the QX-5's noise floor was free from power-supply–related spuriae, other than a touch of 180Hz at a vanishingly low –120dB in the left channel, and even lower in the right.

Increasing the depth of a dithered 1kHz tone at –90dBFS from 16 to 24 bits dropped the noise floor by 20dB (fig.6), suggesting resolution of almost 20 bits. The waveform of an undithered tone at exactly –90.31dBFS was symmetrical (fig.7), with the three DC voltages described by the data very well defined. With undithered 24-bit data, the QX-5 output a clean sinewave (fig.8), despite the very low signal level.

As shown in fig.4 and confirmed in fig.9, the Ayre has relatively high levels of distortion present in its output, though again, the highest-level harmonic, the third, is still low in absolute terms. Commendably, the distortion didn't rise when the very high 100k ohm load was replaced with the punishing 600 ohm load. The distortion signature also remained constant at different signal levels, indicating that the circuit's transfer function isn't modulated by the signal. This lack of modulation is something that I suspect correlates with an ease in a product's sound quality.

When tested for intermodulation with my usual equal mix of high-level 19 and 20kHz tones, the QX-5's behavior depended on which filter had been selected. With the Measure filter (fig.10) intermodulation was respectably low, given the analog circuitry's lack of loop negative feedback; the 1kHz difference product lay at –79dB (0.01%). With the Music filter the difference product lay at the same level, though the picture is confused by the presence of many aliasing products (fig.11). Fortunately, music almost never has high energy at the top of the audioband.

The QX-5 offered superb rejection of word-clock jitter, with the odd-order harmonics of the undithered low-frequency, LSB-level squarewave in the 16-bit Miller-Dunne J-Test signal all lying at the correct levels and with no sideband pairs visible (fig.12). This graph was taken with AES/EBU data; with Ethernet and USB data, neither of which should be prone to jitter interference, the picture was identical except for a slight broadening at the base of the spectral spike that represents the high-level tone at one-quarter the sample rate with USB data (fig.13). This broadening, which will be due to very low-frequency random jitter, should have no audible consequences, but it should be noted that I did find the QX-5 to sound slightly more authoritative with AES/EBU and Ethernet data than with USB data. With 24-bit J-Test data (fig.14), there were no spuriae to be seen!

Overall, Ayre's QX-5 Twenty digital hub offers superb measured performance.—John Atkinson