Digital Processor Reviews

Superficially, the DAC-R looks identical to Rega's original DAC, which Sam Tellig reviewed in the May 2011 issue. However, the DAC-R replaces the DAC's Burr-Brown PCM2707 USB receiver chip with a higher-performing XMOS chip. I measured the DAC-R with my Audio Precision SYS2722 system (see www.ap.com and the January 2008 "As We See It"). Sources were S/PDIF on TosLink and coaxial from the SYS2722, and USB from my 2012 MacBook Pro running on battery power. Macintosh's USB Prober utility reported the Rega's product string as "XMOS USB Audio 2.0," and confirmed that the USB port operated in the optimal isochronous asynchronous mode. Though specified as being 192kHz capable, the TosLink input locked to datastreams with sample rates only up to 96kHz. However, Rega does warn, in the DAC-R's manual, that TosLink connection above 96kHz will depend on the link used, and I was using a cheap plastic TosLink cable. The coaxial S/PDIF and USB inputs operated correctly up to 192kHz.

The maximum output level at 1kHz was very similar to that of the original Rega DAC (footnote 1) at 2.176V. The output preserved absolute polarity and was sourced from a fairly low impedance: 595 ohms at high and middle frequencies, rising inconsequentially to 663 ohms in the low bass. The impulse response with Filter 1 (fig.1) is a conventional time-symmetrical, linear-phase, half-band type; with Filters 2 and 3, the impulse response is a minimum-phase type, with all ringing following the impulse (fig.2).

With 44.1kHz-sampled white noise at –4dBFS, Filter 1 gave a rapid rolloff above the audioband, reaching the stopband noise floor at 24kHz (fig.3, red and magenta traces), and suppressing the ultrasonic image at 25kHz of a full-scale tone at 19.1kHz by almost 110dB (fig.3, blue, cyan, footnote 2). Filter 2 behaved virtually identically on this test, though with 10dB greater suppression of the 25kHz image (not shown). Filter 3 is specified as being an apodizing type, and its output spectrum with these signals (fig.4) confirmed that spec, the ultrasonic rolloff reaching the stopband noise floor at precisely half the sample rate (fig.4, vertical green line). A more conventional plot of the DAC-R's frequency response with data sampled at 44.1, 96, and 192kHz (fig.5) suggested that the ultrasonic bandwidth was more restricted than usual at the higher sample rates. With 96kHz data, for example, the response is down by 6dB at 31kHz, which is very similar to how the Rega DAC behaved.

Channel separation at 1kHz was superb, at 105dB R–L and 115dB L–R, these respectively decreasing to 79 and 88dB at 20kHz due to the inevitable capacitive coupling between channels. The audioband noise floor was clean, other than some very low-level spuriae (around –123dB) at the power-supply frequency of 60Hz and its odd-order harmonics, these most likely due to magnetic interference from the transformer.

These spuriae can be seen in fig.6, which plots the DAC-R's output spectrum as it reconstructs a dithered 1kHz tone at –90dBFS with first 16-bit data (cyan and magenta traces), then 24-bit data (blue, red). The increase in bit depth drops the noise floor by 12dB or so, suggesting that the DAC-R's resolution is ca 18 bits—easily enough for it to correctly reproduce the stair-step shape of an undithered tone at exactly –90.31dBFS (fig.7). With undithered 24-bit data (fig.8), the result was a good if slightly noisy sinewave. These last tests were performed with S/PDIF data; repeating them with USB data gave identical results, confirming that the USB input correctly handles 24-bit data.

Like Rega's DAC, the DAC-R offered very low levels of harmonic distortion, even into 600 ohms (fig.9). The second and third harmonics are the highest in level, both lying a couple of dB above –90dB (0.003%). Intermodulation distortion was also extremely low (fig.10). This graph was taken with Filter 3, which suppresses the 20kHz tone by 4dB. This will not audible, however.

Tested for its rejection of word-clock jitter with 16-bit/44.1kHz J-Test data, the DAC-R produced a spectrum very similar to the Rega DAC's (fig.11). While the odd-order harmonics of the LSB-level, low-frequency squarewave are all close to their correct levels (indicated with the green line), there are sidebands at ±120 and ±240Hz.

In our review of the Rega DAC in May 2011, I had ascribed these sidebands to jitter, but I was incorrect. The level of jitter-related sidebands decreases as the frequency of the signal described by the digital data also decreases. But as you can see in fig.12, which compares the spectrum of the DAC-R's output while it decodes data representing a full-scale tone at 10kHz (blue and red traces) with that of a full-scale 6kHz tone (green, gray), the sidebands remain at the same level with the low-frequency tone. This suggests that these sidebands are not jitter related, but are more likely due to there being some 120Hz leakage from the power supply on the DAC chip's voltage reference input. The level of the sidebands dropped by 10dB for each 10dB reduction in signal level, as can be seen in fig.13, which plots the output spectra when the level of the 6 and 10kHz tones is reduced to –20dBFS. With tones at –40dBFS, there are no sidebands at all (fig.14). I conjecture that the DAC-R's power supply has too high an impedance to fully isolate the DAC's voltage reference input when the DAC is reconstructing high-level analog signals. How serious a problem is this? Probably not very, though I did wonder, with the earlier Rega DAC, if this behavior correlated with both Sam Tellig and Jon Iverson finding the DAC to sound warm, even a touch heavy.

Other than that minor issue, the Rega DAC-R offered measured performance that was beyond reproach.—John Atkinson

Footnote 2: This test was suggested to me by Jürgen Reis, chief engineer of MBL.