Digital Processor Reviews

I don't routinely measure the products written about by Stereophile's columnists—Sam Tellig, Michael Fremer, Art Dudley, Kalman Rubinson, John Marks—but when something comes along that catches their and my fancies, I do get it up on the test bench. Such was the case with the Chinese-made Music Hall dac25.2 D/A processor, which Sam Tellig wrote about in the August 2009 issue, as it offered a lot of both features and sound quality for just $599. It even uses a tube—a single 6922 dual-triode—to give it the required degree of audiophile coolness.

Looking inside the dac25.2, its circuitry is carried on a single, large printed-circuit board behind the rear panel, and the power transformer is located between this board and the front panel. A Crystal CS8416 chip is used for the S/PDIF and AES/EBU inputs, allowing the Music Hall to handle data with sample rates up to 96kHz and bit depths up to 24. A Texas Instruments SRC4192 provides user-selectable sample-rate conversion to 96 or 192kHz, while the DAC is a 24-bit/192kHz Burr-Brown PCM 1696 chip, which also performs the digital filtering. As well as the 6922 tube, the analog circuitry features high-performance op-amps in the forms of Burr-Brown OPA2084 and OPA2134 chips.

I looked at the measured behavior of the Music Hall dac25.2 using the Audio Precision SYS2722 system (see www.ap.com and "As We See It" in the January 2008 issue), as well as, for some tests, my Audio Precision System One Dual Domain and the Miller Audio Research Jitter Analyzer. To examine the performance of the Music Hall's USB input, I drove it with my battery-powered MacBook running OS10.4.11 and playing WAV files using Bias Peak 6.2. The dac25.2 identified itself to the computer as "USB Audio DAC," and the AudioMidi setup program reported that the DAC was limited to 16-bit data and sample rates at or below 48kHz—as I expected from its Burr-Brown PCM2704 USB receiver chip.

The dac25.2 successfully locked to S/PDIF data with rates ranging from 32 to 96kHz. All outputs preserved absolute polarity—ie, were non-inverting—and the maximum output level was 2.2V from the balanced XLR jacks, 1.95V from the unbalanced RCAs (the latter 0.2dB below the CD standard's 2V RMS), and 2.6V from the headphone jack. The balanced output impedance was a very high 3900 ohms, though this was the same across the audioband, meaning that there won't be any frequency-response anomalies with preamplifiers of low input impedance. The unbalanced impedance was a more reasonable 580 ohms at high and midrange frequencies, though the fact that it rose to 2k ohms at 20Hz will mandate using a preamp with an input impedance >20k ohms if the balance is not to sound a bit lean.

Similarly, though the headphone jack's output impedance was fairly low at 12.5 ohms in the treble and midrange, it rose to 98 ohms at 20Hz, which will adversely affect the bass quality with low-impedance 'phones, possibly making them sound both light in weight and a touch flabby. And comparing the dac25.2 with the Musical Fidelity V-CAN, ST did indeed feel that the V-CAN got a grip on the bottom end that the dac25.2 couldn't manage. The Music Hall headphone output lacked "balls, if you will"—what I would expect from its rise in output impedance at low frequencies.

The dac25.2's frequency response with 96kHz data (fig.1, blue and red traces) gently rolled off in the top octave, dropping by 0.7dB at 20kHz and reaching –3dB at 40kHz. The response with CD data (fig.1, cyan and magenta traces) followed the same rolloff, but of course dropped like a stone above 20kHz. These responses were taken into the high 100k ohms load; into 600 ohms, the balanced output actually didn't roll off quite as quickly, the output at 20kHz lying at –0.35dB instead of –0.7dB. Balanced channel separation (fig.2) was superb, at >120dB in both directions below 3kHz, and still 110dB at 20kHz.

Fig.1 Music Hall dac25.2, balanced frequency response at –12dBFS into 100k ohms with 44.1kHz sample rate (left channel magenta, right cyan) and 96kHz (left blue, right red). (0.25dB/vertical div.)

Fig.2 Music Hall dac25.2, balanced channel separation (5dB/vertical div.).

Spectral analysis, using a swept 1/3-octave bandpass filter, of the Music Hall's output reproducing a dithered 1kHz tone from S/PDIF-sourced data revealed a noise floor that was free of power-supply or harmonic spuriae with both 16-bit data (fig.3, top pair of traces) and 24-bit data (bottom traces). The drop in the noise floor of around 14dB with the increase in bit depth indicates a resolution of around 18.5 bits, confirmed by FFT-based analysis of the same signals (fig.4). The USB input would handle only 16-bit data, of course, 24-bit data being truncated, with the consequent introduction of quantizing distortion.

Fig.3 Music Hall dac25.2, 1/3-octave spectrum with noise and spuriae of dithered 1kHz tone at –90dBFS with 16-bit data (top) and 24-bit data (bottom). (Right channel dashed.)

Fig.4 Music Hall dac25.2, FFT-derived spectrum with noise and spuriae of dithered 1kHz tone at –90dBFS with 16-bit and 24-bit data (left channel blue, right red).

Plotting the linearity error with 16-bit data gave a trace that was dominated by the recorded dither noise (fig.5). With 24-bit data, any level error was negligible down to –120dBFS. The dac25.2's reproduction of an undithered tone at exactly 90.31dBFS was essentially perfect, unambiguously presenting a symmetrical waveform and the three DC voltage levels (fig.6; see the Playback Designs review in the February 2010 issue for considerably worse performance on this test). Increasing the incoming word length to 24 bits gave an excellent facsimile of a sinewave (fig.7).

Fig.5 Music Hall dac25.2, linearity error, dBr vs dBFS (2dB/vertical div.).

Fig.6 Music Hall dac25.2, waveform of undithered 1kHz sinewave at –90.31dBFS, 16-bit data (left channel blue, right red).

Fig.7 Music Hall dac25.2, waveform of undithered 1kHz sinewave at –90.31dBFS, 24-bit data (left channel blue, right red).

Even with the high output impedance from its balanced jacks, the Music Hall dac25.2 offered very low harmonic distortion. It behaved similarly into both 100k and 600 ohms, with the harmonics predominantly the innocuous second and third and lying at or below –109dB (0.0003%, fig.8). This graph was taken with the oversampling switched off; oversampling, whether to 96 or 192kHz, did increase the distortion, with a regular series now evident above the noise floor (fig.9), but even the harmonic highest in level, the second, was still very low, at –94dB (0.002%). I doubt that this difference in behavior would explain why ST preferred the 192kHz setting "for the most airy, open, relaxed sound," or Roy Hall's preference for no upsampling at all. Conversely, the dac25.2 gave lower levels of intermodulation distortion with oversampling switched on (fig.10), though again, the levels involved are very low.

Fig.8 Music Hall dac25.2, balanced spectrum of 50Hz sinewave at 0dBFS into 600 ohms, 24-bit data, no oversampling (left channel blue, right red; linear frequency scale).

Fig.9 Music Hall dac25.2, balanced spectrum of 50Hz sinewave at 0dBFS into 600 ohms, 24-bit data, 192kHz oversampling (left channel blue, right red; linear frequency scale).

Fig.10 Music Hall dac25.2, balanced HF intermodulation spectrum, 19+20kHz at 0dBFS peak into 100k ohms, 24-bit data, 192kHz oversampling (left channel blue, right red; linear frequency scale).

With its oversampling switched off, the Music Hall dac25.2 offered only modest rejection of word-clock jitter. In the worst case, with the processor fed a 16-bit version of the Miller-Dunn diagnostic data from my PC via 15' of plastic TosLink, I measured 624 picoseconds peak–peak of jitter-related sidebands in the left channel and 629ps p–p in the right, primarily at the data-related frequency of 229Hz. Feeding the data from my MacBook via USB dropped the jitter to 496ps left, 504ps right, but the spectrum was relatively dirty (fig.11), and there was some spectral spreading of the central spike representing the 11.025kHz tone. This is to be expected, given that the PCM2704 chip operates in what is called "isochronous" mode, which hands off the clocking of the data to the host PC. Switching on the oversampling eliminated the data-related sidebands (fig.12) and dropped the measured jitter below the resolution limit of the Miller Analyzer with both S/PDIF and USB data.

Fig.11 Music Hall dac25.2, high-resolution jitter spectrum of analog output signal, 11.025kHz at –6dBFS, sampled at 44.1kHz with LSB toggled at 229Hz, 16-bit USB data, no oversampling. Center frequency of trace, 11.025kHz; frequency range, ±3.5kHz (left channel blue, right red).

Fig.12 Music Hall dac25.2, high-resolution jitter spectrum of analog output signal, 11.025kHz at –6dBFS, sampled at 44.1kHz with LSB toggled at 229Hz, 16-bit USB data with oversampling. Center frequency of trace, 11.025kHz; frequency range, ±3.5kHz (left channel blue, right red).

The Music Hall dac25.2 offers generally excellent measured performance for a very competitive price. Though I would like to have seen a lower output impedance from its headphone jack, that feature does distinguish it from its similarly priced competition in the forms of the Cambridge DacMagic and the Musical Fidelity V-DAC.—John Atkinson