HRT Music Streamer HD USB D/A processor Measurements
Sidebar 3: Measurements
My testing of the HRT Music Streamer HD was primarily performed with 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"), with some tests performed on my vintage Audio precision System One. The sources for the tests were WAV and AIFF files played from my MacBook Pro running BIAS Peak Pro and connected to the HRT with the Music Streamer set to USB2.0. I checked that the sample rate and bit depth were correctly set for each file with the Audio-MIDI setup utility. Bit depth was set to "24-bit Integer" for all tests and I didn't see any way to change it.
The USB Prober utility identified the HRT as "Music Streamer HD 2.0," with a manufacturer ID string of "www.hirestech.com 2013 REV 1.0," operating in Isochronous asynchronous mode. Audio-MIDI, however, identified the HRT as "USB 2.0 Audio Out." The Music Streamer HD operated correctly with USB data sampled at all rates from 44.1 to 192kHz, including 88.2 and 176.4kHz.
The HRT's maximum output level at 1kHz was 4.77V from its balanced jacks, 2.38V from the unbalanced, the latter 1.53V higher than the CD standard's 2V RMS. Both outputs preserved absolute polarity (ie, were non-inverting), the XLR jacks being wired with pin 2 hot. The balanced output impedance was 99 ohms, the single-ended 50 ohms (both figures include 6' of interconnect).
The Music Streamer Pro's frequency response followed the same basic pattern at all sample rates: a gentle rolloff beginning at 10kHz, reaching 0.3dB at 20kHz, with then a sharp drop-off just before half the sample rate (fig.1). The channels were matched to within 0.1dB. Fig.2 shows the behavior of the HRT's reconstruction filter. The red trace is the FFT-derived spectrum of a full-scale 19.1kHz tone sampled at 44.1kHz. Though the second, third, and fifth harmonics can be seen, there is no trace of the primary image of the tone at 25kHz (44.119.1kHz). The blue trace was taken with 44.1kHz-sampled white noise at 4dBFS. It reveals that the reconstruction filter rolls off very rapidly above 21kHz, reaching 110dBFS by 23kHz. However, the rate of rolloff slows dramatically above that frequency, something I also saw in the first sample of the iFi iDAC, reviewed last month, though that product uses a different DAC and digital filter. The noise floor begins to rise above 50kHz, perhaps due to the noise-shaping used by the HRT's Burr-Brown DAC chip.
Channel separation (not shown) was excellent, at 125dB in the midband and still 115dB at 20kHz, and the Music Streamer HD was very quiet. If "HD" stands for "High Definition," then the Music Streamer HD earns its name. The top two traces in fig.3 were taken by sweeping a 1/3-octave bandpass filter from 20kHz down to 20Hz while the Music Streamer HD decoded 16-bit data representing a dithered 1kHz tone at 90dBFS. The traces peak at exactly 90dBFS, suggesting minimal linearity error, which was confirmed by additional testing (not shown). The top traces in fig.3 actually show the spectrum of the dither noise used to encode the signal, there being no harmonics or power-supply spuriae visible.
With 24-bit data (fig.3, middle traces), the noise floor drops by up to 24dB, suggesting that the DAC has around 20-bit resolutionsuperb performance even when not taking into account the Music Streamer HD's price, and easily enough to resolve a 24-bit tone at 120dBFS (fig.3, bottom traces). Repeating the test with a modern FFT technique (fig.4), I had to increase the vertical scale to 160dB in order to uncover the 24-bit noise floor (blue and red traces). However, this does unmask some very low-level distortion harmonics.
With its superb resolution and low noise, the Music Streamer HD's reproduction of an undithered 16-bit tone at exactly 90.31dBFS was perfect (fig.5): the waveform is superbly symmetrical about the time axis, the three DC voltage levels described by the data are readily resolved, and the Gibbs Phenomenon "ringing" on the tops and bottoms of the waveform is not obscured by noise. This ringing also indicates that the HRT's reconstruction filter is a conventional linear-phase type. With undithered 24-bit data, the result is an excellent sinewave (fig.6).
At low frequencies (fig.7), the distortion signature comprises a collection of harmonics that decrease with order, with the second harmonic the highest in level, at 100dB (0.001%). The picture changes at higher frequencies, however, with the third harmonic now the highest in level (fig.8), at 90dB (0.003%). And looking back at fig.2, the third harmonics of the 19.1kHz tone in that graph lie at 66dB (0.05%). So though the distortion is low at all frequenciesand, commendably, unaffected by reducing the loading to 600 ohms!there appears to be a relationship between the distortion spectrum and the signal frequency. I have no idea if this will have any bearing on the HRT's sound quality.
When it came to intermodulation distortion, an equal mix of 19 and 20kHz tones sampled at 44.1kHz, with the waveform peaking at 0dBFS, has the same dirty-looking spectrum I found with the HRT Music Streamer Pro, which Jon Iverson reviewed in February 2011, and which uses the same PCM1794 DAC chip. Repeating the test with 96kHz-sampled data gave the spectrum shown in fig.9. The spectrum is significantly cleaner overall, in that the noise floor is no longer obscured by enharmonic products, but a large number of intermodulation products spaced at 1kHz are visible. Reducing the signal level by 6dB dropped the level of the intermodulation products in the midband considerably, but the components at 1, 18, and 21kHz were reduced only by the same 6dB as the stimulus. It is very unlikely for music to contain high-level tones at the top of the audioband, but perhaps this particular test signal reveals some mathematical limitations of the digital filter.
An asynchronous USB connection should be free from jitter problems; the Music Streamer HD's reproduction of 16-bit/44.1kHz J-Test data is excellent (fig.10), there being no jitter-related sidebands visible around the 11.025kHz tone, and all the odd-order harmonics of the 229.6875Hz LSB-level squarewave lie at the correct levels. With 24-bit J-Test data there appeared a pair of sidebands of unknown origin, at ±150Hz and a level of 135dBFS (fig.11).
Overall, HRT's Music Streamer HD measures significantly better than the Music Streamer Pro. There is, however, still that peculiar behavior with high-level 19 and 20kHz tones to contend with. As for why AD ultimately preferred the sound of the Halide HD: Though the Halide has less measured resolution, it does use a slow-rolloff reconstruction filter, something that listeners do seem to like.John Atkinson