Columns Retired Columns & Blogs |
PS Audio NuWave DSD D/A processor Measurements
Sidebar 3: Measurements
I measured the PS Audio NuWave DSD with my Audio Precision SYS2722 system (see www.ap.com and the January 2008 "As We See It"), feeding its S/PDIF input data from the SYS2722 and its USB input data sourced from a battery-powered MacBook Pro running Pure Music 2.0. The S/PDIF input accepted data sampled at all rates from 44.1 to 192kHz. Apple's USB Prober utility identified the NuWave as "PS Audio USB Audio 2.0" and confirmed that it operates in the optimal isochronous asynchronous mode. Mac's AudioMIDI utility stated that the NuWave's USB input would accept 24-bit integer data at all sample rates from 44.1 to 352.8kHz, though PS Audio lists sample rates up to 192kHz in its specifications.
The NuWave DSD's maximum output level at 1kHz with both DSD64 and PCM data was 5.46V from its balanced jacks, 2.73V from its single-ended jacks, and both sets of outputs preserved absolute polarity (ie, were non-inverting). The output impedance was low, at 198 ohms balanced and 99 ohms single-ended, both values consistent at all audio frequencies and confirming the specified impedances. The impulse response with 44.1kHz data (not shown) indicated that the digital reconstruction filter is a conventional FIR type, with symmetrical ringing to either side of the impulse, but with fewer taps than is usually the case. Wideband spectral analysis with 44.1kHz-sampled white noise (fig.1, red and magenta traces) reveals that this filter rolls off very slowly above the audioband, which means that the image at 25kHz of a full-scale tone at 19.1kHz is suppressed by just 10dB (blue and cyan traces).
Fig.2 shows the NuWave's frequency response with data sampled at 44.1, 96, and 192kHz. The use of a slow-rolloff filter with 44.1kHz data (green and gray traces) means that the output is down by 3dB at 20kHz. The responses at the higher sample rates also rolled off earlier than usual, though these were flat almost to 30kHz. With data sampled at 352.8kHz, the response was 3dB at 50kHz and 20dB at 120kHz. Channel separation was excellent, at close to 120dB in both directions below 1kHz, and still 107dB at 20kHz.
With dithered data representing a 1kHz tone at 90dBFS, increasing the bit depth from 16 to 24 dropped the noise floor by 12dB (fig.3), suggesting that the NuWave's resolution is close to 18 bits. With undithered 16-bit data representing a tone at exactly 90.31dBFS, the waveform of the reconstructed analog signal (fig.4) clearly shows the three DC levels described by the data, with excellent waveform symmetry and no DC offset. With undithered 24-bit data (fig.5), the NuWave output a well-formed sinewave overlaid with some high-frequency noise.
As anticipated from fig.1, the dominant distortion harmonic was the third. With a full-scale signal, this lay at 63dB (0.07%, fig.6); but with a tone at 10dBFS, the third harmonic dropped to 81dB (0.009%, fig.7). Figs. 6 and 7 were taken from the balanced outputs into a high load of 100k ohms; the spectra didn't change from the unbalanced outputs or into a very low load of 600 ohms. The use of a slow-rolloff reconstruction filter means that when it was tested with an equal mix of 19 and 20kHz tones, the spectrum of the NuWave's output was dominated by aliasing spuriae (fig.8). Actual intermodulation products were low in level, however.
A broad rise in the noise floor around the two fundamental tones is visible in fig.8; the same behavior can be seen in the spectrum of the NuWave's output when it decoded the undithered 16-bit J-Test tone (fig.9). The modulation of the noise floor is high enough to obscure the higher odd-order harmonics of the LSB-level low-frequency squarewave, and this behavior was identical for the TosLink, coaxial, and USB inputs with both 16- and 24-bit data. Fig.10 shows a wider-band analysis with 16-bit USB data; you can see the lower-order harmonics, but the rise of the noise floor obscures them between 8 and 13kHz. I occasionally used to find behavior like this a decade ago, particularly with a particular data-receiver chip from Finland. I don't know what the effect of this behavior will be on sound qualitythe noise modulation is still 110dB down from the signal's peak levelbut it's something the engineer in me doesn't like to see.
Other than that minor issue, the PS Audio NuWave DSD offers respectable measured performance.John Atkinson
- Log in or register to post comments