NAD M51 Direct Digital D/A converter Measurements
Sidebar 3: Measurements
I used Stereophile's loan sample of the top-of-the-line Audio Precision SYS2722 system to measure the NAD M51 (see www.ap.com and the January 2008 "As We See It"; for some tests, I also used my vintage Audio Precision System One Dual Domain.
Both the M51's S/PDIF and USB inputs successfully locked to data with sample rates of up to 192kHz, including 88.2 and 176.4kHz. Although the TosLink format is specified to work only up to 96kHz, the M51's TosLink input did lock to 192kHz data. The Mac's USB Prober utility revealed the M51's product ID to be "NAD USB Audio 2.0" and that the USB receiver was operating in isochronous asynchronous mode (which means that the DAC, not the computer, controls the D/A conversion timing).
The volume control operated in accurate 1dB steps. With the output level set to 0dB, the M51's maximum output level at 1kHz in fixed output mode was 4.75V from the balanced XLR jacks and 2.375V from the unbalanced RCA jacks, the latter 1.5dB higher than the CD standard's 2V, as JI found in his auditioning. The output impedance at low and middle frequencies was 187 ohms from the XLRs, 141 ohms from the RCAs, these figures respectively increasing to 213 and 153 ohms at 20kHz. I had set the output polarity to Positive with the Menu button; even so, the NAD's output from both its XLRs and RCAs inverted polarity (fig.1). This graph also indicates that the NAD's reconstruction filter is a time-symmetrical FIR type.
Fig.2 shows the M51's frequency response with data sampled at 44.1kHz (cyan and magenta traces), 96kHz (blue and green), and 192kHz (gray and red). The output above the audioband rises before plunging to just below half the sample rate. At 192kHz it peaks by 2.5dB at 80kHz; while I doubt this will have any audible effect, it is unusual. Channel separation (not shown) was superb at >125dB in both directions below 1kHz, and still 113dB at 20kHz.
The top pair of traces in fig.3 shows a 1/3-octave spectral analysis of the NAD's outputs while it decoded dithered 16-bit S/PDIF data representing a 1kHz tone at 90dBFS. The traces peak at exactly 90dBFS, suggesting low linearity error, and the noise floor is free from AC- or harmonic-related spuriae. These traces actually show the spectrum of the dither noise used to encode the data. Extending the bit depth to 24 gives the middle pair of traces in fig.3: The noise floor has dropped by up to 30dB, which implies resolution of 21 bits! The M51 is among the highest-resolution DACs I have measured, and readily resolved a tone at 120dBFS (bottom pair of traces).
To remain consistent with the measurements of DAC resolution I have performed since 1989, I used a swept-bandpass technique to generate the traces in fig.3. Repeating the analysis with a modern FFT technique gave a similar picture (fig.4), confirming that there are no harmonic-distortion or power-supplyrelated spuriae present with 24-bit data. Rerunning the analysis with USB data confirmed that the M51 preserves the 24-bit resolution through this input, though now a trace of third-harmonic distortion, at a mind-numbingly low 120dB, is evident (fig.5).
With its very low noise floor and very high resolution, the M51's reproduction of an undithered 16-bit tone at exactly 90.31dBFS was essentially perfect (fig.6), with a symmetrical waveform and the Gibbs Phenomenon "ringing" on the waveform tops well defined. With 24-bit data, the M51 produced a superbly defined sinewave (fig.7).
The level of the M51's noise floor varied with signal level to a greater extent than the norm. The bottom two pairs of traces in fig.8 were taken with 1kHz tones at 40 and 90dBFS; the noise floor is not affected by the change in signal level. However, with a 1kHz tone at 0dBFS, the noise floor rises by up to 30dB, depending on frequency. (While there is a small effect on the level of the noise, due to the Audio Precision's autoranging input circuitry, this is an order of magnitude less than what can be seen in this graph.) I repeated the noise analysis with a 1kHz tone at 0, 3, and 10dBFS (fig.9). This graph reveals that the largest modulation of the noise floor occurs when the signal rises above 3dBFS. It would appear that to obtain that superb resolution of low-level information, NAD's engineers have sacrificed some dynamic range for the very highest-level signals, which will be relatively rare in all music other than hypercompressed pop.
The same phenomenon appeared when I examined the M51's harmonic and intermodulation distortion, for which testing I usually use tones at 0dBFS. Fig.10 shows the spectrum of the NAD's output while it decoded data representing a 1kHz tone at 0dBFS. Some odd-order harmonics can be seenthese are low in level, at or below 84dB (0.006%), and were commendably unaffected when I reduced the load impedance to 600 ohmsbut not only that, the noise floor rises with increasing frequency. This behavior is reminiscent of DSD-encoded data. Reducing the level of the 1kHz tone to 10dBFS gave a more conventional spectrum (fig.11), with the noise-floor components at or below 150dBFS and each distortion harmonic at or below 120dBFS (0.0001%). However, an odd tone can be seen at 1.5kHz at 124dBFS; perhaps an idle tone of some kind? A similar picture can be seen with the high-frequency intermodulation test, with the 19+20kHz signal at 0dBFS (fig.12) and 10dBFS (fig.13).
Tested for rejection of word-clock jitter with 16-bit (fig.14, blue and magenta traces) and 24-bit (fig.14, cyan and red traces) versions of the J-Test data, the M51's output spectra were free from any jitter-related or power-supplyrelated sidebands. With 16-bit data, the odd-order harmonics of the low-frequency, LSB-level squarewaves are at the residual level. Fig.14 was taken with TosLink data; repeating the test with 24-bit USB data gave an identical result, indicating the effectiveness of the M51's asynchronous input mode.
NAD's M51 Direct Digital DAC offers measured performance that is almost beyond reproach. Color me impressed.John Atkinson