Measurements: Sample 2
John Atkinson remeasured the NuForce CDP-8 in February 2011 (Vol.34 No.2):
"The NuForce CDP-8 is a remarkably good CD player, one of the best I've heard in years," summed up Wes Phillips in his November 2010 review of this slim, attractive-looking CD player from California company NuForce. However, when I got the $1500 CDP-8 on my test bench, I found some serious shortcomings in its measured performance. The player's level of background noise and the spectrum of that noise were related to the signalthis is called "noise modulation," and human hearing is sensitive to it. In addition, the NuForce suffered from a very high level of jitter-related spuriae in its analog output, calculated by the Miller Jitter Analyzer to be 3.13 nanoseconds peakpeak. This is one of the highest levels of jitter I have encountered, and about 20 times higher than in the best products I have measured.
NuForce's Jason Lim addressed this matter in his "Manufacturer's Comment": "We are . . . embarrassed and chagrined by the poor jitter measurement. We had felt that the basics of performance were addressed by design, and have been applying our effort at listening and making decisions based on listening. Like Wes, we had not noticed any significant degradation in the sound, so we had no reason to look for jitter issues. The measurement was a surprise to us. However, we have recently made a small change to reduce EMI-RFI radiation that also seems to have effectively eliminated the jitter."
Fig.1 NuForce CDP-8, spectrum of 50Hz sinewave at 0dBFS into 100k ohms, old sample (left channel blue, right red) and new sample (left cyan, right magenta). (Linear frequency scale.)
Lim sent me a sample of the revised CDP-8 (footnote 1), which I ran through my set of tests using Audio Precision's SYS2722 system (see www.ap.com). The new sample's frequency response, channel balance and separation, output impedance, and maximum output level were all identical to the original's. The noise floor of the new sample was significantly cleaner, however, as can be seen in fig.1, the spectrum of the CDP-8's output while it decoded data representing a full-scale 50Hz tone. The blue trace is the original CDP-8's output; both harmonic and 124Hz-spaced enharmonic spuriae can be seen, and the noise floor rises around the spectral spike at 50Hz. The magenta trace shows the behavior of the new sample; there is now no noise modulation or enharmonic spuriae, and the distortion components are all very low in level.
Fig.2 NuForce CDP-8, HF intermodulation spectrum, 19+20kHz at 0dBFS peak into 100k ohms, 16-bit data, old sample (left channel blue, right red) and new sample (left cyan, right magenta). (Linear frequency scale.)
Fig.2 shows a similar comparison for the high-frequency intermodulation test. Again the old sample's spectrum is shown in blue, the new sample's in magenta, and again the spectral spikes around the twin fundamentals have disappeared with the new CDP-8. However, the noise floor at lower frequencies still looks granular.
Fig.3 NuForce CDP-8, high-resolution jitter spectrum of analog output signal, 11.025kHz at 6dBFS, sampled at 44.1kHz with LSB toggled at 229Hz, 16-bit CD data: old sample (left channel blue, right red), new sample (left cyan, right magenta). Center frequency of trace, 11.025kHz; frequency range, ±3.5kHz; linear frequency scale.
Finally, the blue trace in fig.3 shows the spectrum of the original CDP-8's output while it decoded the Miller/Dunn J-Test tone. The spectral sidebands to either side of the 11.025kHz tone are spaced at 124Hz intervals and are at a very high level. By contrast, the new sample offers only 220 picoseconds peakpeak of jitterjust 7% of the original'sand the spectrum is much cleaner (fig.3, magenta trace). Data-related sidebands are now at the residual level, and the only significant sidebands lie at ±83.5 and ±167Hz, these of unknown origin.
So yes, the revised version of the NuForce CDP-8 measures very much better than the version WP reviewed. Wes now has both samples, and will report on any differences in sound quality in a future issue.John Atkinson
Footnote 1: The serial number of the original sample was 0401012; the new sample's serial number was 0301069.