Musical Fidelity CD-Pre24 CD player-preamplifier Measurements part 2
I looked at the CD-Pre24's DAC linearity error using a dithered 16-bit, 500Hz tone that faded from -60dBFS into the noise floor. The left-channel performance from CD was good (fig.9), and was identical for external data and to both channels of the second sample. The good linearity for CD playback and the low noise floor mean that reproduction of an undithered 1kHz sinewave at -90.31dBFS is essentially perfect (fig.10), with the three discrete voltage levels and excellent waveform symmetry readily apparent. When the Musical Fidelity was fed undithered 24-bit data, the waveform was a pretty good sinewave (fig.11).
Fig.9 Musical Fidelity CD-Pre24, left-channel departure from linearity, 16-bit CD data (2dB/vertical div.).
Fig.10 Musical Fidelity CD-Pre24, waveform of undithered 1kHz sinewave at -90.31dBFS, 16-bit CD data.
Fig.11 Musical Fidelity CD-Pre24, waveform of undithered 1kHz sinewave at -90.31dBFS, 24-bit external data.
I was somewhat puzzled when I looked at the CD-Pre24's intermodulation behavior for CD playback. Fig.12 was taken under the same conditions as was the spectrum for analog sources (fig.5), but the spectral lines at 19 and 20kHz show some spreading by comparison. The only possible explanation for this phenomenon is word-clock jitter—yet earlier digital products from Musical Fidelity have been uniformly excellent in this regard.
Fig.12 Musical Fidelity CD-Pre24, HF intermodulation spectrum, DC-25kHz, 19+20kHz at 0dBFS/1V into 8k ohms (linear frequency scale).
Fig.13 is a narrowband spectral analysis of the CD-Pre24's analog noise floor while it decoded undithered data representing a high-level tone at one quarter the sampling frequency, over which has been laid a low-frequency squarewave with a maximum amplitude of 1LSB. As both audio signals are exact integer submultiples of the sample rate, any spuriae apparent in the analog output signal of the device under test will be due to jitter and DAC errors, not to quantization error. What should be a narrow central spike in fig.13 is significantly broadened at its base, just as was seen in fig.12. The actual measured jitter level was a good 365 picoseconds peak-peak, mainly due to data-related sidebands at ±229Hz. But this figure doesn't take random-noise jitter into account, and it is this kind of jitter that is affecting the Musical Fidelity's performance.
Fig.13 Musical Fidelity CD-Pre24, sample 1, high-resolution jitter spectrum of analog output signal, internal CD playback (11.025kHz at -6dBFS sampled at 44.1kHz with LSB toggled at 229Hz). Center frequency of trace, 11.025kHz; frequency range, ±3.5kHz. (Grayed-out trace is similar analysis from PC fitted with RME Digi96/8 Pro soundcard via 1m TosLink.)
Paul Miller, the developer of the Jitter Analyzer I use, has conjectured that this kind of jitter behavior will blur stereo imaging. I wonder also if it correlated with MF finding the CD-Pre24 to lack the Nu-Vista 3D's "bloom and midband richness." When I checked the second sample of the CD-Pre, the jitter dropped to a superbly low 168.7ps, with a pair of sidebands at ±20Hz dominating the measured figure. However, the spreading of the central peak was still apparent, if not quite to the degree shown by the first sample.
Shown grayed-out in fig.13 is a spectral analysis taken under identical conditions with the CD-Pre24 fed external 44.1kHz-sampled data representing the same analytical signal. The level of random jitter has dropped, but there are now high-level pairs of sidebands apparent at ±126Hz and its harmonics, as well as data-related sidebands, all of which sum to give a total word-clock jitter figure of 2.7 nanoseconds, which is poor performance. The second sample was no better in this regard, the measured jitter figure being almost as bad at 2.4ns, with again a single pair of low-frequency sidebands contributing most of the jitter.
Finally, I looked at the performance of the CD-Pre24's A/D converter, which appears to operate at 44.1kHz (footnote 2) and with 24-bit data words available at the unit's digital output. An input level of 2.1V RMS gave a digital level of -1dBFS. However, raising the input level to anything above 2.237V (-0.3dBFS) resulted in the ADC hard-clipping.
Fig.14 shows the ADC's linearity error plotted against absolute level. The increase in positive error below -80dBFS suggests resolution between 15 and 17 bits despite the 24-bit word lengths, with the right channel being about two bits better than the left. The digital-domain spectral analysis shown in fig.15 reveals that the Musical Fidelity's A/D converter produces high-order harmonics. Though these are very low in absolute terms—the 16 MSBs are not affected—their existence implies that the ADC is not properly dithered.
Fig.14 Musical Fidelity CD-Pre24 ADC, departure from linearity (right channel dashed, 2dB/vertical div.).
Fig.15 Musical Fidelity CD-Pre24 ADC, digital-domain spectrum of 1kHz sinewave, DC-25kHz, at -1dBFS (red) and -80dBFS (blue) (linear frequency scale).
Summing up the Musical Fidelity CD-Pre24's measured performance is difficult. The product combines an excellent, virtually bombproof line preamplifier with a good CD player with high dynamic range that is let down somewhat by the random-noise jitter present in its output and the fact that it doesn't apply the correct de-emphasis for the admittedly small number of pre-emphasized CDs. And while the downsampling of high-sample-rate sources will not affect the majority of audiophiles who play CDs, it is still an issue for me. (The problem with handling right-channel data for external digital signals is presumably a sample-specific fault.) The CD-Pre24's A/D converter is okay, but it's not up to the standard I've been led to expect by such relatively inexpensive products as the Alesis Masterlink, which Mikey reviewed in June, or the RME Digi96/8 PAD soundcard, which Wes Phillips reviewed last February.
Given the technical excellence of the last two digital products from Musical Fidelity that have been reviewed in these pages—the A324 D/A converter and the Nu-Vista 3D CD player, both of which offered state-of-the-art performance—I must admit to being disappointed with some of the CD-Pre24's measurements.—John Atkinson
Footnote 2: Peculiarly, even though RME's DIGICheck program measured the sample rate as a true 44.1kHz, the "32kHz" flag was set in the datastream's subcode.—JA