Meridian 508.24 CD player Measurements
Checking the Meridian player's error correction with the Pierre Verany test disc, which has deliberate data dropouts engraved on its surface, revealed a good ability to deal with missing data. The CD2 could play up to track 32 (a 1.25mm gap in the data) without the sound muting, and didn't fail catastrophically until track 34 (a 2mm gap). This is considerably better error correction than required by the official CD standard, but it is interesting to note that Meridian machines from the early '90s actually did better than this.
At 2.54V, the player's output level at 0dBFS from its single-ended jacks was a very audible 2.1dB higher than the CD-standard 2V, while a healthy 5.07V was available from the balanced XLR jacks. The player was noninverting from either output, the XLRs being wired with pin 2 hot. The output impedance didn't vary significantly with frequency, the single-ended outputs averaging a low 47 ohms and the balanced twice that, as expected, at 94 ohms.
The 508.24's frequency response at full level, measured from its balanced outputs, is shown in fig.1 (top traces). It is basically flat from 10Hz to 20kHz, with a hint of droop in the top octave. But at -0.1dB at 20kHz, this is negligible. The unbalanced output was identical, and the lower pair of traces in fig.1 are the response with de-emphasis engaged. This maps the unemphasized response exactly, implying a perfect de-emphasis implementation—commendable, given that I understand Meridian performs de-emphasis in the analog domain. The Meridian's channel separation (not shown) was superb, at better than 100dB below 1kHz, rising to a still excellent 86dB at 20kHz.
Fig.1 Meridian 508.24, frequency response at 0dBFS (top) and de-emphasis response (bottom) (right channel dashed, 0.5dB/vertical div.).
Fig.2 shows a 1/3-octave-smoothed spectrum of the 508.24's output while it decoded data representing a dithered 1kHz tone at -90dBFS. The trace is completely free of power-supply components, harmonic distortion, and idling pattern noise. This is one heck of a quiet CD player! Note that the traces don't quite reach the -90dBFS line, implying a small amount of negative linearity error.
Fig.2 Meridian 508.24, spectrum of dithered 1kHz tone at -90.31dBFS, with noise and spuriae. (16-bit data, 1/3-octave analysis, right channel dashed.)
Repeating the spectral analysis but extending the measurement bandwidth to 200kHz and driving the player with code representing "digital black" gave the bottom pair of traces in fig.3. The DAC appears to mute its output when it detects this signal, so this trace effectively shows the ultimate noise floor of the player's analog electronics. This is below -130dBFS over most of the band, and is 10dB quieter than that in the power-supply hum region! (Ignore the small peaks at 2kHz and 20kHz, which might well be measurement artifacts.)
Fig.3 Meridian 508.24, spectrum of digital silence, with noise and spuriae (bottom) and spectrum of DC signal (top). (16-bit data, 1/3-octave analysis, right channel dashed.)
The top two traces in fig.2 show the player's noise floor in the presence of a signal—in this case, data representing a DC signal from a JVC test CD. Note now the rise of around 12dB in the noise floor in the audio band, and the huge increase in the noise above 20kHz. This is due to the high-order noise-shaping used in the Crystal delta-sigma DAC to achieve as high a resolution as possible below that frequency. While this is not 24-bit performance given the 16-bit CD code used to feed the DAC, the shortfall is academic. And in any case, the "24-bit" nomenclature Crystal uses for its DAC more truthfully reflects the fact that the DAC will accept 24-bit data without truncation, which has positive consequences for sound quality.
Plotting the 508.24's measured output level against theoretical level using a dithered 500Hz tone gave the linearity plot shown in fig.4. The left channel is shown, and any error is inconsequential in subjective terms. The linearity is smooth, and pretty accurate down to below -115dBFS or so, where the player's very low noise floor finally starts to intrude.
Fig.4 Meridian 508.24, left channel, departure from linearity (2dB/vertical div.).
Confirming the implication that this is a very-high-resolution machine, the waveform of an undithered 1kHz tone at -90.31dBFS was reproduced with an almost perfect shape (fig.5): there was very low noise, almost equally spaced level steps, and even more than a hint of the Gibbs Phenomenon "ringing" due to the CD standard's 22kHz band-limiting.
Fig.5 Meridian 508.24, waveform of undithered 1kHz sinewave at -90.31dBFS (16-bit data).
Driving the CD player with data representing a high-frequency "torture" signal—19kHz and 20kHz, each at -6dBFS, which gives a waveform peaking at 0dBFS—gave the spectrum shown in fig.6. All the intermodulation products are extremely low in level; even the highest, the 1kHz difference component, is at -96dB (0.0015%).
Fig.6 Meridian 508.24, HF intermodulation spectrum, DC-22kHz, 19+20kHz at 0dBFS (linear frequency scale, 20dB/vertical div.).
I used the Miller Audio Research analyzer to look at the Meridian's jitter performance. This drives the CD player under test with data representing an 11.025kHz tone at -10dBFS, a 229Hz tone toggling the LSBs on and off while it performs a high-resolution FFT analysis—32,768 points, with 64 FFTs averaged—to examine the noise floor in the analog domain. Stereophile used to examine jitter at the DAC's wordclock pin, using the Meitner LIM analyzer; while this produced consistent results, it is the effect of jitter in the analog domain that really matters. We have therefore switched to using the Miller analyzer.
The absolute timebase error in the Meridian's performance was just +9 parts per million! The black trace in fig.7 shows an analysis of the 508.24's noise floor for a ±3.5kHz span around the central tone. For comparison, the grayed-out spectrum is the noise floor of the Audio Research CD2, which WP reviewed in the April issue. Both players are excellent, but the Meridian's noise floor is 2dB lower overall, digging close to the theoretical -96.3dBFS limit of a 16-bit linear-PCM system. (While the individual elements of the noise-floor "grass" are close to -130dBFS, it is their root-mean-square sum that gives the noise figure typically expressed as a Signal/Noise Ratio.)
Fig.7 Meridian 508.24, high-resolution jitter spectrum of analog output signal (11.025kHz at -10dBFS with LSB toggled at 229Hz). Center frequency of trace, 11.025kHz; frequency range, ±3.5kHz. Grayed-out spectrum is that of the Audio Research CD2.
The Miller Analyzer searches among the noise components for symmetrical pairs of sidebands; these will be due to word-clock jitter. The Audio Research has very low jitter, at 168.6 picoseconds peak-peak. But the Meridian is even lower, at 144.2ps peak-peak. You can see that its data-induced ±229Hz sidebands, marked with red "2s," are around 10dB lower than the equivalent sidebands with the CD2, and contributed just 47.4ps to the total. Only with the higher-order data-induced sidebands, again marked with red numbers, does the Meridian's jitter rise to equal the Audio Research's performance.
As with the Audio Research, note the slight spreading of the central peak; this may be due to low-frequency, random-noise jitter. The Mark Levinson No.39, for example, whose jitter measured with this test is of the same order as the Meridian's, has a clean, narrow peak rising from the noise floor. And there are also some low-frequency sidebands in the 508.24's spectrum, marked with blue "1s" and "3s," that are jitter-related. The "1" sidebands, totaling 52.2ps, are spaced at 120Hz, meaning that they are related to the power supply. But the Analyzer's software engine labels the "3" sidebands "of unknown origin."
It is fashionable in some audiophile circles to deride measurements as not having any connection with sound quality. But when you get a player such as the Meridian 508.24, which measures so close to the theoretical ideal in so many ways, it's hard to resist the temptation to point to its superb sound quality and say, "There!" At minimum, the 508.24's measured performance reveals that Bob Stuart and his talented digital design team at Meridian have done a fabulous job.—John Atkinson