Ayre CX-7 CD player Measurements

Sidebar 3: Measurements

The Ayre CX-7's maximum output level was very slightly higher than specified, at 4.514V balanced and 2.255V unbalanced. The latter is 1.03dB higher than the CD standard's 2V. Both sets of outputs preserved absolute polarity (ie, were noninverting), the XLRs being wired with pin 2 hot. The source impedance from the unbalanced RCA jacks was a low 58 ohms across the audioband, this doubling as expected from the balanced XLRs, which is still usefully low. Error correction, assessed using the Pierre Verany Test CD, was good, the player occasionally dropping out when the gaps in the data spiral reached 1mm in length.

With the rear-panel filter switch set to Measure, the response was flat up to the top octave, which, as can be seen from the top pair of traces in fig.1, then featured a rolloff of a fraction of a dB by 20kHz. De-emphasis error was negligible, as shown by the bottom traces in this graph. With the filter set to Listen, the response above 14kHz or so rolled off more rapidly, reaching -2.5dB at 20kHz. Not apparent on this audioband graph will be the fact that this filter setting doesn't roll off ultrasonic frequencies as aggressively as the Measure setting, which has ramifications that will become apparent later. All of these measurements were made from the balanced outputs; the response from the unbalanced jacks was identical.

Fig.1 Ayre CX-7, balanced outputs, frequency response at -12dBFS into 100k ohms with filter set to Measure (top at 15kHz), with filter set to Listen (middle at 15kHz), response with filter set to Measure and with de-emphasis (bottom). (Right channel dashed, 0.5dB/vertical div.)

Channel separation from the balanced outputs (not shown) was excellent, at better than 105dB across the band in the R-L direction but only 90dB in the other direction. The unbalanced performance was about 5dB worse, but still respectable. Fig.2 shows a 1/3-octave spectrum of the Ayre's output while it decoded data representing a dithered 1kHz tone at -90.31dBFS. No harmonic or power-supply-related spuriae can be seen above the noise floor, this actually due to the dither used to encode the 16-bit data. This is excellent performance, suggesting that the Ayre's true noise floor is very low.

Fig.2 Ayre CX-7, balanced, 1/3-octave spectrum of dithered 1kHz tone at -90dBFS, with noise and spuriae (16-bit data, right channel dashed).

Linearity error was also low, remaining below 2dB to below -110dBFS (fig.3), which, in conjunction with the player's low analog noise, results in accurate reproduction of the waveform of an undithered 1kHz tone at exactly -90.31dBFS (fig.4). Each of the three voltage levels that describe this signal can be clearly seen, with very little noise riding on the waveform.

Fig.3 Ayre CX-7, left-channel departure from linearity, 16-bit data (2dB/vertical div.).

Fig.4 Ayre CX-7, waveform of undithered 1kHz sinewave at -90.31dBFS, 16-bit data.

When it came to harmonic distortion, the CX-7 was good rather than great, something that can be laid at the feet of its lack of loop feedback. Fig.5 shows the spectrum of the player's unbalanced output while it reproduced a full-scale 1kHz tone into a fairly low load of 8k ohms. A regular series of harmonics can be seen extending to the 10th, which lies at the right-hand edge of the graph. Only the second, third, and fifth harmonics lie above -80dB (0.01%), the first two lying at -70dB (0.03%), which should still be low enough to be inaudible. The right channel was, overall, better than the left, at 0.042% THD (actual sum of the harmonics) compared with 0.057%, though it has more fourth harmonic evident (red trace). As expected, the balanced outputs significantly reduced the level of just the even harmonics, leaving the third and fifth highest in level (not shown).

Fig.5 Ayre CX-7, unbalanced, spectrum of 1kHz sinewave, DC-10kHz, at 0dBFS into 8k ohms (linear frequency scale).

With the filter set to Measure, the Ayre CX-7's intermodulation behavior was good rather than excellent (fig.6). The difference tone at 1kHz was the highest intermodulation product visible, at 0.02% in the left channel and less than half that in the right (0.006%), but several higher-order products can be seen. Switching the filter to Listen changes the picture quite dramatically (fig.7). The 1kHz product actually drops in level—to 0.012% left, 0.0008% right—but not only can a strong tone be seen at 24.1kHz (44.1kHz minus 20kHz), but many other aliasing tones and tone pairs are now apparent throughout the audioband, the latter spaced by 100Hz.

Fig.6 Ayre CX-7, unbalanced with filter set to Measure, HF intermodulation spectrum, DC-25kHz, 19+20kHz at 0dBFS into 8k ohms (linear frequency scale).

Fig.7 Ayre CX-7, unbalanced, with filter set to Listen, HF intermodulation spectrum, DC-25kHz, 19+20kHz at 0dBFS into 8k ohms (linear frequency scale).

I rush to make it clear that I heard nothing in my auditioning that could be laid at the feet of this filter behavior. Still, I don't like to see it. Both of these graphs were taken from the unbalanced output jacks; the balanced behavior offered a 1kHz intermodulation product that was about 10dB lower in level, but the same level of aliasing products with the filter set to Listen. Peculiarly, the intermodulation performance was best when the CX-7 was driving the demanding 600 ohm load.

I assessed the Ayre CX-7's rejection of word-clock jitter using the Miller Analyzer, which examines the spectrum of a player's analog output for symmetrical sidebands around a high-level tone with a frequency at exactly one quarter the sample rate. The driving signal also contains a low-frequency squarewave at the LSB level, which is a worst-case situation for exciting jitter. The CX-7 did very well on this test, producing just 159 picoseconds of peak-peak jitter—among the lowest results I have measured. Fig.8 shows the jitter spectrum, which primarily consists of data-related sidebands (red numeric markers), these all very low in level. Some low-frequency sidebands can be seen at ±15.6Hz and its harmonics (purple markers), but these are also very low in level.

Fig.8 Ayre CX-7, unbalanced, high-resolution jitter spectrum of analog output signal (11.025kHz at -6dBFS sampled at 44.1kHz with LSB toggled at 229Hz). Center frequency of trace, 11.025kHz; frequency range, ±3.5kHz.

To sum up, the slow-rolloff Listen setting of the CX-7's reconstruction filter trades off better time-domain performance against the introduction of some aliasing-related spuriae. However, it is fair to note that anyone bothered by this measured behavior can simply switch the filter to its Measure setting. Its relative lack of transfer-function linearity aside—and again, this was not something that I was aware of in my auditioning, unless it contributes to the vibrant character—Ayre's CX-7 is a model of modern CD-player design.—John Atkinson