Cary Audio Design CD303/300 CD player Measurements
With its integral volume control and its choices of balanced and unbalanced outputs, tubed and solid-state output circuitry, and 1–17.4x upsampling ratios, Cary's CD303/300 was considerably more complicated to measure than a conventional CD player. To simplify matters, therefore, I set the volume control to its maximum (indicated "73") for all measurements, and mainly checked the measured performance with either no upsampling or the maximum upsampling (768kHz), with additional spot checks.
Assessed with the Pierre Verany Test CD and with me monitoring the error flags in the digital output with RME's DIGICheck software, the Cary had the best error correction I have found. It played tracks with gaps in the data spiral up to 2mm in length without either glitches or flagged errors, and, even with gaps of 2.4mm in length, muted its output only occasionally.
The CD303/300 preserved absolute polarity from all four of its outputs (the XLR jacks are wired with pin 2 hot). The maximum output voltage at 1kHz was different for the solid-state and tube output stages, at respective values of 6.236V/3.113V and 5.072V/2.589V, balanced/unbalanced. The difference between tube and solid-state is a very audible 1.6dB in favor of the latter, which will invalidate direct comparisons unless compensated for. The output impedances were also very different. The unbalanced solid-state output was sourced from a low 99 ohms, this doubling as expected to 198 ohms from the balanced jacks. However, the tube output stage appeared to have very high source impedances: 4k ohms at 20Hz, 3.5k ohms at 1kHz, and 2.6k ohms at 20kHz, these unbalanced and all much higher than the specification. The partnering preamp (or power amplifier, if the Cary's own volume control is used) needs to have an input impedance of at least 47k ohms if the response is not to be tilted up.
Like Cary's CD303/200, which Brian Damkroger reviewed in May 2004, the '300 didn't apply the appropriate de-emphasis with pre-emphasized discs (fig.1, top pair of traces above 2kHz). Fortunately, such CDs are very rare these days. The response of the solid-state output was ruler-flat (fig.1, middle traces), but rolled off very slightly—down 0.5dB at 20kHz—from the tube output stage, with a slight, 0.05dB channel imbalance evident (fig.1, bottom traces above 10kHz). Channel separation was excellent from the solid-state output (fig.2, bottom traces), but about 18dB less good from the tube stage (top traces). Both sets of traces in this graph reveal decreasing separation with increasing frequency due to capacitive coupling between the channels. However, with crosstalk below 90dB below 15kHz, even the tube stage is good enough not to compromise soundstaging.
Fig.1 Cary Audio Design CD303/300, unbalanced solid-state output frequency response at –12dBFS into 100k ohms, with de-emphasis (bottom below 2kHz) and without; unbalanced tube output response (bottom above 10kHz). (Right channel dashed, 1dB/vertical div.)
Fig.2 Cary Audio Design CD303/300, balanced channel separation with tube output stage (top) and solid-state output stage (bottom). (Right channel dashed, 10dB/vertical div.)
Both the tube and solid-state outputs have very low noise floors, and the spectrum of a dithered tone at –90dBFS from the tubed output is dominated by the recorded dither noise (fig.3). This was identical to the spectrum from the solid-state output and no power-supply or harmonic distortion products are visible. Similarly, the graph of linearity error (fig.4) really shows only the dither noise recorded on the test CD. No level errors are visible.
Fig.3 Shahinian Diapason & Double Eagle, system electrical impedance (solid) and phase (dashed) (2 ohms/vertical div.).
Fig.4 Cary Audio Design CD303/300, balanced left-channel departure from linearity, 16-bit CD data (2dB/vertical div.).
Switching in various upsampling ratios had no effect on any of these measurements, nor did it change the Cary's reproduction of an undithered sinewave at exactly –90.31dBFS (fig.5). This graph clearly shows the three voltage levels used to describe this signal, and no waveform asymmetry is evident. This is excellent digital-domain performance.
Fig.5 Cary Audio Design CD303/300, balanced waveform of undithered 1kHz sinewave at –90.31dBFS, 16-bit CD data upsampled to 768kHz.
In the analog domain, the two output stages differed considerably when it came to distortion. With the Cary driving a full-scale 1kHz sinewave into 100k ohms, the spectrum of the balanced tube output (not shown) featured a third-harmonic component at –66dB (0.05%) and a second-harmonic at –80dB (0.01%). This is not a high distortion level, but switching to the unbalanced solid-state output, driving a low 4k ohms (fig.6), dropped the third harmonic to –112dB (!), leaving the second harmonic the highest in level at just –96dB (0.0015%). However, as the signal level dropped, the tube stage began to approach the excellent linearity of the solid-state stage, and by the time the signal lay at –90dBFS, the two offered identically low distortion (fig.7). The only significant difference between the two stages at this level was the very slightly higher level of power-supply spuriae from the tubed stage; but at –110dBFS, these will have no subjective consequences.
Fig.6 Cary Audio Design CD303/300, unbalanced solid-state output stage, spectrum of 1kHz sinewave at 0dBFS into 4k ohms, DC–10kHz (linear frequency scale).
Fig.7 Cary Audio Design CD303/300, unbalanced tube output stage, spectrum of 1kHz sinewave at –90dBFS into 4k ohms, DC–10kHz (linear frequency scale).
Only when it came to high-frequency intermodulation distortion did the tube stage stumble a little. Driving a full-scale mix of 19kHz and 20kHz tones into 100k ohms (fig.8), the solid-state output produced a 1kHz difference product at just –96dB (0.0015%), with higher-order products around the same, negligible level. By contrast, the tube stage (fig.9) produced a 1kHz product at –60dB (0.1%) and higher-order products at –50dB (0.3%). I don't believe this behavior will be audible with music, but it does suggest that the tubed stage has to work hard with this signal, even into the benign 100k ohm load.
Fig.8 Cary Audio Design CD303/300, balanced solid-state output stage, HF intermodulation spectrum, 19+20kHz at 0dBFS peak into 100k ohms, DC–24kHz (linear frequency scale).
Fig.9 Cary Audio Design CD303/300, balanced tube output stage, HF intermodulation spectrum, 19+20kHz at 0dBFS peak into 100k ohms, DC–24kHz (linear frequency scale).
Finally, I used the Miller Audio Research Jitter Analyzer to search for the effects of digital word-clock jitter in the CD303/300's analog output. There were no significant differences between the tube and solid-state output stages when it came to jitter, nor were there when the upsampling was switched in. The Cary produced just 176.5 picoseconds (peak–peak) of jitter sidebands (fig.10), with the data-related components (red numeric markers) very low in level. The highest-level jitter sidebands lay at ±15.6Hz (purple "1" markers), ±818Hz (purple "6"), and ±3100Hz (purple "15"). I have no idea what these are due to, but perhaps of more subjective import is the shaping of the noise floor in this graph on either side of the central 11.025kHz tone. Though modest in level compared with the similar noise-floor modulation seen in the Linn Unidisk SC processing external S/PDIF data (see fig.11 in that review), it may result from the massive amount of digital signal processing going on inside the player.
Fig.10 Cary Audio Design CD303/300, unbalanced solid-state output stage, 768kHz upsampling, 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). Center frequency of trace, 11.025kHz; frequency range, ±3.5kHz.
Overall, the Cary CD303/300 offers excellent measured performance. While its tube output stage is not as good as its solid-state output when it comes to channel separation, distortion, and drive capability, these factors should not compromise its sound quality.—John Atkinson