dCS Puccini SACD playback system Measurements
I examined the measured behavior of the dCS Puccini using the Audio Precision SYS2722 system (see www.ap.com and "As We See It" in the January 2008 issue), as well as, for some tests, my Audio Precision System One Dual Domain. To examine the performance of the U-Clock's asynchronous USB input, I drove it with the USB 2.0 output of my MacBook running OS10.4.11, playing WAV files using Bias Peak 6.2.
The Puccini had the best error correction I have encountered, not suffering from occasional glitches in its audio output until the gaps in the data spiral of the Pierre Verany Test CD reached 3mm in length. This is even better than the top-line dCS Scarlatti player, which Michael Fremer reviewed last August. With 4m gaps in the data, the player muted its output once per revolution, but was still able to track the data spiral. The Puccini's S/PDIF input locked to audio data with sample rates of 32, 44.1, 48, 88.2, and 96kHz. The maximum output level was to specification, at 1.97V or 6.05V RMS from both the balanced and unbalanced jacks, and both outputs preserved absolute polarityie, were non-invertingwith the XLRs wired with pin 2 hot. The source impedance from the balanced jacks was an extremely low 3 ohms at all audio frequencies, rising to a still very low 52 ohms from the unbalanced jacks, as specified.
To test the Puccini's performance as an SACD player, I used the Sony test disc. The red trace in fig.1 shows the response with SACD playback with Filter 1, the factory default setting. The output starts to roll off relatively gently above 40kHz, reaching 3dB at 70kHz and 10dB at 93kHz. To my surprise, given that it is specified as rolling off a little earlier, Filter 2 had an identical response, though Filter 3 (magenta trace) and Filter 4 did progressively curtail the Puccini's ultrasonic output. These filters are also operational when the Puccini upsamples CD data to DSD. Looking at the player's output while it played a CD with a series of single-sample, full-scale positive pulses revealed that these filters are conventional finite-impulse response types, there being an equal amount of ringing before and after the pulse (fig.2; this waveform was captured using a digital oscilloscope that lacked an input antialiasing filter, so that the 'scope's own impulse response would not obscure that of the Puccini).
Fig.1 dCS Puccini, SACD frequency response at 3dBFS into 100k ohms with Filters 1 & 2 (red), Filter 3 (magenta), Filter 4 (gray). (1dB/vertical div.)
Fig.2 dCS Puccini, CD upsampled to DSD with Filter 1, response to single sample at 0dBFS (4ms time window).
Fig.3 shows the Puccini's frequency response with PCM data fed to its S/PDIF input. Other than with 32kHz-sampled data (green trace), the output is flat in the audioband, with then a steep rolloff above 19kHz for CD data (magenta) and above 40kHz (96kHz data). For reference, the rightmost red trace shows the response for SACD playback using Filter 1. Channel separation (not shown) was superb, at better than 100dB in both directions in the audioband.
Fig.3 dCS Puccini, SACD frequency response at 3dBFS into 100k ohms (red trace above 40kHz), external data frequency response at 12dBFS into 100k ohms at 96kHz sample rate (red), 44.1kHz (magenta), 32kHz (green). (1dB/vertical div.)
For consistency with my tests of digital components, I examine resolution by sweeping a 1/3-octave-wide bandpass filter from 20kHz to 20Hz while the device under test decodes dithered data representing a 1kHz tone at 90dBFS. The top pair of traces below 6kHz in fig.4 show what happens with the Puccini decoding 16-bit data from CD: the trace peaks at exactly 90dBFS, suggesting minimal linearity error, while the noise floor is free from harmonic- or power-supplyrelated spuriae. These plots actually show the spectrum of the dither used to encode the data. Switching to the S/PDIF input and increasing the word length to 24 bits gave the middle pair of traces in fig.4: the noise floor has dropped by 16dB or so, suggesting close to 19-bit resolution.
Fig.4 dCS Puccini, 1/3-octave spectrum with noise and spuriae of dithered 1kHz tone at 90dBFS with 16-bit data (top), 24-bit data (middle at 2kHz), DSD data (top trace at 20kHz), dithered 1kHz tone at 120dBFS with 24-bit data (bottom at 1kHz). (Right channel dashed.)
The top traces above 6kHz were taken with the Puccini playing an SACD. The high level of RF noise inherent in DSD encoding leaks past the bandpass filter's "skirts," making it look as if the noise floor is rising in the top two octaves, but this is a false impression. There is a touch more LF noise with DSD data than with 24-bit PCM data, however. The bottom pair of traces in fig.4 shows the spectrum with the Puccini fed dithered 24-bit data representing a 1kHz tone at 120dBFS. The tone is readily resolved, with again no harmonic or supply-related spuriae visible, other than the suggestion of some very-low-level spectral content at 180Hz.
Fig.5 repeats the spectral analysis of the tones in fig.4, but now using an FFT technique and a linear rather than a logarithmic frequency scale. Again, increasing the bit depth from 16 to 24 drops the noise floor by around 16dB.There is no rise in the noise above 6kHz with DSD data (gray and green traces), but a spectral line is visible at 625Hz with DSD data (green trace) that is not present with PCM data. This tone was also present with the dCS Scarlatti: When, after its publication, I discussed the Scarlatti review with dCS engineer Andy McHarg, he told me that this tone will be eliminated in a forthcoming firmware update for first the Scarlatti, then the Puccini. As it lies at 120dBFS, this idle tone will be subjectively innocuous.
Fig.5 dCS Puccini, FFT-derived spectrum with noise and spuriae of dithered 1kHz tone at 90dBFS with 16-bit external data (left channel cyan, right magenta), 24-bit data (left channel blue, right red), DSD data (left channel green, right channel gray).
Looking at the Puccini's linearity error with my usual 16-bit fade-to-noise test, all the graph showed was the effect of the recorded dither noise, so I haven't published it. With 24-bit data, the linearity error was negligible down to below 120dBFS (also not shown). With its very low levels of analog noise and linearity error, it was not surprising that the Puccini's reproduction of an undithered sinewave at exactly 90.31dBFS, which is described by just three DC voltage levels, was essentially perfect (fig.6). Increasing the word length to 24 bits or changing to SACD playback of a tone at the same level gave the same result: a superbly defined sinewave, despite the very low signal level (fig.7).
Fig.6 dCS Puccini, waveform of undithered 1kHz sinewave at 90.31dBFS, 16-bit data (left channel blue, right red).
Fig.7 dCS Puccini, waveform of undithered 1kHz sinewave at 90.31dBFS, 24-bit data (left channel blue, right red).
The Puccini's output stage appears to be bombproof. Whether it was driving the benign 100k ohm load or the punishing 600 ohm load, the harmonic distortion in its output was very low and composed of the second and third harmonics (fig.8), which are subjectively innocuous even at levels 1000 times higher than they are in the Puccini. Similarly, intermodulation distortion was also extremely low, even into 600 ohms (fig.9).
Fig.8 dCS Puccini, spectrum of 50Hz sinewave at 0dBFS into 600 ohms, 24-bit data (left channel blue, right red; linear frequency scale).
Fig.9 dCS Puccini, HF intermodulation spectrum, 19+20kHz at 0dBFS peak into 600 ohms, 24-bit data (left channel blue, right red; linear frequency scale).
Used as a CD player, the Puccini was excellent at rejecting word-clock jitter. With the U-Clock connected to its word-clock input and playing a CD carrying the Miller/Dunn J-Test signal, there were no sidebands identifiable in the player's analog output, and the jitter level was below the resolution of the Miller Jitter Analyzer. Using the player's internal clock did give a measured figure of 105 picoseconds peakpeak of jitter, though this is at the very limit of resolution of the Miller Analyzer. With the Puccini fed external PCM data, there was some dependence on the source. In the worst case, using my lab PC with its RME soundcard to output, via 15' of TosLink cable, data that I then converted to coaxial S/PDIF with an antique Sonic Solutions format converter, I managed to measure 450ps pp of jitter, with most of the energy in pairs of sidebands at ±60Hz and ±229Hz (fig.10). This is still low in absolute terms, however; and using higher-quality S/PDIF sources again dropped the jitter below the analyzer's own resolution limit.
Fig.10 dCS Puccini, high-resolution jitter spectrum of analog output signal, 11.025kHz at 6dBFS, sampled at 44.1kHz with LSB toggled at 229Hz, 16-bit external data via coaxial S/PDIF. Center frequency of trace, 11.025kHz; frequency range, ±3.5kHz (left channel blue, right red).
With USB data from my MacBook, the effectiveness of dCS's asynchronous operation can be seen by comparing fig.11 with fig.10. The measured jitter level was just 79ps pp, due to a single sideband pair at ±60Hz (with, again, the caveat that this figure is at or below the analyzer's limit of resolution). The dCS Puccini U-Clock joins Ayre's QB-9 and the Wavelength Cosecant in demonstrating the advantages of using the USB interface in asynchronous rather than adaptive isochronous mode.
Fig.11 dCS Puccini, high-resolution jitter spectrum of analog output signal, 11.025kHz at 6dBFS, sampled at 44.1kHz with LSB toggled at 229Hz, 16-bit USB data. Center frequency of trace, 11.025kHz; frequency range, ±3.5kHz (left channel blue, right red).
In all aspects of performance, the dCS Puccini offers state-of-the-art measured performance.John Atkinson