Oracle CD player Measurements
This stunningly beautiful player had a maximum output of 2.7V, which is both higher than the 2.4V specified and higher than the CD standard's 2V RMS. The output impedance was a low 57 ohms across the audioband, and the player inverts absolute polarity.
The Oracle's frequency response, with (top) and without (bottom) pre-emphasized data, is shown in fig.1. With a regular signal, the response is perfectly flat. With a pre-emphasized signal, however, a large (-1dB) error appears throughout the treble, which will be audible (on those few discs that have been pre-emphasized) as a more distant, somewhat recessed tonal balance.
Fig.1 Oracle, frequency response with (top) and without (bottom) de-emphasis at -12dBFS (right channel dashed, 0.5dB/vertical div.).
It was hard to assess channel separation due to the presence of ultrasonic noise—this noise made the Oracle's output waveform look distinctly fuzzy on an analog oscilloscope—but separation appeared to be better than 80dB in both directions. The noise floor in the audioband was also higher than I had anticipated. This can be seen in fig.2: The top traces are a 1/3-octave spectral analysis of the Oracle's analog output while it decoded data representing a dithered 1kHz tone at -90dBFS. The spectral peak at 1kHz just reaches the -90dBFS line, suggesting low linearity error. But the noise floor is about 5dB higher overall than that of the Arcam FMJ CD23 I used as a reference (lower traces). Note also the presence of peaks at 240Hz, 120Hz, and 60Hz, which are due to residual power-supply noise.
Fig.2 Oracle, 1/3-octave spectrum of dithered 1kHz tone at -90dBFS, with noise and spuriae, 16-bit data (top); bottom traces are the Arcam FMJ CD23 under identical conditions (right channel dashed).
Repeating this spectral analysis but with "digital black" data and extending the measurement bandwidth to 200kHz (fig.3) reveals the ultrasonic noise to be due to the noiseshaping used by the delta-sigma DAC to achieve sufficient baseband resolution. Even with a bandpass filter tracking the 500Hz stimulus, there was still enough noise present to foil any attempt to assess linearity error (fig.4). The rise in the trace below -80dBFS in this graph is due to the decreasing signal/noise ratio, not to linearity error. As fig.2 suggests, the Oracle's linearity is probably good. The ultrasonic noise content also completely obscures the waveform of an undithered 1kHz tone at -90.31dBFS (fig.5). This should consist of just three voltage levels, and is reproduced as such by D/A chips with low noise and good linearity (fig.6).
Fig.3 Oracle, 1/3-octave spectrum of "digital black," with noise and spuriae, 16-bit data (right channel dashed).
Fig.4 Oracle, left-channel departure from linearity, 16-bit data (2dB/vertical div.).
Fig.5 Oracle, waveform of undithered 1kHz sinewave at -90.31dBFS, 16-bit data.
Fig.6 CardDeluxe, waveform of undithered 1kHz sinewave at -90.31dBFS, 16-bit data.
Harmonic distortion was low in level, even into the punishing 600 ohm load, at better than -70dB (0.03%), and consisted entirely of the sonically benign second and third harmonics (fig.7). High-frequency intermodulation distortion (fig.8) was a little higher than I have measured from the best CD replay systems, but, with a 1kHz difference component around -76dB (0.015%), is nothing to be worried about.
Fig.7 Oracle, spectrum of 50Hz sinewave, DC-1kHz, at 0dBFS into 100k ohms (linear frequency scale).
Fig.8 Oracle, HF intermodulation spectrum, DC-22kHz, 19+20kHz at 0dBFS into 600 ohms (linear frequency scale).
To assess a player's jitter, I play a test CD-R with a high-level 11.025kHz tone with the LSB toggling on and off at 229Hz. I then perform a narrowband FFT analysis with the Miller Audio Research Jitter Analyzer. Fig.9 shows the spectrum of the Oracle's output as it decodes these data. The absolute level of jitter was low at 268 picoseconds peak-peak, with most of the jitter energy present in three pairs of sidebands. One pair, marked in the graph with red "2s," lies at ±229Hz and is therefore clearly signal-related. The other pairs, indicated with purple "4s" and "10s," lie at ±1.16kHz and ±3.14kHz. I have no idea what these are due to, or what might be their effects on the player's perceived sound quality. But I wish they weren't present.
Fig.9 Oracle, high-resolution jitter spectrum of analog output signal (11.025kHz at -6dBFS with LSB toggled at 229Hz, CD data). Center frequency of trace, 11.025kHz; frequency range, ±3.5kHz. (Grayed-out trace is that from Linn CD12.)
The grayed-out trace in fig.9 is a spectral analysis of the Linn CD12's output taken under identical conditions. The sample rate of the Jitter Analyzer is 32kHz, meaning that noise present in the player's output with a frequency higher than 16kHz will be rejected. With the Oracle's ultrasonic noise eliminated, you can see that its noise floor is actually a little lower than the Linn's. The Scottish player, however, has almost half the measured jitter of the Canadian, and has a narrow central peak. The spreading of the Oracle's 11.025kHz component will be due to the presence of random low-frequency jitter, and might be expected to broaden the soundstage at the expense of image precision.
Finally, the Oracle offered superb error correction, coping with gaps of 2mm in the data spiral on the Pierre Verany Test CD without skipping or muting.—John Atkinson