Wadia 861 CD player Measurements
The Wadia's output impedance was a low 58 ohms from the unbalanced RCA jacks, this doubling as expected from the balanced XLRs. (Because a balanced connection has a separate driver for each phase of the signal, the drivers impedances appear in series and must therefore be added to get the overall output impedance.) The player preserved absolute polarity from both sets of outputs, with the XLR wired with pin 2 "hot." Its error correction was superb, coping with gaps in the CD data spiral up to 1.5mm in length without audible glitches.
The maximum output level at 1kHz was 2.13V with Filter A from both balanced and unbalanced outputs, this dropping by 1dB to 1.9V when the filter was switched to type B. Filter C raised the level by 0.2dB compared with B, to 1.94V. These differences may be small, but because they cover the entire audioband, they will be audible. Crosstalk (not shown) was unmeasurable at low and midrange frequencies, but rose a little at the top of the audioband, to 113dB (L-R) and 101dB (R-L), both figures still superb.
The frequency response with Filter A (fig.1, top pair of traces) shows the classic top-octave droop (-3dB at half the sample rate) typical of Wadia's DigiMaster algorithm, which trades off top-octave flatness and stop-band rejection against much better time-domain behavior. With a pre-emphasized signal (fig.1, lower traces) the response has a slight presence-region rise evident but is otherwise identical.
Fig.1 Wadia 861, Filter A, CD frequency response at -12dBFS, without emphasis (top) and with emphasis (bottom). (Right channel dashed, 0.5dB/vertical div.)
Because the review sample had a digital input board, I was able to examine its frequency response with higher sample rates than the CD's 44.1kHz. The top trace in fig.2 shows the response at a 96kHz rate (of necessity, it includes the response of the A/D converter used). It was the same regardless of the filter selected, and shows the top-of-the-passband droop characteristic of the DigiMaster algorithm. However, because of the higher sample rate, the audioband rolloff is negligible, allowing the user to have his cake and eat it too: flat audioband response and excellent time-domain performance. This graph also shows the 44.1kHz response with Filters B and C, which are both flat in the top audio octave, though differing slightly in level, as mentioned before.
Fig.2 Wadia 861, external input frequency response at -12dBFS with (from top to bottom): 96kHz sampling, Filter A; 44.1kHz sampling, Filters A, C, and B (right channel dashed, 1dB/vertical div.).
To examine a D/A converter's absolute resolution, I compare the spectrum of its output when decoding a dithered low-level 1kHz signal with the CD's 16-bit resolution with a similar spectrum taken using dithered 24-bit data. While the 16-bit noise floor will be dominated by the presence of the dither noise, this noise will be reduced in level by 48dB with the increase in word length, unmasking the DAC's own noise level.
The result is shown in fig.3. At high frequencies, the noise floor has dropped by 10dB, implying performance close to 18 bits, which is good. (Even the best processors I have measured don't get more than another couple of bits' worth of resolution.) But almost no difference can be seen at low frequencies, the Wadia's own noise dominating the measurement and implying resolution closer to 15 bits.
Fig.3 Wadia 861, 1/3-octave spectrum of dithered 1kHz tone at -90dBFS, with noise and spuriae (from top to bottom): 16-bit CD data, 24-bit external data (right channel dashed).
I found a similar story when I did similar spectral analyses using "digital black" (not shown). Other than my surprise at the slightly-higher-than-expected lower-frequency noise floor, the only real consequence is that the Wadia's digital volume control should not be used to apply more than about 12dB of attenuation in critical listening situations. However, with its relatively low maximum output level (MOL), this will not be an issue, and suggests that Wadia's design team has paid attention to balancing maximum output vs resolution vs the volume-control implementation for real-world situations. (Yes, you could drop the noise floor by 12dB and improve the ultimate resolution if you increased the MOL to 8V, but given real-world power-amplifier sensitivities, the volume-control attenuation and attendant loss of resolution would have to be correspondingly greater.) Internal switches can be used to increase the MOL, by the way, which will increase the unit's resolution accordingly, as the lowest-level information emerges from the analog noise floor.
Linearity error assessed with a dithered 16-bit, 500Hz tone was very low (fig.4), but with slightly more noise apparent than with the very quietest players. Nevertheless, the 861's reproduction of an undithered 1kHz sinewave at -90.31dBFS (fig.5) was essentially perfect, the three discrete voltage levels easily apparent. Increasing the word length to 24 bits gave a reasonable facsimile of a sinewave (fig.6).
Fig.4 Wadia 861, right-channel departure from linearity, 16-bit CD data (2dB/vertical div.).
Fig.5 Wadia 861, waveform of undithered 1kHz sinewave at -90.31dBFS, 16-bit CD data.
Fig.6 Wadia 861, waveform of undithered 1kHz sinewave at -90.31dBFS, 24-bit external data.