47 Laboratory 4715 D/A processor & 4716 CD transport Measurements
Shipping damage prevented me from measuring the sample of the 4715 D/A converter that had been auditioned by Art Dudley (serial no. 80072). Yoshi Segoshi of 47 Laboratory's US distributor therefore arranged for me to borrow, at very short notice, a sample (serial no. 80030) from Manhattan audiophile Steve Yagerman. My thanks.
I used a standard 75 ohm S/PDIF cable to connect the 4715 to the 4716 CD transport, using what I understood to the latter's appropriate RCA jack. The transport offered excellent error correction, coping with gaps in the CD's data spiral up to 2mm in length without audible glitches. The 4315's maximum output at 1kHz was 0.1dB below the CD standard's 2V RMS. This was for the left channel; the right channel was another 0.1dB below this. The DAC's output impedance was quite high, at 2.3k ohms across the audioband. This would suggest that a preamp with a high input impedance would be optimal. However, as will be seen later, there are other things at work here.
Looking inside the 47 Laboratory components, I was struck by the extent to which the designer has been able to eliminate circuitry that others would think important. The separate power supplies, for example, each consist of a transformer and a single diode, while a look inside the 4315 left me gasping at both the minimal circuitry and at the time-consuming care with which it had been assembled—some wires were hand-soldered to the individual pins of the surface-mount chips!
The 4715 inverts absolute signal polarity, and without either digital or analog reconstruction filters, the raw DAC output is fed directly to the output jacks. This can be seen in fig.1, the waveform of a full-scale 1kHz sinewave. What would, with conventional CD playback, be a clean sinewave constructed by the low-pass filters (see my August 1986 article on the subject) is instead broken by 44 stairsteps. It does look like a sinewave, but at high frequencies, the data-point sparseness results in something more like a squarewave (fig.2). It's important to note that the harmonic content that differentiates between 44.1kHz-encoded sine- and squarewaves lies above 20kHz for signal frequencies above 6.7kHz, so the difference should not be audible. Again, however, it is fair to note that there are other factors to take into consideration.
Fig.1 47 Laboratory 4715, waveform of 1kHz sinewave at 0dBFS.
Fig.2 47 Laboratory 4715, waveform of 20kHz sinewave at 0dBFS.
The 4716's data output does pass on the pre-emphasized data flag, but the 4715's minimal analog circuitry means that it cannot provide the necessary de-emphasis. As a result, the small number of pre-emphasized CDs will suffer from a severe treble boost (fig.3, top pair of traces). The vast majority of CDs, however, will be reproduced with a slight top-octave rolloff (fig.3, bottom traces). Channel separation (not shown) was only fair at 55dB (L-R) and 61dB (R-L) at 1kHz.
Fig.3 47 Laboratory 4715, frequency response without (top) and with (bottom) de-emphasis, at -12dBFS into 100k ohms (right channel dashed, 0.5dB/vertical div.).
Fig.4 shows the 1/3-octave audioband spectrum of the 4715's outputs while it decoded dithered 16-bit data representing a 1kHz tone at -90dBFS. The noise floor seen is primarily due to the dither; commendably, considering the 4715's half-wave-rectified topology, power-supply spuriae are absent. However, some second-harmonic distortion is apparent in both channels, while the left channel shows some negative linearity error. Increasing the data word length to 24 bits (fig.5) minimizes the second harmonic, but the third through ninth harmonics make appearances, due to the Philips DAC chip truncating the incoming data to 16 bits and thus eliminating the 24th-bit dither. This will not be an issue for CD playback, but it will be if the 4715 is used with hi-rez data sources. A similar but wider-band analysis performed with the DAC being fed 16-bit digital black data (fig.6) reveals the presence of energy in its output at the CD's 44.1kHz sample rate and its odd harmonics.
Fig.4 47 Laboratory 4715, 1/3-octave spectrum of dithered 1kHz tone at -90dBFS, with noise and spuriae (16-bit data, right channel dashed).
Fig.5 47 Laboratory 4715, 1/3-octave spectrum of dithered 1kHz tone at -90dBFS, with noise and spuriae (24-bit data, right channel dashed).
Fig.6 47 Laboratory 4715, 1/3-octave spectrum of digital black, with noise and spuriae (16-bit data, right channel dashed).
Plotting the 4315's output amplitude against absolute signal level gives the traces of linearity error in fig.7. As expected from fig.4, the right-channel behavior is excellent, the left channel less so. The effect of noise in this graph is less than I would have expected from the minimal power supply and the passive I/V conversion. However, the lack of any output filtering results in a poor-looking waveform with an undithered 1kHz tone at -90.31dBFS, with sharp spikes of ultrasonic energy evident at the zero-crossing points (fig.8).
Fig.7 47 Laboratory 4715, departure from linearity, 16-bit data (2dB/vertical div., right channel dashed).
Fig.8 47 Laboratory 4715, waveform of undithered 1kHz sinewave at -90.31dBFS, 16-bit data.
Without an output buffer, the DAC chip is exposed to the vagaries of the outside world. To my surprise, it produced higher levels of harmonic distortion when driving high impedances than the usually more demanding low impedances. This is illustrated in figs. 9 and 10, taken with the 4715 driving a full-scale low-frequency sinewave into 100k and 600 ohms, respectively. Into what would be considered the "safe" load (fig.9), a veritable picket fence of potentially audible distortion products is evident, with the third harmonic the highest in level at -52dB (0.25%). But into the very low load (fig.10), the harmonics either disappear or drop below -80dB (0.01%), with the second now the highest in level.
Fig.9 47 Laboratory 4715, spectrum of 50Hz sinewave, DC-1kHz, at 0dBFS into 100k ohms (linear frequency scale).
Fig.10 47 Laboratory 4715, spectrum of 50Hz sinewave, DC-1kHz, at 0dBFS into 600 ohms (linear frequency scale).
This perverse behavior was also evident when I tested the 4315 for high-frequency intermodulation: lower load impedances reduced the level of classic intermodulation products. However, the absence of an output low-pass filter means that the mix of 19kHz and 20kHz tones I use to test this performance aspect results in aliased tones being dumped back down into the audioband, which can be seen in fig.11. The audibility of this behavior is unpredictable, as it will very much depend on the music's spectral content. However, I do wonder if it is associated with the softness of the 4715's sound noted by AD.
Fig.11 47 Laboratory 4715, HF intermodulation spectrum, DC-25kHz, 19+20kHz at 0dBFS into 8k ohms (linear frequency scale).
Despite the designer's claims, an area where the 47 Laboratory gear did trip over its own feet was in its rejection of word-clock jitter. I used the Miller Analyzer to search for sidebands in the 4715's analog output that would result from jitter, among other things, while the 4716 transport played a disc containing a high-level 11.025kHz tone over which had been laid a 229Hz squarewave at the LSB level.
The absolute jitter level was a high 1.47 nanoseconds peak-peak. The resultant narrowband spectrum is shown in fig.12; it can be seen that almost all the jitter components lie at 229Hz and its harmonic multiples to either side of the 11.025kHz tone (red numeric markers). I suspect that this, too, contributed to the audible softening of the 4715's presentation. But apart from these and the low-frequency jitter sidebands (purple circles), the 4715's noise floor is actually a lot cleaner than I expected.
Fig.12 47 Laboratory 4715, high-resolution jitter spectrum of analog output signal, 4716 CD transport via 1m, 75 ohm S/PDIF link (11.025kHz at -6dBFS sampled at 44.1kHz with LSB toggled at 229Hz). Center frequency of trace, 11.025kHz; frequency range, ±3.5kHz.
In fig.12, the low-frequency sidebands indicated with a gray "5" lie at the ±75Hz subcode frame frequency; the ones indicated with purple numeric markers lie at ±17.6Hz and its harmonics. These sidebands appear to be associated with the 4716 transport. When I used the 4716 to drive other processors, these sidebands appeared in their analog output spectra; if I drove the 4715 with my usual PS Audio Lambda transport, they disappeared from the 47 Lab's output spectrum (though the data-related sidebands were as strong as before). Though the noise floor in fig.12 is about 6dB higher in level than that of the best CD playback systems I have measured, it is lower than the noise floor of other players that use passive I/V conversion—the Balanced Audio Technology VK-D5, for example.
The 47 Laboratory components had idiosyncratic measured behavior that, while not as bad as I would have expected in some cases, certainly suggests that they will have readily identifiable sonic characters. And a preamp with a lower-then-usual input impedance would appear to be essential to getting the most linear performance from the DAC.—John Atkinson