dCS Purcell D/D converter Measurements part 2
Setting the Purcell to upsample the CD data to 24-bit/96kHz and driving the Elgar with a double-speed AES/EBU link almost doubled the jitter level, to 325ps. This is still low, but reflects the fact that the effect of jitter in the datalink increases with increased sample rate. I was surprised, therefore, to find that changing the link between the Purcell and the Elgar to a double AES/EBU link, each carrying one channel of data at 48kHz, increased the measured jitter to 377ps.
In both cases, the levels of data-related jitter were unaffected by the Purcell, the increases in jitter being due to the introduction of higher-frequency sidebands. This can be seen in the grayed-out trace in fig.1: the spectrum of the Elgar's analog output when fed CD data upsampled by the Purcell to its maximum rate of 192kHz. The jitter level was now 373ps, with sidebands visible at ±675Hz, ±743Hz, and ±810Hz, indicated with magenta ovals. While these are significantly higher than with the Elgar driven direct, it is difficult to estimate what, if any, subjective effect these sidebands will have. Jonathan definitely preferred the sound of CDs upsampled to 192kHz, which strongly suggests that the effect of the Purcell's digital filter outweighs any degradation due to the increased word-clock jitter and its different spectrum.
Incidentally, I repeated this measurement with the Elgar's output set to "high" but with its volume control set to "-10" to bring the analog level back down to what it had been before. The spectrum was identical but the absolute level of the jitter dropped to 218ps.
How about when a much less expensive DAC is used, with an input receiver whose jitter reduction might not be as good as the Elgar's? In recent issues I have reviewed computer soundcards with digital outputs and have found that the Musical Fidelity X-24K D/A processor did seem sensitive to jitter datastreams. Costing $599 when it was available, the Musical Fidelity DAC uses a standard input receiver based on the Crystal CS8414 chip.
The black trace in fig.2 shows the jitter spectrum taken from one of the X-24K's analog outputs under identical circumstances as the black trace in fig.1, with the processor fed an S/PDIF datastream sourced from a PS Lambda transport. The absolute level of the word-clock jitter is actually lower than the Elgar's, at 163.8ps peak-peak, and is dominated by data-related sidebands. The pair of sidebands at the fundamental frequency of ±229Hz are indicated with red "4" markers, for example, and contribute 78ps to the total.
Fig.2 Musical Fidelity X-24K, high-resolution jitter spectrum of analog output signal (16-bit/44.1kHz data, 11.025kHz at -6dBFS with LSB toggled at 229Hz). Center frequency of trace, 11.025kHz; frequency range, ±3.5kHz. Source is PS Lambda CD transport via 6' Apature S/PDIF cable. Grayed-out trace is with dCS Purcell inserted into the datastream, set to upsample to 24/96, and connected to X-24K via a true 75 ohm S/PDIF datalink.
The effect of inserting the Purcell into the datastream is shown by the grayed-out trace in fig.2. The jitter level has risen only slightly, to 195ps, and the contribution of the ±229Hz sidebands has been reduced to 42ps. However, the low-frequency end of the noise floor has risen, as it did in the grayed-out trace in fig.1 (though to a greater extent), and a number of additional sidebands have appeared: at the AC supply-related frequencies of ±60Hz and ±120Hz and their harmonics, and at multiples of the subcode-related frequency of ±75Hz. Thankfully, these are all at low levels.
So the answer to my third question, too, is "no": The dCS Purcell does not appear to make the situation worse by introducing significantly higher levels of word-clock jitter, at least not with these two D/A processors. However, before purchasing a Purcell, it would still be a good idea to audition it with your own processor.—John Atkinson