RME Digi96/8 Pro computer soundcard Measurements
As I almost exclusively used the RME Digi96/8's digital outputs to feed outboard digital processors, I first examined the card's jitter performance. To do this, I used CoolEdit 2000 to drive the card with an analytical 44.1kHz-sampled 16-bit WAV file. This consists of a high-level high-frequency tone at one quarter the sample rate (11.025kHz at -6dBFS), with the least significant bit toggled at 229Hz. The Digi96/8 Pro's digital outputs drove a Musical Fidelity X-24K D/A converter, and this unit's analog output was fed to a Miller Audio Research Jitter Analyzer, a "virtual instrument" running on a National Instruments PC card. The analyzer performs a high-resolution FFT spectral analysis on the reconstructed analog signal and searches for jitter-generated sidebands on either side of the 11.025kHz tone.
The result, using a 15' TosLink datalink, is shown in fig.1. The noise-floor components lie around the -128dBFS level, about 3dB higher than the best 16-bit D/As I have tested, and the overall dynamic range is 91.3dB. The overall jitter level is a low 248 picoseconds peak-peak. Although some power-supply-related sidebands can be seen in this graph at ±60Hz and ±180Hz (indicated with brown numeric "2" and "3" markers), most of the jitter is data-related and comes from the strong sideband pair at ±229Hz (red "4").
Fig.1 Musical Fidelity X-24K driven by RME Digi96/8 Pro via 15' TosLink, 44.1kHz sampling, high-resolution jitter spectrum of analog output signal (11.025kHz at -6dBFS with LSB toggled at 229Hz). Center frequency of trace, 11.025kHz; frequency range, ±3.5kHz.
Replacing the TosLink S/PDIF connection with a 6' length of Illuminati 75-ohm electrical datalink gave the spectrum shown in fig.2. The jitter level has increased to a still-good 686ps. The 229Hz sidebands have risen from 187ps to 610ps, while the high-frequency harmonics of 229Hz have also risen in level. High-level sidebands related to the subcode frequency make an appearance (indicated with green numeric markers), but so do a large number of spurious tones (blue markers). As a result, the dynamic range has decreased to 73.2dB. Replacing the Illuminati S/PDIF link with AudioQuest Digi Pro gave the same result.
Fig.2 Musical Fidelity X-24K driven by RME Digi96/8 Pro via 6' Illuminati electrical S/PDIF datalink, 44.1kHz sampling, high-resolution jitter spectrum of analog output signal (11.025kHz at -6dBFS with LSB toggled at 229Hz). Center frequency of trace, 11.025kHz; frequency range, ±3.5kHz.
This behavior surprised me, as the RME's electrical data outputs are specified as being isolated by pulse transformers (these obvious on the circuit board). I would have thought that these would have minimized the effect on the datastream of any electrical interference from the hostile PC environment. Inserting a Meridian 518 digital processor in the datastream eliminated all the spurious noise spikes in the Musical Fidelity's analog output (fig.3), at the expense of increasing the level of the 229Hz sidebands and imposing the Meridian's characteristic rise in the reconstructed signal's noise floor around the fundamental. The dynamic range improved to 83.6dB—note the 1dB drop in the random analog noise-floor components—but the jitter now measured 1152ps.
Fig.3 Musical Fidelity X-24K driven by RME Digi96/8 Pro via 6' Illuminati, Meridian 518 digital processor, and 1m AudioQuest Digi Pro electrical S/PDIF datalinks, 44.1kHz sampling, high-resolution jitter spectrum of analog output signal (11.025kHz at -6dBFS with LSB toggled at 229Hz). Center frequency of trace, 11.025kHz; frequency range, ±3.5kHz. Grayed-out trace is the same as fig.2, without Meridian or AudioQuest.