S/PDIF cable simulator: jitter, loss and echoes

S/PDIF carries bits, but the quality of the clock the receiver recovers depends on the received waveform. A poorly matched or overly long cable increases jitter, which can degrade digital-to-analog conversion. On a short, properly matched link (75 ohms) the effect is usually negligible. The most common case where a cable really changes the sound is electrical: a coaxial cable ties the grounds of the devices together and can create a ground loop (hum), which optical or an isolation transformer removes. The tool quantifies the jitter and losses of each cable; ground loops are not modelled.

Jitter is the deviation of the signal transition instants from a perfect clock. The receiver rebuilds its clock from these transitions, through a PLL that attenuates their fast part. The lowest published mean threshold is 10 ns RMS (individual means of about 7.5 to 13 ns), for a full-scale 20 kHz sine with 17 kHz jitter (Benjamin and Gannon, 1998); with random jitter on music, none of the 23 listeners of Ashihara et al. (2005) detected it at 250 ns.

The eye diagram overlays all the signal transitions onto a single period. The more open the central eye, the better the signal. The vertical opening measures the noise margin (distance to the decision threshold, the green line), the horizontal opening measures the jitter margin (stability of the transition instants). On the left, the Belden 1694A over 40 m keeps a wide-open eye: the receiver tells 0s and 1s apart without ambiguity. On the right, the simulator's non-standard RCA cable over 40 m: loss and picked-up noise close the eye and errors appear. Over 2 m, it keeps a wide-open eye.

With a 75 ohm cable between a 75 ohm output and input, there is no echo: length is only limited by loss and noise, beyond 100 m for a good coaxial cable. With a mismatched cable, such as a non-standard RCA cable, the echo returns 2L/v after each edge. As long as this delay is shorter than one code cell (about fifteen to twenty metres at 44.1 kHz depending on the cable), it delays nearly every edge by the same amount; beyond, it adds data-dependent jitter. There is no "magic" length: for a given cable, shorter remains better.

Coaxial ties the grounds of the devices together (a ground loop is possible) and suffers echoes if it is not 75 ohm; a good cable reaches beyond 100 m. Optical isolates the devices galvanically and removes the loop through that cable, at the cost of a short range (0.2 to 5 m for the Toshiba TOTX147/TORX147 module pair) and of the modules' own jitter, which is not modelled: the simulator shows an ideal optical eye. The tool simulates both families.

For signal reliability, a 75 ohm coaxial connection on BNC is the most robust: low attenuation over length and well-controlled impedance. RCA works too: its connectors are not 75 ohm, but they are a few centimetres long, too short to reflect a significant part of the edges (under 1 % with 20 ns edges, up to about 2 % with 5 ns edges). Optical TOSLINK becomes the safest choice when there are ground loops or an electrically noisy environment, thanks to its galvanic isolation, but its range depends on the modules (a few metres for common audio modules). There is no universal winner: the tool compares these cases with numbers.

It works over a few metres, but it is the wrong cable. An ordinary RCA audio cable has a poorly defined impedance, often far from 75 ohms (about 42 ohms for the simulator's “spaghetti” cable). The mismatch creates reflections that distort the edges and add jitter: the effect stays small over a short run and grows with distance. A true 75 ohm cable, video or digital coax, costs barely more.

Not if both are true 75 ohm cables. Over 10 m, a Belden 1694A shifts the edges by about 0.05 ns peak to peak (without noise; 0.014 ns RMS with the simulated noise of a living room), hundreds to thousands of times below published audibility thresholds. We found no peer-reviewed study comparing two S/PDIF cables blind; related tests show nothing audible at these levels: measured jitter thresholds are hundreds of times higher, two CD sources are not told apart overall (Stereophile calls the result ambiguous) and an Ethernet cable is not recognised by 6 people out of 7 (a rough test that its author calls inconclusive). A higher price can buy sturdier construction or better shielding, not better sound.

They are two variants of the same frame format. S/PDIF (IEC 60958, consumer) runs over unbalanced 75 ohm coax at about 0.5 V, or over optical fibre. AES/EBU (AES3, professional) uses a balanced 110 ohm pair with XLR plugs and a 2 to 7 V level, which lets it cover longer distances in noisy environments. The channel status data also differ.

Almost all S/PDIF outputs go through a transformer or capacitor that blocks DC, forming a high-pass filter whose cutoff usually lies between 2 and 20 kHz. Frame preambles remain free of DC, but their three-cell step, longer than any step in the data, shifts the signal baseline at every subframe (baseline wander), with a part that depends on the data. Since the receiver compares against a fixed threshold, this shift moves the detections and creates data-related jitter. The tool simulates it as an option, with an adjustable cutoff frequency.

The sensitivity matrix quantifies the relative influence of each physical parameter (length, impedance, propagation velocity, attenuation, shielding) on each quality metric (jitter, CER, eye opening). It reveals at a glance the limiting factor of a given cable, so you know what to act on first.