What is jitter in digital audio, and does it matter?

Timing errors, with the data intact

Jitter has nothing to do with bits going missing. Audio sent over USB, a network or S/PDIF either arrives as the same numbers that left the file or fails audibly, with dropouts and clicks. What can vary is the moment each of those numbers becomes a voltage.

A DAC converts on the edges of a clock. If those edges aren't perfectly evenly spaced, each sample is reproduced slightly early or slightly late, and a value that was right for one instant is played at another. The result is a small analog error, and once the conversion has happened it can't be removed further down the chain.

How much timing error it takes

For a sine wave, a timing error produces an amplitude error proportional to how fast the wave is changing at that moment. Worked through, the error relative to the signal is 2π × f × Δt, where f is the frequency and Δt the timing error.

Putting numbers in settles most arguments about scale:

Two things follow. Jitter matters more as frequency rises, by 6 dB per octave, so its effects show up in the treble first. And the unit that decides relevance is the picosecond, which is the range in which converter clocks are commonly specified.

The error also has a shape. Periodic jitter, from a power supply ripple or a clock interfering with another, adds pairs of sidebands on either side of each tone. Random jitter raises the noise around the tone. Both rise and fall with the music, which sets them apart from ordinary hiss, and a measurement reads jitter from the sidebands around a test tone.

Where jitter gets into a digital chain

Only at the conversion clock, so the question for each input is whether the incoming timing reaches that clock.

With asynchronous USB it doesn't. The DAC's own oscillator is the master, the computer is asked for data as the DAC needs it, and samples wait in a buffer until the local clock calls for them. The source's timing is discarded by design.

With S/PDIF, over coaxial or TOSLINK, it can. The clock is embedded in the data stream and the receiver recovers it, usually with a phase-locked loop, so a noisy transport or a degraded signal can turn into timing error at the converter. I²S carries its clocks on separate conductors, which removes the recovery step, and a DAC that uses those clocks directly inherits their quality.

Some DACs refuse the incoming timing altogether. The Henry Audio DA 256 buffers its coaxial and optical inputs and plays them out on its own precision clocks, the same ones its USB input uses, and its review names timing as the best thing about it, most audible on coax and optical.

What a DDC or reclocker changes

A digital-to-digital converter sits between a source and a DAC. It takes USB from a computer, buffers the data and sends it on as S/PDIF, AES/EBU or I²S timed by its own clock. The Denafrips Hermes 12th does this with an oven-controlled crystal oscillator, a FIFO buffer, two BNC inputs for an external master clock and two I²S outputs. Its review heard it lift a veil without adding any color of its own, and lists as a con that it won't fix a bad DAC.

The Douk Audio U2 Pro does the same job for $70. Its review heard tighter timing and a fuller, more detailed sound, at the cost of some air and a narrower soundstage, and notes that how much it helps depends on how much jitter your system has. Its stock oscillator is listed as 24.000 MHz ±20 ppm. A ppm figure says how far the clock's average frequency may sit from nominal, which changes pitch by an amount far too small to hear. Jitter is the short-term wobble between edges, specified separately as phase noise or in picoseconds.

A reclocker can only help where the DAC after it depends on the timing it receives, over S/PDIF without a buffer of its own, or over I²S. A DAC that already buffers and reclocks every input leaves it much less to do.

Can you hear jitter?

Large amounts, yes. In a blind test by Ashihara and colleagues, trained listeners needed several hundred nanoseconds of random jitter before they could hear it on music. Converter clocks are commonly specified in picoseconds, thousands of times smaller, which is why many engineers regard jitter as solved in competent designs.

Where it gets argued is the residual difference between good digital sources. Reviews on this site report hearing DDCs change a system, the Hermes as a clear improvement and the U2 Pro as a trade-off. Measurements can show whether a device lowers jitter at a DAC's output, and they can't say how audible a change that small is to you. If you can record the same passage from two sources, the ABX blind test plays them level-matched and tells you whether you can pick them apart. The DAC guide covers converters whose own clocking decides most of this.