USB versus coaxial versus optical: does the DAC input actually matter?

Last week a reader emailed me three screenshots of the same track measured off the same DAC, switching only the input: USB, coax, optical. The differences were real but tiny, and every single one of them was explainable by something boring — not by "USB sounds digital" or "optical is pure light so it's cleaner," which are the two claims I hear most often on forums. The short version: the input matters less than the implementation behind it, and almost nobody tells you that plainly.

I've spent a fair chunk of the last few years pulling DACs off shelves at home in Perth and running the same album through every input they offer, mostly out of stubbornness. The results are consistent enough that I feel comfortable being blunt about it here, even though it'll annoy people who've built a whole identity around a $400 USB cable.

What each input actually is, mechanically

Coaxial (RCA, using SPDIF) and optical (TOSLINK, using the same SPDIF protocol over light) are both fixed-clock formats. The transmitting device — your streamer, transport, or computer — packages audio data with an embedded clock signal, and the DAC has to recover that clock to know when to sample each bit. This is the source of "jitter" as a real, measurable phenomenon: if the clock recovery is sloppy, timing errors creep into the reconstructed waveform. TOSLINK has historically been slightly worse for jitter because the optical-to-electrical conversion at the receiving end adds its own timing smear, though a well-designed receiver chip largely irons that out these days.

USB is different in kind, not just degree. Since USB Audio Class 2.0 became standard, most competent DACs use asynchronous mode, where the DAC's own clock — not the source device's — governs playback timing. The DAC tells the source "send me data at this rate, on my schedule," and the source complies. This is the single biggest reason USB implementations, when done properly, sidestep the jitter arguments that dominate coax and optical entirely. The clock quality question moves from "how good is the transmission path" to "how good is the DAC's own oscillator," which is a much easier problem to solve well.

Where jitter actually shows up in real gear

Here's the thing that gets lost in translation: modern well-engineered DACs, across all three inputs, reduce jitter to levels that are inaudible by any published research I'd stand behind. The Audio Engineering Society has published work on jitter audibility thresholds for decades, and the numbers required to hear anything at all are well beyond what a competently designed 2020s DAC produces on any input, sync or async. The RME ADI-2 DAC FS, which I keep coming back to because it publishes its own measurements rather than asking you to trust marketing copy, shows near-identical jitter performance across coax, optical and USB in its own documentation — differences exist, but they're buried well under any audibility threshold.

Where jitter actually becomes audible is at the cheap end: a $40 DAC with a basic receiver chip and no reclocking stage, fed by a source with poor SPDIF output impedance matching, on a long or poor-quality cable. That's a real scenario. It's just not the scenario most people fret about when they're comparing a $2,000 streamer's coax output against its USB output.

The sample rate and channel count ceiling nobody mentions

This is the one genuinely practical difference and it trips people up constantly. Optical TOSLINK is capacity-limited — it typically tops out around 96kHz/24-bit and can't carry multichannel PCM or many high-resolution formats at all. Coax SPDIF stretches a bit further, commonly to 192kHz, though implementations vary. USB has no such ceiling in practice; it'll happily carry 384kHz, DSD, MQA-era decoded streams, whatever your DAC and source both support. If you're feeding a DAC hi-res files or DSD and you've got it wired via optical, you may be silently sample-rate-limiting yourself without any error message telling you so. That's not a sound-quality argument, it's a "your files aren't actually playing at the rate you think" argument, which matters more.

I'd point you at our piece on sample rate and bit depth explained if you want the fuller picture on whether those higher numbers matter once they do get through — the honest answer is "less than the format matters for capacity, more than nothing."

Galvanic isolation: the genuine coax and optical advantage

I don't want to make this sound like USB wins on every axis, because it doesn't. Optical has one real, physical advantage: it carries no electrical connection between source and DAC at all, because it's light through a fibre. That means zero possibility of ground loop hum or noise coupling between a noisy computer power supply and your DAC's analogue stage. Coax gives you an electrical connection but many well-designed coax inputs use a transformer for isolation, achieving something similar. USB, by contrast, carries both data and often power over the same cable, and if a DAC doesn't isolate its USB input properly, noise from a laptop or NUC can genuinely leak into the analogue output as a background hiss or hash.

This is the actual reason some people report "USB sounds worse" on cheap DACs — it's real, but it's a grounding and isolation engineering problem, not an inherent property of the USB protocol. Decent DACs, including the iFi Audio ZEN DAC 3 (check price) and the Chord Electronics Mojo 2 (check price), build in USB isolation specifically to deal with this. If you're running a DAC off a noisy desktop PC and hearing hiss that disappears when you switch to optical from a dedicated streamer, that's diagnostic information, not proof optical is categorically superior.

What I actually do at my own desk

I'll admit a small bias here: my main desktop rig runs USB into an RME ADI-2 DAC FS (check price), mostly for convenience — one cable, full sample rate range, no capacity ceiling to think about. But when I set up a dedicated streaming box like the WiiM Ultra (check price) or a Cambridge Audio CXN100 (check price) feeding a separate DAC, I'll often use coax if the run is under a couple of metres, purely because it removes the USB driver and power-noise variables from the chain entirely and I've never measured it losing anything to USB on gear built this decade.

The mildly contrarian bit, since I promised I'd give you one: I think the entire "best digital cable for your DAC" cottage industry is solving a problem that competent DAC design solved years ago, and the money is far better spent moving up a tier in the DAC or streamer itself. A genuinely better DAC or streamer will do more for your sound than any cable swap on a good input, full stop.

When the input choice genuinely does change the sound

To be fair to the folklore, there are real scenarios where switching inputs changes what you hear, and they're worth naming plainly so you can diagnose your own system instead of guessing.

First, ground loop hum — if your source and DAC are on different power circuits or in different rooms, a galvanically isolated optical or transformer-isolated coax link can eliminate a hum that a USB connection introduces. Second, sample rate mismatches — if you're streaming hi-res or DSD through optical and getting silent dropouts or automatic downsampling, that's the capacity ceiling biting you, not a sound-quality issue but a very real practical one. Third, driver and clocking quirks on some Windows PCs with USB Audio Class 2.0 — occasionally a specific USB chipset and driver combination introduces audible glitches that a coax link from the same source, via a dedicated audio interface, avoids. None of these are "USB sounds worse because it's USB." All of them are specific, fixable engineering issues.

The practical takeaway for buying and setup

If you're building a system around a desktop DAC/amp or a standalone streamer, pick your input based on capability and isolation, not mythology. Use USB when you need the full sample rate range or you're running DSD, and buy a DAC with proper USB isolation if it's sitting near a noisy computer. Use coax for a clean fixed-rate connection between a transport and DAC where ground loops aren't a concern. Use optical specifically when you need galvanic isolation between two boxes on different power circuits, and just keep half an eye on whether your files exceed its bandwidth ceiling.

None of this is exciting, and that's rather the point. The genuine engineering here was mostly solved a decade ago by good implementation; what's left is mostly people re-litigating an argument that doesn't need re-litigating. Spend the cable money on room treatment instead — it'll do more.

Frequently asked

Small notes worth having in one place before you go rearranging inputs on your own rig.