Why is USB audio so often preferred over AES/EBU or S/PDIF?

[klettermann]

In certain circles, especially those using audio computers and servers, USB is clearly the favored connection to a DAC. Okay, that might make sense for a PC-to-DAC connection. But it then gets applied to streamers and other appliances that already have other digital outputs. In fact, the fancier the DAC, the more likely it is to have S/PDIF and AES inputs. Yet these guys will unfailingly go for USB. I read one case where a guy with an expensive Lumin streamer was advised to add another endpoint so he could connect to a similarly expensive DAC by USB. Simply using the existing BNC connection wasn’t even considered.

This is puzzling. Why devote enormous attention to fixing USB’s apparent flaws with better outputs, isolators, regenerators, endpoints, and so on? If USB is inherently superior, why is so much remediation needed? A large USB-remediation ecosystem has developed to improve the supposedly superior interface.

By contrast, AES/EBU and 75-ohm S/PDIF over BNC seem to attract almost no corrective effort. Is that because they are technically inferior, or because they are mature interfaces that generally just work? I’ve been using both forever and have never felt they were somehow lacking. Clearly I'm m issing something.

Have a great day, and cheers.

 

[Roland68]

In certain circles, especially those using audio computers and servers, USB is clearly the favored connection to a DAC. Okay, that might make sense for a PC-to-DAC connection. But it then gets applied to streamers and other appliances that already have other digital outputs. In fact, the fancier the DAC, the more likely it is to have S/PDIF and AES inputs. Yet these guys will unfailingly go for USB. I read one case where a guy with an expensive Lumin streamer was advised to add another endpoint so he could connect to a similarly expensive DAC by USB. Simply using the existing BNC connection wasn’t even considered.

This is puzzling. Why devote enormous attention to fixing USB’s apparent flaws with better outputs, isolators, regenerators, endpoints, and so on? If USB is inherently superior, why is so much remediation needed? A large USB-remediation ecosystem has developed to improve the supposedly superior interface.

By contrast, AES/EBU and 75-ohm S/PDIF over BNC seem to attract almost no corrective effort. Is that because they are technically inferior, or because they are mature interfaces that generally just work? I’ve been using both forever and have never felt they were somehow lacking. Clearly I'm m issing something.

Have a great day, and cheers.

Most people don't realize that USB 2.0 was a thoroughly inferior and unsuitable interface for audio.
It was only through ongoing developments—such as XMOS and Amanero—that it eventually became an interface on par with SPDIF/AES.
The rest is just the usual audio marketing that fools many buyers. Uninitiated users, in particular, see an advantage in the potentially higher data rates, but... well.
People overlook the fact that a vast amount of music is recorded using SPDIF/AES; yet, suddenly, that’s supposedly insufficient for playback? That really highlights the lack of technical understanding involved.

And before anyone brings it up: USB 3.x, 4.x, and so on wouldn't be any better—they would simply introduce new and additional problems.

 

[HalSF]

In 2026, the subjective soup of digital interconnect sound-quality “preferences” is something unrelated to audio science.

There’s a thread over at Steve Hoffman Forums right now devoted to the sound of optical cables that dives head first into this madness of connoisseurs and artisanal taste testers perceiving the most insane variations and flavors attributed to digital cables, especially those marketed as “audiophile.”

All competently manufactured, robust, properly spec’d digital cables used with well-engineered gear are going to sound the same.

Of course, that’s only after 11,000 hours of cable break-in [sarcasm emoji].

 

[Nutul]

Most people don't realize that USB 2.0 was a thoroughly inferior and unsuitable interface for audio.
It was only through ongoing developments—such as XMOS and Amanero—that it eventually became an interface on par with SPDIF/AES.
The rest is just the usual audio marketing that fools many buyers. Uninitiated users, in particular, see an advantage in the potentially higher data rates, but... well.
People overlook the fact that a vast amount of music is recorded using SPDIF/AES; yet, suddenly, that’s supposedly insufficient for playback? That really highlights the lack of technical understanding involved.

And before anyone brings it up: USB 3.x, 4.x, and so on wouldn't be any better—they would simply introduce new and additional problems.

If you allow me... everything but USB depends on the source clocks. USB makes everything cleaner and easier, by allowing asynchronous data transfer of the audio packets, hence:
1. - eliminating jitter (whatever it is nowadays a problem or not)
2. -avoiding any - ANY - interruption in the flow due to buffering and/or RFI / EMI getting in the way

The recording technique has nothing to do with the reproduction technique. On the contrary: jitter induced (by spdif) at recording stage needs not to be amplified (by spdif again) at playback stage if this can be avoided, right? Not only, the clock, in such transmissions, needs to be extracted from the data-flow... not only data is affected by all kind of interference, but the clock also)
USB can somehow be affected, but has checksum and retransmission algorithms that eliminate everything you could classify as bad-signal on a conventional serialized audio tranfer such as spdif (be it coax or toslink)

That is why USB is your choice when connecting digital audio devices. It is like having the music already all inside your DAC, leaving the how-it-got-there a no-question.

ETA
USB 2.x 3.x 4.x has nothing to do with UAC1 or UAC2, it's just speed that's involved, hence more audio buffered on the destination device that is already correct and perfect at destination. BTW, coax/optical AND USB are both serial data transfers, just USB offers integrity control over it, while the others don't...

 

[Bleib]

USB outputs are quite common from many devices, optical used to be more common from soundcards and motherboards but are increasingly more rare. Coax is the least interesting as here cable design can mess up a with ohm-levels

HDMI obviously has both picture and many channels of uncompressed audio, but it's not needed for mere 2 channels

 

[DavidShe]

USB provides multi-channel audio, MIDI, interface control, firmware updates, power and potentially even more.

USB does not embed an audio clock in the data stream the way AES/EBU or S/PDIF do. Instead, modern USB audio interfaces typically use asynchronous USB, where the audio interface controls the timing. This is usually an advantage.

 

[Nutul]

Nobody could ever bring my attention away from USB, especially when the alternatives are optical (limited bandwidth) or coaxial (impedance matching problems) or I2S (never intended for inter-device transmission (inter-chip, instead, where paths are short. indeed SHORT) or any other data+clock on the same line are involved.

 

[Redacted]

Cheap and found everywhere.
Another solved problem for consumers!

 

[HalSF]

Making astute comments about various technical advantages and use cases and protocol history for different digital connections for audio, USB or otherwise, is obviously perfectly fine, as long as these considerations don’t drift into audiophile woo-woo about sound quality.

 

[klettermann]

Nobody could ever bring my attention away from USB, especially when the alternatives are optical (limited bandwidth) or coaxial (impedance matching problems) or I2S (never intended for inter-device transmission (inter-chip, instead, where paths are short. indeed SHORT) or any other data+clock on the same line are involved.

Coax and AES only have “impedance matching problems” if the wrong cables or connectors are used. Proper 75-ohm BNC and 110-ohm AES are mature, solved interfaces; AES also provides balanced transmission. So what would USB actually improve between a streamer and DAC already equipped with competent AES/BNC?

 

[Multicore]

I was gonna write something like: Why do you care what "these guys" "in certain circles" do, or about "one case where a guy"? I thought it was fairly well established that some audio hobbyists are weird.

But instead of that, given some replies you already got that I think maybe aren't sarcastic jokes, ...

I’ve been using both forever and have never felt they were somehow lacking.

Lots of people have. Myself included.

 

[klettermann]

I was gonna write something like: Why do you care what "these guys" "in certain circles" do, or about "one case where a guy"? I thought it was fairly well established that some audio hobbyists are weird.

But instead of that, given some replies you already got that I think maybe aren't sarcastic jokes, ...

Lots of people have. Myself included.

I confess to getting an odd kind of gratification from browsing "certain circles" where "those guys" debate such issues. It's kind of like reading a flat-earth or intelligent-design blog. And yeah, to my surprise I guess that even here it's not a solved problem.

 

[Nutul]

Coax and AES only have “impedance matching problems” if the wrong cables or connectors are used. Proper 75-ohm BNC and 110-ohm AES are mature, solved interfaces; AES also provides balanced transmission. So what would USB actually improve between a streamer and DAC already equipped with competent AES/BNC?

clock dissection from the audio signal.

 

[john61ct]

I am happy with any of the four digital media as the INITIAL output from a preamp or streamer into the signal processing section.

Same with analog, no audible differences.

HOWEVER once signals are getting split up (crossovers, upsampling) or filtered EQ / DRC'd by separate devices, then signal CLOCKING becomes critical.

I do not want sink devices that do D/A or A/D conversion to rely on slaving to signal clocking, so I standardise on BNC Word clocking for those.

I wish for multi-channel transport / conversion that USB-out were more common among DDCs, without having to insert RPi to get host mode.

ADAT is a great alternative, but all too rare in affordable consumer-targeted gear.

The DSPi project looks very promising for that.

 

[SSS]

Most people don't realize that USB 2.0 was a thoroughly inferior and unsuitable interface for audio.
It was only through ongoing developments—such as XMOS and Amanero—that it eventually became an interface on par with SPDIF/AES.
The rest is just the usual audio marketing that fools many buyers. Uninitiated users, in particular, see an advantage in the potentially higher data rates, but... well.
People overlook the fact that a vast amount of music is recorded using SPDIF/AES; yet, suddenly, that’s supposedly insufficient for playback? That really highlights the lack of technical understanding involved.

And before anyone brings it up: USB 3.x, 4.x, and so on wouldn't be any better—they would simply introduce new and additional problems.

Statement without underlying facts. USB is a widely used standard and works perfectly for audio reproduction. Of course not suitable for highest sample rates which are necessary for the recording and mixing desks. For reproduction the 24 bit and 96 k sample rate is fully sufficient for good reproduction. And computers (which I use for reproduction) have USB as standard available, all other would need specific interfaces. So please don't blame the "uninformed" customers.

 

[xanalog]

And before anyone brings it up: USB 3.x, 4.x, and so on wouldn't be any better—they would simply introduce new and additional problems.

Argh!?
As you requested, I will not be bringing up USB3.x/4.x...
..., but I am that you are trying to shut down a wrong and contentious statement (you made) about advancements in interface improvements (via versioning) are useless since ' —they would simply introduce new and additional problems'.

I don't know how else to put this more kindly and in PC; but that kind of logic would land me square in the middle of 1950s, when tubes were vogue and transistors would've simply introduce new and additional problems.
Argh!?

 

[DWPress]

Probably one of the biggest advantages is multichannel and bi-directional processing. HDMI and ADAT can carry multichannel but it's more rare to use it in both directions.

Example: My Okto DAC8 Pro can accept a clocked AES input from my ADC and send it to my Mac mini where it is processed into 8ch including XO and DSP and then send it back to the Okto for 8ch out to amps and subs. All in real time with minimal latency.

So, especially in pro audio, USB is used quite a bit and for good reason.

 

[klettermann]

OK, all this pretty well clarifies why USB can be useful: it’s ubiquitous, computer-native, supports async operation, multichannel, bidirectional audio, control functions, etc etc.
But what about the USB-remediation industry? If async USB is already the technically cleaner solution, why the huge ecosystem of USB cards, isolators, regenerators, reclockers, special endpoints, special cables, linear supplies, filters, and all the rest? That seems less like a settled superior interface and more like a very tweakable anxiety platform.

My guess is that USB had enough real historical and implementation-specific issues - noisy PCs, bus power, ground leakage, weak early receivers, RF leakage or whatever - to give the category plausibility. Maybe those edge cases became generalized into a permanent audiophile market: “USB is best, but only after you fix it.” By contrast, proper AES/EBU and 75-ohm BNC S/PDIF are mature and boring. They have limitations, but in a normal two-channel streamer-to-DAC setup they generally just work. There isn’t much product ecosystem to build around “use the correct impedance cable and move on.”

So I’m not arguing that USB has no advantages. It plainly does. I’m wondering about the leap from those advantages to the USB-tweaking culture favored in "certain circles."

 

[Redacted]

I have 3 tin-foil hats to send in the post to today's lucky winners!

 

[DWPress]

why the huge ecosystem of USB cards, isolators, regenerators, reclockers, special endpoints, special cables, linear supplies, filters, and all the rest?

Many of these are useless tweaks with, again, use cases (reclockers) in pro audio. Specifically look at some of the reviews Amir has done on the isolators sold to audiophiles - you may as well add some cable lifters. Cables can make a difference especially for long connections but otherwise rarely make a difference for audio needs.