USB and i2s debate

Theoretically, no, you’re not. USB is a perfectly fine interface for moving digital data. While theoretically it might be possible to make a device that sounds in some way differently depending on which digital input is used, it would really be an extra effort to make what amounts to a broken device.

Honestly, I am not sure why I2S connection between devices is even a thing. I2S is n electrical protocol for moving bits inside a device.


In the real world, USB does not always sound as good as other digital inputs. The digital interface world is not a simplistic as you suggest. I2S is used because it sometimes has benefits over other digital interfaces. You can play the “broken device” card but that again is a simplistic view. USB implementations vary. Some are much better at galvanically isolating the DAC from the USB input than others. Digital interfaces are not just bits…

1 Like

We’ve been over this before… Yes, a particular device might sound different (better or worse) on different inputs. Give me two weeks, I’ll design one for you that sounds like crap on any given input. Another week, and there will be a switch to select which input will be bad, too :rofl:

But I still hold that if a device does not work well with an input, assuming that the source is operating within interface specs, it is broken. This is proven, if anything, by existence of any number of devices that work just fine from any digital input without needing any fancy external gizmos.

I2S… it can work, but being an interface that has never been designed to be used between devices (note that there isn’t even a standard I2S plug, like there is with USB) its advantages are super questionable. Yes, you can make USB work badly, but, as noted, it just means that the USB interface has been subject to some severe cost cutting (or outright incompetence, can never discount that) as even some quite reasonably priced devices can produce output from USB that is identical to that over I2S or any other input.


What? Do you think all these companies that make I2S-compatible devices do so because they are stupid? What is the motivation? If I2S didn’t provide anything useful it would not be used by so many companies.

This idea that devices with less than ideal USB interfaces are broken is not proven…it’s just an opinion.

I2S was developed as an on-board bus and does not even have error detection (other than USB). This is problematic as it was never intended to work over cables and there can be errors. I suppose they can be mitigated, but still.

I don’t know why hifi companies still use this ancient specification for something that it was not designed for. Because it can be marketed as something special compared to “mundane” USB? It’s not as if hifi companies selling useless stuff is unheard of. What do you think are the advantages?

Huh? Surely if a device offers a USB port it should be well implemented?


Of course the companies are not stupid. This hobby has enough people with sufficiently deep pockets and sufficiently shallow understanding of physics to support not one, not two, but multiple manufacturers of directional cables (some selling for tens of thousands of dollars per foot), risers for said cables, magic fuses, quantum dots, CD Demagnetizers, and God knows what else. From the manufacturer’s POV it is quite smart (if not too ethical) to sell audiophiles yet another set of sources, DACs, and cables. It’s good business and has avsolutely nothing to do with any (nonexistent) advantages of an electric protocol that even has a quite convenient backstory of having been designed, at some point in prehistoric times, to move digital audio data. The point that it is not designed to move data over interconnect cables between devices can be conveniently overlooked.

Umm, no. If the device does not work properly with valid, to-spec input, it is broken. You can call it badly designed, or cost-optimized (but devices we’re talking about tend to be quite expensive), but really it is broken by design. It does not properly work as a USB receiver. Especially when a $100 box can handle any spec-compliant USB input with no ill effect, so there isn’t even an excuse that designing a proper USB connection is somehow complicated or super expensive.


Have to agree. That’s what “broken” means – it doesn’t work properly. More of a “meaning” than an “opinion”. To what degree the brokenness is important – that’s an opinion.


Guy, there are plenty of audio devices out there with USB interfaces that transfer all the bits correctly yet also allow a bunch of electrical noise into the device making it sound worse than other better implemented digital inputs on the device.

So, the USB interface is not broken yet it sounds worse. This is why bits are not just bits….

Bits are just bits. That there are plenty of devices that are badly or cheaply designed and might sound doesn’t mean anything other than that there are enough people willing to pay for them. There are also plenty of devices that sound absolutely fine when fed any spec compliant USB signal.

It is even funnier when the proposed alternative is a protocol that is not even designed for connecting devices and involves more of noise adding electrical conversions to get to the DAC itself than USB.

1 Like

The protocol that uses Ethernet via RJ45 or HDMI connectors is not raw i2S. So making assumptions about its performance are immediately faulty if the belief is that the protocol for moving a signal no more than 20cm across a board and internal wiring is the same as that between discrete components. Applied correctly it is HDMI over LVDS. LVDS launches any protocol applied to it as a differential signal with all of the benefits it implies. Greater distance, higher voltage, noise cancelling properties of a differential signal etc.
i2S over HDMI exists because USB wasn’t very good. Data can exist uncorrupted in very noisy environments. Environments which a printer or HDD are impervious to but a DAC with a very low noise floor in the analogue domain might not be. Today with lessons learned about how to apply USB in more sensitive environments it is not hard to question why i2S over LVDS exists but ten years ago if you wanted the bit rates offered by USB over the legacy DAC inputs but with the noise performance of those legacy inputs, i2S/LVDS was the only potentially universally applicable way to go.


Yeah, that would be Ethernet, not i2S.

Possibly, And not to forget that USB didn’t exist when i2S was invented (as an on-board bus). But I thought the question was why i2S is used today and what it advantages would be over a competent modern USB implementation.

You know what I meant!

i2S/LVDS is competent by default. USB is not though designers and manufacturers are finally figuring it out. i2S’s Achilles heel is it isn’t standardised so has never been guaranteed to work between manufacturers. Something USB didn’t suffer from. One isn’t better than the other by default. They have their pros and cons.

I didn’t

I would not say that for something that does not have error detection when connecting devices over longer distances with cables. (It’s fine for an on-board bus).

I’m still waiting for what are the pros of i2S for device connection in modern times

1 Like

Isn’t the key caveat, “applied correctly”? Applied correctly, doesn’t USB make more sense than some non-standardized makework audio input system? And if there’s no standard, how does one know that it’s been “applied correctly”?

1 Like

i2S is there for the same reasons SPDIF and AES/EBU remain. As a legacy connection many sill prefer over USB.

@Bill_Janssen It makes sense when you listen and isn’t quite clearly inferior to other inputs to the point where the manufacturer recommends alternative inputs.

Listen? As always, listen with whose ears, mind, and imagination?

But I’d be more sympathetic to this argument if I’d ever seen any evidence that noise somehow seeps in via the differential signalling on USB cables; that, as you put it,

But I’ve never seen that evidence. You’d think that if there was a real problem, someone would have documented it. Sure, if you’re in the production room of an old-fashioned web press newspaper, or in a factory with solenoid punch presses going, maybe there would be interference. Of course, neither of those would have a very low noise floor in the analogue domain.

Is there any evidence of this noise problem other than anecdotal?

1 Like

I thought that there were claims of actual technical advantages. That isn’t one


Yup, it’s different, and needs to go through more electrical conversions before it becomes the interchip I2S inside the DAC than USB does.

Is there any actual proof of that?

But how would this be relevant now, when output of any well-designed DAC with USB input is demonstrably independent of what (within USB spec) you feed it on input?

Ah… that must be the reason pretty much everything connected externally is connected over USB – it’s incompetent!

No. But that isn’t because it is not true. It is because no one is financing this sort of research for public consumption or peer review. If it is being done, it is being done informally by the usual suspects who sell their ideas as products or to manufacturers.

Now the thing here is I am not knocking USB in its better implementations, most of which are more recent. But it has been AWOL for quite a long period of the HD era and from humble beginnings as a proprietary PS-Audio interface i2S/LVDS has been allowed a foothold. Once it is there it has earned its place at least until USB becomes consistent enough across all of its implementations. USB has very obvious advantages and by any metric it should and will win out as the most suitable connection. But as with other legacy connections, perhaps i2S has earned its place as one of a number of still popular connections.

The same argument could be made for any hifi snake oil though, or any other claim really. It’s very weak.
(If we accept this kind of argument, what foothold do we have to reject “There is no evidence for chemtrails, but not because it is not true but because no one is financing this sort of research for public consumption or peer review”?)

I addition, one would think that someone trying to sell something would have more success if they had evidence that what they are selling fixes a real problem