• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Study: Is I²S interface better for DACs than S/PDIF or USB?

Point 1) Wait, did I miss something? USB to I2S converter?? Why did you do that? You are supposed to be testing I2S vs USB? What you tested was a USB to I2s converter against USB. So, it can only be worse than USB, or at best, equal to considering you are already going through USB.
Point 2) I think that people are confusing the words "sounds better" as opposed to technically better. I agree with Paul's theory that removing unnecessary multiplexing from a signal should be better, but what he didn't articulate clearly was that you may introduce new problems with clock timing over distance (he mentioned it without explaining it technically). I respect Paul but he does have a vested interest in I2S in this case. As a technical purist, I see technical benefit in I2S, but I would never claim to be able to perceive it with my aging ears.
Point 3) I absolutely respect this site for its attempt to quantify what has been subjective, but by now we have gone beyond the ability to perceive these differences and are moving into obsessive purist ideals, and that is what the USB vs I2S debate is all about, which protocol gets the bits from A to B with the least errors.
Point 4) I would love for this test to be run again without a USB to I2s converter.
 
  • Like
Reactions: Zog
is there a way to capture the bitstream at the end of the process and compare it to the original audio file? That would be quite definitive.
 
is there a way to capture the bitstream at the end of the process and compare it to the original audio file? That would be quite definitive.
Yes. That's effectively what RME's ADI-2 DACs do when you use their Bit Test File. Any computer with a suitable interface would do - say a Pi. It won't tell you much though - all the interfaces are capable of transferring the bits correctly, and the argument becomes about clocking, jitter, ground loops etc. and how they are handled or might affect sound and measurements. Those are really implementation issues, and from time to time we see DACs that have a measurable problem on one interface but not another because of that implementation.
 
Point 1) Wait, did I miss something? USB to I2S converter?? Why did you do that? You are supposed to be testing I2S vs USB? What you tested was a USB to I2s converter against USB. So, it can only be worse than USB, or at best, equal to considering you are already going through USB.
Point 2) I think that people are confusing the words "sounds better" as opposed to technically better. I agree with Paul's theory that removing unnecessary multiplexing from a signal should be better, but what he didn't articulate clearly was that you may introduce new problems with clock timing over distance (he mentioned it without explaining it technically). I respect Paul but he does have a vested interest in I2S in this case. As a technical purist, I see technical benefit in I2S, but I would never claim to be able to perceive it with my aging ears.
Point 3) I absolutely respect this site for its attempt to quantify what has been subjective, but by now we have gone beyond the ability to perceive these differences and are moving into obsessive purist ideals, and that is what the USB vs I2S debate is all about, which protocol gets the bits from A to B with the least errors.
Point 4) I would love for this test to be run again without a USB to I2s converter.
USB always becomes i2s in conjunction with a DAC.
Nothing was done here other than testing the USB route of a DDC against the USB route of a DAC.
SPDIF/Toslink and AES are also generated from i2s and converted back to i2s in the DAC. This also applies to BT routes to the DAC.

The only thing that would be interesting would be a test with a device that can output i2s without USB. But even there the only difference will be interference/interference in the signal, as @somebodyelse already said.

Bit precise comparisons were made in the 90s with jukebox CD systems for industry, data backup and the music industry and even then there were no problems, not even with a 100 € audio CD player with a digital output. This has also been a solved problem for over 30 years and has since been checked many times by manufacturers, universities and engineers.
 
Returning to the initial question, I wanted to reframe the issue from a different perspective, particularly from a conceptual standpoint that is often completely overlooked.

Leaving aside the S/PDIF issue for a moment, since when digital audio first emerged, it was the only available interface, I would like to focus on the comparison between I2S and USB. As we will see, this is a comparison that essentially has no reason to exist.

The fundamental concept is that a PC or streamer must provide to the DAC (Digital-to-Analog Converter) with a lossless file sourced either locally (from an HDD/SSD) or from the web. This simple observation highlights that the PC and the DAC must perform two distinct roles, and in these roles, it is beneficial to avoid redundancies and overlaps.

We know that I2S was theoretically conceived to improve upon the synchronous S/PDIF protocol by extracting the clock and transmitting it separately. Indeed, in all types of I2S, it is easy to identify a pin labeled CLK or SCK.
This theoretical idea of using a specific BUS, one that is also used internally within DACs to transmit digital signals, reveals that I2S:

  • Was designed exclusively for internal connections
  • Has no error control
  • Does not even define a standard or specifications for external connections
  • Lacks galvanic isolation
I2S is supposed to solve a problem that, in reality, was resolved at least 25 years ago. Its marketing is based on the notion that, in ambiguous tests with too many variables (such as the type of streamer/PC, DAC, digital input, or balanced/unbalanced analog output), one can measure instrumental improvements where the threshold of audibility does not even come close to being reached.

On the other hand, asynchronous USB:

  • Is jitter-free,
  • Does not have an associated clock in its digital stream, and the pinout does not include it,
  • Is not galvanically isolated (the only known problem).

Any PC, anywhere in the world, can always copy any file millions of times to an external USB drive without the possibility of error, even using an external cable (defined by precise specifications) up to 5 meters long.
This should be a fundamental point that clearly illustrates that a PC or streamer provides a DAC with a file that is always absolutely correct and error-free. The file remains correct, even when it travels thousands of kilometers and through thousands of connections from Qobuz/Tidal servers reach to any streamer or PC.

Thus, it is unclear why, from a theoretical and technical perspective, a streamer or PC, which does not even need to extract any file from any disk medium (like CD/SACD), should require an additional interface to deliver a file to a DAC that is already digitally correct through the I2S protocol, which doesn’t even provide a standard for external interconnection cables or error correction!

All measurements we can perform to support one interface over another (USB vs. I2S) should only confirm that if there are differences, they are solely induced by the DAC. Therefore, we would only be measuring a DAC that is not functioning correctly if we find better measurements using the I2S interface.
In fact, at most, I2S should match USB because the DAC always receives a digitally correct file in its bit-by-bit sequence through the USB port.

A properly designed DAC with balanced outputs should be able to address ground loop problems and other issues present on the USB port, and even for system-related noise problems. After all, no one has ever seen a file copied with errors to a USB drive due to these problems! Expecting a PC/streamer to be equipped with this additional interface so that the DAC "works better" remains a technological folly, because if such an interface were indeed necessary for some improbable reason, it should always and only be implemented within the DAC itself, rather than requiring the streamer to provide it.
Moreover, within its architecture, a modern DAC already utilizes the I2S bus,
but this its another deal!
 
We know that I2S was theoretically conceived to improve upon the synchronous S/PDIF protocol by extracting the clock and transmitting it separately.

That part is also a myth.

I2S was mainly conceived to make the integrated circuits used in audio gear simpler and cheaper, while still keeping the performance uncompromised.

The I2S protocol makes it significantly easier to design the part of the chip that receives and transmits the signal(s), but requires two additional PCB traces. A compromise well worth it.

Yes, S/PDIF was designed to repurpose a connector almost as old as reproduced audio itself, but integrity of the data was never really an issue.

I can only imagine that the idea of "improvement" was spun in the minds of the people at PS Audio, and they just made it a thing. They probably thought of how RGB+sync offers actual improvement in analogue video transmission, and they completely ignored the fact that digital audio transmission is a very different beast.
 
That part is also a myth.

I2S was mainly conceived to make the integrated circuits used in audio gear simpler and cheaper, while still keeping the performance uncompromised.

The I2S protocol makes it significantly easier to design the part of the chip that receives and transmits the signal(s), but requires two additional PCB traces. A compromise well worth it.
I was referring to I2s as an external interconnect interface (PS Audio) as an alternative to S/PIDF which is a synchronous protocol
 
I can only imagine that the idea of "improvement" was spun in the minds of the people at PS Audio, and they just made it a thing.
Who? PS Audio didn't invent I2S or its use as an intercomponent connection.
Wikipedia:
"There is no standard interconnecting cable for this application. Some manufacturers simply provide three BNC connectors, an 8P8C ("RJ45") socket or a DE-9 connector. Others like Audio Alchemy (now defunct) used DIN connectors. PS Audio, Musica Pristina and Wyred4Sound use an HDMI connector.[7] Dutch manufacturer Van Medevoort has implemented Q-link in some of its equipment, which transfers I²S over 4 RCA connectors (data, MCK, LRCK, BCK)."
 
He didn't imply that they invented it, only that they inferred that it was superior using poor reasoning.

When he says "made it a thing" he doesn't mean they made I2S, he referring to the idea of it being better.
Unfortunately, that is not entirely correct, as is @widemediaphotography conclusion.

PS Audio were neither the first nor the only ones to experiment with external i2s or i2s over LVDS transmission, just the best known.
The approach was the same for everyone: higher bandwidth over 192 kHz, less jitter with SPDIF and less jitter/contamination with USB. The other argument was that many source devices had to convert i2s to SPDIF at the output anyway and back to i2s in the DAC. Not having to convert the data stream twice but transmitting it directly should be understandable to any technically savvy person. In practice, that doesn't matter anymore. We mustn't forget that USB was really bad as an audio interface back then, and SPDIF wasn't at the same level as it is today.

The funny or ironic thing is that the efforts of PS Audio and the emergence of high-quality DDCs from other manufacturers, together with Amanero and XMOS, have led to an enormous development of USB audio transmissions. The current level can only be described as excellent.
The jitter reduction in SPDIF has also developed enormously since then. This can be seen in the extremely reduced jitter over SPDIF, for example in the SMSL DACs with the self-developed CK-03 clock and transceiver replacement.

So we owe part of the current state of development in this area to this advance of external i2s.
 
Do we, I can’t see any benefit in I2s ( apart from its original intended use) some manufacturers added it simply pandering to audiophile nonsense.
Keith
 
Do we, I can’t see any benefit in I2s ( apart from its original intended use) some manufacturers added it simply pandering to audiophile nonsense.
Keith
It's not about i2s, but about the further development of USB audio through the whole DDC/i2s story. A lot of development work was put into reducing the jitter of the USB and SPDIF interfaces for these devices.
These further developments then gradually spilled over into the DACs.
However, we are talking about further development in the last 20 years, with the biggest leaps in development taking place from 2010 to 2020.
Many people didn't notice this at the time because USB audio wasn't such a big issue back then, others are simply too young for it.
For the same reasons, many people don't know (or don't want to know) that USB and USB audio were once really bad and full of problems.
 
Who? PS Audio didn't invent I2S or its use as an intercomponent connection.
Wikipedia:
"There is no standard interconnecting cable for this application. Some manufacturers simply provide three BNC connectors, an 8P8C ("RJ45") socket or a DE-9 connector. Others like Audio Alchemy (now defunct) used DIN connectors. PS Audio, Musica Pristina and Wyred4Sound use an HDMI connector.[7] Dutch manufacturer Van Medevoort has implemented Q-link in some of its equipment, which transfers I²S over 4 RCA connectors (data, MCK, LRCK, BCK)."

True. Although, I see PS Audio as the company that has most "aggressively" been pushing for I2S to be the next big thing, and ultimately being responsible for the myriad of DACs from Chinese brands having some sort of HDMI port with I2S support. But I could very well be giving PS Audio way too much credit.
 
True. Although, I see PS Audio as the company that has most "aggressively" been pushing for I2S to be the next big thing, and ultimately being responsible for the myriad of DACs from Chinese brands having some sort of HDMI port with I2S support. But I could very well be giving PS Audio way too much credit.
Audio Alchemy, Camelot Technology and Sonic Frontiers, all now defunct, were the ones who move I2S into the intercomponent sphere. I wrote two DIY articles about it in AudioAmateur 3/1995 and 1/1997.
 
Audio Alchemy, Camelot Technology and Sonic Frontiers, all now defunct, were the ones who move I2S into the intercomponent sphere. I wrote two DIY articles about it in AudioAmateur 3/1995 and 1/1997.
Audio Alchemy made a fairly inexpensive set of devices with DSP and even jitter reduction before these things became sexy, with mini-DIN I2S connectors. Camelot tech came later and put a premium price on some similar technology. I still have some of the devices made by both somewhere in storage.
 
There is no need to involve AMANERO or anyone else developing specific USB Audio interfaces, because from my point of view, they have no reason to exist. The philosophy behind using a PC, streamer, or Raspberry Pi is to provide to the DAC with a clean DATA file input, which is already inherently jitter-free, directly through a USB port or any port a PC uses to exchange data with peripherals.

How to handle these files at an audio level is solely the responsibility of the DAC, and the need to add another interface, such as I2S or USB Audio, inside a PC seems completely redundant to me.

From the USB output of the source, there may be unwanted signals that affect the audio, caused by things like ground loops or system issues. However, it is always the exclusive task of the DAC to implement the best method to eliminate all unwanted signals or reduce them to a level of clear irrelevance.

A lossless audio file travels thousands of kilometers and crosses thousands of connections before always arriving intact at my PC or streamer. Are we really suggesting that a decent DAC can't filter out all unwanted signals? USB data is more than enough—DACs should be capable of doing what DACs are meant to do!

For example, CHORD Audio converts the USB signal to optical to cleanse it of everything that isn't DATA. I'm not saying that Chord's approach is the right one or that they make the best DACs, but even my SMSL DO300 has no problems with USB, even when using PCs that are 30 years old!

In summary, we should only provide a DAC with a lossless file sent from the software player. It should be the DAC's job to make it sound good, not the interfaces on the PC or streamer side;)
 
  • Like
Reactions: EdW
It should be the DAC's job to make it sound good
dunno... should be the DAC's job to accurately reproduce the 'described' digital input signal.
Whether or not that 'sounds good' to someone is another matter.
 
Last edited:
Back
Top Bottom