15 USB/SPDIF converters shootout

[elcorso]

Are there any DACs at this time, at any price, whose USB input is not bettered by a first-rate external USB to S/PDIF converter?

 

The exaSound e20 Mark II has this kind of extraordinary USB input.

Tried a lot of USB interfaces by his SPDIF input and the bass 'sounds' like bump and not deep enough: Much more jitter?

 

The Playback Designs (Series 5) USB input is only bettered by his own PD USB-X USB Interface trough his HD-15 cable/connector.

 

According to Paul Hynes the BNC in the Metrum Octave DAC could be replaced with a good and real BNC, since it's 75Ohms internally (by in series resistors).

 

Roch

[elcorso]

I'd think that you would have to know that their specs are inaccurate, otherwise I suppose it could be considered libel if those assertions were untrue...

 

The 110s have a thin conductor though polymer components to maintain 75 ohms through 1GHz. The metal bodied units only maintain this characteristic impedance through 200MHz, I believe is what they are spec'ing.

 

No one in communications will construct and athena, cable or repeater based on RCA, but (good) BNC. And now in the audio hobby we are talking about DSD.

 

There are another things besides pure specs, like contact tightness, isolation (from noise & humidity), etc. Even good 75 ohms cables are different from 50 ohms.

 

Manufacturers are very stingy?

 

Roch

[Audio_ELF]

Errrr - exactly what DAC's you know 100% for sure do I2S internally?

 

I hate to tell you but most in fact do SPDIF internally. They use an inexpensive chip that is available that has both analogue and SPDIF out - but no I2S and simply feed it into the SPDIF. USB in the past was an after thought which is why external converters were usually universally way above internal implementations. That is slowly starting to change - but as of now don't expect any but the very best USB DAC's to convert to I2S.

Bill ... that description sounds very dated.

 

Most DACs offering 24/192, even the lower priced ones, now use the XMOS chip. The reference design (which I imagine most manufacturers would use) for this offers both i2s and SPDIF output. Prior to this, the 24/96 async devices often used the TAS1020B which also output i2s as it's native option. It's only devices with adaptive (non-async) USB - usually limited to 16/44.1 and 48 - that would use an "inexpensive chip that [..] has both [only] analogue and SPDIF out".

 

Thats not to say that they DON'T use SPDIF internally, but it's not down to the chip's capability if they do.

 

Eloise

[Audio_ELF]

No one in communications will construct and athena, cable or repeater based on RCA, but (good) BNC. And now in the audio hobby we are talking about DSD.

 

There are another things besides pure specs, like contact tightness, isolation (from noise & humidity), etc. Even good 75 ohms cables are different from 50 ohms.

 

Manufacturers are very stingy?

Straying off topic and possibly controversial, but it would be interesting to know IF DACs where more attention is paid to proper 75ohm connections, etc (as per Roch's description above) if cables - where they are also constucted to proper specification rather than being "boutique audiophile" - and USB/SPDIF converters are less relevant?

 

Eloise

[Superdad]

Errrr - exactly what DAC's you know 100% for sure do I2S internally?

 

I hate to tell you but most in fact do SPDIF internally. They use an inexpensive chip that is available that has both analogue and SPDIF out - but no I2S and simply feed it into the SPDIF. USB in the past was an after thought which is why external converters were usually universally way above internal implementations. That is slowly starting to change - but as of now don't expect any but the very best USB DAC's to convert to I2S.

 

Thanks

Bill

 

Sorry Bill, not only is Eloise right about all the current and past async-USB chip implementations being I2S out, but virtually all the adaptive-USB audio input chips--even the cheapest $3 TI PCM2706 from 2004--provide I2S output. To do it with S/PDIF you have to ADD parts. So other than maybe some ancient pre-existing DACs where the wanted to add in USB in a hurry and were really sloppy...

[bhobba]

Sorry Bill, not only is Eloise right about all the current and past async-USB chip implementations being I2S out, but virtually all the adaptive-USB audio input chips--even the cheapest $3 TI PCM2706 from 2004--provide I2S output.

 

The Burson HD160, for example, used a Tenor TE7022 which converted to SPDIF and internally fed it to its SPDIF input - and that was a pretty common way to do it. Their new Conductor however from what I can gather does it via I2S using the new async Tenor chip.

 

It is changing but I am pretty sure other than the new stuff like the Conductor they all converted to SPDIF first - it was easier.

 

I know a tech who has pulled a lot of this stuff to pieces. If I remember will check with him tomorrow and see what he says.

 

Thanks

Bill

[Audio_ELF]

The Burson HD160, for example, used a Tenor TE7022 which converted to SPDIF and internally fed it to its SPDIF input - and that was a pretty common way to do it. Their new Conductor however from what I can gather does it via I2S using the new async Tenor chip.

 

It is changing but I am pretty sure other than the new stuff like the Conductor they all converted to SPDIF first - it was easier.

 

I know a tech who has pulled a lot of this stuff to pieces. If I remember will check with him tomorrow and see what he says.

The Tenor TE7022 can natively output i2s (I believe) so if Burson converted first to SPDIF its not the fault of the chip!

 

Eloise

[Audio_ELF]

The Burson HD160, for example, used a Tenor TE7022 which converted to SPDIF and internally fed it to its SPDIF input - and that was a pretty common way to do it. Their new Conductor however from what I can gather does it via I2S using the new async Tenor chip.

 

It is changing but I am pretty sure other than the new stuff like the Conductor they all converted to SPDIF first - it was easier.

 

I know a tech who has pulled a lot of this stuff to pieces. If I remember will check with him tomorrow and see what he says.

The Tenor TE7022 can natively output i2s (I believe) so if Burson converted first to SPDIF its not the fault of the chip!

 

Eloise

[bhobba]

The Tenor TE7022 can natively output i2s (I believe) so if Burson converted first to SPDIF its not the fault of the chip!

 

You berlieve? In fact it can - but with a bit of mucking around - the standard board didnt have the pinouts which is why the manufaturers I am aware of didnt use it - they simply connected the output it had to the SPDIF.

 

Its only of recent times more care has been taken with USB and DAC's like the Conductor are appearing that do direct I2S. Eventually all will do it that way as well as async but right now there are quite a few about that took the easy route.

 

Thanks

Bill

[Superdad]

Its only of recent times more care has been taken with USB and DAC's like the Conductor are appearing that do direct I2S.

 

Recent times, like since about 2008? I will venture that the majority of (USB input equipped) DACs released since then take the I2S off the chip (even with the old adaptive 2706 series) and skip S/PDIF.

This is not at all to say that their USB implementations sound as good as what the same DAC offers via its S/PDIF input--there is a surprising amount of engineering involved in doing a great sounding USB>I2S board.

 

There are about a half dozen async-USB>I2S boards available to the DIY crowd, each by some very serious engineers, and yet they vary greatly in their sophistication and sound. BTW, very few firms use the Tenor chips anymore; most of the best are using the very fine and programmable XMOS.

Here is a link to the 180(!) page thread on a popular one from Romania.

XMOS-based Asynchronous USB to I2S interface - diyAudio

 

It is the one I have installed in my multi-power-supplied, NOS PCM1704K DAC with discrete output stage. I use Audirvana Plus with iZotope to upsample to 24/176.4

[junker]

I'm not disputing that BNC is superior to RCA for S/PDIF, but many consumer level products only offer RCA. And in that case I'd prefer to have connectors as close to 110 ohms as possible and a cable length >= 1.5m. The WBT also has pretty contact pressure and the threaded metal connector that helps with shielding.

 

I'd think that the performance advantages with BNC are probably more critical than at the bandwidths we are using for digital audio. I'm always pleased when I see BNC being used, however, such as in the Berkeley DAC converter and Alpha 2 DAC.

 

No one in communications will construct and athena, cable or repeater based on RCA, but (good) BNC. And now in the audio hobby we are talking about DSD.

 

There are another things besides pure specs, like contact tightness, isolation (from noise & humidity), etc. Even good 75 ohms cables are different from 50 ohms.

 

Manufacturers are very stingy?

 

Roch

[barrows]

This discussion is getting ridiculous!

1. True 75 ohm BNC connections are far superior to using RCA, but, the line has to be properly terminated and designed to maintain as close to 75 ohm impedance as possible, this is not trivial, and there will always be some degree of impedance mismatch (and hence some reflections), but no one should ever use RCA connectors for SPDIF.

2. I am not aware of any DAC which converts USB to SPDIF internally, this is nonsense. Even most first generation USB DACs like the original PS Audio DL-III had I2S output from the (cheap and terrible) TI adaptive USB receiver chip to the ASRC and then on to the DAC chip.

 

What matters for USB quality: everything, all the details matter, including some very important factors like PCB layout. This is high speed PCB design where every little detail matters. For the "features" one would want to have in a USB interface here are some points:

asynchronous operation, separate power supplies (best would be separate transformers) for oscillators/output driver (clean side) and processor (dirty side), isolation between processor/computer side and oscillator/output side, re-clocking of the data stream on the output side, very low phase noise oscillators (Crystek CCHD series, NDK, or better), short and impedance controlled I2S connections between the USB receiver and DAC chip.

Those are just some of the things to look for in a really good, internal USB receiver solution. But, even if a USB interface has all of those things, it still could suck if the features are poorly implemented: say power supplies with noise, or poor PCB design/layout.

I do not know the exact USB implementation in a lot of DACs, but I do know that Ayre, Wavelength, and Aesthetix do it right as above.

Also, consider that a DAC with a single input actually has a performance advantage over DACs with multiple inputs. Multiple inputs require some way to switch them, usually done by either a relay, or FET switch (switch in a chip IC), or a DIR with mutiple inputs. Either way, any kind of switch will distort a high speed digital signal.

[ar-t]

I suppose as much as I believe you... Do you have a legitimate reason to not?

 

http://www.wbtusa.com/pdf/0110.pdf

 

Yes, because the laws of physics don't have a special exemption, wrt high-end audio.

 

Anyone can say anything in their poop sheets. They can even print it, on the part itself.

 

I have a nice collection, of "110 ohms Digital AES/EBU" cables, that despite being marked on the jacket, are not 110 ohms. All made by companies that you guys would recognize.

 

And a handful of USB cables that are supposed to be 90 ohms, but are not.

 

I even have a collection of mic cables, some made by Belden (who is big enough to know better) that do not measure what the catalog says they are.

 

Ah............but if you dig deeper...............they show the conditions they measure them under. (Which is to attach one of the twisted leads to the shield, which means it is not longer a twisted pair, with a shield over the pair.) So, if you measure it in the way that no one uses, then yes, it measures what the catalog says. But, when measured the way they are actually used then they do not meet the catalog spec. (Sometimes they show data for that configuration, but they use the other one for the published value. HUH?)

 

I can say I am Emperor of the North Pole, simply because I wrote it on the side of a water tower. Doesn't make it so.

[ar-t]

Also, here is the international patent on the design:

 

WOW! It has a patent. Well, that certainly settles it.

 

I have a buddy at TI who got a patent on something that most reasonable folks would call a bug. He turned it into a feature, and got a patent for it.

 

Someone tried to 'splain to me that WBT has some special construction method, where they do something strange wrt to the outer conductor, and when you use the male and female mated together then it measures like 75 ohms.

 

"So what happens when you mate it with a conventional RCA, like 98% of the people are going to do? I'll tell you: the pair will now measure 35-40 ohms, just like every other RCA pair."

 

That was the end of that conversation.

 

I'm not the one making claims to go counter to the laws of physics. Folks like Canare and WBT have the burden of proof on their shoulders, since they are making them. When I see someone who is independent that can show valid test data...............well, when that happens, we'll talk about it. Until then, it is just codswallop.

[ar-t]

This discussion is getting ridiculous!

 

Yes. Sorry to have contributed to it. But, sometimes you have to tell it like it is.

 

I'm not touching the internal SPDIF part. That makes my brain hurt. Oy vey.

 

OK, I'm done. Enjoy!

[CatManDo]

WOW! It has a patent. Well, that certainly settles it.

 

The patent only protects a very specific aspect of the connector, defined by the claims. (page 8 of the PDF)

 

https://data.epo.org/publication-server/pdf-document?pn=1570554&ki=B1&cc=EP

 

And of course the patent examiner does not check if this plug is actually better than a conventional one.

[ar-t]

That isn't their job. (Whether they do their proper job is debatable.)

 

Decades ago, I was an AES member. The only benefit I saw was the patent reviews, they periodically published. One particular contributor was equally skeptical. One of his reviews covered a speaker, in a polystyrene enclosure. His conclusion was the patentable feature was it used a 1/4" phone jack.

 

Don't throw patents out, as proof of something. Unless you want to degrade your position.

 

OK.......now I am out. Peace, man.

[junker]

So you are saying you've personally tested the WBT connector to be non 110 ohm? If so, then what characteristic impedance did you find?

 

My point with WBT is that they clearly give the spec and I would like to see something specific and credible that says otherwise. Just saying it isn't so because you happen to have a loom of substandard non-spec cables doesn't really address the specific question about the WBT 110's characteristic impedance.

 

And again most consumer products are unfortunately using RCA, for better or worse. So are you saying all RCA-RCA digital cables are equally unsuitable or reliable in typical digital audio environments?

 

And PS: is your buddy in Richardson? I worked at Kfab a while back. A lot of fun in Dallas.

[dallasjustice]

Let's just say Pat knows what he's talking about. But to expand on this point, I want to give you a single data point about the impact of impedance mismatch.

 

Pat came over to my listening room one time. Of course, I fed him grilled chicken and was forced to view his family photo album. (I guess it's sort of like facebook for people older than say 30.) Pat brought with him some converters for me to test by listening. The purpose of the test was to see whether I could subjectively hear any differences in any of his converters. Of course, i couldn't. But as part of the test we were intentionally using an impedance mismatched connection; bnc out and rca s/pdif into my DAC. Pat interposed a "pad device" (my term, not his) that had variable levels of attenuation. The idea was to figure out whether any amount of attenuation on the impedance mismatched s/pdif signal would help to reduce the reflections to the point where I could hear a difference. Pat's "pad device" had a button that he could press to change the attenuation very quickly so there were only seconds between plays. Bottom line, I couldn't hear any significant differences with increased amount of attenuation. I should have been able to hear "less jitter" due to few reflections, but I couldn't. Maybe my system sucks; Maybe I am deaf; who knows? Or just maybe there are some gears made by smart engineers that reduce the effect of these known cable reflections due to impedance mismatch. I am not going to say which engineers know how to do this better than others. But I can guess.

 

So you are saying you've personally tested the WBT connector to be non 110 ohm? If so, then what characteristic impedance did you find?

 

My point with WBT is that they clearly give the spec and I would like to see something specific and credible that says otherwise. Just saying it isn't so because you happen to have a loom of substandard non-spec cables doesn't really address the specific question about the WBT 110's characteristic impedance.

 

And again most consumer products are unfortunately using RCA, for better or worse. So are you saying all RCA-RCA digital cables are equally unsuitable or reliable in typical digital audio environments?

 

And PS: is your buddy in Richardson? I worked at Kfab a while back. A lot of fun in Dallas.

[junker]

Cool info. Thanks!

[DanRubin]

This is great stuff.

 

I'm still a bit puzzled about something:

 

1. If USB implementation is super critical (barrows and others), and

2. DAC built-in USB is almost always USB to I2S, and

3. External converters are typically used as USB to S/PDIF; then

4. Does it follow that DAC built-in S/PDIF to I2S implementation is much easier and nearly universally better than #2 above

 

I hope this makes sense. It seems that USB is the stumbling block in all of this. Or is it the computer that's the stumbling block, such that direct S/PDIF out from the computer would also be a big compromise?

 

Dan

[barrows]

Dan:

 

"1. If USB implementation is super critical (barrows and others), and

2. DAC built-in USB is almost always USB to I2S, and

3. External converters are typically used as USB to S/PDIF; then

4. Does it follow that DAC built-in S/PDIF to I2S implementation is much easier and nearly universally better than #2 above"

 

Much easier for who? It might be easier for the music lover to not have to worry about adding an external converter... But, it is not easy or trivial for a manufacturer to build a really good USB interface into a DAC (although it is getting easier for them as they learn how to do it right, and since the XMOS processor came on the scene). Universally better, no, because manufacturers do not always build in a "perfect" USB interface into their DACs. I will say this: if a manufacturer builds in a technically "perfect" USB into their DAC, then that DAC will perform better (technically) using its internal interface than it will by using any external converter which uses SPDIF.

 

"I hope this makes sense. It seems that USB is the stumbling block in all of this. Or is it the computer that's the stumbling block, such that direct S/PDIF out from the computer would also be a big compromise?"

 

I would not consider any one thing a "stumbling block", everything matters, every little detail of how a DAC is engineered and built. Every detail is a stumbling block, there are thousands of places to get things "wrong". There is also another consideration when people start talking about using external converters: some people may actually prefer the sound of something which is actually technically flawed (increased jitter), as a certain spectrum of jitter components may have a euphonic effect in their system context. One example of this: say a person's system actually has a flaw which causes an emphasis on perhaps transient attacks, combined with a bit of upper midrange boost-this person might prefer a digital source which clouds over details just a bit, which could result in moderating the problems already inherent in the analog side of their system, making things sound "better" from a listening perspective.

 

Unless one can actually measure jitter components (very, very difficult to do) one is going to have a hard time proving the superiority of one interface to another, all one can really say is that a given interface sounds better in one's own system. Hence, I think subjective shoot outs of USB interfaces as in this thread should only be considered as very rough guidlines, and one must do their own auditioning to fiond what will work in their system.

[DanRubin]

Barrows: I can see that I wasn't clear. What I intended to say is that doing built-in S/PDIF to I2S is much easier for DAC manufacturers to do a good job of than it is for them to do built-in USB to I2S. I say this because most of the USB DACs that also offer S/PDIF input seem to sound better when fed S/PDIF from an external converter than when fed USB directly. So it follows that their S/PDIF to I2S is (must be) better than their USB to I2S. And this despite the fact that S/PDIF is, as we all know or thought we knew, an antiquated POS protocol.

 

Which leads to my point about USB as the stumbling block. There must be something about receiving and converting USB specifically that makes it the great challenge here because, apparently, once you successfully convert it to S/PDIF, you're in pretty good shape with most DACs -- despite the limitations of S/PDIF. Not to mention how much variation there is in the "successfully convert" part.

 

I get that "everything matters." But it seems to matter more with USB than with S/PDIF or Firewire or ...

 

Dan

[barrows]

Dan:

 

"Barrows: I can see that I wasn't clear. What I intended to say is that doing built-in S/PDIF to I2S is much easier for DAC manufacturers to do a good job of than it is for them to do built-in USB to I2S. I say this because most of the USB DACs that also offer S/PDIF input seem to sound better when fed S/PDIF from an external converter than when fed USB directly. So it follows that their S/PDIF to I2S is (must be) better than their USB to I2S. And this despite the fact that S/PDIF is, as we all know or thought we knew, an antiquated POS protocol."

 

I disagree. Most DACs (at least the ones I like) do not sound better from their SPDIF input at all, regardless of the SPDIF source. I also feel it is quite difficult to make an SPDIF input perform well. Take a look at the jitter measurements made by John Atkinson (stereophile.com) on DACs. Most DACs he has tested recently show better performance (less jitter) from their USB inputs.

 

"Which leads to my point about USB as the stumbling block. There must be something about receiving and converting USB specifically that makes it the great challenge here because, apparently, once you successfully convert it to S/PDIF, you're in pretty good shape with most DACs -- despite the limitations of S/PDIF. Not to mention how much variation there is in the "successfully convert" part."

 

Once again, I totally disagree. Most of the good DACs around these days perform better via their USB inputs. But, certainly some manufacturers have not yet made the effort to have a really good internal USB interface, perhaps those are the DACs with which you are more familiar. I do think that there are plenty of people using external USB-SPDIF converters who are actually getting worse performance (but preferring the resulting sound in a flawed sytem) than if they were using the internal USB interface within their DAC. Generally speaking, excellent USB performance is available using the XMOS processor in a good implementation, and some folks do their own proprietary USB interface as well. Their is also a very good pre-built USB interface board available to OEMs from ABC-PCB in Switzerland which uses XMOS, any DAC manufacturer could use this board to get excellent USB performance, and the board can be taken to even higher levels of performance through external clocking.

I suspect that many of the people who prefer using an external converter are actually prerring the sound of higher jitter levels...

[Paul R]

No one in communications will construct and athena, cable or repeater based on RCA, but (good) BNC. And now in the audio hobby we are talking about DSD.

 

Oh I don't know - give me a nice ST Fibre connection any day.... (grin)