The State of USB Audio by Alan Taffel

[manisandher]

My understanding is that JA feeds a 11.025kHz signal at –6dBFS, sampled at 44.1kHz with LSB toggled at 229Hz. Now, the harmonics above and below the fundamental 11.025KHz look pretty consistent and predictable to me... a very nice mirror image actually! But these aren't jitter-related, right?

 

But for sure, the Wavelength's high noise floor doesn't help in making a comparison.

 

Look, I'm certainly no expert in jitter measurement and would love to learn more... including whether it really matters after a certain point. I too suspect that other things become far more important., e.g. the output stage etc.

 

Mani.

 

[idiot_savant]

on the stereophile website:

http://stereophile.com/features/1208jitter/

 

I do agree people do get obsessed about jitter, however it is something that you can measure, and therefore improve.

It's also the most obvious cause of differences between interfaces - for the reviewer to establish that S/PDIF and firewire are superior, carrying the same data, we know for a fact that the USB interfaces he is testing have much higher jitter than is available, so why not even mention that better implementations exist?

 

The interesting thing about the Wavelength jitter ( that you can see ), is that it is mains related ( therefore probably nothing to do with the interface ).

 

your friendly neighbourhood idiot

 

[manisandher]

Thanks for the link - very interesting.

 

The most salient comment being: "There is no consensus about what levels of jitter in a digital product's output are acceptable..." - JA.

 

It was interesting to read that some 'experts' believe that jitter levels should be below 100ps for 16bit and 10ps for 24-bit audio! News to me...

 

So, was Alan's aversion to adaptive USB due to the effect jitter was having on the sound? Is async USB really that much better than the adaptive USB in say the Benchmark (which measures pretty well according to JA)?

 

If the answers to these questions are 'Yes' and 'No' respectively, then I think Alan's conclusion was fair.

 

But perhaps we'll never know...

 

Mani.

 

[The Computer Audiophile]

Alan's conclusions about individual products are totally fine with me. If he measures/looks at jitter specs and compares specific products I'm cool with that. But, it's his comments below that I can't see any justification for because they are inclusive of products he has no experience with. It's similar to listening to Wilson Sasha speakers and saying they are the best and no other speakers are currently as good, even though you haven't listened to any Magico speakers. It just doesn't add up.

 

(taken from Clay's earlier post)

 

"Q: How does USB sound?

A: I regret having to report that, at this stage in its development, USB does not rise to the level of an audiophile-quality interface"

 

Note: this last bit (USB does not rise to the level of an audiophile-quality interface) was the penultimate statement of the article, being shown in bold red letters.

 

"Q: How does USB compare to S/PDIF?"

A: There IS no comparison. USB sounds much worse"

 

note: the emphasis on 'IS' was the authors

 

[Wavelength]

Mani,

 

FYI... for the second time.. CG is probably the person anywhere that I asked about jitter measurements. He does this for a living.

 

I would take what he says as fact. He is one of the people I acknowledged in my design for the Cosecant because his knowledge about jitter, power supplies etc are beyond words.

 

Thanks

Gordon

 

[manisandher]

Can you ever make a generalization without trying everything?

 

Perhaps the 'fundamental physics' behind this situation help?

 

(For example, I would happily attest that, although having driven only a few, mid-engined cars are more nimble than front-engined ones - because of my experience, but also because the 'physics' tells me so... though I've never driven a McLaren SLR.)

 

The closest I can get to the 'physics' of this situation are the jitter measurements in Stereophile, where spdif 'beats' both adaptive and async USB.

 

Mani.

 

[manisandher]

Gordan,

 

I took absolutely no issue with CG's initial post... except the reference to the stereophile measurements.

 

So I should take as 'fact' that the Wavelength's async USB jitter figures are better than the Benchmark's spdif jitter figures, as measured by JA in Stereophile?

 

A simple 'yes' or 'no' would suffice...

 

Mani.

 

[cfmsp]

 

"The closest I can get to the 'physics' of this situation are the jitter measurements in Stereophile, where spdif 'beats' both adaptive and async USB."

 

To stick with your example above - i.e., nimble-ness of mid engined cars - jitter measurements would seem to be the relative equivalent to the Road & track skid-pad measurements (do they still do these? I've not read R&T for over 25 years) of yore. They might provide a clue, but no one (I know of) would rely on them as anything more than that as to the relative 'nimble-ness' of a car's overall handling capabilities.

 

YMMV, of course,

clay

 

 

[idiot_savant]

Hi Mani,

 

you keep returning to this assertion that "SPDIF beats Async USB" measurements, despite people attempting to explain to you that the benchmark uses an ASRC, which makes it impossible to measure jitter in the same way as with devices that don't - as I asserted the Julian Dunn technique measures symmetric sidebands around the fundamental, excited by the 229Hz square wave. The benchmark has lots of tones mingled in with the square wave ( if you closely examine the FFT the spikes from the square wave will appear exactly 578Hz apart ). These MAY be jitter related products, and they almost certainly will not be symmetrical, and as to whether or not they are preferable in the audio output is another question.

The Wavelength measurements are very hard to compare, because the noise floor is too high, but we CAN see mains related spuriae, which certainly have nothing to do with async, adaptive, phase of the moon etc.

Perhaps when one of the companies who use a solid state output with async is measured ( dCS, Ayre ) we can have a hope of meaningful comparisons.

 

As for car analogies, the article is the equivalent of the author test driving a selection of yugos, and declaring that all european cars are slow and unreliable, in spite of claiming to know of the existence of porsche and ferrari,

 

your friendly neighbourhood idiot

 

 

 

[The Computer Audiophile]

Speaking of Asynchronous USB DACs, by the end of the week I should have the dCS async USB Puccini / Paganini combo, to add to my current review stock of Ayre QB-9, and Wavelength Proton :~) Quite the hat trick of Async technology.

 

Maybe I should do a "State of USB DACs" article and pronounce all USB DACs as fabulous and that the technology is so much better than anything else despite the fact that I have other adaptive USB DACs in my listening room that I'm choosing not to listen to before the review gets published. That would be quite comical.

 

[manisandher]

I_S,

 

You said, "... the article is the equivalent of the author test driving a selection of yugos, and declaring that all european cars are slow and unreliable, in spite of claiming to know of the existence of porsche and ferrari."

 

Very good!

 

Yes, I'm well aware that the Benchmark uses ASRC. What I didn't know was that this then invalidates any jitter measurement comparison with DACs that don't use ASRC. Interesting stuff and thanks for the explanation.

 

Mani.

 

[rlodad]

I know what my most important question regarding this is: How do the Ayre, Wavelength and DCS pieces sound with USB inputs compared to your (and many other "experts") new reference- the Berkeley. We have all price ranges, get to it Chris- inquiring minds want to know! (And if you have too much work to do, I be happy to listen to any/all of them for you...)

 

:^)

 

 

 

[The Computer Audiophile]

I love it, a sense of humor pops up out of all this jitter :~)

 

[omahapianist]

I didn't want to comment until I read the article and while Alan is certainly entitled to his thoughts and opinions, I'm wondering if he did a slight disservice in describing the sound of USB-streamed audio. Most of the time, the differences between two components are subtle at best but you sure wouldn't know that from the words and phrases chosen to describe USB audio:

 

"Synthetic (and used again as "a telltale synthetic glaze"), cardboard-flat, "plastic" quality, sloppy timing, pale, washed-out, shrill, etc."

 

Dem is fightin' words! :-)

 

I haven't read such strong descriptives about a device since the heydey of audio! The disservice here is that there are a large number of audiophiles who enjoy the wonderful sound of USB-connected DAC who may now be a little "spooked" as to whether they are hearing what Alan has noted. So what does this say to those who may (or may not have) a somewhat similar set-up as highlighted by Alan who thinks their sound is [insert positive, uplifting terminology].

 

1. They have no idea that such audio/(digital?) grunge exist in their connection (ignorance is bliss, indeed!).

 

2. They read a previous article heaping praises on a particular USB unit and cable and bought into it convinced that audio nirvana has been achieved (the placebo effect is bliss, indeed!).

 

3. The article has triggered a stage 3 audiophilia nervosa attack causing everything to become misaligned and distorted which is on the same scale as listening to music on the headphones supplied with a ipod, listening to music on a mass-produced receiver with the tone controls turned up to max, and eating packaged soup.

 

I did like the car reference described earlier but to place USB-audio in the real world of listening, imagine comparing the performance characteristics of a Porsche 911 Turbo to a Ferrari F355. Both cars are supreme in what they do but let's say the Turbo rips 0-60 in 3.9 sec to the F355's 4.2 sec run. Now, describe how "washed-out", pale, and "plastic" the Ferrari is just because it's "slower". Pretty ridiculous!

 

The fact is, both cars achieve very similar, REAL WORLD, performance numbers. However, the Porsche may do it in a slightly more relaxed manner that's driver-friendly when needed in parking lot situations but can turn into a raging bull at the drop of a throttle. The F355 may do it in a more, driver-focused method that requires a steadier hand but rewards the driver with road-handling nirvana and hair-raising acceleration.

 

It's about preference, not "right or wrong". Alan's choice of words, while mildly complimentary of USB at times, do not resonate as strongly as his non-complimentary ones. The sound difference between USB and S/PDIF is NOT dramatic. But, if he was describing the difference between, say, a $2,500.00 pair of bookshelf speakers and the speakers built into most laptop computers.............

 

I mean, what does "plastic" sound like??

 

Randall

 

[icebreaker]

"As for car analogies, the article is the equivalent of the author test driving a selection of yugos, and declaring that all european cars are slow and unreliable, in spite of claiming to know of the existence of porsche and ferrari"

 

Sometimes analogies really are useful. Well said! :>)

 

[johnnyturbo]

seems to me we - the CA community - have done an admirable full dissection of the taffel article. based upon reading the CA responses, i think we can agree on several conclusions:

 

1. the taffel article was ill conceived and poorly researched

2. the article was poorly written

3. the editor, robert harley, did not do his job: he should have sent the draft article back to taffel for a massive re-write or, alternatively, 'deep-sixed' it in the "seemed to be a good idea at the time" pile

4. taffel and harley share and deserve the blame for publishing the article and should take responsibility for their journalistic failure. it is fair to say that their reputations have suffered and that of TAS has taken a solid hit amidship.

 

perhaps it is now time to put l'affaire taffel/TAS 'to bed' and to move on and followup on the host of substantive posts submitted as part of the thread.

 

[cfmsp]

 

I like analogies in general, and this one in particular....and though I've never considered any of these, if I had a Benchmark DAC 1, or the Audio Research, or the Bryston, it might ring a little hollow.

 

 

 

 

 

 

 

[icebreaker]

Using the USB on the Bryston DAC is like never getting your Ferrari into gear.

 

[The Computer Audiophile]

"perhaps it is now time to put l'affaire taffel/TAS 'to bed' and to move on and followup on the host of substantive posts submitted as part of the thread.

 

Maybe you're right. We've given Taffel more press than deserved. Heck, Google him and this thread is fourth on the list!

 

http://www.google.com/search?q=Alan+taffel&ie=UTF-8&oe=UTF-8&hl=en&client=safari

 

[cfmsp]

 

Unlike Mr Harley and Mr. Taffel, John Atkinson seems to have a much better handle of the State of USB Audio.

 

here's an excerpt of a post on the S'phile forum, where John was asked if there really was 'no free lunch' in USB Audio.

 

http://forum.stereophile.com/forum/showflat.php?Cat=0&Number=71177&page=0&fpart=1&vc=1

 

 

"The USB Standard wasn't conceived with streaming high-quality audio in mind, so DACs using conventional USB receivers feature high jitter.

 

The Benchmark DAC 1 gets around that problem by using a sample-rate converter chip but the true solution is to emulate what happens in a FireWire link, which is to slave the PC to the DAC, which is to operate the USB link in asynchronous mode. As I said above, the latest DACs from Wavelength, Ayre, and dCS work in this manner, with a drastic reduction in jitter - see, for example, fig.13 at http://www.stereophile.com/digitalprocessors/wavelength_cosecant_v3_usb_digitalanalog_converter/index5.html

 

Unfortunately, none of these products are cheap enough to be regarded as a "free lunch."

 

The Wavelength Cosecant was reviewed in July, the dCS Scarlatti USB DAC is reviewed in our August issue, and the Ayre QB-9 USB DAC in our October issue."

 

enjoy

clay

 

 

 

[manisandher]

Earlier, I had said, "The fact that Alan did not include an async USB DAC isn't a big deal in my opinion."

 

Having read through this thread once again (and the external links), I now see that this is wrong - it actually IS a big deal. Thanks to all who've helped to explain why this is the case.

 

Mani.

 

[PeterSt]

I have mixed feelings on all this. I did *not* read the article, but by now I hope I can guess more or less what's in there. I'll try to respond to some things which got stuck in my mind after reading the thread.

 

Windows2000 AFAIK does not contain the mixer stuff as how it is in XP. Generally, people appreciate W2K as more lean, and therefore better sounding (FWTW).

IMHO it is a bit strange to force for a certain hardware environment if it is about a DAC, while no hardware impeeded issues seem to be present. Also, when an asynchronous DAC is said to be immune to software influence, why to tune the hardware to it ?

But then maybe I'm not on par with what Gordon thinks about this these days, and maybe he now says that software will influence (note : when software does, it is always hardware related ... says me).

 

Nearly any system younger than, say, 10 years will perfectly play audio. This does not say, though, that a system with e.g. more than one processor (core) can't be utilitized to do that better. Again, how this is DAC related is beyond me at this moment, at least in the case of asynchronous.

 

Then, I am not sure what to think of stuff like the Benchmark - that having less jitter because of the applied principle, while part of the principle seems to be that it needs 24 bits for input (hence have 16 and it doesn't work), that by itself causing some software not being able to use it. Is that better because some - but not everyone can use it and it even sounds better, or is it worse because not everybody can use it ?

IMO this it is the most stupid assumption of such a manufacturer that the OS will do that resampling for you automatically, in the mean time just not *allowing* bit perfect playback. Yea, think about it twice if needed.

 

I never found an USB DAC sounding better than SPDIF. Of course I didn't listen to all of them, but it seems that somehow it just doesn't work (out). Odly enough USB users seem to be rare for my own (XX) environment, but those who could choose and compare, always decided for SPDIF.

To name one stupid thing : I never saw an USB DAC showing emotion. I own one which can switch between USB and SPDIF, but sadly internally USB is converted to SPDIF making the comparison apples and oranges.

 

"If it measures good and sounds bad, -- it is bad. If it sounds good and measures bad, -- you've measured the wrong thing." Daniel R. von Recklinghausen

 

Wrong.

If it measures good and sounds bad, you've measured the wrong thing.

If it sounds good and measures bad, you don't know what to listen for.

I'm not trying to be funny here.

 

I notice that people tend to take Pro devices and treat it like a DAC. My statement would be if you compare things and something like a Fireface wins sound wise, you just don't know audiophile DACs. And audiophile DACs are not those sold by normal audio shops, but they sound good. A Fireface does not (at all).

I never heard a Weiss, but there is no reason for me not to think the same : the history is Pro. All IOW, I'd grant this mr. Alan. There are no audiophile Firewire DACs.

 

USB, firewire and ethernet all have the capability of providing excellent data with excellent timing

 

Ok, let me say something stupid here (what Peter, again !?) : Firewire should win because it is the only means that will support I2S natively. With that I mean that the in fact one chip that arranges for all and everything, will (yes *will*) output I2S. And right before the chip ... it's data.

Add to this that Firewire just carries SPDIF if you want (this is hard to understand for those not owning Firewire boxes), and you'll see that there are quite some benefits here, one of them being each normal PC carrying Firewire.

But all Firewire does is teasing us ...

 

2c-Peter

 

[idiot_savant]

The TAS1020B which is the USB chip used for just about all USB DACs that support 24/96 ( including both adaptive and async implementations ) has a USB input and a "codec output"

The codec output can be programmed to be AC'97, AIC or I2S...

 

your friendly neighbourhood idiot

 

[cfmsp]

 

 

Peter,

I'm a little surprised by such mixed opinions on Firewire.

 

You seem to acknowledge (if that's not too strong a word) that Firewire is theoretically the preferred interface, with which I agree, and yet you appear to discredit ALL Firewire devices (including the well respected Weiss) on the basis of a single RME model and/or a bias towards all things 'pro'.

 

clay

 

 

 

 

 

[rlodad]

My take- Stereophile doesn't listen fast enough or write fast enough. That's why we have Chris!

 

:^)