Stello U3: async USB to coax & AES/EBU transport <Impressions updated>

[leeperry]

Hi guys, It seems that no one posted impressions about the brand new Stello U3 yet, so I'll give it a shot

 
I should state that I'm from team anti-purple prose, so I will happily discuss any technical aspect of this new transport and will also try to translate my real world impressions into words. I've got an audio engineering diploma, and I've worked as a recording and mastering audio technician/engineer in several parisian studios. I'm not going to write pages of idiophile prose, but I'll do my best to share what I've learned when listening to this nice little device.
 
This said, Here's what it looks like:

Its dimensions are 100 x 36 x 100 mm, making it small enough for either a nomad or sedentary use.
 
On the back, we've got 3 connectors: 1) 480Mbit USB 2.0 input 2) RCA coax output 3) XLR AES/EBU output: 
 

 
On the front, there's a LED that will light up when connected to a computer:
 

 
I've run all those tests using a short RCA/RCA male adapter from Viablue, that IME provides a better sounding alternative to cables(even pricey ones, such as WireWorld's) as it provides a shorter signal path and a lower eletrical attenuation. I believe Audioquest sell the exact same adapter, prolly sourced from the very same OEM manufacturer.
 
You can read coax cables reviews that claim different jitter measurements depending on which direction the cable was connected, I don't own pricey equipment to measure jitter but I can easily tell that this adapter provides a clearer sound to my ears than any cable I've tried so far.
 
Getting back to the subject, I've found a very interesting page about those new XMOS USB audio devices on a foreign forum, so I'll just borrow its links.
 
Basically, the U3 promises to provide $900 performance for half the price:
 
-USB Audio Class 2.0 compliant, providing 24/192 support in MacOS X without drivers. You can also install a USB Class 2.0 driver on Windows(from XP to Seven), like what Wavelength currently does. So we're not talking about half-working proprietary WDM drivers here, and whenever m$ will wake up USB Class 2.0 will be natively supported by Windows as well.
 
It will also work natively in Linux, as Windows is pretty much the last mainstream OS that doesn't support the 2.0 Audio USB Class out of the box.
 
-Using the XMOS async USB 2.0 chipset, the exact same one that can be found in the $900 Wavelength WaveLink HS 24/192 and some Ayre USB DAC's: http://www.xmos.com/products/development-kits/usbaudio2
 
The XMOS chipset strong points over the "jellybean" PCM2704 or ubiquitous Tenor TE7022L are:  1) it uses an asynchronous USB protocol 2) it supports all those sample rates natively: 44.1/88.2/176.4/48/96/192kHz 3) It bases its timings off two discrete clocks.
 
The whole point of asynchronous USB is to relieve the DAC from having to (re)clock the data coming from the host.
 
-two low jitter discrete clocks, for pinpoint 44.1 and 48kHz multiples(that will avoid the nightmares of the Musiland's that couldn't output spot-on sample rates).
 
-built-in PSU filtering from the computer, making it perfect for laptop users as there's no need for an external PSU.
 
-no useless outputs like toslink, only coax and AES/EBU. Currawong recently posted a very interesting link where you can see why toslink is a terrible idea, and why all coax outputs are not born equal: http://lampizator.eu/LAMPIZATOR/TRANSPORT/CD_transport_DIY.html
 
-supporting 44.1/88.2/176.4/48/96/192kHz in 24bit...honestly, the lack of 88.2kHz of the TE7022L Tenor chip is becoming more and more annoying IME.
 
All summed up, this seems like the ultimate transport for those that are not willing to waste a grand on their transport...and it's got the nice added value of providing AES/EBU which opens a whole bunch of new options, being a more solid and professional protocol than S/PDIF

 
So, you now know a bit more about what's in the box and what it can do for us.
 
First, I will install the drivers, which are the exact same ones that are bundled with the Wavelength WaveLink HS 24/192. I would guess that when you pay to use the XMOS chip, you also get a licence to the Thesycon USB Class 2 driver: http://www.usbdacs.com/hs24192/hs24192.html
 
It's important to realize that XMOS didn't try to reinvent the wheel and write a WDM model driver from scratch. The Thesycon driver is only necessary because Microsoft didn't do their homework, it basically adds the support for the 2.0 USB Audio Class Microsoft should have added by themselves. So the way the XMOS chipset communicates with the host computer fully abides by an official USB specifications white paper, and this will be as flawless as using a Tenor 24/96 chip(that uses the older 1.0 USB Audio class) on XP/Vista/W7. So don't expect the usual glitches, stutter, blue screens and what-not that are usually associated with companies deciding to write WDM drivers from scratch.
 
So the drivers are now installed, and here's what happens when you click on the tasktray icon:
 

 
The Thesycon USB Class 2 driver is not tied to the XMOS chipset and it would work with any USB Class 2 compliant controller, so it's telling us what chipset/revision we run. You can also see an option to update the firmware of the XMOS chip, and an indicator of the current sample rate.
 

 
This section is pretty irrelevant for the U3, but it tells us about the clock source and the current sample rate.
 

 
This allows us to play around with the streaming buffer size and ASIO latency.
 
Indeed, the nice thing about those drivers is that they provide native ASIO support, meaning that you won't need to rely on kludges such as ASIO4ALL.
So I believe I've covered the technical details and software sides...now let's discuss the most important: how it sounds

 
I've compared the U3 to two other units: My beloved Firestone Bravo(that uses an isochronous Tenor TE7022L controller, reclocked by WM8804 for a lower jitter) and a $40 HA-Info U2(that uses the Tenor TE7022L chip as well, but not reclocked this time).
 
I've also used 3 different DAC's with coax inputs, one running the ancient CS8414 S/PDIF controller, one with the newer DIR9001 and one with the newest WM8805.
 
I've been comparing the sound of those 3 DAC's and 3 transports for a little while, and I can only confirm what I read on a DIY audio forum a while ago. Two guys in two different locations were running the same DAC and had the ability to roll the coax input controller using a DIP package. They were mostly comparing CS8414 to DIR9001. They both came to the conclusion that CS8414 provided richer tones and a wider SS, and that DIR9001 sounded narrower but tighter. I can only echo those impressions, and I personally kinda prefer the sound of CS8414 to DIR9001. The sound is more laid back, but "richer" as in less "colored". I can easily hear why many companies go back to CS8416 these days(especially in "PDUR=1" low jitter mode).
 
WM8805 on the other hand is really unforgiving, you really must have nothing to hide from this little man and have a very mellow sounding DAC output stage otherwise it will make your ears cringe. Some will call CS8416 "analog sounding" and I can hear some truth to that. Digital perfection can easily end up with a prefty hefty price tag to pay IMHO.
 
I would also like to stress that these audible differences between S/PDIF controllers could easily account for the diametrically opposite opinions you can sometimes read on audiophile forums concerning the very same transport. This, and the fact that all DAC chips will also react differently to the incoming jitter. Some manufacturers choose to use ASRC in order to put an end to this problem, which doesn't come for free and will also color the sound(we all know how upsampling can be fun for a day, and gradually become unbearable one week later). Some manufacturers using the "holy" Sabre DAC's have made clear that they decided to disable its internal ASRC, as they didn't like the sound. This could easily account for all the ppl calling the Sabre's clinical and lifeless.
 
I've spent several days of critical listening and my impressions about these 3 transports will take those 3 different DAC's in account. I will try to tell how each of them sounds in a broader way:
 
-HA Info U2: It's not bad for the price, but it really can't match the clarity and tightness of the two other contestants. There's an obvious lack of resolution, which becomes even more obvious when playing 24/96 audio. The wah-wah guitar in the intro of the Shaft SACD is mangled and sounds like you're hearing it from the next room, or that you're wearing hearing protections. I would guess that for $40 you get what you paid for: very minimal PSU filtering, and that's what it sounds like. Experience has taught me that the "you're only as strong as your weakest link" saying is very real, and this transport could/would/will easily cripple the potential of any serious DAC connected to it. Of course, I'm talking in an audiophile headphones environment, using prima calidad recordings. If you're running some cheap speakers in a warehouse, This U2 will make you very happy....but if you are an audiophile seeking the most truthful way to play back your favorite songs, you can look elsewhere.
 
Kingwa from Audio-GD stated that he encountered clock synchronization problems with one of the Sabre DAC's, and Firestone told me that they couldn't feed the I2S output of the TE7022L chip directly to the WM8804 coax chip in the Bravo, because it was out of specs...and that they were forced to get it through DIR9001 first. Vladimir even posted a diagram of what happens in the Bravo. A wild guess would be that the built-in coax output of the Tenor TE7022L chip might need some improvement too. Galaxy Corp(the maker of the USB Tenor chips) also wrote some proprietary ASIO drivers, but they are terrible...it's a glitching feast, and I'm not the only one saying so. I also talked to the coder of those drivers, but he clearly told me that he didn't care whatsoever about bug reports. This company likes to support as many features as possible to look good in the datasheet, so native ASIO/I2S/coax are all there...but in practice, it's another matter altogether

 
-Stello U3: Ah, this is far better! Very impressive 3D soundstage depth and width, sound clarity and tightness are also extremely versatile...even after listening to this baby for days, I would have a hard time calling this transport "colored". In the 3 different DAC's I tried, it always sounded like a perfect link without any audible drawback. Shaft sounds like being in the studio cabin with the breathtaking lively wah-wah and loud hi-hat sounding as true to life as can get. Isaac Hayes low and deep voice fills the sound stage, never overpowering the strings or bass guitar. I can easily guess that Stello spent a lot of time finetuning this little thing and didn't skimp in the PSU filtering design. Movies also sound very impressive, especially surround effects. Audio is merely an illusion, and we're alway looking for new ways to fool the human brain even further....well, this is clearly the most "transparent" transport I've heard so far.
 
-Firestone Bravo: Well, I don't really like the sound without the Supplier linear regulated PSU. The latter provides a much more dynamic and clearer sound. But the PSU costs almost as much as the Bravo, so this is a rather high hidden cost. The Bravo sounds good, not as clear and versatile as the U3, but still good. One could argue that the sound is colored, as this is clearly sounding more euphonic and less versatile than the U3. Bass isn't bloated per se, but it isn't as tight or chirurgical in the way it allows the brain to retrieve tiny little details and room ambiences/walls reverbs.
 
Audio is a nasty hobby as there's always something better sounding out there...and until you've heard it, you never really get to realize what you had been missing all along

 
Its drawbacks should also be clear in your mind by now: no 88.2kHz or >96kHz support whatsoever. More and more HD audio starts showing up in 88.2kHz(Dan Lavry claims that 88.2kHz is pretty much the best choice for digital audio, being the next "standard" sample rate over the 70kHz he advises as being optimal).
 
So I've tried really hard to find faults in the U3, as I'm always seeking excuses to avoid buying new equipment...but the U3 seems quite flawless to my ears. The worst part is that I borrowed it from a friend, and that I have to give it back to him. I'm trying to go back to my Bravo, but I don't see much future in that

 
Thanks for reading, and I hope I didn't bore you too much. I merely tried to share what I read here and there and what I heard myself. I would be happy to discuss even further about that hopeless OCD for the holy grail of digital audio we all share on this forum.
 
I know there are naysayers who will say that S/PDIF is made of 0's and 1's and that this is all placebo....well, you can look up this link(kindly provided by Currawong), where you can easily understand that there's far more happening in a S/PDIF connection than what you might be willing to believe: http://lampizator.eu/LAMPIZATOR/TRANSPORT/CD_transport_DIY.html
 
If everything sounds the same to you...well, you're very lucky. To the unlucky few, I hope this thread will have provided at least some interest in serious coax USB transports.
 
Peace,

 

[uelover]

Will be interesting to see how this stack up against audiophilleo since they are priced similarly.
 
Audiophilleo allows one to plug the converter directly into the BNC port, saving money for coaxial cable.

 

[leeperry]

Will be interesting to see how this stack up against audiophilleo since they are priced similarly.
 
Audiophilleo allows one to plug the converter directly into the BNC port, saving money for coaxial cable.

 
From what I can see the Audiophileo 2 comes with a female BNC, and you can plug in either a male/male BNC/BNC or a BNC/RCA adapter...you can do the exact same thing on any transport, using this kind of adapter for instance: http://www.audio-hifi-shop.de/images/viablue_rcamalemale_b.jpg
 
That's the very one I currently use on my Firestone Bravo, it sounds much better than any coax cable I've tried(even over a fancy $200 1 meter WireWorld cable...that's the day I lost faith in coax cables ^^).
 
The Audiophileo 2 costs $579 on the official site: http://www.audiophilleo.com/buy.aspx
 
The U3 can be found for $445: http://soundapproach.com/aprilmusicstellou3usblink19224bit-p-2922.html
 
And I wonder what kind of drivers the Audiophileo 2 uses, as they claim "The default Windows XP, OS X, and Linux drivers support up to 192 kHz, 24 bits; for Windows Vista and Windows 7, just download and install our custom driver if you use sample rates above 96 kHz."
 
That's weird to read that it's more compatible with XP than with W7?! It seems to be using the USB Class 1.0 on Vista/W7, but some very smart custom drivers on XP? I don't see how that'd work tbh and it's a shame they didn't do the opposite, who honestly cares about XP these days(MacOS and Linux being very small niche markets). I personally prefer the USB Audio Class 2.0, coz as soon as m$ will consider doing their job, it won't require any optional USB Class driver to be installed on Windows whatsoever and it's already natively supported by MacOS X and Linux.
 
I've found a subjective comparative review of the Hiface and the Audiophileo 2 here, that'll give me some food for thoughts about the audible differences between serious transport: http://www.computeraudiophile.com/content/Audiophilleo-Vs-Hiface-Evo-My-Tale
 
I will be comparing my friend's Stello U3 to my Firestone Bravo, which is using a completely different vision: isochronous USB but 50ps WM8804 reclocking at the very last stage. Do all roads really lead to Rome? I'll be soon to find out and share my impressions here, but the Bravo uses the non-88.2kHz Tenor chip either way...so it'll be losing points as I've got a few digital SACD rips that are 24/88.2, and it's really annoying me to no end to be forced to upsample them

 

[uelover]

Maybe I can use a USB A-B adapter and that Coax Male-Male adapter so that I can go cable-less =D
 
From the description on soundapproach:
 
Platform Support:
Customized Windows Driver : Windows XP, Windows Vista, Windows 7
Natively supported by Apple Mac OS X
 
It seems that one would require customized drivers even for win xp?
 
 
Nonetheless, I like the way how U3 supports AES/EBU although I am not sure if that will represent an improvement over direct coax bnc/rca input. A good AES/EBU cable costs a bomb too.
 
Yeah, I have read that 'review' before. The guy prefers the Audiophilleo to EVO. Audiophilleo 2 used to cost USD499 before it was raised to USD579 which I think is a little overpriced.
 
Nonetheless, with all the explanations and comparisons against other brands on their website as well as the many positive feedbacks it has received, I think that it is a fairly competent product.
 
I am looking for an alternative that is more reasonably priced. Had U3 under my radar for some time but the lack of detailed information, user impressions, comparisons, as well as availability outside the US is keeping me at bay.
 
 
Let us know how U3 performs with the olimex isolator sitting between it and your PC. =)

 

[leeperry]

Well, the best cable is no cable that's for sure: less jitter/attenuation and way cheaper too.
 
The XMOS chipset requires 480Mbit USB operations AFAIK, so ADuM4160 will be a no go.

 

[uelover]

Hmm have just received Olimex USB isolator.
 
I assume that U3 will work with the isolator, just that it will lose the async feature???

 

[leeperry]

The Olimex ADuM4160 dongle is a very nice toy, I'm sure you'll be enjoying it thoroughly(especially when feeding it an external PSU). I don't think the XMOS chipset will enjoy 12MBit/s operations, it prolly won't work at all.
 
BTW, my friend has finally received his U3 and he just lent it to me a few hours ago. Pics(I've hidden the serial#): 

 
So far, indeed I'm very impressed by the SQ and yes it does sound quite different from my Firestone Bravo! To the point that I believe I will switch too, hah....audio is a pretty nasty hobby as there always seems to be something better sounding out there

 
I'll be doing some critical listening through the week-end, and will update the OP accordingly. The Windows drivers also support native ASIO, always a nice touch.

 

[uelover]

The U3 is real cool!
 
I can't seem to buy the viablue coupler anywhere online though. Seems like they only sell it within germany.
 
The Olimex really cleans up the sound by quite a fair margin. I am sure if there is any ways to get a galvanic isolator onto the U3 the sound would be much better.
 
How heavy is the U3? Will it strain the female rca plug on your DAC??
 
As much as I am wary about powering the device directly via the dirty USB power, I am interested to hear how this thingy perform =)

 

[kr0gg]

omg. you mean that at it's price the U3 doesn't have galvanic isolation from PC?

 

[leeperry]

Originally Posted by uelover /img/forum/go_quote.gif
 
I can't seem to buy the viablue coupler anywhere online though. Seems like they only sell it within germany.
 
The Olimex really cleans up the sound by quite a fair margin. I am sure if there is any ways to get a galvanic isolator onto the U3 the sound would be much better.
 
How heavy is the U3? Will it strain the female rca plug on your DAC??

I bought mine here, and they had no problem shipping outside Germany: http://www.audio-hifi-shop.de/product_info.php/products_id/410
 
The U3 is rather heavy, but of course I always put it on top of of a pile of books/CD's so it won't strain the RCA connector of my DAC's.
 
Don't get too hung up on ADuM4160, some companies sell audiophile USB filters: http://www.odysseyaudiohk.com/eng/UNR.htm
 
I've had some emails with them and they made it clear that messing with the data lines was a big no-no, as ADuM4160 is meant to be used within industrial environments. All we really need is galvanic isolation on the coax output, and the U3 has it as you can clearly see a pulse transformer on this picture: http://soundapproach.com/images/cat_149/U3_board.jpg
 
There is no way to filter 480Mbit USB atm, and I personally prefer to keep the data lines alone(filtering always colors the sound to my ears, much like upsampling)...and tbh, the power isolator in your Olimex dongle is rated at 75 mV ripple, hardly audiophile grade: http://www.alldatasheet.com/datasheet-pdf/pdf/160257/AIMTEC/AM1D-0505S.html
 
There's no free lunch to be had from those industrial isolators, either they output a high level of EMI/RFI(ADuM5000 comes to mind) or stellar jitter(from 75 to 150mV ripple).

the U3 doesn't have galvanic isolation from PC?

The U3 runs in fully async mode and has a galvanically isolated coax output, so that would render ADuM4160 useless when it comes to what we care for in computer audio: avoiding ground loops with the dirty computer ground.

 

[uelover]

I have just realized that the viablue plug is not a true 75ohmz impedance plug. I am looking for a male XLR-XLR coupler though if I can find it then U3 will be awesome! Good AES/EBU are way overpriced.
 
 

 

[leeperry]

I have just realized that the viablue plug is not a true 75ohm impedance plug.

IMHO, there's also a lot of commercial fluffing in this department...many experts claim that RCA cannot be 75Ω by design, and that even those shiny non-soldered Canare plugs aren't *really* 75Ω: http://www.bluejeanscable.com/articles/75ohmrca.htm
 
Some other ppl claim that even BNC is hardly ever true 75Ω in audio because that would drastically increase RMA: http://hifiduino.wordpress.com/2010/03/30/some-audio-wisdom/

At frequencies encountered in word clock and SDIF2/3 connections, the impedance mismatch problem at the connector is academical

 
And a 75Ω connector isn't quite enough! If it's soldered to a cable, then you can be 99% sure that it won't be true 75Ω in the end.
 
The BNC connector on the Musiland 01USD was measured as being very close to 75Ω, and as you can see the connector isn't soldered: http://assets.head-fi.org/5/58/5896b76e_IMG_0165.JPG
 
Solder increases resistance...it's like those DAC's that have gold plated RCA plugs on the outside but not on the inside ^^
 
I also tried to use a 75Ω BNC coupler w/ two RCA adapters, meaning two of these and one of that but I ended up still preferring the sound of the Viablue. This company knows what sounds good IME.

 

[uelover]

Maybe I will grab one from viablue to try. Still saving up for the usb/coax converter.
 
Buying a korean product from US (in the case of Stello) is pretty crappy though. Don't seem to find it selling anywhere else.
 
 
The performance of Berkeley Audio Design Alpha USB may be worth checking out but it is way too pricy.
 

 
http://www.berkeleyaudiodesign.com/products4.html
 

 

[Currawong]

Maybe if we ask nicely, the guy who got the measurements from the Audiophilleo might be able to do the Stello as well?

 

[leeperry]

Originally Posted by uelover /img/forum/go_quote.gif
 
Buying a korean product from US (in the case of Stello) is pretty crappy though. Don't seem to find it selling anywhere else.

Stello have a list of all their resellers worldwide available on their website, my friend simply got in touch w/ the nearest one and there you go..he made a special order as this is a new product, waited a bit et voilà!

Maybe if we ask nicely, the guy who got the measurements from the Audiophilleo might be able to do the Stello as well?

Not sure who you're referring to? I saw a test on 6moons I think that was giving the most expensive Audiophileo($1K?) having been measured as low as 10ps jitter, and the Halide at 700ps

 
OTOH, you can find a lot of subjective comparative reviews that give the Halide as much better sounding than say the Hiface...and some measurements gave the Hiface almost as good jitter-wise as the holy Lynx2 internal soundcards(that are notorious for the tightness of their clocks) but some other measurements gave the Hiface a really really high jitter.
 
And that's only one part of the story, as jitter can have patterns that will also influence the decoding....the second biggest factor being the chip taking care of decoding S/PDIF into your DAC. The most widely used being CS8414/6, DIR9001 and WM8804/5. IME, they all react differently to the transport and the cable/adapter being used. You can see in that PDF that CS have created a new "low jitter" mode in the legacy CS8416 chip(but then its max sample rate will be capped to 108kHz instead of 192kHz): http://www.cirrus.com/en/pubs/appNote/AN339REV1.pdf
 
And the third factor is the DAC chip itself, that can also react very differently depending on the S/PDIF decoding chip...they all claim "superior jitter resilience" but CS8416 in legacy 192kHz mode will make it react differently from WM8804 for sure. Anyway, my point is that a jitter figure outside its context doesn't mean much on its own, and can't really become a decisive data to take into consideration IMHO. The same way THD and SNR won't tell you how a DAC/amp will sound in your rig. As usual, there's only one way to find out.