[borrego]

Product Information: http://www.xmos.com/products/development-kits/usbaudio2
 
Pros: $149, comes with ADC/DAC capability, fully programmable, pin outs for SPDIF (for adding a co-axle output board)
 
Cons: Limited 24bit/96kHz bandwidth of the TOSLINK output. DIYer can add a co-axle SPDIF boad (not from XMOS) and unleash the 24bit/192kHz capability.
 
Purchase Reason: Low cost USB Audio 2.0 (i.e. Async USB Audio) transport replacement for the good quality but "drop-outs" frequent Audio-gd Digital Interface.
 
Whom is it for: Budget minded DIYer who does not mind handling a bare circuit board and having TOSLINK output
 
Whom isn't it for: Anyone who need a readily finished case and co-axle SPDIF output for his/her USB Audio Transport
 
Where to Purchase: Digikey
 
What else I bought to use with it: Lifatec Silkflex Glass Fibre TOSLINK cable (http://www.lifatec.com/toslink2.html)
 
Case I currently used with it: A cardboard packaging box (originally came with my Mofi case purchase) with cut-outs for the TOSLINK and USB cables
 
Board Inputs: USB 2.0 B-type input, 3.5mm audio in
 
Board Outputs: TOSLINK (bandwidth limited to 24bit/96kHz) and 3.5mm line out (decode up to 24bit/192kHz)
 
OS I am using it with: Windows 7 32bit and 64 bit (WASAPI). The device also supports Mac OS.
 
Drivers I am currently using: Thesycon 1.26 drivers downloaded from Wavelength website, with modified inf files (manually adding the XS1-L1 device id)
 
Troubles I encountered: The Theyscon driver (even the one downloaded from the XMOS site) wouldn't install probably. Thanks to the helpful XMOS technically support pointing me to use the USBVIEW utility (http://www.ftdichip.com/Support/Utilities/usbview.zip) to uninstall all previous Windows USB audio devices. Since that the drivers installed properly and has been working perfectly.
 
Quirks: After the Windows driver is installed, there is just a single "Speaker" output option in the Windows sound device panel. There is not a dedicated SPDIF output option in the Windows sound device panel as usual with other USB audio transport devices. But the TOSLINK output does indeed work and output bit perfect SPDIF signal. According to the XMOS technical support, the board outputs I2S and SPDIF at the same time and cannot be individually turn off.
 
Special Thanks: To XMOS technical support, and leeperry who pointed me to the links of Theyscon 1.22 production drivers
 
DAC I use it with: The original Audio-gd NFB-1 (ES-9018 Sabre32 version, with 80Mhz TCXO upgrade)
 
Computers I use it with: Dell Latitude D520 (Windows 7 32 bit), HP Probook 4310s (Windows 7 64 bit)
 
Burn in period: The TOSLINK output circuit section of the board does not really have any capacitor requiring burn-in. I think the burn-in period was mainly for the TOSLINK receiver circuit section of the NFB-1ES which I didn't use previously. The sound "settled" around 15 hours of music play time.
 
Audio quality compared with the Audio-gd DI w/ 74AHC04 inverter chip upgrade:
 
1. Bass/Mid/Treble Balance: The XMOS board is more linear. It is like a straight line comparing to the Audio-gd DI's curve. I think the XMOS board is more neutral
 
2. Details/Instruments Separation: About the same with the "Audio-gd DI at its best" (with Fidelizer optimized OS, computer not running anything else except the audio player). Beats the Audio-gd DI when the computer is being used for any other tasks like web browsing.
 
3. Sound Stage: If the Audio-gd DI is 4:3 (width/depth), then the XMOS reference design board is more like 5:2. The sound depth of the XMOS board increased with more burn-in.
 
4. Stability/Consistency: XMOS beats the Audio-gd DI hands down. There is no drop-outs ever with the XMOS board. I don't experience and audible quality increase by using Fidelizer
 
Possible/Future Upgrades: Battery powered USB cable, lower ppm crystals (the stock crystals are 30ppm parts), co-axle SPDIF board, a proper case
 
Alternative: $349 XMOS XS1-L2 USB Audio 2.0 Multi Channel Reference Design (http://www.xmos.com/products/development-kits/usbaudio2mc)
 

 

[leeperry]

you can easily update to 1.43.3: http://www.head-fi.org/t/562732/stello-u3-async-usb-to-coax-aes-ebu-transport-impressions-updated/45#post_7900730
 
too bad about the lack of coax...

 

[borrego]

Quote:

you can easily update to 1.43.3: http://www.head-fi.org/t/562732/stello-u3-async-usb-to-coax-aes-ebu-transport-impressions-updated/45#post_7900730
 
too bad about the lack of coax...

I noticed the Lindermann 1.43 driver files are substantially smaller in size (wonder if any feature is missing). Also I like the original 1.26 Thesycon control panel which I can see the streaming buffer duration. I am currently using the 8ms setting.
 
Do you hear an improvement upgrading driver from Thesyncon 1.22/1.26 to Lindermann 1.43?
 
I think TOSLINK is actually better than non galvanic isolated coax (as acceptable as 24bit/96kHz). And one do not have to mind with matching 75ohm connectors.
 
 

 

[borrego]

I am also thinking to upgrade to these "Super" super TX/RX TOSLINK modules for the XMOS board and the NFB-1: http://parts.digikey.com/1/parts/1571444-transmitter-opt-mod-20-125mb-s-totx1701-f.html
 

 

[DarknightDK]

Is this the same circuit used in the Stello U3? Which would be better? the Stello U3 or this?

 

[leeperry]

 

I noticed the Lindermann 1.43 driver files are substantially smaller in size (wonder if any feature is missing). Also I like the original 1.26 Thesycon control panel which I can see the streaming buffer duration. I am currently using the 8ms setting.
 
Do you hear an improvement upgrading driver from Thesyncon 1.22/1.26 to Lindermann 1.43?
 
I think TOSLINK is actually better than non galvanic isolated coax (as acceptable as 24bit/96kHz). And one do not have to mind with matching 75ohm connectors.

In no particular order:
 
-I haven't tried the 1.43 Lindermann because the DaMagic+ installer states 1.43.3 and tusbaudio.sys has a different binary content. I also prefer the layman words for the latency.
 
I only have the release notes for 1.22 but each new version has had tons of bug fixes and 1.43.3 is a year older. Placebo or not, I prefer the sound of 1.43.3 on the Stello Eximus DP1 indeed.
 
-Jitter over toslink is stellar at +1ns, sometimes even almost 2ns. This kills all the potential of the XMOS chip IMO.
 
-Buying a $27 Toslink connector is about polishing a turd. An eval board is just that, your money would be far better spent on this board for instance: http://www.diyaudio.com/forums/digital-source/188902-xmos-based-asynchronous-usb-i2s-interface.html
 
Your eval board has poor clocking(prolly in the +100 ppm region), no isolated coax/I2S outputs, evaluation drivers that bleep every 2 mins, poor +5V filtering....you've only scratched the surface of what the XMOS chipset can do for you, the rabbit hole goes much deeper

 

[borrego]

Quote:

-Buying a $27 Toslink connector is about polishing a turd. An eval board is just that, your money would be far better spent on this board for instance: http://www.diyaudio.com/forums/digital-source/188902-xmos-based-asynchronous-usb-i2s-interface.html
 
Your eval board has poor clocking(prolly in the +100 ppm region), no isolated coax/I2S outputs, evaluation drivers that bleep every 2 mins, poor +5V filtering....you've only scratched the surface of what the XMOS chipset can do for you, the rabbit hole goes much deeper

TOSLINK aside, I think the $149 reference design board stack up pretty well to the Stello U3 for 1/3 of the price. The 2 pins crystal it is using actually easier use of external clocks. I have already identify a 1ppm clock + "current transfer" type coax board I can add to it for another $50. The "evaluation" drivers only started beeping after 1 hour, and it is an non issue as I can use the wavelength driver. For the its power supply section, I believe it is no more worse than the Stello U3.
 
The Stello U3 is by all means a much more completed product. For me I enjoy the DIY process.
 

 

[leeperry]

TOSLINK aside

 
Feeding a $1K Sabre DAC like yours w/ Toslink is far from optimal to say the least, I'd take a mobo coax output over any kind of Toslink

 
If you wanna go DIY, that 99€ board on diyaudio seems hard to beat IMHO. At least it's got more serious clocking, but you'll have to come up w/ your own outputs wiring. There are other DIY XMOS boards, but they're pricier IIRC.

But that's right: Toslink is very much aside, I'll give you that(scroll to the bottom of the page): http://lampizator.eu/LAMPIZATOR/TRANSPORT/CD_transport_DIY.html

 
I believe your DAC uses WM8805 on its coax input, and when I compared coax and glass toslink inputs in the Firestone BRAVO WM8804(same chip as 8805, but only stereo) reclocker: as expected it ended up being a clear knock out.

 

[borrego]

Quote:

Originally Posted by leeperry /img/forum/go_quote.gif
 
If you wanna go DIY, that 99€ board on diyaudio seems hard to beat IMHO. At least it's got more serious clocking, but you'll have to come up w/ your own outputs wiring. There are other DIY XMOS boards, but they're pricier IIRC.

 
I don't quite understand how you find 99€ diyaudio board has more value. Just the omission of the XMOS programming and firmware flashing capability would be enough to disqualify it as a project kit. In my opinion the XMOS technical support alone would worth $50.
 
The XMOS reference board can accept 2 pin external clock signal easily (for both the processor and signal master clocks). I can easily (and inexpensively) pair external 0.5ppm clocks and SPIDIF module boards to use with it. Quite likely my finished work would have the clock & SPDIF sections similar to the Audio-gd CD7 transport, depending whether I can get a hold of an Audio-gd ICAC module. 
Quote:

Originally Posted by leeperry /img/forum/go_quote.gif
 
Feeding a $1K Sabre DAC like yours w/ Toslink is far from optimal to say the least, I'd take a mobo coax output over any kind of Toslink

 
I believe your DAC uses WM8805 on its coax input, and when I compared coax and glass toslink inputs in the Firestone BRAVO WM8804(same chip as 8805, but only stereo) reclocker: as expected it ended up being a clear knock out.

 
No Sabre32 DAC would ever require external SPDIF receiver chip because the ES9018 DAC chip got it built-in (though my NFB-1 got some serious PLL circuit). The Sabre32 "Time Domain Jitter Eliminator" SPDIF receiver behavior is kind of strange: Stringent on SPDIF signal rise/fall spec (thus I experienced drop out from my Audio-gd DI)  but can tolerate relatively high jitter. It could be the reason why I find the NFB-1 TOSLINK input not that unpleasant.
 
Too bad my Audio-gd NFB-1ES does not have I2S input pins on the circuit board. Otherwise I would just stuff the XMOS board inside the spacious NFB-1 case and have direct I2S connection.

 
 

 

[leeperry]

AFAIK, his 99€ board comes as PNP...doesn't it? he does provide unlocked drivers and a fully working firmware FWIR, and it's got good clocks to boot. You can use commercial drivers found on the web w/ your eval board, but you don't own a license per se.
 
Yes, I remember Kingwa stating that the S/PDIF input of those Sabre chips was more or less fubar and that he had to add some third partry circuitry in order to get it working...I wanted to look it up but this DAC has apparently been phased out from his website.
 
You know, ppm resolution, jitter, impedance....electronics are not LEGO, you need the right measuring tools and knowledge in order to get them working in an optimal way and this is mandatory to provide them w/ a proper supportive design. Slapping a 0.5ppm clock on an eval board w/o drastic R&D and measurements is blind failth at best. The wires will pick up noise, interferences, and the jitter seen by the XMOS chip could easily be worse than the stock jellybean clocks.
 
Jitter resilience is just that: algorithms to make up for it....get it as low as possible in the first place, and the SQ will improve substantially...especially w/ such a crappy protocol as S/PDIF(and even worse in its horrid Toslink form)

 
Either way, I didn't mean to threadcrap your topic...I just think that you haven't quite heard what a proper XMOS implementation w/ better clocks and isolated coax could do for you. If you enjoy toslink off your eval board, then more power to you!

 

[borrego]

Quote:

AFAIK, his 99€ board comes as PNP...doesn't it? he does provide unlocked drivers and a fully working firmware FWIR, and it's got good clocks to boot. You can use commercial drivers found on the web w/ your eval board, but you don't own a licence per se.
 

 
You may not seems to have any programming experience base on the above comments. Do you know with the XMOS tool I can write my own driver?
 
Quote:

You know, ppm resolution, jitter, impedance....electronics are not LEGO, you need the right measuring tools and knowledge in order to get them working in an optimal way and this is mandatory to provide them w/ a proper supportive design. Slapping a 0.5ppm clock on an eval board w/o drastic R&D and measurements is blind failth at best. The wires will pick up noise, interferences, and the jitter seen by the XMOS chip could easily be worse than the stock jellybean clocks.
 
...

 
What kind of wire do you expect me to wire the external clock? If it is the case like you said than how the high end DACs use external clock?
 
You seems to be a educated consumer and definitely a nice/helpful guy. But I don't see why you think the XMOS $149 board is a bad design. IMO, at most it is a basic design, but readily to accept upgrade with pin outs. Have you read its schematic?
 
Well, may be I should put this thread to the DIY section.

 

[leeperry]

 

Do you know with the XMOS tool I can write my own driver?

No idea, I use commercial boards that come w/ a proper end-user license.
 

 
What kind of wire do you expect me to wire the external clock? If it is the case like you said than how the high end DACs use external clock?

No idea either, I'm not an EE. But that's the difference between kitchen DIY and commercial products. Either you do things blindly and hope for the best(but in the end Murphy's law will very much apply), are an EE and have access to the right tools...or pay for the R&D commercial companies are providing you w/. Ppl often believe that DIY is so much cheaper, well it depends on how you look at it ^^
 

Originally Posted by borrego /img/forum/go_quote.gif
 
I don't see why you think the XMOS $149 board is a bad design.

 
Coz toslink is junk, and installing your own connectors opens a whole new window to a million problems as mentioned above. You cannot get true 75Ω impedance if there's soldering onto the connectors, because solder increases resistance. My point is that low jitter doesn't come for free

 

[borrego]

Quote:

 

No idea, I use commercial boards that come w/ a proper end-user license.
 

No idea either, I'm not an EE. But that's the difference between kitchen DIY and commercial products. Either you do things blindly and hope for the best(but in the end Murphy's law will very much apply), are an EE and have access to the right tools...or pay for the R&D commercial companies are providing you w/. Ppl often believe that DIY is so much cheaper, well it depends on how you look at it ^^
 
 
Coz toslink is junk, and installing your own connectors opens a whole new window to a million problems as mentioned above. You cannot get true 75Ω impedance if there's soldering onto the connectors, because solder increases resistance. My point is that low jitter doesn't come for free

I finally understand how you evaluate the XMOS reference board. Basically you just declare it as a bad product just because it has TOSLINK, and not even bother to find out what other options it provides because you are not an EE or programmer. With your understanding, TOSLINK is like, bad power supply, which would contaminate every other circuits on the same device.
 
With your understanding every piece of equipment that comes with a TOSLINK connector socker is junk (even if it does have other input/output pin options).
 
Yeah, the iPhone4s is junk because  it cannot stream data when connecting to a 2G GSM network (even if can connect to HSPA+) . The Canon D60 DSLR is junk because its built-in flash unit is not powerful enough (even if it has hotshoes for external flash). Fair enough.
 
What are you trying to say on soldering SPDIF connectors? Are you confused the coax cable connectors with coax/BNC sockets? Can you point me to a single commercial device which does not use soldering with its coax/BNC SPDIF sockets?
 

 

[leeperry]

you're way oversimplifying what I said...I only said that jellybean clocks and toslink are a disgrace to what the XMOS chipset can provide.
 
look at the Musiland 01USD, its BNC connector isn't soldered and it's got almost perfect S/PDIF specs: impedance, bandwidth, voltage, rise times...a chinese magazine posted its measurements, they're pristine! someone recently pasted them in the 01USD thread.
 
RCA simply cannot work for S/PDIF...it will never have the right impedance, no matter what....and as good as the 01USD can be on a specs pov, its clock synthesis is horrid. The perfect product does not exist, or maybe it does but there's a good reason why the top transports cost so much ^^
 
this said, let us know how your clocks and connectors transplantations will go if you like

 

[borrego]

The soon to be "New Home" of the XMOS reference board, just received from the builder. Unfortunately I am super busy at works, with long overseas trips and won't be able to finish it for at least another 2 months
 
R transformer, dual stage linear power supply, dual 0.5ppm TCXO clock boards, dual "current mode" SPDIF output boards. The XMOS board will be placed behind the BNC/Co-axle sockets and connecting to the clock boards via coaxle cables. All together, including the XMOS board and all shipping costs, is still USD100 cheaper than an Audiophileo 2.