Open-source USB interface: Audio Widget

Hi UnixMan,

Both the Hiface an the Legato are USB to SPDIF interfaces. The jitter is most likely coming from the SPDIF link - thus the great engineering effort to reduce jitter.

The audio-widget is NOT an SPDIF interface, and despite requests to add SPDIF out, we are reluctant to do so precisely because of the jitter of the SPDIF interface.

The digital part just supplies the i2s samples to the DAC. The DAC is clocked by a < 1 ps jitter XO.

However, no amount of theoretical arguments will settle this jitter issue. We will wait for someone to do measurements to do listening tests :)

Alex
 
The SQ of hiface interface depends much on the external power supply, "the flea" circuit or LifePO4 battery both can deliver much better sound while direct feeding 1.8V/3.3V with shunt regulator is the best.

The AW-1 and Audio-Widget with ADP151 both deliver decent sound as a good start point, they also have connectors to external power supply bypassing onboard ones (3.3V for DAC and XO) to provides extra choice for DIYer.

I am not sure how I2S signal looks like generated from MCU, IMHO it may benefit from reclocking the I2S using master clock from XO.


yes. But not only. :)

Also the digital board may need some improvements. That's why I mentioned the Legato (which is just an USB to s/pdif UAC1 async interface, see here: http://www.audiofaidate.org/forum/viewtopic.php?t=8408 ).

Before hearing that one, I was (naively...) thinking that with an async connection and a good local clock one could basically forget about the digital domain and all of its problems. As I discovered the hard way, it's definitely not that simple. Even with an async link, there is still a long way to go. Have a look at the interesting measurements made by my friend and colleague George "Joseph K":

DIYHiFi.org • View topic - Some jitter analysis - and beating of a dead horse..

To obtain such amazing results, the Legato uses a lot of care in power supplies, clock, signal routing, layout, etc. Most (if not all) of the tricks used there have been previously documented by Joko (the Legato designer) on either this or that other forum, and perhaps similar strategies and care may/should be applied to the audio widget as well.


I definitely second your points. :)

BTW: for what regards PSUs, have a look at "Salas shunt reg." and/or Joko designs on this forum. ;)
 
Last edited:
Hi,

I have validated the USB9023 RevA boards and am satisfied with the results. This revision contains the following modifications:

-The Si532 clock has its own ADP151 low noise regulator.
- I have replace a number of ceramic caps with polymer aluminum caps to improve the sound quality.
- The board now has jumper selection for power supply sources. You can use USB supply for all sections or power 1 or all regulators from an external 5vdc supply. You may also bypass the on-board low noise regulators and power with a 3v3 supply of your choosing.

The addition of the polymer aluminum caps and the additional regulator have bumped the cost of a kit to $85 + shipping + paypal fees. You can view an assembled board on my website I have a limited quantity of parts for kits and will most likely not do another run of this revision.

Contact me directly if you are interested in a kit.

Regards,
George
 
New hardware: AB-1.1

Hi guys,

Finally, the first 10 working AB-1.1 units have arrived with me. More
are on their way. Pictures and schematics are attached.

The boards have been tested with 24bit 44.1 and 48ksps asynchronous USB audio on driverless Win7/64. The boards will do 24bit 176.4 and 192ksps on Linux (native) and Mac (native) and Windows (experimental drivers).

So what sets this kit apart from the existing AB-1?
* It now comes in a pretty case. It even has a nice logo on it :)
* There are OS-CONs at important power loads.
* The DAC is now ES9023.
* There are headers to easily put in alternative power supplies.
* The front has a programmable R+G LED.
* Output filter caps not mounted, feel free to experiment with your own!
* Clocks have much improved MUXing, power and layout.

AB-1.1 uses Golledge precision XOs for good clock quality.

I'm _almost_ ready to ship them out. Unfortunately, my design for the
rear panel was a bit off. I'm redesigning it now for a better fit. It
will probably be a couple weeks before the rear panel and remaining
kits arrive. Let me know if you want to start playing with it ASAP,
and I'll send the rear panel when they arrive with me.

In the meantime, don't hesitate to place your orders! The price is
still USD120 + shipping. With a tracking number shipping is USD63(52)
international(Europe). Without a tracking number it is 14(9). Please
paypal to borge.strand@gmail.com.

USB-I2S module only is USD70 plus shipping. For each AB-1 you may have
bought from me before there is a USD30 discount.


Cheers,
Børge
 

Attachments

  • AB-1.1_20110818.zip
    453.5 KB · Views: 175
  • DSC_0098d.jpg
    DSC_0098d.jpg
    86 KB · Views: 672
Here's a small correction. With the box the AB-1.1 is now over the weight limit, so the shipping without tracking number is USD18 to Europe and USD31 to the rest of the world. Sorry about that.

Shipping with a tracking number is still USD53/62.


Børge


Hi guys,

Finally, the first 10 working AB-1.1 units have arrived with me. More
are on their way. Pictures and schematics are attached.

The boards have been tested with 24bit 44.1 and 48ksps asynchronous USB audio on driverless Win7/64. The boards will do 24bit 176.4 and 192ksps on Linux (native) and Mac (native) and Windows (experimental drivers).

So what sets this kit apart from the existing AB-1?
* It now comes in a pretty case. It even has a nice logo on it :)
* There are OS-CONs at important power loads.
* The DAC is now ES9023.
* There are headers to easily put in alternative power supplies.
* The front has a programmable R+G LED.
* Output filter caps not mounted, feel free to experiment with your own!
* Clocks have much improved MUXing, power and layout.

AB-1.1 uses Golledge precision XOs for good clock quality.

I'm _almost_ ready to ship them out. Unfortunately, my design for the
rear panel was a bit off. I'm redesigning it now for a better fit. It
will probably be a couple weeks before the rear panel and remaining
kits arrive. Let me know if you want to start playing with it ASAP,
and I'll send the rear panel when they arrive with me.

In the meantime, don't hesitate to place your orders! The price is
still USD120 + shipping. With a tracking number shipping is USD63(52)
international(Europe). Without a tracking number it is 14(9). Please
paypal to borge.strand@gmail.com.

USB-I2S module only is USD70 plus shipping. For each AB-1 you may have
bought from me before there is a USD30 discount.


Cheers,
Børge
 
Hi guys.

I just stepped over this thread. I didn't really follow how things were developping. Though it seems to be a very promising project on the first glance.

Two - to me - very imptamt questions.

1. Do you consider glavanic isolation for your USB interface respectively I2S ?
I think it would be a very good idea to do so. You might consider it.
See also EXA2UI or Arye QB9

2. To me and many others it would be very important to have a
mulitchannel I2S interface for a Linux machine to setup an active system
based on brutefir.
From what I understand you interface is stereo only.
Any plans to go multichannel? Time line?


Thx

Cheers
 
Two - to me - very imptamt questions.

1. Do you consider glavanic isolation for your USB interface respectively I2S ?
I think it would be a very good idea to do so. You might consider it.
See also EXA2UI or Arye QB9

Just curious here - do you consider galvanic isolation a good idea from the point of view of sound quality or for other reasons? Other reasons might be for example ability to run with a large voltage difference between the two grounds.

I ask because if its sound quality reasons, I'm not certain that galvanic isolation by itself provides confidence of good sound. There are two issues with it - the degree of isolation provided at RF, and the clock jitter introduced as a consequence of providing the isolation.
 
Member
Joined 2004
Paid Member
There are several good solutions to get good isolation. ADI has some and Silabs has a good capacitive solution as well http://www.silabs.com/products/power/isolators/Pages/Si86xx-Digital-Isolators.aspx. But getting low jitter on the audio side will require careful design since the clocks should be on the isolated side, although a good transformer interface for the clocks won't degrade the jitter/phase noise. The Silabs part has a peak jitter of 350 pS, very low for what it is but too high for a reference clock. The optocouplers aren't better. That's why the design is a little tricky, requiring resyncing the data on the isolated side, with the correct clock.

Anything that reduces the coupled and radiated noise is a good thing.
 
Just curious here - do you consider galvanic isolation a good idea from the point of view of sound quality or for other reasons? Other reasons might be for example ability to run with a large voltage difference between the two grounds.

I ask because if its sound quality reasons, I'm not certain that galvanic isolation by itself provides confidence of good sound. There are two issues with it - the degree of isolation provided at RF, and the clock jitter introduced as a consequence of providing the isolation.

Yep. You gotta point. The question is what's the lesser of two evils.
EXA2UI for example is well aware that their isolator does come with a rather average built-in jitter performance of 250ps if I recall it correctly.
It seems though that this compromise is well worth it - even if you run it on your I2S lanes.
Of course we shouldn't forget that the EXA guys do have a buspowered device which makes things worse. Raytech has shown a way how to get
that device sounding best by applying local PS. Obviously having a good local PS in place is another great success factor.

Jkeny who's been intruducing a LiPoFe battery supply on the Hiface made a big step forward even without going the isolation route.
People claim though that the EXA device sounds better than the HiFace. I'm not sure how it would compare to the Jkenny modded HiFace.
It would be nice to figure out what the isolation really does. With the isolation usally comes the change on the power scheme. In this case you change 2 variables at the same time!?!?

As long as I've been running USB, I've been running Opticis optical USB with great results. There was no way to run my system without isolation.
The big advantage I had with the Opticis, which unfortunately is USB 1.1. only, was that I could put it in front of the receiver chip and reclocker.
It wasn't sitting on the I2S. Perhaps a better choice with solutions like the EXA2UI device would be than to feed the isolators by the receiver chip and run those
isolators into a reclocker. But that's more complex and of course needs a seperated PS scheme. ( BTW: My squeezebox setup without isolation transformer on the
SPDIF link wouldn't be acceptable. It's a night and day difference. There's something good about isolation. ;) )

However, the final result might also depend on the DAC. E.g. the Sabre is supposed to do reclocking on it's own. If a rather clean or less complex jittered I2S signal
is received, the Sabre might be able to live with that "little higher" jitter. The jitter just would have to stay below a certain threshold. Results could turn out to be perfect in such a case.
Other, more jitter sensitive, DACs might have a big problem with such a setup. Sidenote: We might should try to identify one day which type of jitter is causing most of the headache.

To me it's not easy to judge what's the best solution for know. There are a lot of variables in the game. One thing I know for sure.
All USB solutions I'm aware off havn't managed to decouple 100% from the PC induced mess. I hope we'll get this under control one day.


However. It's good to see some USB options popping up nowadays.

A little off-topic:

I just read over AA somebody (ackcheng) is reporting that hooking up his RME PCI-E card to an optical PCI-E extender has been a huge success. Just 900$. ;)

Maybe that optical USB extender at 385$ is a nice alternative to try. I'm not sure though if it is 100% USB-Audio compliant.

Cheers
 
Last edited: