USB 3 streaming audio from XMOS?

Technical discussions around xCORE processors (e.g. General Purpose (L/G), xCORE-USB, xCORE-Analog, xCORE-XA).
rostec
New User
Posts: 3
Joined: Tue Nov 27, 2018 4:52 pm

USB 3 streaming audio from XMOS?

Postby rostec » Sat Dec 01, 2018 8:34 pm

When will we see a USB 3.0/3.1 streaming audio processor from XMOS?

USB 2.0 is little by little becoming a bottleneck in larger systems due to its limited bandwidth. Does anybody know about any plans?
User avatar
infiniteimprobability
XCore Legend
Posts: 1120
Joined: Thu May 27, 2010 10:08 am

Postby infiniteimprobability » Tue Dec 04, 2018 11:12 am

There is no schedule currently, because it requires a major change at the physical layer (MGBT support) rather than a "plain old 480Mbps PHY" which is already on chip. But it would be interesting to understand the use case for this to help build a business case.

What application do you see that needs such bandwidth?
User avatar
Caleb
Experienced Member
Posts: 65
Joined: Thu Apr 04, 2013 10:14 pm

Postby Caleb » Wed Dec 05, 2018 12:51 am

Remember that there is s difference in max data rate of USB 2.0 and USB 2 Audio Class. UAC2 can do 1024 bytes every 125us, which is ~80Mb. If you develop your own audio driver for transferring audio via bulk data then you can transfer a lot more data.

I have not looked into USB 3 specs lately but I don't recall that there is a USB 3 audio class. Even if there is, is there a 3rd party driver (one that you don't have to develop yourselves)? It seems to me that your first bottleneck is a custom driver.

Someone once suggested that a high channel count USB device could be implemented using an adapter in your box: PCIE or Thurnderbolt -> multiple USB2 interfaces. Can you then adapt multiple USB2 PHYs to a current XMOS IC? Note that the XMOS L1 designs require external PHY. Perhaps it would be possible to adapt multiples - or add a second PHY to one of the newer XMOS ICs that already have one PHY built-in. Maybe a 2-core device could have an external PHY connected to the 2nd core (tile / whatever it's fashionable to call it now). Then maybe your host computer's UAC2 driver can handle multiple UAC2 devices and aggregate channels from those multiple devices.
rostec wrote:When will we see a USB 3.0/3.1 streaming audio processor from XMOS?

USB 2.0 is little by little becoming a bottleneck in larger systems due to its limited bandwidth. Does anybody know about any plans?
User avatar
Wavelength
Member++
Posts: 21
Joined: Mon Sep 24, 2018 9:56 pm

Postby Wavelength » Thu May 23, 2019 3:21 pm

Caleb,

Hope things are well out at Grace!
UAC2 works for anything above the FS spec. There really is no reason for newer spec until there is something worthwhile to add.
The nice thing about USB 3 over 2 is that you have dedicated RX and TX channels which means no turnaround and less cable issues, at least from my testing.
I have a Thunderbolt license and yes a number of companies in the PRO industry are doing just that. They will have a USB2 port and a Thunderbolt port and basically the the Thunderbolt port which exits as PCIe is then feed to a USB2 HOST controller and then the DEVICE USB2 port is muxed to either the external USB connector or the internal HOST controller.
But with the introduction of USB-C and either USB3 or Thunderbolt I would think there is going to be more and more push for the use of USB3. Especially since most newer pads and phones are sporting those as well and the EU is pushing for USBC for all!

Thanks,
Gordon
Wavelength Audio, ltd.
Traktorist3d
Junior Member
Posts: 4
Joined: Tue May 28, 2019 7:06 am

Postby Traktorist3d » Tue May 28, 2019 7:32 am

Hello everyone, just write in this thread, since the conversation went about the throughput of USB 2.0. There is a wonderful DIY DAC v2.5.2 / DSC2 DAC http://puredsd.ru/, which sounds good for real money in a very natural and detailed way. There is a desire to make a multi-channel DAC based DAC. It turns out you need a transport that will support the transmission of 8 DSD channels to DSD512. If I'm not mistaken, USB 2.0 is not capable of providing such bandwidth.
I would like to ask if there is a possibility based on xmos to implement a transport that will receive data via ethernet and provide for 8ch DSD recapture. Even with the help of ethernet, we can get a decoupling of the DAC from the PC.
Here's an example fee http://www.pawpaw.hk/waiyu/ArticleList6b.asp?menu2=55 XE216-512-TQ128
can it use 8ch dsd over ethernet? Well, or I understand it through USB 3.0 bandwidth would be easy enough ..
User avatar
akp
Respected Member
Posts: 281
Joined: Thu Nov 26, 2015 11:47 pm

Postby akp » Tue May 28, 2019 3:55 pm

For 8 channels of DSD512 in DoP won't you need something like 270 Mbps? I'd be interested to know if the XMOS RGMII can handle that bandwidth streaming. Perhaps it can.
Traktorist3d
Junior Member
Posts: 4
Joined: Tue May 28, 2019 7:06 am

Postby Traktorist3d » Wed May 29, 2019 2:16 am

Well, it is permissible and dsd256), in general, is it real? And there are people who would take up the implementation of such a device?

Who is online

Users browsing this forum: No registered users and 1 guest