XMOS as base for an AVB endpoint

[Wouter]

Status: Under development
License: BSD

IEEE is developing a standard for real time audio / video streaming. This may become a worldwide standard and is very interesting technology for the professional AV scene.
With three XC-2 kits I'm building an AVB network. With the standard source code developed by XMOS it is already working. Now I am researching the possibility's to implement other protocols to manage the AVB endpoints and also at a protocol for real time control. The main goal will be something like a speaker, MIC and some user controls all implemented in one housing with one Ethernet connector.
Image: Two AVB endpoints cross wired with the interface PCB's attached to/from speakers and cd-player
Images:

[Light]

This project can go over Wi-Fi to do streaming too, right?

[Wouter]

Yes one of the most important protocols is the timing protocol (IEEE802.1AS). 802.1AS is based on IEEE1588v2 with extensions for Wi-Fi networking. I haven’t looked and this functionality yet.

[errordeveloper]

hey ..that's good, i'd build one too, but still couldn't save any cash to buy two xmos boards :(

if you are interested in control protocols ..i got this work in progress proposal for an advanced protocol for audio related stuff ..
i can drop you a brief description of my ideas, let me know on errordeveloper at gmail dot com :)

[ashleyb]

IEEE802.1AS is already handled by the XMOS code, what XMOS haven't implemented yet are all the other protocols needed for AVB, namely the standards required for stream reservation and management.

Their implementation is also currently lacking some flexibility in the way 1722 streams are handled, in that it really only works for audio streams at this stage.

[Wouter]

You are absolutely right, alto there is some work done on the Stream Reservation Protocol IEEE802.1Qav in the AVB source. I have recently contacted XMOS and they are working on a new release of the AVB source. There are no implementations yet to easy manage an AVB endpoint with something like SNMP because everything in 1722 packets is set by adjusting a variable and recompile.

I also agree that the V in AVB isn’t implemented at all at this stage.

[tknesel]

We are also working on an AVB network using two of the XDK and XAI. I'd be very interested in learning more about your control protocol. We are looking at a couple of things including OSC.

[Wouter]

What we mean with a real time control protocol is different from a management protocol like OSC and SNMP.

I'm watching the AVB working groups very closely. And they have been working on AVBC. It looks just like SNMP only the message's are better to understand and media related. In my opinion it is not smart to create another protocol if you could solve it with the existing protocols. There are documents send to the IEEE form people whom share that opinion. So there is not a standard yet.

The real time control protocol we use is :
http://en.wikipedia.org/wiki/MambaNet

[Light]

I will help to check that out.

[AntOFF]

Personal I compare OSC with SNMP where a difference is that OSC is easier to recognize for humans but SNMP will fit better in 'code' and is more widely used in networked equipment

I guess both rely on a client-server model, so a client have to initiate actions where a server responds to. The MambaNet model it multi-master, so any node may initiate an action.

This is exactly the difference between:
Management/configuration and real-time control.

Over AVB.
The basic idea is really nice because standardization is very important to make devices working together. I have my doubts on the direction they go; this compared to existing protocols like: CobraNet, LiveWire, Dante. But these last protocols are not a open standard :(

My main concern is that for AVB you require special AVB switches.