All devices on an AVB network have a clock which is automatically 
synchronized using the Precision Time Protocol (PTP). However, this does 
not mean that every AVB talker on the network automatically uses a 
synchronized sampling rate for the audio stream it sends ("Media Clock") 
[1]. For example, one device might generate an 48 kHz media clock with 
an external oscillator; another device might derive an 48 kHz media 
clock from its PTP clock; yet another device might synchronize to an 
ADAT lightpipe input; etc. While these sampling rates are nominally the 
same, the clocks are not synchronized.

To generate AVB streams with synchronized media clocks in multiple 
devices, a media clock has to be generated in one device and distributed 
to the others. At the moment, the only way to distribute a media clock 
is to act as a talker and provide a stream (which does not neet to 
contain any actual audio channels) to other devices, which may then 
recover the media clock from this stream and generate their own talker 
stream with the same media clock. In the XMOS AVB reference 
implementation, if an input stream is connected to the listener, the 
media clock is automatically recovered from this stream and used for the 
talker [2].

For IEEE1722a a stream format for the distribution of a global media 
clock has been proposed [3].

[1] Understanding IEEE's deterministic AV bridging standards, 
http://www.embedded.com/print/4008284
[2] XMOS AVB Media Clocks, http://xcore.github.io/sw_avb/clocking.html
[3] 1722A Global System Clock Streams (aka Media Clock Streams), 
http://grouper.ieee.org/groups/1722/contributions/2013/1722a_rsilfvast_Global-System-Clock-Stream-Principles_15-Apr-2013.pdf