Looking at the 5G demo, you say you’re doing 1ms transmission. I’m curious to know how you’re doing it exactly? Are you just doing AES67 with a network that has a very low jitter? Are you using a network driver in the Xenomai layer?
I’m not sure where you read “1ms transmission” regarding 5G.
1ms is about the latency on our boards, then for any kind of network transmission we still need to stick with the rules imposed by the speed-of-light and networking issues
Typically, we get aroud 10 to 20ms roundtrip latency depending on the conditions.
We’re not using AES67 because that’s a protocol that was designed for LAN network assuming perfect network conditions without packet drops. For internet-based streaming (including 5G), we have a proprietary protocol which does some error concealment in case of missing packets.
We are using Xenomai network driver only experimentally in one of our Intel boards, not on the RPi3 / 4.
Interesting! I read the title of the section that says “1 milliseconds, 100 miles”, without actually reading the content of the paragraph! 10 to 20ms of audio round trip latency is already quite impressive!
Out of curiosity, are you doing anything beyond adding a bit of FEC?
The whole thing is still in development and prototyping phase. Currently, we have both simple FEC with some redundancy and post-error correction (i.e., error concealment when packets are dropped using simple signal extrapolation algorithms). But we have plenty of ideas to evolve the whole thing and it’s one of the main topics we’re focusing internal research this year.
Out of interest, roughly when will you be including this functionality in the OS? Ive got a project on the go that this could be very useful for.