Triple radio device mesh networking - full duplex wireless mesh


Even though I am a long time OpenWRT user, this is my first post in this forum.

Today I was reading an article/advertisment regarding mesh networking with 3 radio devices and the following configuration.

1st radio: 2.4 GHz as AP for clients
2nd radio: 5GHz for mesh ingress
3rd radio: 5GHz for mesh egress

As the article describes, this configuration have minimal throutput losses between hops with losses of 4% on 10 hops in lab environment and 40% on 10 hops in real world setup.

I am familiar with 802.11s and batman-adv, since I have deployed some small scale mesh networks, but I can't seem to understand if this kind of configuration is possible with OpenWRT.

So my questions.

  • Is it possible to set two 5GHz radios dedicated for meshing one as sending and the other as receiving? (2 different channels used)
  • Is it possible to do this with dual stream radio? 1 stream for only receiving and the other stream for only sending? (single channel used)

P.S. Sorry for my english, it is not my native language.

Your English is very good!

If I were doing it (and I will be shortly, as I just received a triple-radio device), I would:

  • Mesh on 5 GHz channel N for all APs
  • Provide client access on a different channel on each of the APs

That way there isn't contention between clients and the mesh backhaul.


A more sensible, at least more conventional, approach might be:

1st radio: 2.4 GHz as AP for clients
2nd radio: 5 GHz as AP for clients
3rd radio: preferably 5 GHz as backhaul for the mesh (or at least simple WDS/ 4addr)

I don't quite see how you'd split ingress and egress otherwise.


Maybe I didn't write down properly what I am looking for.


Is this possible?

How will that scale when you have more than two mesh nodes.

Your wasting the 2x5GHz band on interconnect and limiting clients to 2.4GHz

@slh suggestion above seems a better approach

1 Like

I am mostly concern for throutput, current mesh confings over 4 hops is a dissaster

How will you arrange radio channels with more than two nodes. You only have two 5GHz radios per node

So do you have a link to the article

here is a configuration with 7 nodes that I just made

and here is the article / advertisment ( I am not sure if I am allowed to post it)

Notice how not all nodes can connect to any other node in your diagram: if two nodes both transmit on channel 52 and receive on channel 36, then they are never going to communicate; and you cannot have more than two nodes with different configurations if you have just two channels.

That idea seems to be viable only when you can configure your mesh in a zebra-like setup, where each node can only communicate with nodes of a different color... seems very limiting for the general use-case of a mesh network, and a hell to configure anyway.

Besides, you are not offering any 5GHz access to your devices...

If you insist in this idea, I would experiment (and "experiment" means that I have still never tried anything remotely similar to this) with two separate mesh networks, one on each channel (or even one on each band), and then bond the networks together on each node.


You will also have to somehow allow only one of the networks to transmit.

I made this diagram on the fly without much thinking on it, I know there are some limitations but with propper planning can be avoided.

My purpose is not to serve many clients but to have a rock solid mesh backhaul with minimal losses over multiple hops.

Well, if someone know if it is possible to configure interfaces like STA Mesh Point and AP Mesh Point would be excellent, else if there is another way.

In 1st post I also asked if this can happen on single radio with multipple streams

It seems like maybe adhoc mode would be good, and some routing rules, as long as a static mesh works for you.

1 Like

After reading the article I understood it differently to you.

The low latency network say uses two mesh networks but sends data on the mesh that has the correct direction.

For example if you had a large mesh organised in a ring you could have one mesh assigned counterclockwise data and the other clockwise. Then each node would need to know the topology of the ring and send data in the direction of the destination using the appropriate mesh.

So your not assigning tx / Rx pair per node but a structured topology with assigned data direction.

No the article definitely is about getting full duplex via multiple channels, not wireless repeating on same channel.

This is excactly what routing protocols such as batman-adv are doing.

That's what I think as well.

1 Like

Ok will read again

I quote a paragraph on another brochure of the same company

One must overcome the half-duplex nature of 802.11 protocols to solve the throughput and latency problems discussed above. As such, the number of radios dedicated to handle wireless handsets and the mesh backbone

In the multi-radio or "structured mesh" approach, there are several dedicated link interfaces where at least three radios are used per network node, including one radio for wireless client traffic, a second radio for ingress of 802.11a wireless backhaul traffic, and a third radio for egress of 802.11a backhaul traffic. This approach to wireless mesh networking offers significantly better performance than either the single or dual radio approaches. It allows for dedicated mesh backhaul links that can transmit and receive simultaneously because each link is on a separate channel.

I think you could accomplish this using a more fully connected adhoc mesh but setting route metrics. Then if a link disappears you still get a fallback link even if it means repeating instead of full duplex.

Ok, so far of what I have understand, for starters to find a way to create a full duplex wireless bridge.
I guess I can set to redirect incomming and outcomming traffic from one interface to another.

Is there any other way?

edit: how is Ubiquiti AirFiber achives full duplex?