BLE mesh capacity estimate in 1-to-1 and 1-to-many connections

[Krastanov]

Are there any publications, from Nordic or academia, that give estimates on what would be the maximum capacity of a BLE mesh network?

I believe there is a hard limit of about 32000 nodes, but what about limits related to the capacity of the RF spectrum?

For instance, consider a large lecture room with a minimum number of relay nodes (maybe only one).

How many non-relay nodes can receive a broadcast message from the relay before the network becomes unreliable?

How many of the non-relay nodes can send a message to the relay (per unit of time) before the network becomes unreliable?

[endnode]

In principle when you use BLE advertisements and passive scanning for propagation of information it doesn't matter if it's mash network or not. There are 3 dedicated channels and until packets start to interfere massively it will work. Standard length of full ADV packet is ~0.5ms and switching time to recognize two packets by the scanner is negligible in this case. As usual advertiser sends packets every 100ms you will have very low interference even with few hundreds of devices (assuming devices sends the same packet on all 3 channels during one interval which is typical behavior of embedded BLE systems today).

This is just back of the envelope 2-minute calculation, unfortunately I don't have any serious research paper on this topic in hands (but probably some basic radio signalling theory and statistics would do the job so it's not even for any serious thesis work).

[rolandash]

Normally 'clear of air' principle is used in most RF IC PHY, purely depends on randomness of transmitting to avoid collision. So the only way to achieve more number of nodes is longer interval between 2 transmission. Depends on collision level you expected you can reach number of thousands with interval about 1~2 seconds.

However, longer interval can also limit the real time demand on data collection.