So in my ideal network (centralised command and control) I have one co-ordinator, and then all the other devices are nodes. The devices are mains powered but there may be 100s, possibly even thousands, within a large event space.
I see the number of 250 nodes mentioned as the limit for a zigbee mesh network, I assume this is stated as thing start to deteriorate after this? What happens as network size increases, latency goes up, packet loss goes up, there becomes a need to designate roles in the network (I don't want to have to do this).
How does Thread fare in this situation?
This all depends on how you use your network. For network where you need low latency, high reliability and nodes are generally in communication reach of each other then 100 may be a good number of nodes.
ZigBee uses two methods of routing. AODV and Source Routing. For command and control networks you will use AODV routing which can be very bandwidth munching. It it is only sensors reporting back their readings then you can use Source Routing which is not that heavy on bandwidth and you can have several hundred of nodes.
For reliable networks of more than 100 nodes, you need to make sure that your nodes will be deployed at distance from each other so they will be say 10-20 nodes in range of each node. Or if that cannot be guranteed then you have to use some clever ways to overcome these routing mechanism shortfalls.
This is from my personal experience with ZigBee other community members experience may be different than mine,
we have 3860 of nodes in one large mesh network. we use EM250.
we tested 2 years our mesh network before all fatal error find.
see our debugger stand
Today we use it in AMR/AMI solutions.
Larger network is lower final price of solution.
the goal is. Who is bigger?
@Manuel Malagon Thank you
I also have the same question。
The system was developed before I got here and with the hardware and ZigBee stack it was originally designed, there is a limitation of 20 nodes per coordinator/router which does not scale well at all since we are planning on deploying the system with networks of 100 or 200 nodes as minimum and it could be up to 1000. I'm trying to change everything to Silicon Labs Zigbee stack using EFR32MG12 chips but it's not clear on the documentation I've read if I can form a network with a minimum of 100.
The traffic is not much, but it's special. Because the network consists of "mobile" nodes, this is, the nodes may and will go out of range often so they will not be leaving the network properly. Because of this, we are transmitting a couple of bytes every 60 seconds or so from each node and use these bytes as a "keepalive" for the system. These keepalive bytes from the end-devices are to let know the coordinator that the end-device is still on the network because the node does not know when is going to be leaving the network and therefore it can't leave it properly.
This will be the general scenario and network activity. Also, since each node is logging data, then once a day the coordinator will query the data from each node, this is the heaviest activity because those logs consist of 3,800 bytes. But that is only once a day.
So I need to know if I would be able to have a network bigger than 20 nodes per coordinator because this alone will justify the change to all-silabs. The way the system works right now, with the keepalive mechanism, was dictated by the current hardware and firmware stack chosen at the beginning and it is not written in stone. If changing to silabs means also a better way to have a large network then that also would be really nice.
It's great to see everyone's contribution from unique experiences. As multiple users (ETL, Embeenpo, YK, to name a few so far) have alluded to, a Zigbee network can in theory handle a lot more than 100 nodes. In reality, reliable networks larger than hundreds of nodes may be rare but can be done depending on many factors (traffic pattern, application design, etc.). Keep in mind that network size is but one of a multitude of network performance indicators.
If it's of interest to anyone, we have published a collection of application notes and comparisons on mesh network performance:
Hi Manuel Malagon,
You should definitely not be limited to 20 nodes in a Zigbee network unless there are specific design and resource constraints not apparent in your description. Hopefully all the comments on this topic have given you a general idea, but if you run into any problem that's not easily resolved by existing forum posts please feel free to start a new thread with details of the problem and we'll help you tackle it :)
@YK and @Yuping Xiao
The network was originally designed using XBee Pro S2C modules and those have a 20 nodes limitation I don't know why, and this is confirmed by Digi themselves. I don't know if that is because they are using an EM357 chip or there is another reason.
This nodes limitation, among other things, are the reasons I want to get rid of XBee modules and go directly with Silicon Labs. And the maximum number of nodes alone is a good justification for that.