Whats makes a strong mesh?


#1

I always thought i had a strong Zigbee mesh (- the fact that they are Zigbee lamps) I don't seem to have any issues (- some oddity a few months ago when a device was about to fail). Most of my devices are Osram lamps X 7 with 4 Philips hue spots (tor replace the oddity issue). I also have 2 x Osram buttons, 2 X ST presence sensors and a Exbee. Hub is C4

Today though, don't know why but i decided to look at the Zigbee logs and most off the RSSI values are 80+ so shouldn't i be having issues?

my kitchen has the 4 Philips spots then there is a doorway (no door) to the dinning room with a pendent chandelier lamp, in that room is the coat cupboard with the hub (ground floor centre of the house stairs up next to it) this leads to the hallway with the front door with 2 more pendents in. Off this is the front room with the Xbee. Then 1st floor is bathroom bedroom 2 (LarB) and 3 (SmB) and each bedroom has a Zigbee pendent. Then up to top floor bedroom 1 (MB) with 2 zigbee spots on front window.

question is for those of you with "strong" meshes how many devices do you have within a room and whats the kind of values you get?


#2

Most people say that osram smart bulbs, like many zigbee bulbs when paired in ZHA mode, do a bad job of repeating in a zigbee network.

That may or may not mean in your case that you have a “problem” per se, especially if you haven’t really been having actual problems.

But I would think that’s having some influence on how your zigbee mesh is routing. Could you use the Xbee to get a better visualization of your mesh?


#3

I'm not an expert in the field, but the example screenshot in Hubitat's Zigbee log docs show RSSI values quite similar to yours, albeit generally slightly better, and I can only imagine those were taken from the functioning hub of a staff member. Nearly all of your LQIs are excellent. I'm not sure if the RSSI is an indicator of a problem since the docs only tell us that "close to 0 is good" but I don't think actual 0 is possible in the real world, and the docs don't provide any indication as to what acceptable values may be (and they really suggest that the difference could be in either direction, presumably as if the signal would spontaneously amplify...).

In any case, if the only Zigbee devices you have on this network are bulbs, you're likely fine. The issue with bulbs is what was mentioned above, but it's generally described as a problem with bulbs dropping messages from non-bulb devices (I theorize it's because they are meant to work on Zigbee Light Link networks, where the demands are less, but being joined to a Zigbee Home Automation network like Hubitat overwhelms their comparatively limited resources). But I'm not sure an occasional total loss would affect your RSSI, since it seems to me that this value can be calculated only when something is truly received.

But I'm also curious what typical real-world RSSI (and LQI, I guess) values look like. I'm not sure if my mesh is objectively "strong," but I have mostly Zigbee devices. I also have a second Zigbee network (Philips Hue) and my own 2.4 GHz Wi-Fi network, plus who-knows-how-many of my neighbors' networks bleeding into my house. LQI values I see in logs since I've had this post up range from about -51 (Xiaomi temperature/humidity sensor a few feet away from a Tradfri outlet) to -78 (many are around this value, including an ST ADT water sensor sitting on the floor a few feet above an Xbee, my Hampton Bay Fan Controller that is right next to the same Xbee, and an Iris v2 motion sensor in my basement that has to go through either a wood floor or partial brick wall to get to a router). Most of mine are in the -70s. I have a Zigbee router in about every other room. I know some people go all-out and put one or two in every room, but I'm pretty happy with my network as-is, and with a two-story house I think I benefit from having routes available in multiple directions (though I'm surprised the ones just above the Xbee, generally an attractive route according to my scans, don't seem to be as strong as I'd expect). My LQIs are almost always 255, though I see one 250 and 254 in my recent logs.


#4

Those values look very similar to what I get.

My hub is located in the bottom corner of my basement in a 3 story house. I have about 60 zigbee devices (most of which are xiaomi stuff) and only 4 Ikea outlets as repeaters (3 main floor 1 upstairs).

They all report fine. The furthest out I have is 2 tilt sensors on the garage doors. I also have one contact sensor in the mailbox but the repeater is on the other side of the wall from that. So it goes through the mailbox and brick wall but hits the repeater fine.

I get notifications if any of them have not reported in in at least 3 hours so I will know if any are giving issues.

It did take me a long time to get it this stable though. I tried many things but now am happy that I did.


#5

Unlike RSSI, LQI's take into account link error rates. A LQI value of 200 or better supposedly corresponds to an 80% or better probability of a packet arriving intact (not requiring a retransmit) and can probably be considered a 'good' link. If you're seeing slow performance you may want to pay attention to those with LQI <200.

As far as my network's RSSIs, most run in the low -60's to -80's with a few outliers. Anything that routes through the Zigbee plug which is a foot from my hub shows an RSSI of -34 which is the strongest seen on my network. My 'worst' link is showing LQI of 134 and RSSI -84; it's an Osram bulb over my kitchen sink. In spite of the shockingly poor link quality it still seems to be working fine (I'd definitely notice its failure to operate as it is controlled by a motion sensor in a high traffic area).


#6

I just want to share that I know one or two Zigbee stack still took a short cut and calculate LQI straight from RSSI without considering other factor such as error rate. There is an advantage for this type of primitive LQI calculation in regard to path cost calculations. However, In this type of devices, lower LQI does not mean it has bad connections. Lower LQI value in this type of devices may still be in the operating range of its sensitivity specification to receive good packet.

I also observed some stack also weighted too heavily on error rate to calculate LQI. In my home Zigbee mesh I see the same type of device that is 2 feet away that has the same LQI (255) with a device 10 feet away. They should not have the same LQI. They should have LQI that is in the range between 200 and 255 to indicate that they have good connection but one has lower cost to route than the other.

Here is some of the example that I found about LQI calculation. TI Z-stack by default also use RSSI to calculate LQI although one can override it with their own implementation.

I suspect that we are not in the time where device has standardized the way they calculate LQI values. I am hopping that this is improved over time.


#7

Good information; that probably explains why my kitchen sink bulb works so well in spite of its bad link numbers.


#8

All good to hear :grin:

I do know about this but as I'm only repeating lamps it's shouldn't be a issue, but that's the reason I asked. As although I plan to add a few more I'm not planning on adding as much as it looked like I needed, so I would need other stuff ie repeaters or sockets which would then potentially make it worse as the lamps would them become unreliable trying to repeat them.

I have tried but it's not as clear as I hoped, would be good if hubitat could populate something like it with real names.

Never knew this one a thing will swat up on it, thanks :wink: