I recently setup a ZIgbee network of approximately 25 Aqara leak and temp sensors. Coordinator started off as a Skyconnect, but due to range and disconnect issues I switched to a Sonoff. I have 8 Third Reality smart plugs for routers spread throughout the house. No matter what I do the visualization shows pool connection quality and I continue to get dropouts. I feel the amount of powered routers I have is overkill so i’m at a loss what to do other than throw $300 in Aqara sensors in the garbage. Are they the root issue? I have manually tried to pair endpoints to the closest router. Some are literally on the other side of the same wall as the router and I still get disconnects. Ironically the sensors furthest away from the routers appear to be the most stable. I have used a Wifi analyzer to clear interference with my 2.4ghz wifi as much as possible.
At this point I’m at a loss and looking for suggestions. Thank you.
Edit: (Thanks @dproffer )
System is ZHA
Coordinator is TI based Sonoff
Coordinator Firmware: Z-Stack 20210708
My connection problems (Aqara temperature sensor with Sonoff dongle using Zigbee2MQTT) seem to have almost gone away since I moved the Zigbee network to channel 20 - might be worth a try if you haven’t done this yet.
Some info you should share that may help folks help you:
what zigbee system, Zigbee2MQTT or ZHA or ???
Which Sonoff device for your coordinator, TI or Silicon Labs chip based?
What firmware rev in your coordinator dongle?
Share a good picture of your network visualization.
Some thoughts from reading your post:
‘disconnects’ is a problematic word when it comes to zigbee end devices of any kind. Each type of end device sleeps and wakes up on a schedule that the firmware developer burned into the device. You and your coordinators/routers have to deal with that fact. For you, that is getting your head around the fact that you will have end devices that are ‘disconnected’ often when you do a ‘point in time’ look at your network. This does not mean the end devices are not functioning correctly. Unfortunately, without a ‘look across time’ you really do not know.
Aqara / Xiaomi device (some) have some unique issues, no doubt, of this if you google this any other zigbee related forums. That said, I will be optimistic that you can get something working for the coin you have invested.
IMHO, leak sensors are a real challenge. Does not matter the brand, technology or other aspects. Fundamentally, there purpose is to sleep for long periods of time to conserve battery life (we are talking years here ideally) and then wake up and function on seconds notice of a leak, with human expectation of 100% reliability. That is a big ask for any bit of technology, lest on that was purchased for a very low cost. Oh and lets not forget that testing these leak detection devices is VERY hard. Just finding them a year on is hard!!!
Sounds like you are doing a ‘join via’ function on whatever network software you are using. That said, it is an unfortunate black hole as to whether a given device model will ‘stay’ with a given route/connection. And yes from my experience and I think from reading of others experiences, some Aqara devices are more likely to not ‘stay’. I played around with several Aqara temperature sensors a while back and one of them would not ‘obey’ no matter the firmware or what I did, I suspect that it had a different PCB revision or some other anomaly that I could not identify. But, it is now at the bottom of my ‘dead tech’ drawer, never to be resurrected again. How you ‘remove’ and ‘reset’ end devices from an operating zigbee mesh network, can in my experience, affect how the end device ‘joins’ the mesh the next time. Do a hard hard hard reset of end devices after removal and before addition.
Far from the scale you are dealing with, however in this Zigbee2MQTT network with a TI based coordinator, I have five Xiaomi devices that work fine. See picture below.
I have had good success with the 3rd Reality plugs in my networks, that said, not by plan, I have a number of different router devices around my zigbee network. And while I do not have any quantifiable data to support this statement : different combos of end devices and routers do seem to act differently.
You might try introducing a different router device into the mix. Don’t go crazy to start. But make sure it is a different zigbee chip in the test router device. For example, buy another Sonoff dongle, either the TI P or the Silicon Labs E and flash it with the zigbee router firmware. Place in somewhere on your network and then ‘join via’ this router device a couple of your end devices and see what happens over time.
Thank you for the thoughtful and detailed reply. I have edited my original post with the additional info. I had not thought that the sensors could just be “sleeping”. I thought they would all phone home enough to not time out with the coordinator, but that is interesting. Tough part is knowing if it’s sleeping or not going to be there for you when you need it.
Yes that is the tough question. As I said, the zigbee system does not offer, IMHO, enough tools to really assess the stability of a zigbee mesh network.
Looking at your network picture, again the problem is this is just a snap shot in time, it looks ‘relatively’ normal, or in the words of Mel Brooks in Young Frankenstein … Abby Normal
