search

Koenkk / zigbee2mqtt

Zigbee 🐝 to MQTT bridge 🌉, get rid of your proprietary Zigbee bridges 🔨
https://www.zigbee2mqtt.io
GNU General Public License v3.0
12.04k stars 1.67k forks source link

CC2530 + CC2591 needs hardware cycle for every pairing #3891

Closed SaturnusDJ closed 4 years ago

SaturnusDJ commented 4 years ago

Using this as coordinator with Z-Stack_Home_1.2 (source_routing).

First off, https://github.com/Koenkk/Z-Stack-firmware/tree/master/coordinator here it says it can handle 5 direct children. Why do I see it having 6 direct children?

But the real issue: pairing is a mess. It seems for every pairing the USB connections needs to be remade and Zigbee2MQTT restarted. Anyone has a clue why?

Edit: it could also be an interval after which pair is not possible until hardware cycled.

Koenkk commented 4 years ago

Can you share some more details about your network (how many routers/end devices)? And where do you pair those devices (close to the coordinator or routers?)

I think this issue occurs because the child device table is full, after power cycling, this table is cleared and populated again as devices reconnect, so until those devices reconnected some can still directly join the coordinator.

SaturnusDJ commented 4 years ago

Thanks.

At that moment (after starting from scratch with the new adapter) I was trying to connect the first end devices via routers. At first from meters distance, then moving closer and closer until direct contact. At one point I even removed the hardware antenna from the coordinator to prevent end devices and new routers from directly connecting with it, if that is how it works? It did not help. In the end it somehow came together, I wish I knew exactly how but it was with lots of re-pairing. What was especially weird was how the end devices (all HS3SA as I started with those) did pair, but their smoke reporting feature did not come through even though the devices were reported as successfully configured sending the configure command also did not work. It took those re-pairs to finally get them to report if smoke was detected (test button to trigger). After that, I joined more and more routers which still connected with the coordinator (I thought 5 was max...) and also a bit more end devices which connected via routers.

This is the network now: network

Considering 16 more light bulbs and 20 Xiaomi sensors end-devices.

Koenkk commented 4 years ago

To get further with this we need to have a sniff of the problem, would you be able to provide such? (https://www.zigbee2mqtt.io/how_tos/how_to_sniff_zigbee_traffic.html)

SaturnusDJ commented 4 years ago

I would do it, however the problem is gone. After those hours of re-paring in the end it worked. No hints to what the solution was. I kept removing those HS3SA devices until they configured and worked correctly. Maybe they are picky and/or what you said about the coordinator table.

But even now, when adding new routers they always also show to have a link with the coordinator. Is that really the case? As the max. should be 5 according to the firmware page.

The network seems good, so much better than CC2531. New devices join quickly and in the visualisation I have not seen unconnected devices yet, as was often the case with CC2531.

SaturnusDJ commented 4 years ago

@Koenkk Apparently the problem is not gone. While MCCGQ11LM, WXKG11LM, WXKG12LM, I think joined without problems, HS3SA (earlier), RTCGQ11LM, WSDCGQ11LM, MFKZQ01LM have a lot of trouble. Maybe the ones willing to join was luck. The MFKZQ01LM all of a sudden did join... I can't find logic behind it.

Flashed the old CC2531 for sniffing. What specifically should I look for now?

A lot of data is coming in. For example Many-to-One Discovery: With Source Routing Or: ___ _pipe_zboss_sniffer 23-7-20 13_41_48

This is the network size now:

Zigbee Map - Home Assistant - Mozilla Firefox 23-7-20 13_44_52

SaturnusDJ commented 4 years ago

From trial and error by re-pairing the MFKZQ01LM it seems: Beacon kind of messages mean a new device is near. Device Announcement means a new device is going to join.

After finding this out I saw the WSDCGQ11LM device not wanting to join also not even triggering that beacon messages. Replaced battery and it paired immediately. 🤦‍♂️ Seems like pairing requires quite a bit more of energy compared to just being in a network and sending out updates once in a while.

Now this makes sense for some other devices which are running on 1-1,5 year old batteries. But...I think this is not the fix for this problem if it is still present because the HS3SA devices had a near full battery.

Will report again if new trouble occurs.

SaturnusDJ commented 4 years ago

Trying more sensors of type WSDCGQ11LM. 3 fails, 1 join.

Let me know how to proceed with sniffing.

Edit Finally all 10 or so joined. I have no idea how. It could be:

Koenkk commented 4 years ago

A typical joining procedure looks the following:

image

SaturnusDJ commented 4 years ago

Thanks, I will report back with logs once new Zigbee devices have trouble joining. 👌

SaturnusDJ commented 4 years ago

Well, a WSDCGQ11LM became half alive, back from dead, so results are in already:

Capturing from __ _pipe_zboss_sniffer 23-7-20 21_32_25

More on it: Am I correct it tries to join via device ending with af:52? This is a Philips Hue Lightstrip Plus. Just now I tried again and it succesfully joined via a Blitzwolf BW-SHP13.

Looking at my network map...no end node is connected via a Philips device. Strange...

Koenkk commented 4 years ago

.....end?

Now the transport key should be send by the coordinator or router (I forgot who is responsible for sending this when the device joins via a router). Two options here:

image

Firmware: CC2530_CC2591_source_routing_join_debug.zip

Am I correct it tries to join via device ending with af:52?

yes

SaturnusDJ commented 4 years ago

More proof of the problem being the Philips Hue devices.

There was trouble joining IKEA LED lights. After some tries, 4 of the 6 joined. The other 2 stayed problematic until the Hue devices were unplugged from power. Also I saw at least one failure when the light tried to join via the Hue light, same like before.

Can that debug-firmware be flashed without having to re-pair/re-setup the devices/network?

For the next paring, if not that firmware, I will pair next to those Philips Hue devices, them being on at first, off afterwards.

Koenkk commented 4 years ago

Can that debug-firmware be flashed without having to re-pair/re-setup the devices/network?

Yes, no repairing is required

SaturnusDJ commented 4 years ago

Would it be possible the Philips Hue devices are not integrated fully because I did not reset them (https://www.zigbee2mqtt.io/information/touchlink.html)? Instead I removed them from within the Hue app and they automatically joined the Z2M network.

Koenkk commented 4 years ago

Don't expect to, both should give the same result.

github-actions[bot] commented 4 years ago

This issue is stale because it has been open 30 days with no activity. Remove stale label or comment or this will be closed in 7 days

Morphy99 commented 3 years ago

I have this issue with pairing not working until I physically power off the CC2530. Restarting or rebooting my RPi didn't resolve it. I had it while running the default 20190608 firmware and now with source routing 20190619 fw on my new instance. Both were running connected to a RPi4 GPIO. When this happens I don't see any other issues with the network. I've had it when I've been joining multiple devices and it fails after 1 or 2 or when I enable joining and try to add 1 it can also do it.

SaturnusDJ commented 3 years ago

Today I switched to Zzh. Pairing routers went fine, pairing end-devices after a bunch of ten turned out a hell again. I did not check what the problem was. Nevertheless I can report that I see three end-devices connected to a Philips Hue light. I am not 100% sure if this happened before but this was extremely rare here. Maybe Zigbee 3.0 is the reason for it happening.