RTOS/CC2652R: CC2652R ZNP stability problems

In order to not fill this thread with offtopic zigbee2mqtt questions, could you contact me directly on telegram? (@koenkk)

Yes, you can get all serial traffic by running with DEBUG=* npm start

I have the same result on the CC1352R1 LaunchPad with the XDS110 disabled. The CC1352R connects to the Ubuntu mini-board via the separate USB-UART adapter. ZNP worked just over two days before it hung. From it there is no UART or radio messages, except for MAC Ack. The Zigbee configuration is very simple, only 2 devices: Aqara LED light and Mijia door sensor. I attach the log files archive to the message, maybe it's interesting for someone.8637.logs.zip

Are you aware why the Link Status period for the ZNP devices is set to 1 minute again?

I've never provided a C1352R firmware, did you build it yourself?

Yes, I applied https://github.com/Koenkk/Z-Stack-firmware/blob/d70c4f14ad3f70b04757c04cd4ceee8d418bdc22/coordinator/Z-Stack_3.x.0/firmware.patch to the znp project for CC1352R and built the firmware myself.

I'm still not able to replicate this issue with my setup. Do you have a way to generate a RAW data log for the UART communication between the ZNP and the mqtt server? I would like to inspect all MT commands from your tests in order to see if we can find something that is messing up with the MT module.

I have been running a test that reproduces network topology like yours using two temperature sensors and one light from the Texas Instruments sample apps but using Z-Tool commands to configure everything instead of zigbee2mqtt.io and I'm not able to reproduce your issue with this setup. I need you to repeat this test using zigbee2mqtt.io and also to provide me the instructions to do the same.

You will need to compile 3 routers as follows:
2 Zigbee routers with temperature sensor sample app
1 Zigbee router with light sample app

Join the 3 routers to the network and add binds for them. The temperature sensors will star to send reports and you will be able to toggle the light periodically (I did it every 2 seconds). Please give it a try and tell me how it goes. If you are able to reproduce the issue with this it means that something within zigbee2mqtt.io is messing up with the ZNP and we need to find what it is.

I'm afraid I don't have to correct hardware to build such setup.

I will try with my setup but then with the TI Linux gateway instead of zigbee2mqtt.

What firmware did you use together with Z-Tool? With my firmware Z-Tool cannot find the device, is additional setup needed?

I'm using the ZNP project with the changes you posted here:

https://github.com/Koenkk/Z-Stack-firmware/blob/dev/coordinator/Z-Stack_3.x.0/firmware.patch

Here is the configuration that you need for Z-Tool to recognize the com port:

Thanks Jose that was the configuration I was missing. Will build a test setup now.

Managed to setup the coordinator and pair and control all devices using Z-tool. Will monitor if it keeps stable.

Hey Koen,

Do you have any updates to provide given your investigation?

Regards,
Ryan

Hi Ryan,

The setup with CC2652R + Z-tool is now working stable for 9 days, which makes me assume that it will also remain stable.

Note that this setup is not completely identical (e.g. zigbee2mqtt registers more endpoints etc.) 

I'm a bit lost in what the issue could be. On the CC2531/CC2530 with Z-Stack 1.2 Home and 3.0 everything works really stable for many users. 

Regards,

Koen

Can you try to replicate the issue using the zigbee2mqtt and just a temperature sensor sending reports to see if you can get the error and then do the same with Z-Tool.

I'm running more tests know, I've reduced the message interval (greatly) and things have been stable for 5 days know, will keep an eye on that.

I also saw that there is a new simplelink release (SIMPLELINK-CC13X2-26X2-SDK_3.10.01.11), I use 

SIMPLELINK-CC13X2-26X2-SDK_3.10.00.53 know. Has the NVOCTP driver issue been fixed in this release?

No, this was simply an update for the RTM of the CC1352P.  No bug fixes or new features were implemented.

Regards,
Ryan

Sorry for no updates lately, I might have found an easy way to reproduce the issue, currently investigating this.