CC2630: Intelligent retry behavior when network lost

Hi Jose,

many thanks to your suggestion, it sounds promising. However I have difficulties in reproducing the behavior.

If I understood right then changing

if (startMode == MODE_REJOIN && scanCnt++ >= 5 )

to

if (startMode == MODE_REJOIN && scanCnt++ >= 2 )

should reduce the amount of retires on the same channel. But I could not see any change with a sniffer (yes I recompiled the EndDevice project).

How exactly do I restart the EZMode? From the SampleSwitch original code this was done with the key right press:

    if(keys == KEY_RIGHT)
    {
#if defined (ZCL_EZMODE)
        // Start EZMode Commissioning
        {
            zclEZMode_InvokeData_t ezModeData;
            // only bind on the on/off cluster
            static uint16_t clusterIDs[] =
            {   ZCL_CLUSTER_ID_GEN_ON_OFF};

            // Invoke EZ-Mode
            ezModeData.endpoint = SWITCH_EP; // endpoint on which to invoke
                                             // EZ-Mode
            if( (savedState == zstack_DevState_DEV_ZB_COORD)
                || (savedState == zstack_DevState_DEV_ROUTER)
                || (savedState == zstack_DevState_DEV_END_DEVICE) )
            {
                ezModeData.onNetwork = true;   // node is already on the
                                               // network
            }
            else
            {
                ezModeData.onNetwork = false;  // node is not yet on the
                                               // network
            }
            ezModeData.initiator = true;        // OnOffSwitch is an initiator
            ezModeData.numActiveOutClusters = 1; // active output cluster
            ezModeData.pActiveOutClusterIDs = clusterIDs;
            ezModeData.numActiveInClusters = 0; // no active input clusters
            ezModeData.pActiveInClusterIDs = NULL;
            zcl_InvokeEZMode(&ezModeData);

            LCD_WRITE_STRING("EZMode", LCD_PAGE2);
        }
#elif defined (ZSTACK_MANUAL_START)
        Zstart_discovery();
#endif // ZCL_EZMODE
    }

    // update the display
    Switch_updateLcdDisplay();
}

So I added something like this in our code:

            if (hdrStatus == zstack_ZStatusValues_ZSuccess && reqStatus == 0)
            {
            // OK
            }
            else {
                // we already made 3 attempts for user notification, see APSC_MAX_FRAME_RETRIES in f8wConfig.cfg
                sos_state = sos_state_SEND_FAILED;
                // avoid exhaustiv channel scans

                zclEZMode_InvokeData_t ezModeData;

              ezModeData.endpoint = RB4_ENDPOINT_SOS;
              ezModeData.onNetwork = true; // node is already part of a network
              ezModeData.initiator = true;
              ezModeData.numActiveOutClusters = 2;
              ezModeData.pActiveOutClusterIDs = rb4_g_outputClusters;
              ezModeData.numActiveInClusters = 2;
              ezModeData.pActiveInClusterIDs = rb4_g_inputClusters;
              zcl_InvokeEZMode(&ezModeData);
            }

But I could not see any change in the behavior regarding beacon requests in different or the original channel. So, how do I restart the EZMode?

Hi Simeon,

The changes I posted are only visible when the node is in rejoin mode (had a network before) and I leave the buttons ins the application like in the default sample application. Try to just use a clean fresh project and just add the changes, then start a network with a coordinator and join the end device. If you perform a power cycle in the coordinator and then press the right button to run the EZ mode again. That's what I did, please tell me if works for you. After that we can work on other changes that you want to include.

Regards,

If I understand correct this would stop the status message interval until the ZDO_JOIN_CNF_CBID gets executed. So after a parent loss (out-of-reach), it would stop the status messages until the enddevice re-joined the network again (realignment?). This would require the user to trigger the button. However we need the status message to be sent every 4 hours, whether the device lost its parent or not in previous transmissions. In case the parent is in-reach again, the status message should "just work".

We have an application running behind the receiver of the status messages. In case the status message is missing for 72 hours an alarm is generated. Therefore the status messages must be continued without user intervention.

I will try  your suggestion with ZDO_RegisterForZdoCB( ZDO_JOIN_CNF_CBID, &Your_App_Callback); on the clean SamplSwitch project. Maybe it does exactly what we want. If not, please assist me further. Thank you in advance!

Hi,

to be honest I don't understand at all how ZDO_RegisterForZdoCB( ZDO_JOIN_CNF_CBID, &Your_App_Callback); will help me with the problem of channel scans. This just provides a mechanism to issue regular status messages. We already have such a mechanism implemented. How does this circumvent the channel scans we want to avoid?

Hi Simeon,

If I understand correctly your requirement is to send the status message every 4 hour even if the EndDevice doesn't have a parent. That behavior is not compliant with the Zigbee specification, an EndDevice cannot broadcast messages to the network as a routers does, an End device can only talk to other devices in the network through his parent including the broadcast messages, so to send the status message you must search for a parent first.

Regards,