Trouble with nodes going offline after a while

  • I have a mysensors network setup, my config is a WemosD1 mini for a gateway, RFM69 radios for comms. Using the latest stable version of mysensors, the new RFM69 driver

    The network works fine, i have 4 sensors in total. All battery powered, some are a mix of boost converters and direct power, one is 1mhz and the others are 8mhz.

    My problem is that after a while the all the nodes one by one go offline, they come back when you hit the reset button and occasionally they come back to life after the contact sensor is triggered but it's a bit like a game of whack a mole keeping them all online.

    Another way I have had issues is if the gateway goes offline for a few hours the nodes never recover so after sorting the gateway I have to manually reset every node.

    I am using my controller and have set the system to alert me whenever a node stops communicating for 30 mins. Each node is set to transmit some data every 10 mins as a heartbeat (rx strength).

    Do i need to do something on these to enable the watchdog timer? I can upload the sketches if required but for the most part they are the example BME sensor and contact switch ones.

  • @ryanmt I have experienced this exact same issue and not yet found a definitive solution, but......

    1. Smooth power supplies are essential. Adding extra ceramic/electrolytic caps may help.

    2. Bad connections. This is a real pain to track down as although the dupont connections look fine, they can expand and contract with temperature changes and this can cause issues as well.

    Solder as much as you can to eliminate bad crimp connectors.

    3.Boost converters should be avoided if you can as many are poorly filtered and are noisey both in their own power output lines and radiated noise from the charge pump switching.

    I now send sensor data every 5 minutes regardless to if it has changed at all instead of heartbeat. Heartbeat only shows the node is alive, sending data also shows the sensor state and is for me a better option.

    I am not sure if it is the GW or some implementation in mysensors that affects things this way, but if it is the latter jkandassa will find it and fix it.

  • Thanks for the reply.

    I'm using the easypcb for the nodes which have a cap on the radio to aid with stability.

    I have checked my connectors, i'm pretty confident (and experienced) in putting bits together, to have bad connections on 4 nodes would be very unlucky!!!

    Re the boost converters - So far the bootsted ones are more stable than the 1mhz direct drive from the battery ones.

    Tis odd. I think i need to get my head around the watchdog timers better to reboot the arduinos automatically.

    I have tested a node without mysensors too - just running a temp sensor and OLED, that has been reliable.

  • @ryanmt I had these issues again today - I am using pro mini with nrf24l01+ so it's not them as we are using different kit...

    If I ever find a solution I will let you know.

  • @skywatch

    hey. i was having the same problems from time to time - i found out that its mostly wiring - therefor i am a great fan of easypcb and other boards that let one mount the radio directly.
    and also mounting a fan on my gateway (rpi3b) helped a lot with that. other than that repeaternodes...

    also install myscontroller - makes debugging way nicer and more efficient because you see in realtime whats going on... you might not have just one problem but more at the same time.

  • @zen85 Thanks for the feedback - I do have a node where pin D9 was a bit duff - It would work for ages and then suddenly the node ceased sending data. I eventually tacked it down and 'squished' the connector with pilers and it's been solid ever since. I also have a big heatsink and temperature controlled fan on the pi.

    Now it has done its' testing I will build it into an enclosure and solder all wires.

  • A bit of bug tracking today and I noticed 2 things that would cause problems with the mysensors network.

    1. I accidently swapped values on a node and it was sending data as fast as it could - this would eventually crash the gateway - so always check all nodes are sending data as often as you think they are!

    2. Loose connections - been here before but a bad connection on pin9 of promini/uno will cause the node to 'die'.

    Just thought I'd update on this as promised.

  • Plugin Developer

    I've had so many wiring issues that I always use either the "Nano Wireless Expansion Board", where the NRF module just plugs in, or I use devices that have the radio built in, like the RF-Nano.

    Use the MyDebug function to check on your radio. And generally debug your sketch to make sure no race conditions occur 🙂

  • @alowhum I actually orderd some of those nano boards by mistake (thinking they would work with promini). I don't have a nano so not much use, but I will get one today to try.

    Regarding bad connections, dosen't using that board double the failure points?
    With cable to the node you have the cable joint and node soldering joint.
    With this board you have the cable joint, node soldering joint, socket connection and socket soldering connection. - Just a thought! 😉

  • Plugin Developer

    Not all failure points are the same. The Dupont wires break easily. Wires embedded inside hardware don't. Since I use these boards I've never had any trouble.

  • @alowhum I know - I was being pedantic (unusual for me) 😉

    I have had really bad batches of dupont cables from online sellers, so much that I bought a crimp tool and connectors and now make my own, but only for testing and prototyping. Never for finished items as they are all soldered.

  • Thought I would update this thread, still struggling but made some progress.

    I added the watchdog which helped somewhat, so far I have found a few things which I have fixed or investigated further on my efforts to make things more reliable.

    My gateway crashes after a few days - unsure why, suspect its a power issue so I have ordered a standalone 5v to 3.3v power supply to run the radio. My setup is a Wemos D1 mini. I am wary of changing too many things at once but I am considering wiring directly to a PI for a gateway.

    My side gate sensor used to drop off and come back after an hour or two. I noticed the TX was hovering around 90-100% so I thought it could be signal related, I built a mains powered repeater with the RFM69HW and placed it nearby. The side gate sensor favours the direct connection to the gateway but it seems to be more reliable. This sensor is a 1mhz/direct off the battery sensor.

    My BME280 sensor crashes a lot, I think this is related to the temp sensor rather than the node, seems to stop returning measurements except the battery voltage for a few hours then stops entirely. This is a sensor with a boost converter and when it crashes it never comes back, has to be reset. I am going try a different library for reading the BME280 when I get chance. I have also observed it gets less reliable as the battery voltage drops below 2.5v.

    Next up with annoying failures is my front gate sensor, previously most reliable it stopped working for 12hrs before magically reconnecting today, no reset required. It was rock solid till the battery voltage dropped to 2.5v, it's working now but I am going leave it to see if as the voltage drops the sensor becomes more unreliable. This sensor has plenty of signal strength to spare so I don't think the relay will help, its probably only 8m from the gateway.

    The boost converters I am using are the cheapest ones commonly available from china. To continue the experiment I have ordered some slightly more expensive ones to see if the performance changes.

    I do have the capacitors in place, but am wondering if its worth experimenting with the values.

    People mentioned soldering/connectivity issues. I don't think thats my issue here. All my sensors are EasyPCB Rev9. Most of my sensors have no wires (ie no sensors, they're just testing their existence) or simple stuff like contact sensor or relay control

    This is the relay I made, its got a custom PCB under the sensor with a HLK-PM03, fuse, thermal fuse and a few other bits. Debmowsk designed the circuit originally I just rejigged it so it can be stacked on top of the easypcb rev9






    Far right of the graph shows the period the front gate sensor was offline for.

Log in to reply

Suggested Topics

  • 7
  • 1
  • 3
  • 4
  • 3
  • 12
  • 2
  • 2