Hi
I just did some testing with my new RPI gateway with 868mhz RFM69HW and dipole antenna, located at the top floor in my 3 floor house. Walking around with a battery node (also HW, but wire antenna) I tested every corner and floor of the house and the connection is solid everywhere, no need for repeater nodes.
Just for fun I took the node outdoor and was a bit surprised when I discovered that I had quite good connection inside my garage as well, that is some 60m away and with houses in between that I was certain would block the signal (see picture). So, it seems I don't need the RFM95 system I planned for the garage after all! I'm very pleased with the result

@NeverDie I use the NanoVNA-H. It's originally an open source project so there is a lot of info about it and how to use it, and it's cheap.

• https://nanovna.com/
• https://www.youtube.com/watch?v=THc5ZTxQUSg

I like the pogo-pin programming adapter!

Yes, you are probably right. I both cases the RPI still have connection with the router and the deamon is (apperently) still active but not communicating.
I have a wifi mesh network, and I have seen that clients are sometimes disconnected and reconnected. That might have something to do with it.

Thanks, and thanks for the tip, I will try to use MY_RFM69_NETWORKID.

@electrik I made the gateway dipole from some 1.4mm solid copper wire that I pulled from a cable. After tuning it with an antenna analyzer, I ended up with each element 80.4mm long for 868mhz. This is a bit shorter that the teoretical length, I'm not sure why.

I just did some more testing, took my node (with a push button) and mobile phone for a late walk. At 160 meters away from my house the connection is still good. At >180 meters away it randomly lost connection. The gateway and node was by no means in line of sight, there were houses, bushes and trees in between.

This is of course not very scientific as the range depends on a lot of factors, but the result is pretty good anyway

@evb, I really don't know why you get that high consumption of 133uA, but there is defenently someting wrong. Sorry that I cant help.

I use my own pcb design but it's basically the same as a mini pro without reg and led + RFM69HW. All my 328 chips is sourced from china and working fine.
I just measured on of my older designs with 8mhz/no BOD and I get about 6.7uA in timed sleep (WDT active), including 1M/470k voltage divider.
With the same setup without 1M/470k voltage divider and with external interrupt wake-up (sleep forever) I have about 0.8uA.

This was measured with two different multimeters with uA range. Sure, in this range a multimeter is not as accurate as the uCurrent or similar instruments, but IMO more than good enough for this kind of measurement.

If you are able to solve your main problem, I have a few other tips for further tweeking:

• Don't use bootloader with BOD, you don't need it for a sensor that is continous connected to the battery. Saves 20-25uA.
• If you power from 3V battery, don't use the suggested 1M/470k voltage divider for battery monitoring, you can do battery monitoring with code only (search here on the forum for it). Saves 2uA.
• You could perhaps gain a tiny bit more by setting all unused pins to LOW OUTPUT or LOW INPUT.

Some very good info on ATmega 328 power consumption here: https://www.gammon.com.au/forum/?id=11497

@evb 133uA is a lot. Why do you use the lowpower sketch? There are lots of ways to reduce consumption in deep sleep, and the Mysensors library sleep function does a pretty good job of it. For best results you should use a bootloader with fuse settings for 8mhz internal resonator and no BOD. 1mhz is even lower, but not so much to gain from it.

I like the pogo-pin programming adapter!

Pin-change-interrupt is not supported by the Mysensors library. I have used some diodes my self when I need multiple external interrupt sources, but there is also a workaround here: https://forum.mysensors.org/topic/6352/pin-change-interrupt-on-any-pin

@henry-north According to the serial output from the gateway you do have communication between the gateway and the node.

I think there is some issue with the Homeassistant persistence file, or perhaps you changed the child type in the node scetch after it was presented first time as pointed out by @BearWithBeard? If so it will no longer match the data in the persistence file and you will get the "cannot add child" warning, and (I think) not se the sensor in HA. Try to delete the persistence file (mysensors.pickle). Then first restart HA, then restart/present gateway and then reset/present the node(s).

You can change the persistance file to a readable format by putting this in your configuration:

persistence_file: '/config/mysensors.json'

Makes it much easier to see what's going on.

@Sebex The quickest way is to upload a new bootloader that use the Atmega 8mhz internal oscillator, like this one: https://github.com/MCUdude/MiniCore
As serial timing will be less accurate with the internal oscillator you might also need to include #define MY_BAUD_RATE 9600 in your scetch.

@waspie Did you try using Serial.print for step by step debugging to narrow in on the issues? For example; in the function readVcc(), what is the actual returned value when running on batteries? and so on...

@kiesel Have you tried to replace the sensor with a pull-up resistor?

@NeverDie I use the NanoVNA-H. It's originally an open source project so there is a lot of info about it and how to use it, and it's cheap.

• https://nanovna.com/
• https://www.youtube.com/watch?v=THc5ZTxQUSg

I just did some more testing, took my node (with a push button) and mobile phone for a late walk. At 160 meters away from my house the connection is still good. At >180 meters away it randomly lost connection. The gateway and node was by no means in line of sight, there were houses, bushes and trees in between.

This is of course not very scientific as the range depends on a lot of factors, but the result is pretty good anyway

@electrik I made the gateway dipole from some 1.4mm solid copper wire that I pulled from a cable. After tuning it with an antenna analyzer, I ended up with each element 80.4mm long for 868mhz. This is a bit shorter that the teoretical length, I'm not sure why.