Best posts made by frapell

Thanks a lot! that's the only thing I haven't tried... I assumed that since the raspberry pi was working fine, the internal power regulator was providing enough juice to the module through the gpio pins...

This is quite weird, I replaced the power supply, everything started worked fine. So I said, let's try the old power supply again, it should be broken... Everything is working fine too... This is quite insane.

Also, I found https://www.raspberrypi.org/forums/viewtopic.php?t=190584 and I ran that in my rpi, and I get 0x0 with both power supplies

# vcgencmd get_throttled
throttled=0x0

According to https://forum.libreelec.tv/thread/17860-how-to-interpret-rpi-vcgencmd-get-throttled/, 0x0 means all is fine with the power supply, so not sure what is going on...

I will leave it running for some hours and see what happens...

@skywatch Yeah, I have spent a lot of time fighting those as well, however in my case, I have designed a special PCB, which connects directly to the GPIO pins in the rPi, which should avoid these issues:

As per the power supply, I have been using a PSU from the same people I buy the rPi from, which is a 5VCC, 3.1A, for months until this started happening. After @Yveaux suggestion, I tried my pixel charger, which should deliver 5V - 3A, and as I said, it worked fine for about 10 hours straight, but then stopped working again. Furthermore, the rPi is not reporting undervoltage or throttling, so not sure if it is the PSU at all... I don't want to take appart the PSU, but if there are no more ideas on what else might be, I guess I'll have no choice...

@skywatch Ahhh, makes sense now thanks!

@ejlane I got the specifications from the PCB manufacturer (Attaching below), and the copper thickness is 35 microns...
I managed to get the module with the antenna pin right next to the connector, so the trace length would be just 4mm, for what I was able to find, my case for a single sided PCB would be a "Coplanar wave guide" and plugging all of those values in a calculator I found in KiCAD, it seems that I am fine with a 1 mm wide trace, with a 0.25mm gap with the GND around it.

Hope I got it right! Thank you for your suggestions!

@ejlane @Yveaux @skywatch So, I have found exactly what was going on, and had nothing to do with the hardware... Basically, one of my nodes was jamming the radio channel

How I discovered the issue:
There are 2 more flags, besides MY_DEBUG which are MY_DEBUG_VERBOSE_RFM69 and MY_DEBUG_VERBOSE_TRANSPORT_HAL which I couldn't find a way to add, so I did it by editing the configure script and where it says

if [[ ${debug} == "enable" ]]; then
    CPPFLAGS="-DMY_DEBUG $CPPFLAGS"
fi

I changed it for

if [[ ${debug} == "enable" ]]; then
    CPPFLAGS="-DMY_DEBUG -DMY_DEBUG_VERBOSE_RFM69 -DMY_DEBUG_VERBOSE_TRANSPORT_HAL $CPPFLAGS"
fi

Then re-ran the script and recompiled the mysgw binary. The important flag in my case was MY_DEBUG_VERBOSE_RFM69.

When mysgw was running, and everything was working fine, I was seeing lines with DEBUG RFM69:CSMA:RSSI=-102 popping in once or twice (with different values for RSSI) and everything continued normally. Then, when the whole thing was wedged and nothing worked, I noticed that message pretty much going on forever

Sep 09 18:37:11 DEBUG RFM69:CSMA:RSSI=-52
Sep 09 18:37:11 DEBUG RFM69:CSMA:RSSI=-52
Sep 09 18:37:11 DEBUG RFM69:CSMA:RSSI=-53
Sep 09 18:37:11 DEBUG RFM69:CSMA:RSSI=-52
Sep 09 18:37:11 DEBUG RFM69:CSMA:RSSI=-52
Sep 09 18:37:11 DEBUG RFM69:CSMA:RSSI=-53

And on and on and on.... forever... Clearly there was a loop somewhere for whatever reason printing this message. As it turns out, there's a function that the RFM69 driver calls before sending a message https://github.com/mysensors/MySensors/blob/2e00bf6a10f76d6aaa1999e12313237bc3edabd3/hal/transport/RFM69/driver/new/RFM69_new.cpp#L369-L375 that pretty much checks if there is noise in the RF channel, before actually sending anything...
I was doing some modifications to one node I have, where I replaced the arduino with another one, which I thought was a 3.3v one, burned it as so, but it was a 5v. For whatever reason (maybe the different clock speed? dunno...) when doing that, the node will send garbage through the radio... I have no idea how it manages to init the radio... doesn't really matter, the thing was sending something in the same channel constantly...

And that is why the gateway was receiving data from other nodes, but was never able to send back... this other node was preventing it because it would never shut up... I have now replaced the arduino with a 3.3v one, checked in the serial monitor after burning and re-installed, and everything is back to normal and working...

Thank you all for your help and suggestions, I will nevertheless build the new PCB with the suggested changes and replace the ones I am currently using.