I’m having a problem with multiple devices having disconnects and requiring our router to power off the modem to reboot to bring connection back up. Logs show
RSRP is -94 and SINR is 16 and RSRQ is -11.
Connects to Band41 and speeds are good.
When this issue occurs, the LTE interface loses connectivity to the tower but there is no indication on why. The signal level is not very high but also not near a potential disconnect level.
Do you have any ideas on additional logging I could setup to determine why this is happening?
Could it be because of insufferable power supply when downloading a large file the modem tries to aggregate two 5G bands or more? You can rule that out by temporarily disabling all 5G bands except N41.
I’m having similar problems, but in my case the signal is really bad (all carriers in our area are poor, due to NIMBYism blocking the building of any antenna towers near us).
Anyway, I did the research with the help of snowgum, who told me about the official Hardware Design Guide for the modem.
According to the guide, the RM520N-GL draws a max of 3.7V at 5.0A = 18.5W. It requires a minimum of 3.7V at 3.0A continuous current = 11.1W. That’s with no sag in the voltage or current.
As far as I know, no USB to M.2 board puts out a rock-solid steady 3.7V at 3.0A.
I am currently using an MCUZone USB+Ethernet to M.2 board, and it disconnects all the time. I can’t prove that it’s the power supply because the poor signal is always an issue for me, but others here have stopped using the MCUZone boards and said that rework.network boards are better. But even those only claim a max of 10W power supplied to the modem.
I’m hoping that with enough demand, some mfg. will make a board that supplies a clean, rock-solid 20W to the modem to meet all of the modem’s demands.
We found you can somewhat work around this in 2 ways:
better antenna / antenna placement - this reduces the TX power needed
disable 5GNSA - the combination of LTE and 5GNR with poor signal maximises the power draw. I haven’t tested 5G SA - but I would hope it is also better.
Of course the real fix is to supply enough current. In our case this means using an M2 daughter board on our SBC and making sure we have chunky USBC PD power available.
Follow up on my situation: I replaced the faulty MCUZone board with a rework.network board (the one with Ethernet and dual SIM) and the connection became stable (stays in 5G NSA all the time, even with poor signal).
Funny thing is, the rework.network board brags that it supplies a maximum of 3.8V at 3.0A, but this is the MINIMUM required for the RM520N-GL to work. It works, but I am leery that it is right on the edge.
Today, I found that they have this newer board that claims to supply 20W here. I wish I had bought this instead:
Going forward, I will buy the latter one because it claims to provide 20W, even though it doesn’t have as many features. Most likely, they will update the one with Ethernet and dual SIM to supply 20W, and then that is what I will buy.
I’ve metered the power on our setup and (including an octocore ARM CPU and h264 encoder) we rarely exceed 11w - But adding an attenuator on the 5g TX antenna pushes it higher.
Yeah, the signal here is so bad for both 4G LTE and 5G on T-Mobile that it is/was effectively the same as having an antennuator on a good signal. Ours is literally a rare edge case (we’re on the very edge of the coverage area where dropped calls are routine). If we moved the entire setup just 1/2 mile towards the nearest antenna, it would jump from 1 bar to 4 bars of signal and our RM520N-GL’s power consumption would drop dramatically.