I noticed someone else also talking about this in another thread:
I was driving around the past few days and was disappointed that the gateways in my city received so few packets from the One running the default tracking software. I then wrote my own code to make sure the RN2483 is driven correctly (code on my github page).
That didn't help, so I decided to program a serial bridge into the One so that I can manually send commands to the RN2483 from my computer and then see what the responses are. The same set of commands were also sent to a bare RN2483 module. A gateway 3km away picked up the bare RN2483 (with a random piece of wire as antenna), while the Sodaq One with it's provided antenna wire was not picked up.
This led me into investigating the RF path on the One. I noticed the RF Low and RF High outputs are connected to each other with 0 ohm resistors. This can't be good as the different matching circuits for the two RF chains will interact and cause a mismatch. I therefore removed the 0 ohm resistor closest to the RF Low output on the RN2483 as I am only using the RF High output (868MHz). After this the One was still not received by the gateway that received the bare RN2483.
Adding extra "stub" lines to the RF path also changes the impedance, causing less energy to flow to the antenna. On the One there are 0 ohm resistors connecting both the UFL connector and the pad for an SMA connector. As both are connected to the RF path, the one will form a stub while the other is in use. As I'm only using the UFL connector, I removed the 0 ohm resistor going to the SMA footprint. After this modification the One was received by the gateway.
Back side of the Sodaq One after modification:
Having all these 0 ohm resistors on the One connected can't be correct. Is this a bug in the final design, or are users expected to remove the unnecessary 0 ohm resistors themselves? Removing these 0 ohm resistors are not that easy.
EDIT: 2016-08-06 12:05 UTC
I just came across more people mentioning that their Sodaq One performs less well than their other Lora nodes: