First one : RSSI returned by base station seems to be expected received level in dBm. This value is always (I have tried several distances between device and base station) RF power - free space losses. Of course, RF power measurement is done in lambda62 pin 1 with a level meter.
Second one : SetPAConfig and SetTxParams are set to emit 14 dBm. I have checked hexadecimal values sent to SX1262. I have done some other tests with 0 dBm. Only one difference : RSSI returned by base station is 14 dB lower.
I come back. I have done some measurements with my HP 8592B.
Center 868 MHz, span 10 MHz
This measurement was done with a coaxial cable directly connected between Lambda62’s pin 1 and 2 (Rf output and ground) and when power output is set to 14 dBm. Of course, antenna was removed. Power was very similar to power I have found with my level meter.
I have checked my LoRaWan library and sent to serial console all transactions between CPU and SX1262:
I’d inspect one of your lower power units under high magnification for possible solder issues within the module, particularly the main IC and the RF switch.
Another thought might be wrong components in an RF network on some of them. Since you have a spectrum analyzer you could set one up with low power output in a closed-cabling (non-radiating) situation and frequency sweep it over several hundred megahertz. A good one should show networks optimized in the 868 and/or 915 range, one with a bad network might show a different shape or have its best performance elsewhere.
All modules have the same RX sensitivity. I will investigate, but I have checked solder issues with my binocular (20x). I haven’t found solder issues and I have tried to solder a coax before RF switch to check TX power, but without success
Even if I had to create a software scalar network analyzer by using another as a receiver - which would mean I’d be seeing through the network of that one, too. But if the receive side was a “good” one seeing the difference between a frequency sweep of good > good and suspect > good might be informative.
Or do it manually with 50 MHz steps from say 600 - 1000 MHz.
You are actually driving both inputs of the switch, right? I found that with the receive side undriven I get better sensitivity than with no antenna connected, but much worse than I got by actually driving the receive side switch enable.
