Directional Antenna Recommendations for line-of-sight

I am building some data collection devices using WisBlock. I have a RAK2245 mounted 30 meters high with 4.5dBi omnidirectional antenna, with direct line of sight to the destination locations.
On the WisBlock I am using small whip antennas with 5dBi gain (I distrust this gain spec).

I am able to achieve successful Join and bidirectional communication from 2.8km distance (direct line of sight) but the RSSI is reported as -120 in ChirpStack, and LoraSNR between (-8) and (+8). Also moving the location slightly, I was unable to connect even though there was direct line of sight.

At another location only 2.0km from the RAK2245 (again direct line of sight), I was able to see the Join Request arriving in ChirpStack, and Join Accept being sent out, but the WisBlock unit apparently is not receiving the replies from the RAK2245 so it never sees that the Join was accepted.

I was hoping a directional antenna at the end nodes (Wisblock) would improve the situation. However, I tested this directional antenna designed for 900MHz frequencies (I am in the USA, so we use the 915MHz band for LoRaWAN). However, the RSSI only went to -115 which was very little improvement over the tiny antenna. So at this point I’m not sure if this antenna is not suitable, or the omni antenna on the RAK2245 is not the right choice.

Please note that the RAK2245 and omni antenna are installed on the side of an elevated water tower, about 1 meter away from the metal tank. The antenna is directly visible from both test locations, and is not obscured by the tank at those angles.

First, do you have any suggestions for improving range and/or signal quality for line-of-sight connections? I have heard 20-40km is possible line of sight, and I’m only trying to achieve 3km, but getting very low signal strength reported in ChirpStack. What am I doing wrong?

Second, can you recommend any specific antennas available at Amazon or other USA resellers? I notice that it is quite difficult to find antennas that actually meet their listed specifications on Amazon. Maybe there is another source in USA that provides antennas true to their specs?

Thank you for any suggestions or guidance.

Best wishes,
Ryan

Few things I can say about this:

  1. You’ve tried two different antennas and seem to have the same result. Do you have an extra device to check it performance and do comparative test?
  2. Have you tried reorienting or test in a different position the directional antenna? Your directional antenna is multiband and no VSWR stated or any useful specs. Probabaly you can try our directional antenna dedicated for SubGhz application
  3. Do you have access to VNA or even nanoVNA to double-check the resonant freq of your antenna both for the device and the gateway?
  4. Would there be a chance to position the gateway antenna freely and further away from the tank even just for a quick test to isolate if it affects it? Highly unlikely but we can’t really say.

I have no experience running 20-40km from gateway to device (we got one before but just by accident and not intentional). There are many claims that can be achieved anyway. You can try to max out the power, antenna gain, and spreading factor. The next question is how consistent will it be? How critical are your applications for this? If a cellular network is available, won’t that be more practical? Or won’t you consider a satellite backhaul if it is very remote?