If you already have optimised the local position of node and gateway, there are a few things that you can do. (Can the node placed higher for example.)
One solution is that you get the gateway and the node up to the legal maximum. Has the node an antenna that goes to the max. and is the max range in the direction of the gateway…
Another is that you add directional antenna’s on both sides and on both sides lower the send power to the maximum allowed.
TTN nodes are normally running at the maximum permitted power in the first place. So adding gain antennas at Gateway and node is pointless, you need to reduce the transmit power to remain legal so the range\distance goes back to what it would be without the so called gain antennas.
Whilst the use of gain antennas at both ends wont give you extra range\distance, it will reduce the coverage of the gateway to a very narrow arc. An omni directional antenna would give the same range\distance but with 360 degrees of coverage, so directional gain antennas have severe dissadvantages.
Any hack that needs both a non-standard gateway and non-standard node will, if connected to the TTN community network, harm that network. (TTN might elect the non-standard gateway for downlinks of other standard nodes, which might not work then.)
I don’t see a problem at the node, but you’re right that a gateway with this configuration could give problems.
To operate properly the node needs to have two way communications as well. So if you add a gain antenna you have to reduce the nodes transmit power, which negates to supposed advantage of the gain antenna, i.e there is no distance advantage of fitting the gain antenna to the node in the first place.
One dissadvantage would be that if the gateway the directional antenna on the node is pointing at goes down, the packets from the node will probably not be picked up by other nearby gateways in different directions.
I meant to emphasise “both gateway and node”, but just to be sure let me rephrase for future readers:
Any hack for a gateway on the TTN community network that only works for specific nodes, is harmful. (Same goes for, e.g, single-channel and dual-channel forwarders, even if those work fine with some non-standard nodes that only use a subset of the channels.)
Also, as covered above: hacks on either gateways and/or nodes on a private network might be harmful for all networks that use the same shared radio spectrum. Same goes for pushing the limits in a private installation, like apparently affects, e.g., Sofia.
Is this the case? I doubt that alle node makers test the node so that transmit and receive are at the maximum. I see wire antenna’s inside a case, default pcb antenna designs,…
I think that it’s not that important in 99% off the nodes, but it could be useful to check for those “borderline” nodes.
Thats a different issue though.
Although the type of node that you mention (wire antennas, PCB antennas) may be programmed for 14dBm, and is radiating less, these are not the type of node where fitting a high gain yagi antenna to compensate for a loss of radiation efficiency is likley to be practical.
Quite often there’s a u-FL connector and then there’s more possible.
A few weeks ago I asked about an “optimal” pcb antenna with a u.FL pcb dipole with u.FL, but to be honest I expected more response with the maximal allowed antenna’s.
BTW. That still something other than a high-gain YAGI 
Yes of course, appologies, they are of course actually Yagi Uda antennas …
Sorry, but English is not my first language and sometimes gives misunderstandings.
@ecosoph
Now I understand your problem. The nodes are not far away but their signals are attenuated by the trees. Therefor you need more antenna-gain to receive their signals.
If you only want to receive a signal imho it is not forbidden to use a high-gain directional antenna. But you must not transmit using this antenna because you may exceed the +14dBm erp limit.
My ideas:
And for which of those suggestions do the nodes and gateway remain TTN compliant ?
That is something to be determined. The solution has to co-exist with the standard TTN configurations without disturbing other nodes or gateways.
But I think it should be allowed to discuss if there are possible solutions. If you don´t find a TTN-compliant solution the usage of TTN or LoRaWAN at all is not appropriate.
When join requests or downlinks are required (with the two gateway option above) how does TTN know not to send the messages to Gateway1 (the one with the gain antenna) that is receiving the strongest signals ? How can TTN possibly know that to send messages to the remote node it needs to send them to some other arbitary Gateway2 that has never received a packet from the node ?
That is a question by the way, I dont know the answer.
I do know the answer. It can’t so it won’t. It will send the response to gateway1 which will not forward it and as a result the node will never get the data. TTN does not implement the notion of receive only gateways with a sending partner. (Yet, they may well accept patches for V3 implementing this, but discuss first before spending a lot of time on creating patches)
BTW, if gateway2 can’t receive data from a node, chances its transmissions get to that node are near zero. So for this use case I wouldn’t bother to even look into modifying the stack.
I agree. So idea number 2) seems not to be the solution.
Lets go to idea number 1): power-reduction (EU868).
One problem we have (in EU) is the limitation of ERP. Assuming an antenna-gain of 12dBd the gateway is not allowed to output more then 2dBm (I ignore the loss of cable and plugs). Then we reach the ERP-limit of 14dBm for g/g1 sub band (8 LoRaWAN-channels). But there is one channel we are allowed to use an ERP of 27dBm - 869.525 MHz in g3-sub band (RX2-downlink window).
What about reducing the TX-power of the gateway to 2dBm in g/g1 (8 LoRaWAN-channels) and using 15dBm for g3 (RX2-downlink)?
That would give us more then 10dB more sensitivity for the receiver of the gateway without exeeding the legal ERP-limits. Additionally we have 13dB more ERP in RX2-window. A nearly perfect balance.
What would be the gain over using 27dBm with an omnidirectional antenna?
I dont see the restriction of ERP as a problem really as its there for a good reason. There is limited bandwidth available in the RF spectrum for this type of application, so to allow for widespread use or sharing of this limited bandwidth its important to keep power levels down, specifically to restrict the coverage to relativly small geographical cells.
Increase the power level from say 14dBm to an effective 27dBm, the range increases by a factor of 4 and that in turn increases the size of the coverage cell by 16 and the possibility of interference (i.e.another gateway in the same area) also by a factor of 16. Now the RX2 is only a single channel and given the extra range you might expect it to be a popular choice, increasing the chance of interference further.
To use RX2 in this way you would also have to change the LoRaWAN standard (This was mentioned earlier in the thread) and the firmaware used by nodes and Gateways.
If your gateway has an output-power of +27dBm the max. antenna gain allowed is 0dBd or 2.15dBi in sub-band g3. It does not matter whether the antenna is directional or omnidirectional.