Hi everyone,
I haven’t found any information about gateway “repeaters” in order to extend the network (I only found this 5 years ago so I thought maybe something new happened…)
I want to cover a mountain. There is no internet connection in most part of it. I was expecting to have a gateway where there is internet connection, and other “nodes/gateways” that are linked to the first one. In this case I would only use one SIM card to cover the whole mountain…
Do you guys have any idea for this ?
Perhaps talk to Nick @descartes or Steven @stevencellist who have been developing around LoRaWAN relay/hub of late and might be looking for early testers…or if needed consider LoRaWAN over satellite? (Lacuna Space et al).
@simmoinard, if you want to get up to speed on LoRaWAN Relay - the LoRa Alliance and Semtech have several briefing articles on this relatively new tech, and if it sounds like a fit, ask away!
And of course have a single point of failure! 
Personally I would always choose extra GW’s over a more convoluted solution. Having additional GW’s gives redundancy (would the GW(s) be at risk from land slip or avalanche? Power outages? Storm damage or snow/ice coverage/impairment?) Having more GW’s also allows greater prospect of nodes optimising down to a minimum SF improving battery life - possible factor if nodes hard to access or if temperatures get low e.g. in winter, such that they can have spare capacity and keep running and reduce field trips.
Ever an issue and case where relay may help, however, dont forget often the best way to cover the face/base of a mountain is from a GW placed at even a small elevation up the face of one opposite - where cellular coverage may also be better…same goes for coverage in river valleys - GW’s on opposite side may be ideal for coverage on the side of interest! 
(I do this in the Thames, Mersey, Trent and Tyne Valleys here in the UK; and in the past have applied same principle to deploying on tower blocks; where an adjacent block hosting a GW can often have easier ‘view’ into neighbour block vs trying to penetrate multiple floors from roof or lower level GW’s! RF, gotta love it! 
)
Ha, exactly why I suggested looking at the relay specs. Already spent hours explaining circles & trigonometry to peeps, if it keeps happening I’ll end up like Roy Neary
Your scenario sounds a bit like the one we’re considering for a client.
We have a major highway in a very tight canyon. There’s no cellular service whatsoever. There’s no power lines. There’s a river just below the highway. Rockslides and snow avalanches are not uncommon in the canyon and when they happen they often block and/or damage the highway. There are static signs with flashing lights and dynamic messaging signs (DMS) at both ends of the worst part of the canyon.
We want to be able to place sensors (slide fences, wx stations, seismic event detectors, water level, etc.) throughout the canyon (about 15 km / 10 mi) and improve the driving public’s safety and awareness of what’s happening on this major route.
There are a number of turnouts/turnaround points on the road in the canyon, so even if an event happens that forces a road closure, drivers already in the canyon would be able to escape. There have been times where enough snow causes vehicles to be trapped, and so having a way to support a very simple emergency call box for drivers to signal out would be great.
If the road is closed, the only ways around are 300 km / 200 mi detours, so having timely status information and spreading the word via DMS will be valuable at the junctions of this highway with other alternate routes.
We’ve been thinking a LoRaWAN network or similar, with backbone nodes that form a ~10 kbps main route, as much path redundancy as practical, and gateways at each end where mains power (or larger solar plants) and cellular (or satellite) are feasible.
Branches off the backbone would go to solar-powered end nodes either along the highway (like temp/dew point, slide fence, flashing lights, etc.) or up the canyon face where there are persistent issues (rock slides, snow depth, runoff).
We’re even thinking of getting state university involvement, to encourage new and innovative low-power/low-data rate sensors that can plug right in to the network.
In the RF modeling I’ve done so far, it’s hard to get backbone node hops of more than 1.5 km / 1 mi due to the twisting of the road and the steepness of the canyon walls. There’s even a few tunnels where the highway builders had no other choice, and the RF links would have to work around those. The tunnels also are in rock that has a bad habit of spalling, so there are chain link liners on the tunnel roof and walls, and of course we’d like strain sensors that can detect a rock spall event or better yet, incipient ones.
Oh, and this is a scenic route where poles and towers would be unsightly, so the whole thing has to be as low-profile and camouflaged as practical.
Similar challenges!
Cheers - Jon
Interesting challenge Jon!
Must confess my attitude confronted by something like this would be to test the art of the possible and JFDI!
Obviously devil in detail as always so lots of questions around how linear canyon is twists/turns, level/slopping/undulating route bed, geology/canyon wall material/mineral content etc.
Given
That suggests a well placed GW at each end only has to reach in/penetrate some 7.5km on average… well in scope for LoRa/LoRaWAN depending on above issues. Is there a place at each mouth - pref at higher elevation - where GW can be hosted - with backhaul connectivity (even if just cellular coverage) and power (if not then go Solar?!). Is there scope for a ~mid point GW (at higher elevation?) with supporting resources?
Job 1 for me would be place GW’s, pile a load of trackers and/or other nodes into vehicle - possibly on various fixed SF’s then drive slowly through and see what works and where any signals drop
Job 2 would be place basic sensors atop say 2-3m high poles/rods and place then at say 1/4km seperation through the valley (maybe get a dozen set up and work in stages/sections around 2.5-4km long) leave for 48hrs in each case and see what if any signals get through - perhaps working in from each canyon mouth. That atleast informs of what might be possible before committing too much effort and resources.
All well and good but find real world often very different with LoRa often suprising c/w any standard legacy modulations assumed for conventional modelling & LOS based. Also
In the early days of LoRa (even before LoRaWAN became a thing, I saw clients considering using the technology to get coverage in mining and quarrying applications and depending on the questions above (geology/canyon wall material/mineral content etc.) it was found that reach was better than expected in some cases (worse in other especially if very dusty with high metal content in dust!), so don’t write all off just 'cause there is a tunnel. The spread spectrum nature of LoRa can help and where reflections can cause problems with e.g. destructive intereference for legacy modulations, those same relections can even assist with LoRa, plus you get refraction/knife edge effects etc. and if there are is metal signage along the route you might get lucky with then reflecting around obsticals - the only way to find out is JFDI as above
I know that some tunnels - e.g various linear and circular particle colliders have LoRa/LoRaWAN systems deployed in them working just peachy And did some of my own tests years ago in tunnels like the Eurotunnel (UK/FR) and through the Alps and just last summer near Geneva with a few surprise ‘hits’. (And that is besides conventional travel tests on e.g. normal trains on various national rail networks - UK, NL, D, F etc.)
Once you determine if/where hits may be possible you can then explore in greater detail either side of locations to see how might be optimsed or how could be adapted as locations for desired sensors, possibly then applying relay to desirable locations to get to points where signals can be extracted.
Not sure your location but would hazard a guess N.America? If so shame as you are limited to SF10, and even then small payloads, where SF11 & SF12 can often benefit from the improved physics and coding maths to get deeper penetration/reflection paths.
Also try switching sides/locations for GW’s and sensors as small changes can have big impact, and may result in say 2 GW at each end etc…
Give it a go - you never know!