Meshtastic to connect remote villages: Deep in the Amazon


This is a second experiment, following Meshtastic to connect isolated villages, using LoRa as emergency communication for remote villages.

The Tukano people are indigenous to the deep amazon jungle, more specifically the Upper Rio Negro region. Together with local collaborators and the Nova Era Institute we’ll be working with them in a 3 year long project to strengthen their autonomy through communication and information technologies and agroforesting projects.

In the first steps of the Tukano Project, the critical needs for 3 communities within the Balaio Indigenous Territory were assessed. The need for ways to communicate between the villages and ideally to the closest town (75Km far) are in times a matter of life or death. They also really need some access to useful information, such as learning materials for the topics that are relevant to them and healthy enterteinment, that are also relevant to their culture. And longer term, they seek food sovereignty.

The first action for the project will be to setup the communication and information infraestructures. We’ll be arriving in the town of São Gabriel da Cachoeira, and will setup a base there, with at least one RPi 4 and one LoRa radio, connected to each other. We’ll have RPis and radios powered by a 150W + 70ah solar setup in the 3 villages as well. Ideally we’d be able to connect this town node to nodes in the villages, creating an emergency communication channel and remote tech assistance using the SSH over LoRa tecnique.

The geography seems helpful, but it’s amazon jungle vegetation, which I’ve never worked with. I’m hoping we can work thru the gigantec trees, but not sure how. Also we’re still seeking permission to put radio nodes on top of the estrategic hills.

The easiest way would be thru a direct 64.3Km link, probably using the same setup as on the previous experiment. A disadvantage is that the road in between the hills would probably not get coverage.

This is a second way of doing it, with a node on a hill 19.3Km after the town and 45Km before the village hill. These links could be done with a moxon or a more powerful omni. The advantage is that there will coverage along the road, which can be very helpful. But the disadvantage is that we’ll be to deal with one additional hill with gigantec trees.



If we choose to ignore the legality of it, would using the 433Mhz radio be helpful in going thru the dense forest? If we find out we can’t go above the canopy, could making smaller ~2Km 433Mhz links through the trees work?


I’ve only experiment with direction and stock omni antennas. Ideally we’d be using more omnis in order to get better coverage and a more mesh-like network topology. I’ve seen the discussion on fiberglass omnis, and it seems they might be a good choice.

I’ll currently looking at these RakWireless 8dBi fiberglass omnis for 915Mhz, but there are cheaper options such as the Zizor 8dBi fiberglass omni for 915Mhz or a generic 8dBi fiberglass omni for 915mhz, also with a 433mhz version if we go decide going for that. Should we expect much better signal quality and durability with the more expensive ones? How far should an omni such as these could go in a line of sight or through the jungle trees?

In this coverage simulation 433Mhz seems to perform much better. Could I realistically expect such coverages with those 8 dBi omni antennas?


I’m inclined to use T-Beams again, but using the new nrf52 boards would be amazing, to take advantange of the near 0 power consumption when idle, the sd card slot and the e-ink display. Just not sure if they’ll start shipping on time or if they’ll be much more expensive. Any updates on them @geeksville?

Else using other nrf52 board, such as the already supported Rak815, for the repeater nodes and T-Beams for end-user nodes.

Energy and casing

I’ll continue using the same solar setup and casing as the previous experiment, even though I didn’t get a chance to really know if they were a good choice, as our Krahô partners took them down when heavy storms started.

That raises a question about lightning protection. Should we be installing lightning rods close to our nodes in order to protect them from strikes? That would considerably raise the price for each node.

Something I’ll add which I think will give more waterproofing to the case are cable glands.


The ability to get/update settings of a remote device using LoRa would be a huge advantage, and I remember @geeksville mentioned this was on the way. Has is happened? Are there docs or any discussions I can follow to understand how this will happen?

Innovations since last experiment

Meshtastic devs are in a row, can’t congratulate yall enough :clap: I’m excited to test the new features that are coming out such as the Range Test plugin, the Store-forward plugin and specially everything related to web integration: Meshtastic over WiFi, Meshtastic web-ui and meshtastic.js (which a have billion ideas for).

Other then these new features all the bug squashing is really exciting, as I experienced some difficulties with the firmware I had at the time (October 2020).

Ordered some more T-Beams just to get back to doing more tests and working on the video-documentations. Let’s get to work :muscle:


Very cool.

My reading of the over all situation is the project will go much better with all the cool stuff in 1.2 BUT only if you can wait long enough for some testing and bug fixes in the 1.2 branch.

900Mhz is often recommended over 2.4Ghz or 5Ghz for point to point links where tress are an issue.

While 433Mhz will have better penetration, your overall bandwidth is going to be reduced for a given db / rf spread.

If you are going to consider working outside of established parameters for your area, I’d consider using the 1 watt radios and gradually bump up your transmit power for your relay nodes. Generally a larger antenna will have better receive sensitivity, and concentrates the transmit energy (‘omni’ is a ‘donut’ shape). Having a higher transmit power can compensate for some of this.

In some areas higher transmit power is blessed IF it is spread over a larger rf spread. With LORA this could help with bandwidth and resiliency (use 500Khz channel bandwidth) .

If your area of service is on one side of a repeater node on a high hill tilting an omni antenna slightly to that side could help. If one ore more sites is well within 120 degree view something like this could work really well. 900 MHz Sector Antennas 120 Deg. Beam Width - KP Performance Antennas

Assuming you can have decent setups on these hills, if possible I would at least try using some off the shelf wifi gear to create point to point links. If you can pull that off you can start to create a regional network that can host locally run web apps using raspberry pi(s). You could have a local copy of wikipedia, a social network, Craig’s list clone, etc…

I wish I could join you guys, I have a hand full of Unifi devices and directional antennas.

1 Like

Thanks @Spor7biker, that’s really valuable feedback.

I did some research on 1 watt LoRa and only found this HamShield, which is for Arduino, requirering to battle a bit with a firmware build.

Are there ampliers out there that I could use to amplify a board supported by Meshtastic?

Using a sector antenna might actually be a great idea.

I didn’t mention on this post, but the plan is to have the community-server setup that I’ve been working for a few years now running on the RPis. It’s basically just a bunch of scripts and docker-compose files to easily setup a bunch of services that could be useful for the communities, specially made for offline environments.

On the hill that connects the three villages I plan to use a LibreRouter and CPEs 510 on the villages to establish a 5Ghz mesh at least between the villages, enabling to use cool p2p apps like Kolibri, Scuttlebutt, Mapeo and others.

On a next phase, with more knowledge about the terrain, and more funding, we could attempt a 5Ghz link between the villages and town.

This is on target to go into 1.2.0

This experiment will probably happen around end of April or May. So hopefully we’ll have time to release and do some serious testing on 1.2.0.

@mc-hamster do you have user story to share explaining how remote access would work? Is like accessing the other’s API though LoRa?

All I know is there’s an admin channel that is secured with a sha256 preshared key. Once you get access to the channel, you’ll be able to remotely configure the settings.

1 Like

I’ve been meaning to look into how hard it would be to wire a 1w lora radio to a generic esp32 device. That could be much more cost effective and flexible.

What are the regulations in the area for 3.6Ghz transmissions? 3.6 was used in the USA but has been phased out due to FCC reallocating frequencies. I bet you could find a lot of cheep gear on the second hand market. If you contacted enough wireless ISPs that have been around for a long time they may donate that kind of gear if you pay the shipping costs.

I would stay away from external amps. They bring there own set of issues. Maybe as a last resort, but getting a 1w radio to work will likely preform better and cost less.

5Ghz will likely have major issues with long distance if the LOS zone is not clear. But the MIMO devices that use horizontal and vertically polarized antennas allegedly handle minor obstacles fairly well. I’d avoid 2.4Ghz in that environment, it is too close to the absorption frequency of water.

1 Like

The higher the frequency, the easier it is to block. 2.4ghz WiFi has a lot more range than 5ghz WiFi. The higher the frequency the more bandwidth & data you can transmit though. Microwave ovens run at 2.4ghz because water absorbs the energy very well at that frequency. Trees & critters or people are mostly water.

They have gotten unamplified (200mw if I recall) WiFi adapters to connect over 120ish miles. But they were basically old 6’ satellite dishes on each end & had to be perfectly aligned. My 50 watt HAM radio running in the 440mhz range really won’t get 20 miles in the woods & mountains, usually a lot less. Expecting a lot more range with WAY less power isn’t reasonable. Maybe with very high gain antennas & precise aiming, but that negates any mesh potential or benefits.

Be very careful about falling into the trap of “a more powerful antenna”. There is no such thing. You only end up with an antenna that will focus a lot better. Think laser beam vs lantern. A .05mw laser will be visible a mile away but a 1 watt light bulb probably won’t. I know which one I’d rather have for looking around a wider area. A high gain antenna will focus all energy into a very small area & reject signal from a wide area. Great for range, not good for coverage over a general region. A 0db omni antenna will have coverage over a perfect sphere. a 3-5db omni will look like a doughnut. A 6-8db omni will look like a pancake. Higher than that it will look like a piece of paper. Try to get 2 picees of paper to line up perfectly edge wise10’ apart with 2 people holding them in the air & that’s what you are moving towards with a high gain omni. If they aren’t perfectly pointed at each other in the same plane, you aren’t going to get any signal.


@Spor7biker good tip on 3.6Ghz, I’ll look into it. Found this document on bands in Brasil, but it’s hard to understand, asking some friends about it.

Maybe for connecting the hill to the villages, if we can’t get past the canopy, using a couple of CPEs 210 which is 2.4Ghz, sector and MIMO might be a better idea then the 510. But might still not be enough to get to the furthest village which will be 7.4Kms away. Maybe worth analyzing the possibility for 900Mhz band with a device like the Ubiquiti Nanostation M9.

Please let me know if you have any ideas on wiring 1W LoRa to esps.

That’s amazing explanation about antennas and gain @Phallon! Finally understand now.

It seems that we really need to get higher then the canopy to make this happen. I’ll ask the local partners about the possibility. If not… Well, start planning on various nodes closer to each other, or give up on connecting to town for now and focus on connecting the three villages to each other.

Long term the region could be very well served with a 10m tower on a well situated hill. But good luck building such a thing on a remote hill. You may be better off waiting for starlink to work in the area.

I’d think esp32 an 1w radio could be as easy as getting an SPI radio, connecting the right pins, and changing the appropriate pin definitions.


Hello luandro :slight_smile:
Ebyte sells some 1W modules based on the sx1262 chip… some communicate by SPI but those don’t have pins to wire them, instead UART modules would be more convenient to use. The problem is that for now meshtastic doesn’t support these :frowning:

So, you could get some UARTs or use a prototyping pcb, put the SPI module and solder its pins to some connectors :slight_smile:

Here the SPI module

1 Like

If you are open to pairing Raspberry PI to 1 W lora radios someone has done it with python code:




We’ve stumbled upon various legal barriers for setting up nodes in key places, as they’re inside nature reserves. So our team decided to have me visit the place asap to really learn about the territory, leaving the long link to the town for a next phase.

On this first phase we’ll be focusing on installing the local intranets (using Pi4s + OpenWRT router) on each village, and having a LoRa network connecting the 3 villages, so they can at least communicate to each other.

My initial idea is to use TTGO LoRa32 433Mhz (as I remember there’s a problem with T-Beams not waking up after battery discharge?) with DK7ZB yagi atennas that @msws built for the fixed nodes. And also TTGO T-Beams for mobile nodes.

Since we’ll be using Pi4s, the Pine64 dongle mentioned here would be ideal, but probably won’t come out in time. So for now I’ll use USB to connect the Pi and the LoRa32 board, and use the python API and a web UI for users to interact throuh their phones when connected to the access point. Is this ready to go @sachaw?

The geography and vegetation won’t help us much on these links, but they are only ~5Km. For those reasons I choose 433Mhz with the yagi. Yall think this might work?


Bureaucracy is really eating up our time, but it’s moving and we’ll soon start buying the equipment.

For the Intranet I’m thinking on getting:

For the Meshtastic LoRa setup:

The Wisblock will be connected to the Raspberry through USB for power and data, so ideally we’d communicate to an app with the board through serial. Options are to create an API that exposes the python API over http for apps to consume or Meshtastic-python/pull/74, to expose Python over http.

The Intranet + LoRa and a TP-Link WDR 3500/3600/4300 running LibreRouter OS will go inside a 20x16x8cm waterproof case connected through a short ethernet cable to each other.

The case with everything will be put on top of a building (house or school) in a pole and kept as high as possible, so that the LoRa radio and the WiFi radio get’s better reach. In case of WiFi both for people acessing the intranet access point as well as future WiFi mesh extensions that might happen.

Both router and Pi will be powered by a 150W panel + 70ah battery + 10a PWM controller solar setup that will go to each of the 3 villages that will receive the intranet.

Is having a 12v fan enough to keep everything cooled while inside a box in the amazon heat?

The solar setup seems a bit to small for all this gear, is a way to sleep and wake the Raspberry?

The Pi will run Docker and through it various services:

I haven’t had a Pi 4 to test yet, so I’m not yet sure how it’ll handle. And besides we have no idea how much use different services will have. Only way is leaving at the villages for some time and try to collect data.

Which is another important topic. Offline data collection. Ideally We’d write a script that publishes OpenWRT collectd data, Pi memory data (CPU, RAM, storage) and Meshtastic statistics to SSB, so that anyone going to town and syncing through the Internet will be seamlessly sending this data through.


The electronics should be able to handle relativity high heat.

Not cooking the battery is the biggest issue.

Running a fan introduces a whole new set of issues and failure points. If there is not a dust filter, and the filter is not cleaned regularly the combination of dust and moisture can really gum things up.


Yeah, fans and electric motors are particularly horrible for producing RF noise.


For heat limiting, consider having an enclosure shroud to act as the primary solar heat absorber and keep solar radiation off of your main enclosure for the hottest part of the day.

This shroud is open to allow airflow in and around it, as well as provides an insulating gap between the hot shroud and your electronics enclosure. It can be made from almost any practical material, but aim for those that do not provide good thermal mass (i.e. metals) as they become heater/radiators. In my experience in Africa, we’d use a UV stabilised plastic sheet, cut and folded to make the shroud.

For example:


Hi, do you have an update on this? I’m interested in doing it as well here in our country to continue educating our learners living in the mountains where there’s no means of internet connectivity or cellular network.

@itsmealex56 I’m working on an update post. Will hopefully have it ready soon!

1 Like

Meshtastic implementation :satellite:

Re-posted from Secure Scuttlebutt. Check First post on community-servers, or the full thread .

Balaio elders drawing a map of their territory

As with most implementations, we start off by holding a community-wide assembly, where we first introduce the idea o autonomy in communication and information infrastructures, and why that’s relevant for them. I asked the elders to draw a map of their territory, and did a beautiful job at it.

Balaio map with sticky notes symbolising LoRa mesh nodes

To my surprise the original plan for the Lora network was targeting the wrong village. It turns out the Balaio village was 13Km away from the hilltop, much further then expected.

The hill at the middle of the territory is called Morro dos Seis Lagos (Six Lakes Hill) and it’s theoretically a national park under the control of ICMBio. So the plan for my visit was to install the community-servers and show how the LoRa network works so that the communities could ask for permission to install a node at the hilltop.

When the villagers realised how simple a LoRa node is and the huge positive impact it could bring, they decided we should just go for it even without any approval.

Young Balaio men opening up the river path with a chainsaw

Together with some of the youth who had experience hiking the hill we came up with a plan on where and how to place the LoRa node. Each of the 3 villages that were benefiting from the project sent representatives. Me and the boys from Balaio and Ya Mirim villages went by boat, while the guys from Parintins went by foot cutting thru the forest and met us at the base of the hill.

Since we weren’t prepared for the installation, I had to recover an old solar panel. It was large and heavy, but would hopefully produce enough energy to keep the radio alive.

Our guides got a bit lost and we miscalculated the trail, so we walked in circles for an additional 3 hours in the jungle, and that’s when the huge panel proved to be a big hassle.

But we finally managed to find a good campsite and the highest part of highest hill in the territory. Turns out none of them had been there before, and they weren’t aware of anyone who had. Felt like true explorers.

Balaio men at the top of the tree fixing the solar panel and LoRa node

The guys from Parintins weren’t prepared for camping, so they left at nightfall. Can’t imagine them cutting back thru the jungle at night, but they made it back safely. We camped next to a beautiful lake, ate some of the dried fish and cassava flour we had and told stories of some of the mythical creatures of the forest. We slept in our hammocks and I confess being a bit afraid of the jaguars, but soon fell asleep.

We went back to the highest spot early morning. They cut the branches of the highest tree and fixed the panel. They fixed the Lora node to a wooden pole and fixed the pole to the tree as well, so it stood higher then any tree.

Very quickly we were able to get signal from the most distant village 13Km away, so the mission was a success :tada:

Our party of explorers right before descending the hill

We had nothing left to eat besides cassava flour, which they put in water. They call that Chibé. Not very delicious, but enough to give us energy for the trip back.

Checking for LoRa singal on way back thru the river

The ride back was pretty smooth and all we talked about was food. Wasn’t able to signal from the hill node as there was huge blind-spot until very close to the village.

When we arrived, an angry Yanomami came to recover the boat. Turns out the boat we had taken wasn’t from anyone they knew. Despite him looking really angry, it seems they have good relations and nothing happened.

The noodles with chicken we had afterwards was one of the best things I’ve ever eaten in my life.

LoRa 915Mhz network coverage simulation made with RadioMobile with the villages to be connected and the first Meshtastic message between villages

The simulation which I only ran after the visit shows that the Balaio village isn’t so great for reception. But we manged to establish communication between the villages. The moment me and Adelina were able to exchange message we felt we accomplished something very important.

I had taken only 5 Lora nodes and unfortunately one of them malfunctioned. I had to take the node from the Balaio village down from the roof so that it could be used with a phone. Despite the community-servers running Meshtastic-web, it wasn’t able to store messages, so conversation logs were never kept.

The network was compromised by this malfunctioning node.


This experiment was the closest we came to successfully using LoRa to connect distant villages. Unfortunately we were just one node short.

I’ll hopefully be visiting Balaio again early 2022, so we’ve already acquired more LoRa nodes and created a simple pub/bot which posts Meshtastic messages to SSB.

Setting up LoRa networks to connect remote villages have proved to be challenging, but we’ve advanced quite a lot in making this technology usable and accessible. Its low financial cost makes this very attractive to many isolated villages, so hopefully we’ll continue trailing this path in the future.