I really love Reece Innovation Solar powered pod product developed by my friend Jose Marcelino. They’ve made a agricultural / industrial autonomous solar powered LoRaWan gateway for a really competitive price. With much more money to extend TheThingsNetwork in my city I would have used a such solution. But as this is just a hobby for me, I’m looking for really low cost solution, something under 500€ per gateway.

My main issue to extend the network is not really to find roof but to find some where I can pass an Ethernet cable and provide the power from it. I have some place where I could deploy new gateway in conduction to be cable-less. The network is not the main issue as most of the time a WiFi network is accessible. Powering is a larger problem to solve. Advantages of outdoor gateway: you have sun available. So, as Reece Innovation did, I decided to make my solar gateway, the main differences are: my will have no LTE communication capability (only WiFi) and it have to cost as less as possible.

Let see what I’ve done

Update

At first, I’m making this post a live post I’m updating when the solution is improved based on my field experimentation. Since the first release, I’ve changed the box, the solar panel, the battery set … So please come back to see what have been updated and don’t hesitate to ask question in comment if you have a doubt about the current state of this experiment.

2020-09-30, New update this time to change the harvester: it has not supported this summer and destroyed the batteries so I made a change with a better component. I also fix the gateway with cables because summer storm had make it falling down. So now that’s fixed. Stay tuned in case of future update.

Solar powered

As a first question: What kind of solar power solution I can use and what is the needed power ?

I’ve found a nice website to measure the required solar power for a certain usage. You can select a zone (here in France), select a Solar panel power (here 100W) and it computes the energy you will get in winter and summer. You can simulate a load (here 3W 24h per day) and see what will happen.

I’ve measured the different low-cost LoRaWan gateways and select the one requiring less energy. I’ve selected TTIG LoRaWan gateway, it have different advantages: It can be powered from and USB port (this will ft well with my solar Kit), it is really low-cost (90€) and the power consumption I measured was 2W.

So according to the website above, the best is to have a 100W solar panel to cover this 2W need. This is because in winter we can consider a day about 1 hour of solar exposition full capacity. This is about 100W as an input. The power consumption will be 24h * 3W (get some margin) 72W.

I’ve found a kit quickly accessible solar kit in the confinement period on Amazon.

It has a solar panel (100W) the need cable to connect to the harvester and the battery and the harvester. This harvester has 2 USB connector. This is a perfect match with my need for the TTIG. The Solar controller is a PWN based battery charger.

The cost was 140€ on Amazon but you can find the same for less than 100€ on Aliexpress.

The only bad thing with this kit is the flexible solar panel, it will not be easy to fix it to the mat. We will see how to deal with this later. Basically I did not deal with that solar panel and chose a rigid one.The solar harvester did not support the summer temperature and did bad things with the batteries so I changed it also. So basically do not use a such kit, I still only using the power cables !

This question answered we can start listing the equipment needed for this outdoor solar powered WiFi – LoRaWan gateway.

Choice of the gateway

I had two options, both low cost, working on WiFi. The TTIG and the MikroTik gateway. The MikroTik option was the easiest one as we have an antenna connector. It was also a more expensive one and as I have to put all the stuff in a box due to the solar harvester the overcost of the outdoor readiness of the MikroTik vs the TTIG was not so interesting. I prefer to use it on a spot where I can find Ethernet/PoE to fully benefit of its advantages.

The decisive element was in fact the power consumption. I made measure at 2W for the TTIG when the MikroTik was 3W. For information here is the MikroTik consumption curve (@24V):

Both can be powered with a solar equipment but the MikroTik requires a 24V battery. Things I do not have in stock.

So basically the TTIG has some good points for this gateway: low cost, low power, 5V powered and everything is available in my personal stock of stuff. Only problem : there is no external antenna connector. Nothing blocking at all.

Bill of material

I’ve based my BOM on components I already had identified for my previous outdoor home made gateways initially based on Laird gateways.

• Gateway: The Things Indoor Gateway (TTIG) – 90€
• Outdoor Antenna: The Taoglass OMB8912.03F21 on Digikey (75€ w Tax)
• Outdoor Plastic box: 400x350x120 from Amazon (26€ w Tax)
• Platic box fixation: Metronic fixing plate from Leroy-Merlin ( 15€ w tax)
• Fixing mats: 2 meters, extendable, Metronic from Leroy-Merlin (13€ w tax)
• Antenna coaxial: It needs a RP-SMA-Female to N-Male cable, 3m. I ordered a Tp-Link TL-ANT24PT3 cable found on Amazon (8€ w TAX)
• Solar Panel 100W ordered on Amazon (88€)
• Solar harvester / controller from Victron Energy 75 |15 on Amazon (91€)
• 12V to USB power converter for the TTIG on Amazon (14€)
• Solar Panel cables (like these one) (20€)
• 2 x Battery 12V 7A on Amazon (2×17€)
• An UFL to Male SMA connector on Amazon (4€)
• 3M “Dual Lock” fixing solution found on Amazon (17€)
• Metronic Wall Mat fixing for Solar Panel fixing solution (19,90€)
• 2x 3 rows Cage Clamp connectors like these one (3€)
• Electrical cable (from my stock)
• 4x electrical crimp terminal like these one (1€)
• 2x 40mm long, large, screws & bolts

The total cost is 519€ all included. The antenna is a bit expensive even if I selected a lower cost one than usual. At this moment I did not do field test with that antenna. It has not been designed for 868Mhz but the Return Loss sounds not bad around 868MHz… I’ll give you my feedback soon.

Hack the TTIG

Ok, for people knowing a bit the TTIG you could ask how I’ll connect the gateway to the antenna as this gateway have an internal antenna and no connector for an external one … The answer is simple, we are going to hack it ! I’ve seen people doing this on Internet, so it should be easy. Spoiler ! it is.

The first step is to remove the grey facade on top and bottom. they are clipped but easy to remove with a screw driver. Then you will see 2 screws to remove.

At this point you will have to separate the two faces. Do not expect to have a clean device at the end of this step, warranty will be broken. No doubt !

You can see on the picture, in the middle, a place to place your screw driver. That’s better than the external part as I did.

So, once the clip has been detached, you can access the gateway content.

If you have some tips to open the TTIG without making it so dirty, please let me know and share it. I really don’t like to destroy my stuff like I did.

Inside the gateway you will see that the internal 868 antenna is connected to the LoRa concentrator with an UFL cable. So it becomes easy to connect an UFL to SMA connector.

Now we need to find where to put this connector. In my usage the best is to put it near the USB power cable. To ensure it will pass around the TTIG 220V power converted, the best was to remove this not needed part. It saves weight, space and can be reuse for a future project.

So now, we need to drill the facade to get the SMA. My choice was to place the drill where one of the cover screw was. We do not really need that screw as the side and facade are clipped.

Before closing the box… always verify everything is still working. Then close it and screw the SMA connector. Here is the result:

So at this point we have our low-cost gateway ready for installation inside the box !

Install all the stuff in the box

Now everything can be installed in the box. The two batteries need to be fixed in the box. For this I’m using the 3M dual lock. This sounds to be an easier way to later remove the battery compared to a previous solution I’ve chosen. The first step is to place the batteries. I’ve decided to let the solar cable to pass between the two batteries. The antenna cable is passing the other side. The two batteries are connected in parallel. On the picture above the battery is not yet connected to the solar harvester: it is waiting for the connected monitoring device.

So once the Monitoring solution is in place in the box, we still have some available space in that box ;). You can see a second USB cable connected on the harvester, it allows the monitoring system to detect a power loss on this connector.

The addition of the Solar panel was one of the most problematic point at the beginning: how to fix it to the mat with a 45° angle ? I finally found a good solution with a mat wall fixing stuff. I also had to drill the solar panel to attach top fixing part.

That way the solar panel is about 55° oriented to the sky and correctly fixed to the mat. The best would be 45° but for this you need to drill some new hole. Let see what is the result with 55°.

I’m using a parasol base to fix the mat to the ground. That the way I think I’ll put it on the roof top to avoid to drill that place to fix the mat. So I still need a way to fix the mat and gateway to ensure it will not become a weathercock.

The way I solve this is to add a metal bar and a hole available on the parasol base. I also add a piece of wood to ensure the metal bar will not move out from the base. I’m not fully confident with this solution but I’m going to give it a try.

Let see what will happen once installed with some wind.

So I’ll investigate more on this and keep you informed of the improvements. If you have any suggestion, the comment are open, so feel free to contribute.

Now I just need to close the box and add the final touch ! Now that one is working on TTN.

Now we can check that the gateway is connected on The Things Network console. I also made a script to monitor the disconnection and track what is happening during the night. This The Things Networks LoRaWan gateway monitoring script is detailed following the link.

Limits of this installation

There are different limits, I will update based on experience and discussions. You need to be aware of these limits if you want to conduct a such project and never forget the experimental aspect of this. You do it at your own risk 😉

Battery

I’ve selected a 2x7A@12V sealed lead-acid battery, mostly because it is what I have in my stock from an old APC. This is not the best choice for different reason:

• The environmental condition of temperature will affect the life of the battery with a risk of freeze when discharge on really low temperature -20°C … This sounds acceptable.
• Low temperature impact on capacity : 0°C will reduce the capacity by 33%.
• The number of possible cycle for the battery is reduced by the discharge level. Here we have 2W*18h discharging = 36W on the 160W available in the battery + risk of high discharge during the winter period. So we are around 25% of discharge and the cycle number is under 2500 cycles according to this graph. But low temperature will impact this largely.

Solar harvester (first version that did not work for a long time)

The model of solar harvested I’m using is able to charge / discharge the battery based on the voltage. It means that once charged the battery will stop charging until reaching a limit. I did not set anything to optimize the cycles and ensure night autonomy. Source www.scubaengineer.com

The battery controller (solar harvester), have 3 settings:

• The battery high level target (after that point it stops charging the battery)
• The battery low level (under that point it shut the charge down)
• The charge restart point (upper that point it switch on the charge)

According to the graph above, with a charge of 3W on 160W (C/50), a 100% charge is about 13.5V. This could be a good target for high level target.

We want a maximum discharge about 20% capacity to avoid large discharge. This is about 12V.

We can position the restart voltage about 40%-50% at rest, this looks like 12.4V.

I’m not yet sure of these parameters, I’ll update them if I find some way to tune them more. The solar controller works with a PWM controller (vs MPPT). It basically have (at least) two mode of charge during the day:

• Bulk mode => charging is continuous because the battery voltage is lower than the target.
• Flicking mode => solar panel is switched ON/OFF to hold the battery voltage

More globally it works like this:

It seems we can see the phase 1 and 2 at least from this picture: sun was fully blue, we should have a nice Gaussian curve but we don’t. so it is related to the power regulation.

We see the different phases :

• 20:00 to 08:00 – night, gateway is powered from the battery
• 8:00 to 11:30 – battery full charging
• 11:30 to 18:00 – regulation / float
• 18:00 to the end – night

After a 3 month this harvester has start to badly work: I saw the battery voltage dropping under 10V when it was programmed to switch of the system under 12V. As a consequence the battery has been destroyed rapidly. In parallel a storm made the gateway falling down with all the stuff directly exposed to the sun with high temperature. So after holidays and 1 month in this bad situation, batteries and harvester was looking dead and I decide to change this. This kind of hardware is really low cost and PWM is not that good. I’ve open the harvester and when I see how it is made and the way the transistors are dissipating temperature I understand the trouble you can have with a such hardware.

Second solar harvester

So I decide to invest more money on the harvester as it is protecting the batteries and this is also an expensive component you need to change sometime so longer life is better. I was looking on a real MPPT solution to optimize the harvesting as we are entering in the winter period and the condition are becoming hard for the system.

I’ve selected this Victron Energy SmartSolar system. It’s not the one I have listed in the BOM. This one have a Bluetooth connectivity to real time monitor it. This is not useful for the gateway but it was looking fun for me and the experiments. Also to be able to compare with what my sensors are producing. So You can use the non-smart version I listed and save 30€. This equipment has a really different quality compared to the previous one. The setting can be made with bluetooth. They are really fine grained. Basically the default setting was looking go to me. I currently have no feedback on this equipment as I just installed it and update the blog post. So later in the year I’ll be able to give you more details about it. Currently I can just say I’m more confident.

Compared to a PWM, a MPPT is optimizing the solar panel performance all along the day. We are suppose to get more energy. I’ll see if it helps during the winter even if I have no data to compare with. Before finding this model I reviewed a lot of harvester on Internet and saw a lot of them indicating MPTT but being PWM for real. I did not found lower cost solution for MPTT.

The mobile application quality is really good, so it looks really good hardware. So the new gateway setup is now like this:

Fields feedback

First field feedback (after 3 months)

These feedback are cumulative from the first days of test and different revisions of the gateway are concerned.

Currently I had a lot of sunny spring days for my test. I’ve tuned a bit the solar harvester device but not so much. The way it is configured sound not so bad for limiting the battery cycles.

As you have seen, I’ve changed the solar panel for a rigid version, low cost and got the same efficiency. This one is more easy to fix to the mat.

I’ve tuned a bit the angle, shadow with trees around in my experimental place and the orientation. I’ve also add a current measurement to be able to compare day after day and hour after hour the quantity of energy available. Confinement is also limiting my ability to access some good measurement tools. By the way, I’ve got the following first results:

• The system is globally working 24/7, with sunny days, in this period of the year, the system was working 21 to 22 hours per day with a shutdown between 05:00 am and 07:00 am with a single battery. the addition of the second one fixed that issue. As sun only touch my Solar panel after 10:00 am, this is not fully optimized. Also I’ve experience sun rising was enough to get the gateway restarting when only 1 battery was installed. It can also be related to the external temperature impacting the battery voltage or the solar panel capturing environmental energy.
• I measured pic of entering energy in the battery around 30W. This is not the 100W of the panel. The gateway consumption is about 3Wh so we are under 30% of efficiency. The sun angle at this period of the year is not optimum and my solution to measure currents is not optimum also so let seen this point later for more details. More precised and automated measurements are on the way.

Second Feedback (after 6 months)

As already explained I had some issue during summer:

• The harvester did not work correctly and dropped battery voltage up to 6V even with a 12V target.
• The gateway felt down after a storm during holidays and stayed on the solar panel side for a month. As a consequence the battery has continue to drop to 0 and was directly exposed to the sun with temperature over 38°C outside, maybe over 60-70°C inside. So the battery were both destroyed after that period.

I’ve explained above the hardware modification made to fix the harvester & batteries issue. To avoid the storm / wind issue I now, added metal guying to maintain the gateway. Here is the current setup: