LoRaTube | Details | Hackaday.io

That’s why a repeater is necessary: to fully exploit the available link budget and reach long distances, the transmitter’s location must be carefully chosen. By placing the repeater at a high point — a hill, tower, or treetop with a clear line of sight — a single module can cover several tens of kilometers without exceeding the legally allowed transmission power. This strategy is at the heart of the LoRaTube concept: keeping a compliant, very low-cost, discreet, and autonomous transmitter (> 5 years), while maximizing coverage through optimal placement.

Why LoRaTube?

Because this project does not use solar panels — it relies instead on alkaline batteries housed in a PVC tube. Most LoRa repeaters (whether DIY or commercial) depend on solar panels paired with lithium-ion batteries, MPPT regulators, and charge management circuitry.

This leads to:

a high cost (often €150 or more);
some mechanical and electronic fragility;
a bulky footprint, making the device more visible and attractive to theft;
and above all, poor reliability in winter: in temperate climates, solar production can drop fivefold between summer and winter, forcing designers to oversize the system just to maintain basic service.

What We’re Proposing – Design Objectives

The goal is to build an ultra-discreet, robust LoRa repeater using widely available and very low-cost materials — specifically alkaline D-size (LR20) batteries and standard 40 mm / 50 mm PVC pipes for the enclosure and mast. The system is designed to run for multiple years in outdoor environments, including locations with no sunlight (under vegetation, in barns, etc.), for under €50 total, including the LoRa module and enclosure.

Note: This is version 1 of the device, and it is not yet final.
It provides around 2 months of autonomy, but it has already helped validate key design choices (mechanical layout, RF range, propagation testing).
A version 2 PCB is currently under development, with improved autonomy and resilience — including a watchdog, hardware timer, and complete shutdown of the Pico and buck converter between active phases.
Feel free to reach out if you have suggestions, improvements, or strong electronics skills — I'd be happy to discuss it.

Note 2: We are currently editing a video showing real-world radio propagation tests with the LoRaTube installed at Suc au May.
👉 It will be added to the logs soon.

______________________________________________________________

🔋 Power Supply

The power supply is based on 18 LR20 alkaline batteries (D size) in series, housed in a compact enclosure less than 50 mm in diameter.
This battery choice offers several practical advantages: LR20 cells are inexpensive (less than €1 each in most supermarkets) and widely available. Each LR20 battery typically provides 12,000 to 18,000 mAh, or 18 to 27 Wh. With 18 batteries, the total energy amounts to approximately 486 Wh, for a total cost of €13.30 (5 packs × €2.66), which translates to just ~€0.024/Wh — a ridiculously low cost compared to lithium alternatives.

For comparison:

A 18650 lithium cell offers ~11 Wh for ~€4 → €0.36/Wh (15× more expensive)
A pack of ten flat lithium cells (~52 Wh) costs ~€20 → €0.38/Wh

Wiring 18 batteries in series raises the voltage to 27 V (18 × 1.5 V). The cost per delivered kilowatt-hour remains extremely competitive: around €13/kWh, compared to over €250/kWh for a solar + lithium setup (including MPPT regulator) — a 20× cost advantage.

These batteries also have very low self-discharge (< 1% per year), enabling several years of operation as long as the current draw stays low.

Another benefit: the batteries fit neatly inside a standard 40 mm PVC plumbing tube.
There’s just enough room for the LR20 cells and a return wire for GND — with an internal diameter of ~36 mm and the batteries themselves measuring ~33 mm.

Advantages:

No exposed cables
A sealed, rugged housing
Fully weatherproof and mechanically robust
A long, slim form factor that allows discreet installation in many environments: tree trunks, rooftops, wooden posts, etc.
And in the event of a nuclear winter? You’ll be glad it runs on LR20s 😄

Preliminary question: can you be both a survivalist and environmentally conscious?

Using LR20 alkaline batteries may seem like an ecologically debatable choice — their main appeal lies in wide availability and an excellent €/kWh ratio. That said, the picture changes when you consider the following:

Producing 1 kWh of lithium battery capacity emits 150 to 200 kg of CO₂ (according to recent estimates).
Typical lithium extraction consumes ~1.9 million liters of water per ton, threatening local aquifers (e.g. the Lithium Triangle, Atacama Desert...).

By contrast:

LR20 batteries have a self-discharge of ≤ 2–3% per year and a storage life of 5+ years with minimal degradation.
In France, over 60% of alkaline batteries are recycled, while no effective recycling system exists yet for large-format lithium-ion batteries.

Advantages:
This is a simple, primary-chemistry solution (zinc and manganese). While not rechargeable, it offers very long field life and a low environmental impact per useful Wh, especially when yearly energy use is minimal.

The electrical design is also greatly simplified: no charge controller, no MPPT, fewer components — resulting in improved reliability and significantly lower cost.

______________________________________________________________

🛠️ Physical Architecture and Material Choices for a Standalone LoRa Repeater

🧰 Mechanics

The mechanical structure is based on a standard 40 mm PVC drainage pipe, whose internal diameter (~36 mm) perfectly fits an LR20 battery and wiring. This clever choice takes advantage of PVC’s strength, availability, and low cost, and minimizes the need for custom 3D-printed parts.

The 3D-printed components are grouped into three functional modules, requiring about 120 g of PLA in total:

The tube base, which provides access to the batteries for replacement and ensures contact with the negative terminal of the first battery in the 18-cell series. To avoid mechanical stress, the usual spring is replaced by a brass strip simply soldered to a wire.

2. The compression mechanism, which keeps the battery stack under pressure using three springs and two cylindrical PLA parts. The springs exert around 1 kg of force over 6 mm of travel, with a maximum travel of approximately 15 mm. An M3 screw, whose head moves freely in the upper part and locks with a nut at the bottom, ensures the whole assembly stays in place — even when the system is not under compression.

3. The intermediate part, which makes contact with the positive terminal of the topmost battery and acts as an axial stop to keep the battery stack compressed. Small vent holes at the bottom allow any condensation or water ingress to escape.

4. The top cap, which mechanically supports the PCB via the SMA connector, holds the antenna, and ensures the system is sealed.

It fits onto the top end of the 50 mm diameter PVC tube.

Mechanical Stress and Constraints

The compression force from the springs, which ensures good contact between the batteries, generates about 3 kgf at the top of the stack for a displacement of ~10 mm. The battery column weighs approximately 2.5 kg, so the load increases linearly — from 3 kgf at the top to nearly 5.5 kgf at the base — ensuring stable contact throughout the stack. This force is negligible compared to the mechanical resistance of an LR20 cell, which is rated at roughly 200 kgf.

Electrical Contact, Oxidation, and Galvanic Compatibility

Contact oxidation is a major concern for long-term reliability, especially in humid environments or those subject to temperature variations.
The internal connections use 2 mm brass rods, chosen for their conductivity, malleability, and ease of soldering.
However, brass naturally oxidizes, which can increase contact resistance over time and eventually cause power loss.

Two mitigation strategies were used to address this:

Solution 1: Tinning the brass tabs
Coating the contact surfaces with a thin layer of tin helps prevent oxidation by chemically stabilizing the surface.
This is a standard practice for electrical connections and solder joints.
It also helps maintain low contact resistance over time.
Galvanic compatibility is good: the potential difference between tin and the zinc-plated steel terminals of LR20 cells is moderate (< 0.6 V), well below the 0.8 V threshold considered safe in non-saline environments.
Solution 2: Applying silicone grease (dielectric grease)
Silicone grease forms a hydrophobic barrier that blocks oxygen and moisture from reaching the metal surface.
It is commonly used in electrical connectors to enhance long-term durability.
However, it is electrically insulating: it must therefore be applied only around or between contact zones — never directly on active contact surfaces, to avoid impeding conduction.

______________________________________________________________

⚡ Electrical Considerations

Internal Resistance

Using LR20 alkaline batteries in series adds up their individual internal resistances.
A fresh LR20 cell typically has an internal resistance of about 0.13 to 0.17 Ω.
With 18 batteries in series, the total internal resistance becomes roughly 18 × 0.15 Ω ≈ 2.7 Ω.

Under load, this causes a voltage drop:

At 110 mA (typical transmission with an E22400T22D module), the drop is
ΔV = I × R ≈ 0.11 × 2.7 ≈ 0.30 V
At 1.3 A (typical for E22400T33D),
ΔV ≈ 1.3 × 2.7 ≈ 3.5 V

These voltage drops reduce the effective input voltage available to the system.
Fortunately, in both cases, there’s still a comfortable margin for the buck converter, which only needs to step down to 5 V.

Actual Current Draw from the Batteries

Both the E22400T22D and E22400T33D modules are powered through a DC/DC buck converter (≈85% efficiency).
Assuming an input bus voltage of ~23 V (after subtracting internal drops from the 27 V nominal), the current drawn from the battery pack is:

E22400T22D (TX at 22 dBm):
~110 mA at 5 V → 0.55 W output
⇒ Input current ≈ 27 mA
E22400T33D (TX at 33 dBm):
~1.0 A at 5 V (values between 850 mA and 1200 mA depending on datasheet)
→ ~5 W output
⇒ Input current ≈ 242 mA

These are modest current levels for LR20 cells: 27 mA to 242 mA remains well within the comfort zone of a standard alkaline D cell.
An alkaline D cell rated at 21,500 mAh (at 25 mA discharge) can easily supply these currents over extended periods. Even an older battery typically retains several amp-hours of capacity under low current draw. Delivering ~0.3 A continuously remains well below its nominal capability.

🌡️ Effect of Temperature on Internal Resistance

The internal resistance of an alkaline battery increases significantly at low temperatures.
Slower electrochemical reactions and reduced ion mobility effectively "solidify" the electrolyte, increasing its internal resistivity.

For example, with an Energizer E91 AA cell (similar chemistry), internal resistance increases from about 0.15 Ω at 20 °C to 0.9 Ω at –40 °C.
By extrapolation, a new LR20 cell (0.15 Ω at 20 °C) could easily reach 0.4–0.5 Ω at –15 °C, or more as it ages.

This rise in internal resistance reduces the voltage delivered to the modules in cold weather, and decreases the effective ability to provide current spikes.

At –15 °C, we can reasonably estimate LR20 internal resistance to be between 0.3 and 0.5 Ω, leading to much greater voltage drops under load.

For example:

At 110 mA: ΔV = I × R = 0.11 × 2.7 ≈ 0.90 V
At 1.3 A (E22400T33D): ΔV ≈ 1.3 × 2.7 ≈ 10 V

This still works — but we’re nearing the limit.
At much lower temperatures, we’d likely see brownout resets, suboptimal LoRa transmission power, accelerated battery degradation, etc.

💥 Role of Supercapacitors

To address this, the design includes two 10 F / 5.5 V supercapacitors connected in parallel, located just under the PCB near the LoRa modules. These components are inexpensive (about €7.50 for five on AliExpress). Their main purpose is to supply energy during radio transmission peaks, offloading the LR20 cells and protecting them from sudden high current draw. (In the upcoming PCB V2, they will also allow the buck converter to be shut off during idle, further reducing standby current — a key factor in achieving multi-year autonomy.)

Indeed, during transmission, LoRa modules can draw peaks over 1 A for several hundred milliseconds (equivalent to ~300 mA from the battery side). Such bursts are stressful for alkaline batteries, especially in cold weather: their internal resistance spikes, causing voltage drops that can trigger microcontroller brownouts if the buck can't keep up. The supercapacitors act as energy buffers, quickly discharging to keep the voltage stable during these bursts. Between transmissions, the batteries slowly recharge the supercapacitors through a current-limiting device — a TPS2553, commonly used in USB power delivery, set here to ~250 mA.

This current smoothing has several benefits:

Improved performance in cold temperatures (less severe voltage drops)
Prevention of unwanted brownout resets
Extended battery life, especially in low-temperature environments
Reduced electrical stress from pulsed current draws

Additionally, the output impedance of the buck converter plays a critical role in 5 V rail stability during transmission.
During a sudden load spike, the control loop cannot respond instantly — its reaction is limited by bandwidth, output capacitance, and capacitor ESR. The instantaneous voltage drop under load is given by:

Properly sized supercapacitors ensure stable device operation and a steady current during transmission phases.

______________________________________________________________

📐 Electronics

Many thanks to PCBWay for their support. Once again, the manufacturing quality is excellent, and the delivery times are impressively short.
It’s a real pleasure to rely on such a dependable partner for projects that require fast, functional prototyping.

Their help made it possible to build the first working version of the device, which I’m sharing here.
A new version of the PCB is currently in development (see final section).

⚠️ Note: there's a mistake in PCB V1 — pins 3 and 4 can’t be mapped to a UART on the Raspberry Pi Pico.
Manual rerouting to pins 16 and 17 was necessary 😬

______________________________________________________________

📻 Radio : LoRa modules

The repeater uses UART LoRa modules from Ebyte: E22400T22D and E22400T33D, rated at 22 dBm and 33 dBm respectively. These modules integrate an internal 8-bit microcontroller responsible for radio management and buffering up to 1000 bytes, offloading the main microcontroller and simplifying frame handling. This design choice mirrors the one made in my original LoRa communicator project, and I chose to reuse the same LoRa module to ensure full compatibility.

I’ve used both the E22400T22D and the E22400T33D V2.1 in this build. One supports up to 22 dBm output power, the other up to 33 dBm.
Interestingly, the less powerful version is theoretically capable of longer-range communication.

⚖️ Spectrum Use, Duty Cycle, and Reception Performance

Although both the E22-400T22D and E22-400T33D embed the same LoRa chip (SX1262/SX1268), their real-world reception performance differs significantly at equal air data rates.

At 2.4 kbps, the measured sensitivity is:

–137 dBm for the 22D
–128 dBm for the 33D

This gap likely stems from the RF design of the 33D, which includes external PA (Power Amplifier) and LNA (Low Noise Amplifier) components optimized for high-power transmission. As a trade-off, the reception chain seems to introduce more noise, lowering effective sensitivity — despite the chip’s inherent capabilities. This illustrates a classic compromise between transmit robustness and receive sensitivity.

📶 Theoretical Link Budget (Ideal Case, No RF Loss)

E22-400T22D: +22 dBm – (–137 dBm) = 159 dB
E22-400T33D: +33 dBm – (–128 dBm) = 161 dB

So, both modules offer similar link budgets at 2.4 kbps, with a slight advantage for the 33D. However, that balance is broken if you exploit the low-speed modulation capability of the 22D: Unlike the 33D, the 22D supports down to 300 bps, achieving a sensitivity of –147 dBm.
Thus:

+22 dBm – (–147 dBm) = 169 dB, a very large link budget.

But as you will see later, based on the radio propagation simulation results (Longley-Rice / ITM model), the practical advantage goes to the 33D — mainly due to its higher transmission power.

⚡ Power Consumption per Frame (100 bytes at 2.4 kbps)

At 2.4 kbps, transmitting 100 bytes (800 bits) takes 0.333s. At 0.3 kbps, it takes 0.0333s.

Module	TX Current	Duration	Charge (I × t)	Energy @ 3.3 V
E22 22D	110 mA	0.33 s	36.3 mC	0.12 J
E22 33D	1 A	0.033 s	33.3 mC	0.11 J

So, at the same transmit power (e.g. 24 dBm), the 22D is 10× more efficient in energy per bit.

That said, higher transmit power shortens airtime, reducing LoRa duty cycle and enabling smoother text exchanges (shorter delays between packets) — beneficial when throughput is prioritized over range.

Later in the article, three main configurations are evaluated through simulation.

Case	Module	TX Power	Data Rate	Purpose
1	E22-400T22D	10 dBm	2400 baud	Regulatory-compliant setup
2	E22-400T22D	22 dBm	300 baud	Max theoretical range for 22D
3	E22-400T33D	33 dBm	2400 baud	Max theoretical range for 33D

A key limitation of these UART-based LoRa modules from Ebyte, including the E22 series, lies in the inability to adjust parameters such as the Spreading Factor. This value is fixed by the manufacturer and cannot be changed, which is the trade-off for benefiting from a fully autonomous UART buffer.

______________________________________________________________

📡 Propagation Simulation

For the propagation simulation, I used the excellent and free online tool developed by Roger Coudé (VE2DBE), available at https://www.ve2dbe.com/. This tool, in development since 1998, is based on the semi-empirical Longley-Rice / ITM model and offers a highly visual and flexible interface for radio coverage prediction. I’m grateful for the quality and accessibility of this resource.

The chosen test site for relay deployment is the Suc au May, a 908-meter-high summit in the Monédières range (Parc naturel régional de Millevaches, Corrèze). This location provides an unobstructed 360° panorama over the surrounding plateau, including clear views toward the Monts Dore, the Cantal mountains, and large open valleys to the northeast and south. Crucially, the area lacks any major telecommunication infrastructure—no digital terrestrial TV, no cellular relay—making it ideal for isolated RF testing. The Freysselines cirque, encircling the Suc au May, removes most topographical obstructions within a 300-meter vertical drop, significantly improving Fresnel zone clearance and RF propagation, particularly toward the south. This is the direction I selected for the long-distance tests. For reference, the city of Limoges lies about 60 km away.

The site also offers easy road access—an essential factor for practical deployment. I was able to leave the relay running for a few hours without permanent installation or degradation of the environment, as shown in the video documentation.

In terms of propagation assumptions:

the GPS coordinates of the emission point were set at LAT 45.472652 LONG 1.844849
the antenna used was a 5/8-wave RETEVIS tuned for 433 MHz with a 2 dB gain.
Losses in TX and RX were negligible (set at 0.1 dB), as the antenna was screwed directly onto the LoRa module. (To remain compliant with local regulations, I inserted a 2 dB SMA attenuator between the E22-400T22D and the antenna).
The relay mast height was 2.5 m, while the receiving communicator was positioned at 1.6 m above ground.

Case 1 : Regulatory-compliant setup

Case 3 : Max theoretical range for 33D

The simulation results based on the semi-empirical Longley-Rice / ITM propagation model show better coverage for the 33D at full power, despite a theoretical link budget that should actually favor the 22D in Case 2. This discrepancy is explained by the model’s inclusion of real-world factors such as terrain profile, diffraction, vegetation, and non-line-of-sight conditions. Unlike a basic free-space path loss calculation, it provides a more realistic assessment of radio performance in the field.

Note: In any case, both Case 2 and Case 3 should be considered post-apocalyptic scenarios only, as they do not comply with current radio regulations.

______________________________________________________________

New Long-Duration System – PCB V2

What needs improvement

Version 1 does not support shutting down the buck converter—only the complete shutdown of the Pico. Logging is not possible in version 1, whereas version 2 includes support for logging on FRAM, along with an RTC for both wake-up scheduling and log timestamping.
Off-the-shelf buck converters are unsuitable for ultra-low power applications. The MP2451 buck used here is selected for its low quiescent current and high-impedance voltage divider.

Features and Component

Thanks to surface-mount components, version 2 introduces a more comprehensive hardware logic. OR-wire circuits are built using low-leakage SOT23 diodes to protect wake-up pins and allow the Pico to identify the wake-up source. Concretely, an independent circuit handles the buck’s wake-up and current regulation. It uses OR-wire logic to drive a pair of N- and P-channel MOSFETs (to avoid pull-up leakage currents). The wake-up is triggered by the state of a MAX8212 comparator: it activates when the supercapacitor voltage (mounted on the PCB’s back side) drops below 4.7 V and shuts off when it rises above 4.95 V. During each wake-up cycle (which occurs every 2 hours), it also triggers the Pico via a second transistor pair, arranged in a totem-pole configuration. This second pair is wired with OR logic to the E22’s AUX pin (wake-up on radio activity), the RTC alarm, the TPL5110 timer, and an optional latch from the Pico itself. A TPS27081 power switch allows rebooting the E22 module if it becomes unresponsive to the Pico.

The BOM is 100% AliExpress-compatible. It includes:

MP2451 buck converter
MAX8212 voltage supervisor
TPS2553 current limiter
TPL5110 low-power wake-up timer
TPL5010 hardware watchdog
TPS27081 power switch
2N7000 and B250 MOSFET transistors
PCF8523 RTC powered by a coin cell
MB85RC256V FRAM memory

The FRAM and RTC are also mounted on headers, enabling separate configuration and data retrieval without desoldering — useful for debugging or extracting logs after long deployments.

The architecture enables precise energy management: the Pico is only powered up during specific tasks such as writing logs, processing received messages, or checking the E22 module during capacitor recharge. The rest of the time, the 5 V bus is sustained solely by the supercapacitors, and the buck converter is shut down to ensure ultra-low power consumption.

Thanks to their low ESR, the supercapacitors act as a robust energy buffer — even under cold ambient temperatures. Wake-up events for the Pico can come from multiple paths: RTC alarms, voltage comparator thresholds, or the TPL5110 timer, all contributing to the robustness of the system.

You’ll find the Gerber files attached at the article.

______________________________________________________________

Autonomy Estimate for Version 2

Available Capacity:
18 × LR20 ≃ 405 Wh (lower bound)

Daily Usage Assumptions:

1 h of continuous listening (LoRa in receive + Pico active)
100 transmissions per day, each lasting ~0.25 s (75 bytes @ 2400 baud)
Remaining time in deep sleep (buck and Pico off, LoRa in WOR ~10 µA)

Power Consumption and Daily Energy

Mode	Duration (h)	Power (W)	Energy (Wh)
Deep sleep (WOR)	23	0.00005	≃ 0.00115
Reception (LoRa + Pico)	1.000	0.145	0.145
Transmissions (100 × 0.25 s)	0.00694	0.660†	≃ 0.00458

† 0.550 W for LoRa + 0.110 W for the Pico during each transmission

Total daily energy ≃ 0.1507 Wh

Theoretical Autonomy

405 Wh / 0.1507 Wh/day ≃ 2,686 days (~ 7.4 years)

Note:

This is an optimistic estimate: it does not account for PCB leakage, actual converter efficiency, battery self-discharge at low temperatures, or the Pico’s ultra-low sleep currents.
Real-world tests (voltage monitoring, radio profiling) are required to validate this autonomy for PCB V2. With the FRAM logs, I plan to record the battery voltage each day.

______________________________________________________________

Conclusion

This project demonstrates that a LoRa repeater can be fully autonomous, discreet, robust, and ultra–low-cost—achieving multi-year operation without any solar panels. By relying on simple, readily available technologies—LR20 alkaline cells, optional supercapacitor buffering, a microcontroller that powers down between events, and a minimalist radio-routing logic—the device can exceed 5 years of autonomy while remaining compact (long, but under 50 mm in diameter), affordable (under €50 total), easy to build (BOM components all found on AliExpress), and deployable in the field—even under canopy or in winter (and yes, even in a “nuclear winter” scenario).

The design philosophy prioritizes resilience and simplicity:

Minimal software dependencies and no heavyweight OS
Configuration by jumper only—no re-flashing needed
Stateless, prefix-based frame routing—no complex protocols or unique IDs
Compatible with end-to-end encryption: the repeater applies only Reed–Solomon error correction to encrypted frames before re-transmitting them.
Hardware fail-safes including multiple wake-up paths for the Pico (RTC alarm, voltage comparator, radio activity, periodic timer), which verify all system states and trigger regular reboots; dedicated watchdog and timer circuits; a power-switch reset for the E22 module; and daily voltage snapshots logged in FRAM for long-term monitoring

Every design choice—from hardwired wake-up logic to supercap buffering—aims to reduce points of failure and streamline long-term maintenance.

Real-world validation is underway. A propagation test video is coming soon—the footage is currently being edited. An additional log will be published alongside this article to present real-world propagation results, which should be interesting to compare with the simulation outcomes.

As for long-term autonomy, it will be assessed once the V2 PCB is assembled. Daily battery voltage and key system events will be logged to FRAM, allowing us to estimate energy consumption based on the supercapacitor recharge cycles and verify the > 5-year autonomy goal. PCBWay is also supporting the fabrication of this second version—thanks to them! I hope minimal iteration will be needed before reaching a fully operational result. If you have any suggestions regarding the PCB or features it should include, feel free to reach out.

Project Details

Why LoRaTube?

🔋 Power Supply

Preliminary question: can you be both a survivalist and environmentally conscious?

🛠️ Physical Architecture and Material Choices for a Standalone LoRa Repeater

🧰 Mechanics

Mechanical Stress and Constraints

Electrical Contact, Oxidation, and Galvanic Compatibility

⚡ Electrical Considerations

Internal Resistance

Actual Current Draw from the Batteries

🌡️ Effect of Temperature on Internal Resistance

💥 Role of Supercapacitors

📐 Electronics

📻 Radio : LoRa modules

⚖️ Spectrum Use, Duty Cycle, and Reception Performance

📶 Theoretical Link Budget (Ideal Case, No RF Loss)

📡 Propagation Simulation

New Long-Duration System – PCB V2

What needs improvement

Features and Component

______________________________________________________________

Autonomy Estimate for Version 2

Power Consumption and Daily Energy