Warum es das MR2 gibt

Why the MR2 Exists

Compact self-sufficient solar LoRa mesh repeater of the 1 Wp class with directly mounted 868 MHz antenna

A complete, solar-powered, self-sufficient LoRa mesh repeater, barely larger than a fist. The antenna sits directly on the enclosure.

The MR2 is a solar mesh repeater board, born from several years of hands-on repeater operation. This is the story behind it.

The Meshtastic Years

With the rise of Meshtastic, the first LoRa repeaters were built – RAK-based, with Li-ion cells. The first limitation came quickly: the solar input of the WisBlock base board only charges between 4.4 and 5.5 volts. Anything more permanently damages the board, according to the datasheet. That completely dictates the panel selection – to this day, people in the forums ask whether the 6 V panel from RAK's own solar enclosure is even safe to use.

The next step was the classic one: external charge controller, LDO, big 6-watt panel. The poor efficiency of the converter chain was compensated with panel area. It works, but it comes at a price: the setup is conspicuous, the wind load considerable, and the choice of locations shrinks further. Of all places, the exposed spots where a repeater belongs from an RF point of view are ruled out mechanically or visually.

Then winter came. Battery dead, site only accessible through third parties, repeater offline for weeks. Plus a mistake that has to be admitted: the panels were mounted at an angle, because everyone did it that way. That vertical south-facing mounting is better for the winter balance – the low winter sun hits it at a more favorable angle, and snow and dirt do not stick – was a late realization.

The fallback solution was grid-powered repeaters on windowsills and balconies at friends' places. Those fail for reasons that have nothing to do with technology: during Advent, someone needs the outlet for the Schwibbogen (a traditional Saxon Christmas arch), in summer for the electric grill. Two days later the battery is empty and the node goes silent.

The Change of Strategy

With MeshCore, the logic was inverted: reduce consumption so far that small panels are enough. Small means inconspicuous, low wind load, and mountable in places where larger structures would never be tolerated. Energy efficiency became optimization vector number one. Number two: stability and freedom from maintenance, because the best repeater is the one you never have to drive out to again.

The target: a self-sufficient solar repeater in birdhouse format that makes it through the year even at a site that only gets diffuse light under foliage in summer. In winter the leaves are gone, and the birdhouse harvests the little available light unshaded.

Black birdhouse mesh repeater on a tree by a field path – barely visible

There is a complete mesh repeater in this picture: black box, tree trunk, half height. (STL files for self-printing – free)

An underrated side effect of compact construction: short antenna paths. Time and again you see installations where repeater and antenna are mounted separately for wind load reasons, connected by several meters of coaxial cable. At 868 MHz, cheap cable attenuates around 0.5 dB per meter, every connector transition adds to that, and every transition can work loose in the wind and draw water. Five meters of cable cost more signal than the step from 5.5 to 8 dBi antenna gain brings. In the MR2 concept, the antenna sits directly on the enclosure: N bulkhead connector, 15 centimeters of pigtail.

Solar mesh repeater with two 1 Wp modules and fiberglass antenna on a mast bracket

Repeater of the 2 Wp class on a mast bracket: two modules facing vertically south, antenna directly on the enclosure. (Enclosure as STL – free)

The January Calculation

A self-sufficient system is designed for the month with the lowest yield. In our latitudes that is January, and for January the low sun elevation matters – hence the vertical south-facing mounting. The yield can be quantified: a vertical, unshaded south-facing panel delivers around 30 kWh per kilowatt peak in Germany in January. Scaled down, that is about 30 Wh per Wp of panel power per month. Against that stands the consumption: just under 1 Wh per day, calculated at 1 Wh, so 31 Wh in January. A 1 Wp panel and the repeater are level in an average January.

There are also bad Januaries. Then the battery counts – specifically, the energy that can actually be extracted: the 1 Wp class repeater carries 9 Ah of LiFePO₄, nominally just under 30 Wh. In the cold, about half of that remains usable, and the cell rarely comes out of a gloomy December fully charged. In real terms, about one week of buffer remains.

The 1 Wp repeater is thus deliberately cut close to the line. What saves the calculation: really cold and really sunless at the same time is rare. Cold spells are usually sunny – then, because of the frost protection, little or no charging takes place, but the repeater runs directly from the sun via the power path. Overcast spells are usually mild – then charging is allowed. If both do coincide, the repeater goes into controlled hibernation and comes back on its own. That is how it is calculated. Anyone who wants certainty takes the 2 Wp repeater: room for eleven 18650 cells, and in an average January the yield exceeds the consumption.

What the Dream Board Should Do

Years of repeater operation produced a list of requirements:

  1. Small.
  2. Efficient, with minimal converter losses.
  3. Wide-range automatic MPPT: any common panel, whether 5, 6, 9 or 12 V.
  4. Frost protection while charging.
  5. Real-time clock on board.
  6. Temperature and relative humidity inside the enclosure – is the enclosure still tight?
  7. Multi-chemistry: Li-ion, LiFePO₄, LTO, Na-ion.
  8. CE certified.
  9. Future-proof: fully software-configurable.
  10. Brown-out protection. An empty battery must not be a nightmare: controlled sleep when the sun falls short, and autonomous restart.

A board like that could not be bought. So it was developed.

The Design Decisions

Every item on the list goes back to a concrete failure from the field. This is how they were implemented:

Converter losses out. The RAK4630 can run in high-voltage or in standard mode. The MR2 uses standard mode, which eliminates an entire regulator stage from the chain. The 3.3 V rail is generated by a TPS62840 with a quiescent current in the nanoampere range; when the battery voltage is close to 3.3 V, it effectively operates in pass-through mode.

Charging the way winter demands. The BQ25798 charge controller has a power path: in frost, the charge current is reduced or blocked, but the repeater's supply continues to run from solar. The workarounds circulating in the scene disconnect the panel in frost instead – then the repeater runs on battery in bright sunshine. Second property: an input range of 3.6 to 24 volts with autonomous MPPT. Practically every panel common in the scene fits, without charge controller and panel having to be matched to each other (anyone who has ever hunted for the right CN3791 knows the problem). The controller determines the optimal operating point itself and tracks it with the outside temperature, because solar cells also change their optimal voltage with temperature.

Batteries that last for years. The MR2 is designed for multi-year continuous operation. First choice is LiFePO₄: significantly more cycle-resistant than Li-ion and thermally very benign. That is an important point, because a battery going into thermal runaway in the woods would be a disaster for the whole scene – no such case has become known so far, and it should stay that way. Added to this are reduced end-of-charge voltages: 4.1 V for Li-ion, 3.5 V for LiFePO₄. Conventional setups charge to 4.2 V and keep the cells at full charge and high temperatures all summer. That accelerates aging considerably. On the debate around cold charging and JEITA profiles, there is a worthwhile field report by the YYCMesh community from the Canadian Rockies; an assessment can be found in the Battery Guide of the MR2 documentation.

Measuring instead of guessing. LiFePO₄ has a catch: the discharge curve is so flat that the state of charge cannot be read from the voltage. The empty battery then comes as a surprise. That is why an INA228 coulomb counter sits on the board and measures the state of charge. The same counter delivers the energy balance over 24 hours, three and seven days. That yields the statement that matters: if the weather stays like this, this repeater will be empty in two weeks. You see that remotely in the app, weeks before the outage, and can optimize the site or prepare for it. If the voltage does fall below the chemistry-dependent threshold, the INA triggers a hardware interrupt: deep sleep below 500 µA, the charge path stays active, the RV-3028 real-time clock wakes the board hourly for a check. The repeater returns on its own after every dry spell, without a ladder and without a manual reset. A BME280 additionally monitors temperature, air pressure and relative humidity inside the enclosure – you want to detect condensation below the dew point early.

Firmware that plays along. The MeshCore firmware for the MR2 consistently reduces the nRF52840's consumption: sleep optimization, USB completely off when nothing is plugged in. All board parameters are configurable via CLI, locally over USB or remotely via the app's admin access. The firmware translates the measured state of charge into equivalent Li-ion voltages so that the app's battery indicator is correct even with LiFePO₄, LTO and Na-ion.

Inside view: Inhero MR2 board with RAK4630 in a 3D-printed repeater enclosure

Inside a repeater of the 1 Wp class: MR2 with RAK4630, short U.FL paths to the antenna mount, plugged-in battery cable with protection module.

LTO and Na-Ion: Same Board, Different Cell

The battery chemistry on the MR2 is set once via CLI (set board.bat). The board supports Li-ion, LiFePO₄, LTO (2S) and Na-ion – with the matching end-of-charge and cutoff voltages from the firmware.

That is more than a datasheet feature. LTO cells charge even at −30 °C and achieve many times the cycle count of Li-ion; the YYCMesh folks use them for remote mountain sites that are unreachable for months. And Na-ion is on the rise: cold-tolerant when charging, thermally benign, and prices are falling with every manufacturing generation. Anyone installing an MR2 today can switch to Na-ion in two years without buying new hardware. One CLI command, a different cell, done.

From Roof Panel to Front Panel

How seriously winter is taken as the design case is shown by the evolution of the birdhouse repeater. The first version carried its two modules on the roof alone. A snow-covered roof paralyzes such a setup for days. Today's version has an additional vertical front module: it catches the low winter sun at a better angle and practically never gets snowed over or dirty. And if it does: five 18650 cells buffer for weeks.

Birdhouse mesh repeater with hoarfrost – early version with solar modules only on the roof

Birdhouse mesh repeater with vertical front solar module under test on a windowsill

Boxing Day, morning and noon: above, the early version with modules only on the roof; below, its successor with a vertical front module under test on a windowsill. This time without a power outlet. (STL files)

Solar repeater of the 1 Wp class with magnetic mount and ZIISOR 3 dBi antenna on a steel structure in January

Repeater of the 1 Wp class with magnetic mount on a steel structure, photographed on January 5. Mounted in seconds, no drilling; on top a 3 dBi fiberglass antenna. (Enclosure as STL – free)

In the Test Lab

The MR2 was meant to become a product, not a tinkering project. So the hardware went to an accredited test laboratory for conformity testing under the EU Radio Equipment Directive (RED 2014/53/EU). Measurements in the anechoic chamber included spurious emissions from 30 MHz to 1 GHz, in standby while charging and in transmit mode at full power. Both measurements: passed. The transmit spectrum shows a single carrier at 868 MHz; everything else sits more than 20 dB below the limit.

Inhero MR2 in the EMC test setup in the anechoic chamber of the test lab

Spectrum measurement of the MR2 in transmit mode: clean carrier at 868 MHz, EN 300 220 limits met

The MR2 in the anechoic chamber and the transmit spectrum at full power: one carrier in the allocated band, otherwise noise far below the limits of EN 300 220.

Result

The MR2 emerged from several years of hands-on repeater operation. The numbers from continuous operation: around 0.98 Wh consumption per day in real repeater service, measured over 24 hours. With that, vertical south-facing panels from 1–2 Wp are sufficient year-round in Central Europe, and one set of cylindrical cells buffers weeks of overcast weather. Driving out still happens – to build, not to repair.

Installation trip in the van with telescopic ladder, mast brackets, birdhouse and hilltop repeaters

Installation trip: telescopic ladder, self-welded mast brackets, birdhouse and hilltop repeaters, toolbox. Right at the front: a complete repeater of the 1 Wp class, barely larger than a fist.

For anyone who wants to dig deeper: the complete technical documentation including measurement series and the Battery Guide lives at docs.inhero.de. The board and the complete kits are available in the shop: Inhero MR2, Hilltop S, Hilltop L and the birdhouse set. The enclosure STLs remain free.

About the Author

The developer of the MR2 has been building electronics since the 1980s, when circuit boards were still called „printed circuits": traces drawn with nitro lacquer on copper-clad base material, etched in ferric chloride, drilled by hand. Two layers was the high art. Later came the degree, Dipl.-Ing. (FH) in Physical Engineering. What remains from that time: measurements count for more than datasheets.

Back to blog