Battery and Solar Telemetry

Report cell voltage and charge state from a remote node, and calibrate the ADC

The most useful sensor on a remote node is not a thermometer. It is the battery gauge. A node that tells you it is at 3.4V and falling gives you a week’s warning; a node that just goes quiet gives you a mystery and a walk up a hill.

The free sensor you already have

Battery voltage needs no extra hardware. Every MeshCore node reads its own supply through an onboard ADC, and reports it as basic device telemetry.

On the Heltec V3 the battery voltage arrives on GPIO1 through an onboard divider.

Check it from the console:

stats-core

Battery voltage and uptime both appear here. And because battery is treated as basic health data rather than gated telemetry, it is available at a lower permission level than environmental readings - which is exactly right. Anyone should be able to see whether infrastructure is alive.

Calibrating the reading

Board-level battery reading uses an ADC multiplier to compensate for the voltage divider fitted to that particular hardware. If your reported voltage is consistently wrong, that multiplier is the thing to adjust:

get adc.multiplier
set adc.multiplier <value>

Valid values run from 0.0 to 10.0. The Heltec V3’s built-in default is around 5.42, reflecting the divider fitted to that board.

How to calibrate properly:

1. Measure the actual cell voltage with a multimeter, at the battery terminals
2. Read what the node reports via stats-core
3. New multiplier = current multiplier × (measured ÷ reported)
4. Set it, reboot, and check again

Worth doing? If you are going to make decisions based on the number - “go and swap the battery when it hits 3.5V” - then yes. A 5% error is the difference between a healthy node and a dead one. If you only care about the trend, the uncalibrated figure is fine.

Reading a LiPo voltage

A single-cell LiPo tells you a lot if you know the shape of its discharge curve:

The awkward truth about that curve: LiPo cells spend most of their life between 3.9V and 3.7V. Voltage is a poor guide to remaining capacity in the middle and an excellent one at the ends. Treat 3.6V as your alarm threshold rather than trying to read a percentage.

In cold weather, voltage sags under load and recovers when idle. A node reading 3.5V mid-transmit on a January morning may be considerably healthier than it looks.

Proper power monitoring with an INA219

The onboard ADC gives you voltage. If you want to know what your solar panel is actually doing, you need current as well - and that means an INA-series sensor.

MeshCore supports INA219, INA226, INA260 and INA3221.

For a solar node the INA3221 is the interesting one. Three channels lets you watch the panel, the battery and the load simultaneously, which tells you the whole story: is the panel producing, is the battery accepting charge, and is the load what you expected?

Wiring an INA219

I2C, exactly like the BME280:

The last two are the difference from a BME280. The INA measures current by sitting in series with the load, across an internal shunt resistor. VIN+ goes to the source, VIN- to the thing being powered. Get these the wrong way round and you get negative current readings - which is actually useful for telling charge from discharge.

Then, as always:

reboot
sensor list

What a healthy solar node looks like

Watch these over a few days and the pattern becomes obvious:

Good: battery peaks near 4.2V in the afternoon, drops overnight to maybe 3.9V, recovers the next day. Panel current positive through daylight.

Marginal: battery peaks around 4.0V, never quite reaching full. Surviving, but with no reserve for a bad week.

Failing: each day’s peak is lower than the last. You have a few days before it goes dark. Either the panel is undersized, something is shading it, or the node is using more than you thought.

The winter reality in the UK: a panel sized for July will not carry a node through December. Size for the worst month or accept a seasonal outage - and a battery telemetry reading is how you find out which one you have, before the node vanishes.

Cutting the load

If telemetry shows you are losing the battle:

set tx 17
set advert.interval 240
set flood.advert.interval 168

Lower transmit power, and advertise as rarely as the ranges allow. Each of these directly reduces transmit time, which is where nearly all the energy goes.

Beyond that, the honest answer is hardware: a Heltec T114 on nRF52840 idles at a fraction of an ESP32-S3’s current draw and is the standard community choice for solar nodes.

Try it Yourself

1. Run stats-core and note the battery voltage. Check it against a multimeter at the terminals. How close is it?
2. Calibrate adc.multiplier if the two disagree by more than a couple of percent.
3. Unplug USB, let the node run on battery for a day, and check the voltage morning and evening.
4. Challenge: wire an INA219 in series with a small solar panel and log panel current across a full day. When does your panel actually start producing?

Common Issues

Problem: Battery voltage reads zero or wildly wrong.

Solution: Check get adc.multiplier and calibrate against a multimeter.

Why: The multiplier compensates for a board-specific voltage divider. A wrong value gives a plausible-looking but wrong number.

Problem: Voltage reads high on USB and low on battery.

Solution: That is expected.

Why: On USB the board is running from the 5V rail through its regulator and the cell is charging. The reading only reflects true battery state when it is actually running from the battery.

Problem: My INA219 shows negative current.

Solution: Swap VIN+ and VIN-, or leave it and remember the sign.

Why: The INA measures the direction of flow across its shunt. Negative simply means current is flowing the other way - which on a battery channel is exactly how you tell charging from discharging.

Problem: The node dies overnight even though voltage looked fine at dusk.

Solution: Check the reading under load rather than idle, and suspect the cell’s capacity rather than its voltage.

Why: An aged LiPo holds voltage but not charge. It reads 4.0V, then collapses the moment it is asked for real current.

< Previous Next >

You can use the arrows  ← → on your keyboard to navigate between lessons.