The goal
A global network, built to give people warning
A storm announces itself long before it arrives. Every discharge sends out a radio pulse that crosses a continent in milliseconds, while the weather behind it moves at the speed of wind. That gap — often tens of minutes — is warning time, and it costs nothing except somebody listening.
Lightning kills and injures people every year, and it does so most readily where there is least warning: outdoors, at work, far from shelter, in regions no detection network covers closely. Dedicated infrastructure will never reach all of those places. A quiet receiver on a windowsill can.
A single station is already useful. It hears storms far beyond the horizon and tells you roughly which way they lie. But one receiver can never tell you where a strike happened — only which line it came along. Turning directions into positions takes several stations, and it takes them far apart. That is why the goal is a global one, and it is the one thing that cannot be solved by building a better receiver.
Everything here is built toward that: the receiver, the detection maths, the direction finding and the server were all made and measured from the ground up, so every number the map publishes can be traced back to something a station actually heard.
Where this stands today. This is the goal, not a description of the present. Right now the network is a handful of stations in central Europe, and most of the world has no coverage at all. It grows one operator at a time — which is why where a station stands matters more than how many there are.
How it works
From a flash to a bearing
A lightning discharge is a current pulse tens of kilometres long, switching in microseconds. It radiates most strongly at very low frequencies — a few kilohertz, far below the AM broadcast band. At those wavelengths the gap between the ground and the lower ionosphere behaves like a waveguide, so the pulse spreads sideways around the curve of the Earth instead of escaping into space.
That is why a receiver in Saxony can hear a storm over the Mediterranean. The pulses are called sferics, and they arrive as a crack a few hundred microseconds long.
Two rods, crossed
A coil wound on a ferrite rod is deaf along its own axis and most sensitive across it. Two rods at right angles therefore see the same pulse at different strengths, and the ratio between them encodes the direction it came from.
Finding the pulse
The station tracks a slow-moving average of the background noise and flags anything rising six times above it. A sferic rings for hundreds of microseconds; a switching spike from a nearby charger lasts one or two samples. Counting how long the burst stays up separates them.
Averaging the whole burst
The bearing is not taken from the loudest single sample, which can be distorted. It comes from the statistical axis of the entire burst — dozens to hundreds of samples — which is far steadier when the signal is weak.
Every bearing is a line, not an arrow. Two magnetic loops alone cannot tell a strike to the north from one to the south — both produce an identical signal. Resolving that would need an electric-field antenna as well. We do not try to hide the ambiguity in the maths: the network is built to work with lines, and when bearings from two stations cross, the ambiguity cancels out on its own.
Turning bearings into places
Why the distance between stations decides everything
When two stations hear the same strike within the same two-second window, their bearing lines are intersected by least squares. The accuracy of that crossing has almost nothing to do with the quality of either receiver — it is set by the angle at which the lines meet.
The further away the strike, the shallower that angle. For two stations 30 km apart, a strike 1900 km away lies within less than one degree of the same direction from both — a difference that vanishes into the measurement noise. Move the stations further apart and the angle opens up.
| Second station in | Distance apart | Error at 1900 km |
|---|---|---|
| Prague | 137 km | 2308 km |
| Hamburg | 363 km | 867 km |
| Copenhagen | 542 km | 582 km |
| Amsterdam | 580 km | 543 km |
| Warsaw | 583 km | 540 km |
| London | 913 km | 345 km |
Measured from the first station near Chemnitz, Germany, assuming five degrees of bearing accuracy.
Five stations in one city locate nothing. Three spread across a continent locate a great deal. This is the reason the project cares more about where a station stands than about how well it is built.
Being straight about it
What this network cannot do
Plenty of detection projects quietly present estimates as measurements. These are the places where ours is weaker than it looks, stated plainly.
- It is not a warning service — do not rely on it for your safety. Protecting people is what the network is being built for, but it is nowhere near that yet: coverage is thin, stations go offline, and a storm nobody hears does not appear on the map. For decisions about your own safety, use your national weather service and the 30/30 rule — if thunder follows the flash within 30 seconds, go inside, and stay there for 30 minutes after the last one.
- Distance is an estimate, not a measurement. A station infers range from how strong a pulse arrives relative to its own noise floor. But a powerful strike far away and a weak one nearby can arrive identically. The figure shown on the map is a useful proxy and nothing more — it has not been calibrated against reference data.
- Two nearby stations cannot locate distant strikes. That is geometry, not build quality, and no firmware update will fix it. Where the crossing angle is too shallow to trust, the server refuses to publish a position rather than inventing a confident-looking one.
- Bearings carry a 180° ambiguity until a second station resolves them, because the stations measure only the magnetic field.
- Coverage is thin. The network is small and young. Wide areas have no station anywhere near them, and a strike nobody hears simply does not appear.
Why say all this on the front page? Because a detection network is only worth anything if you know how far to trust it. Everything above is visible in the data we publish — the signal-to-noise ratio and burst length of every detection are right there in the map popups, so you can judge a reading for yourself.
The hardware
What a station actually is
A compact receiver that sits indoors, runs from a USB power supply and needs no attention once it is set up. It reports over your WiFi and stores nothing locally.
The demanding part is not the electronics but the quiet. A receiver hunting for signals a few microvolts tall sits in a home full of chargers, LED lamps and computer power supplies, all radiating in the same band. Most of the engineering went into keeping that interference out: careful board layout, a tight band-pass before anything is amplified, and a digital filter that removes mains hum before detection begins.
It works. Moving from the first hand-wired prototype to the current board dropped the noise floor by a factor of roughly 3700, and the station now spends its time listening to the atmosphere rather than to the house around it.
Your location is not published. Station positions are rounded to about a kilometre before they ever leave the server, because a receiver usually sits at someone's home. Device addresses are anonymised before storage. Details are in the privacy policy.
Taking part
The network needs distance, not numbers
If you live somewhere far from central Germany, a station at your place is worth more to this network than several more near the existing one. That is not a figure of speech — it is the table further up.
Stations will be available in small batches, assembled by hand. There is no price or date yet, and priority goes to locations that open up the geometry. The best way to register interest, ask what is involved, or just watch the thing develop is the community server.
Join the community
Run a station, compare measurements, or just ask questions.