Calibration

Every earthquake size on the Catches page rests on three numbers. They describe how the instrument turns real ground motion into the counts the computer records, and if they are wrong then every amplitude here is wrong by the same factor — quietly, and in a way no amount of careful software can detect.

• Sensitivity — how many volts the coil produces per metre per second of ground movement. This one sets the scale.
• f₀, the corner frequency — below it the sensor goes progressively deaf. A 4.5 Hz geophone hears a 10 Hz wave well and a 1 Hz wave barely at all, and you have to know exactly where that turn happens to correct for it.
• ζ, the damping — how quickly the mass stops ringing after it is disturbed. It shapes the response right around the corner.

The destination is simple to state: all three measured, on this instrument, and re-measured often enough that we would notice if they drifted. Not taken from a datasheet describing a different unit built in a different year.

Because it is wrong, and we can prove it. The maker's figure for this geophone is 28.8 V/(m/s). Measured against a professional USGS station 1.6 km away, this element actually produces 9.0 V/(m/s) — about 3.2× less than the sheet claims.

It is not a wiring fault, and that is worth saying precisely, because a factor of three is far too big to shrug at. To lose that much signal in the electronics you would need something around 170 Ω sitting across the coil. There is nothing like it: the damping socket on the interface board is empty, the bias resistors are 100 kΩ — a loss of about half a percent — and the digitiser's input impedance is orders of magnitude too high to matter. No divider, no fault.

Nor is it a weak coil. A geophone's output and its coil resistance rise together, roughly as the square root, so an element genuinely three times quieter should measure nearer 40 Ω than 375. This one measures 385 Ω against a 375 Ω specification. The winding is what it claims to be.

So the factor of three is real and still unexplained. Three candidates remain, and the order matters. The two sites are not the same ground — the reference is 1.6 km away, and local conditions at these frequencies can differ by a factor of two on their own. The measurement leans on the two numbers this page admits are guesses: the comparison assumes 5–15 Hz is safely above the corner, which is only true if the corner really is 4.5 Hz. And a copy can match the winding and miss the magnet. Every one of those is answered by putting a known current through the coil, which is what the box in the next section is for.

Meanwhile the practical consequence stands: every magnitude this station reported before that measurement was out by that factor.

There is no single test that gives all three numbers, so the plan uses three, each answering what the others cannot.

1. Compare with a professional station nearby. There is a USGS instrument 1.6 km from here. Record the same earthquake on both, compare the amplitudes, and the ratio gives the sensitivity. This worked: it is where the 9.0 V/(m/s) above comes from, across 46 events.

It also failed, and the failure is worth stating plainly. It could not pin down f₀ or ζ. Between here and there lie 1.6 km of different ground, and ground changes the shape of a wave far more than the sensor's corner does. The two effects could not be separated. More earthquakes would not have helped — the limit is geology, not statistics.

2. Ask the instrument directly. A geophone is a coil and a magnet, and that works in both directions: push a known electric current through the coil and the mass moves. Cut the current and it rings down like a plucked string, and the shape of that ring-down gives f₀ and ζ with no geology in the way at all. A small battery-powered box is being built to do this four times a day, for years, so the answer is a running record rather than a single afternoon.

3. Listen to the instrument's own silence. Disconnect the sensor, leave everything else recording, and whatever the machine still shows is its own electronic noise rather than the ground. That tells you which parts of what this station records are real, and which are the instrument talking to itself.

This one has been done. On 24 September 2026 the geophone was unplugged for 23 hours and replaced with a shorting plug at the far end of the same cable, so the recorder saw its own electronics and nothing else. Three things came out of it.

• About a third of what this station records on a quiet night is the instrument, not the ground. Between 1 and 15 Hz the electronics alone produced 0.50 µV against 0.88 µV with the sensor connected. The ground is still the larger part, but not by as much as you might hope.
• The 1.05 Hz line in the spectrum is the electronics. It was still there, loudly, in every hour of a day with no sensor attached — so it is something inside the instrument, not something in the world. Which part is still unknown.
• The hum from the house heat pump really is the ground. A cluster of lines near 19, 20 and 41 Hz had been blamed on the air conditioning, but on circumstantial evidence. With the sensor gone they vanished — a thousandth of their usual strength — on a 32 °C day when the compressor ran repeatedly. They reach the record as real vibration travelling through the building and the slab.

It also left a puzzle. With nothing plugged in at all, the instrument's own noise still rises and falls by a factor of one-and-a-half over the course of a day. It is far too large to be temperature, and it is not the heat pump. What it is, we do not know yet.

There are now thousands of homemade and hobbyist seismic stations publishing data. Almost none have had their instruments measured; they use the datasheet value, as this one did. If a typical unit is off by a factor of three, as this one is, then an enormous amount of citizen seismic data is systematically wrong in a way nobody has quantified.

Measuring one station properly does not fix that. But doing it with parts costing a few tens of dollars, and writing down exactly how, means anyone else can check their own.

Everything above is the plan. This is the state of it today — and two of the three numbers are still guesses.

So: the scale is real, and the shape of the response is still borrowed from a datasheet. The instrument response published with this station's metadata is marked PROVISIONAL for exactly that reason.

The box that fixes it is built in firmware and waiting on a soldering iron. When it runs, this section changes and everything above it stays as it is.