Ghost Detector

Your phone does not have a magnetometer you are allowed to read. Both major platforms closed that door years ago, for reasons that have nothing to do with us.

What it does have is a compass — and a compass is a magnetometer that has already made up its mind.

The Field Dispersion Meter samples compass heading many times a second and measures how much that heading scatters. Not which way it points: how much it cannot decide. A phone lying still on a wooden table produces a very small number. Bring it near a speaker magnet, a laptop, a fridge, a wireless charger, or the wiring inside a wall, and the number climbs — because those things genuinely bend the field the compass is trying to read.

That is the entire method. It is real, it is repeatable, and you can prove it to yourself in about four seconds by holding your phone against a speaker.

The four channels

The number on the panel is a fusion of up to four independent measurements, and the panel shows you which of them is carrying the reading at any moment.

• DISP — heading scatter over a rolling window, as described above. The main channel, and the one the instrument is named after.
• ANOM — the compass rate of change held against the gyroscope’s measured rate of turn. The compass says the heading moved thirty degrees; the gyroscope says the phone did not move. The gap between them is the interesting part, and unlike scatter it cannot be produced by waving the phone about, because waving it moves both.
• ACCY — on iPhones only, the handset’s own estimate of how wrong its heading currently is. Apple gives this away for free and it degrades near metal and current, which makes it an honest second opinion at no cost.
• ACOU — microphone level, off unless an instrument asks for it. This one does not, so the front page never requests your microphone.

Two channels agreeing is more interesting than either channel shouting, so the fusion rewards agreement. The panel prints the number of channels currently elevated.

What it does not measure

Ghosts. It does not measure ghosts. Nothing on this site measures ghosts, and nobody has ever demonstrated an instrument that does. The Survey has been careful about this since 1974 and sees no reason to stop now.

What we can tell you is that a room with nothing in it should read close to its own floor, and that not every room does. We have written those up in the case files. We have not explained them.

Walk the room first, with the instrument off. Note where the wiring runs, where the appliances are, and where anyone has left a phone charging. Every one of those will move the needle, and an operator who has not looked will spend the evening measuring a fridge.

Throw the master switch. The movement drives to full deflection and falls back — that is the self-test every physical meter performs when power reaches it, and it tells you the needle is free and the scale is whole. It is not a reading.

Stand still for the calibration. Twelve seconds, in the middle of the room, holding the instrument as steadily as you can. It is measuring the floor. The panel will tell you whether you were actually still, and will say so plainly if the window was disturbed.

Then sweep. Turn slowly on the spot, holding the instrument out in front of you, and let the bearing rose fill in. Slowly is the operative word: a fast sweep gives every sector one glance and the rose will refuse to name a direction off that. When the rose has enough of the circle, it will point at whichever bearing has been reading highest.

Follow the bearing and find out what it is. This is the job. Nine times in ten you will find the cause and it will be dull — a router on the other side of a wall, a doorbell transformer, a radiator pipe. Finding it is still the job, and an operator who logs an excursion without walking to it has logged nothing.

The tenth is why the Survey still exists.

If the readings look wrong

Recalibrate. The floor drifts slowly on purpose, because rooms change when somebody turns a laptop on, but a floor established while you were walking about is worth very little. The panel keeps a confidence figure for the calibration and prints it.

The movement

The dial runs 0 to 100 and the number is a normalised magnitude, not a physical unit. Full scale is six sigma above your room’s floor. There is deliberate headroom above the calibrated ceiling: a meter that pins at full deflection throws away the difference between a large event and an enormous one, and that difference is exactly what a case file is written about.

The scale is not linear against sigma, and this is deliberate. The bottom of the range is expanded, so that a quiet room still produces a needle that sits somewhere and moves — a meter whose bottom third is dead reads as broken rather than as calm. The consequence is that a panel reading of fourteen can sit beside a sigma of nought point one quite legitimately. If you want the linear figure, read the sigma; the dial is for watching, and the sigma is for recording.

The small teal marker on the scale is the threshold at which the instrument starts logging. The amber line left behind further up the scale is the drag pointer — the session’s highest reading, held where the needle pushed it, as on a real meter.

Sigma

The second readout is the honest one. It is the distance above your room’s floor, measured in that room’s own spread, and it is the figure that appears in the register and in your case file. Two sigma is nothing. Six sigma is full scale. Anything into double figures has a cause and the cause is usually within arm’s reach.

The trace

Sixty seconds of history under the movement. The needle tells you what is happening; the trace tells you whether it has been happening, which is the difference between a bump and an event. The dashed line is the logging threshold and the shaded blocks are events that were logged.

The rose

A map of the room by direction. Each of the thirty-six sectors shows two figures: a solid bar for what that bearing usually reads and a translucent one for the highest it has read. A sector nobody has pointed the instrument at is drawn as an empty tick on the rim rather than as a reading of zero — the display will not let you mistake an unsurveyed direction for a quiet one.

On a phone the bearings are referenced to true north and the rose is labelled N, E, S, W. On a device with no north reference the rose still works, because a map of a room does not require a compass rose to be useful, and the panel says relative rather than printing letters it would be guessing at.

The three tiers

An event has to cross the threshold and hold before it counts, because knocking the instrument against a table produces one frame at full scale, and that is a bump. What survives is graded:

• Stir — over the threshold, held, but modest. Common. Worth a lamp and a line.
• Excursion — the logged unit of an incident record.
• Breach — a very high peak sustained for a long time. Rare, and the only thing on this site permitted to be red.

After an event closes, the instrument will not open another for a few seconds. Real disturbances arrive in clusters and without that rule one event logs itself eleven times.

The Survey keeps this list because operators who do not have it in front of them tend to report the same eight things every year. In descending order of how often they turn out to be the answer:

• A loudspeaker. Any loudspeaker. The magnet in a small bookshelf speaker will move the needle from a hand’s width away.
• A laptop, particularly a closed one, and particularly around the hinge and the trackpad.
• A wireless charger, whether or not anything is on it.
• A fridge, a freezer, or any compressor. These cycle, which is why the reading comes and goes on a schedule nobody in the room is aware of.
• Mains wiring in a wall, and light switches and sockets in particular.
• Radiators, pipework, structural steel, and reinforced concrete. Old buildings are full of iron and it is rarely where you expect.
• An induction hob, which is in a class of its own.
• Headphones, especially the ones in your own pocket.
• A phone case with a magnetic closure, or a magnetic mount left on a car dashboard. This one catches operators constantly, because the source is travelling with the instrument.
• Another phone. Two phones held together will both read high.

Work through that list before writing anything up. The Survey’s files are more useful than most because the boring explanations were eliminated first, and an incident record that does not say what was ruled out is not a record of anything.

Ghost Detector Lab is an entertainment site operated as the public terminal of the Thinplace Survey. See the disclaimer, the privacy notice, and the method for every instrument on the bench.