Current in a cable, measured through its magnetic field

Because two unknowns stand between the field and the current, and neither can be measured at home. The first is geometry: the field of a two-core cable depends on how far apart the conductors are inside it, which varies by a factor of two or three between cables, and on exactly how far the sensor sits from them. The second is the phone: sensors low-pass their output before anyone sees it, the cut-off is not published, and at 50 Hz that can attenuate the signal several times over. Both are constants at a fixed mains frequency, so a single reading against a known load absorbs both at once. In simulation with a filter cutting the signal fivefold, the calibrated answer was 676 VA against a true 667.

Because no browser on iOS exposes the magnetometer to a web page — not Safari, and not Chrome or Firefox there either, since they are Safari underneath. Firefox on Android does not expose it either. Chrome and Edge on Android do. This is a platform decision, not something the page can work around, and the tool says so rather than showing an empty dial.

It is volt-amps, and the difference matters for exactly the appliances people most want to measure. A magnetic field is made by current; multiplying current by the mains voltage gives volt-amps. For a resistive load — a kettle, an iron, a filament lamp, a fan heater — volt-amps and watts are the same number. For anything with a switching power supply, a laptop charger or a television, the current and the voltage are out of step and the real power is lower, sometimes by a third. The tool reports volt-amps and says so; a meter that printed "watts" there would be quietly wrong on half of the things in a house.

It means this device is sampling on a perfect grid, and that is the one case where the measurement is impossible rather than merely difficult. A 50 Hz signal read exactly 100 times a second lands on the Nyquist frequency, where the amplitude you recover depends on the phase and can be zero; read exactly 50 times a second it folds onto a constant. Real sensors jitter, and the jitter is what makes the two separable. When it is absent the tool says so instead of returning a number that means nothing. Moving the phone slightly, or another browser, usually restores it.

Yes, and it is the job it does best, because it needs no calibration. Plug a heater or a kettle into the socket and switch it on, then hold the phone against each cable leaving the consumer unit in turn. The one carrying that load reads several microtesla while the others sit at the noise. Switch the load off and the reading collapses, which is the confirmation — a high reading alone can be any live circuit.

Because the field of a two-core cable falls off very steeply — as the square of the distance rather than in proportion to it, since the live and neutral currents flow in opposite directions and almost cancel. A centimeter either way can halve the reading. That is why every measurement should be taken with the phone in the same place on the cable, and why the on-and-off comparison is the reliable one: both halves are taken without moving anything.

Only a heavily loaded one, and only close up. The same steep fall-off that makes position matter means a cable behind fifteen millimeters of plaster gives a fraction of what the same cable gives touching the phone. A cooker circuit under load is findable; a lighting circuit generally is not. For finding cables in walls as such, what you want is a detector built for it — this tool is for the cable you can put a phone on.