Investigation

Every check in this account is a real measurement taken by your own device, and each one says which sensor it is reading before it starts. There are five kinds.

• The compass. Facing a bearing, holding one steady, turning through an arc, or sweeping enough of the circle to have covered a room. This is the same heading reading INSTRUMENT 01 is built on. On a device with no north reference the bearing checks report that they cannot be taken, rather than checking you against a number that means nothing.
• The accelerometer. Standing still, and how steadily the device is being held. Nobody holds a phone perfectly still and the check does not ask you to — it asks for twenty seconds of not much.
• Orientation. Flat, upright, or face down on the floor. Measured from the same event the compass comes from.
• The microphone. Loudness only, against your room’s own tracked quiet rather than a fixed threshold — a fixed number marks everything in a kitchen and nothing in a stone church. No audio is recorded, nothing is transcribed, and nothing is sent anywhere.
• The camera. How much of the frame cannot be resolved: how dark a region is, how much it changes between frames relative to its own brightness, and how little structure it holds. That is the resolution floor from INSTRUMENT 04, and it is a measurement of the camera and the room rather than of anything in the room. The picture never leaves the device and no model is loaded.

The tolerances are deliberately wide and every hold drains rather than resetting, at a third of the rate it fills. A cough ten seconds into a twenty-second hold costs about three tenths of a second. A check lost to a badly calibrated magnetometer is a broken experience rather than a dramatic one, and there is nothing here worth being strict about.

The full account of what every reading on this site is made of is in the instrument methodology.

Every check offers to be taken as read, and taking it as read moves the account on. Not some of them and not the easy ones — all of them, on every device, including the ones that need a camera or a microphone. Half of everybody reading this is at a desk, and a desk has no accelerometer in it.

What a skip does not do is produce a measurement. The file distinguishes four outcomes — held, not held, taken as read, and no sensor — and the first two are the only ones that came off a sensor. They are stored separately, drawn differently on the card, and never added together. A run worked entirely on skips is kept, is shown, and is stamped as a simulation, which is exactly what happens to work done against the desktop bench.

An investigation is also service nowhere. It earns no clearance grade, takes no standing record and adds nothing to your hours, because the grades are statements about measured fieldwork and this is written fiction that happens to read real sensors. It gets its own line on your record and nothing else.

Brelling Mill is the Survey’s standing calibration reference and has been since 4 May 1974. The wheel-pit floor has been recorded at 1.83° on every visit since — three hundred and forty-one readings, fifty-two operators, eleven instruments sharing no components. The standing file is The Floor at Brelling, and its determination is that the Survey has been unable to establish why the figure does not vary, and has been unable to establish that it should not.

You are going for the three hundred and forty-second. What your instrument gives you in the pit is whatever your own room gives you, and the account has something to say either way. If you have not taken a baseline before, how to take one is the note to read first.