Two sealed steel housings each hold a smooth steel flywheel. The right-hand one turns at 3000
revolutions a minute; the other is still. In light they are identical, and since
a smooth wheel spinning has the same shape as one at rest, so is their pull: the near-field image
shows two equal blurs.
But each pixel records two tensors, not one. The second, the gravitomagnetic field, is made by
moving mass, and shows the flow inside the housing: a vortex, turning the way the
wheel turns.
- Smallest detail at 3.8 m
- 33cm
- Mass along the centre of view, weighted by 1/distance
- 0 kgm−3Only the room lies along this line, and it is taken away; it gives 438 kgm−3 here.
The gravity side is an idealization: the mass of the known scene along each line of sight,
weighted by the inverse of its distance, blurred by the paper’s resolution law. It is what
an array as wide as the scene is deep would compute, with no noise and without the
arrangements of mass the camera cannot tell apart. “Outline” adds a faint copy of the shapes
in the ordinary picture so they can be matched to masses. “Tidal field” shows the field in
the room, of which the camera reads the value at its own place: the tensor at every point of
a slice through the room, all at the distance from the camera the slider sets. The colour is
the direction of the stretch there and the fine strokes lie along it; the ticks on the
slider mark where the slice meets something, and the controls under “Tidal field” set how it
is drawn. What one pixel reads at the camera itself, in s⁻² and source by source, is under
“Physics”.
Turning off “Room” removes the pull of the floor and walls. Gravity adds up linearly, so a
reading taken with the room empty can be taken away exactly; what is left is everything
that was brought in.