Hear the Chatter

Chatter is a feedback loop between the tool and the surface it just cut. Each tooth cuts a surface left wavy by the previous tooth; if the new waves line up badly with the old ones, the chip thickness — and so the cutting force — starts oscillating, which shakes the tool more, which cuts bigger waves. Past a critical depth of cut the loop feeds itself and the vibration grows until the tool is bouncing in and out of the cut. That's the scream.

A map of spindle speed versus depth of cut, split into stable and unstable regions. The boundary has a scalloped, lobed shape: at most speeds the stable depth is limited, but at 'sweet spot' speeds — where the tooth-passing frequency divides evenly into the tool's natural frequency — the stable pockets reach much deeper. The practical magic is that adding speed can fix chatter that slowing down couldn't.

Regeneration depends on every tooth arriving at the same interval, so the waves from one tooth line up phase-coherently with the next. Unevenly spaced teeth break that: each tooth sees a different delay, no single phase relationship can dominate, and the energy that would feed one growing wave gets smeared across frequencies. In the simulator, toggle it on while chattering — often the same cut simply calms down.

No. The simulator is a single-mode model with fixed, representative stiffness and cutting-force constants — it teaches the shape of the physics, not your setup's numbers. A real machine has several modes that shift with tool stickout, holder, and workpiece. The diagram's lobes, sweet spots, and the variable-pitch effect are all real; the exact millimetres are illustrative.

Two routes. The rigorous one: an impact (tap) test with an instrumented hammer and accelerometer gives the tool-tip frequency response, from which software computes the lobes. The shop-floor one: cut test passes at increasing depth across a range of speeds and mark where chatter starts — you're tracing the diagram empirically. Listening for the chatter frequency and re-running near a sweet spot gets you most of the benefit with no equipment.

The audio is the simulation itself — a time-domain model of the vibrating tool (a delay-differential equation with the jump-out-of-cut nonlinearity), sampled and played back. Nothing is a recording. The buzz is the tooth-passing forcing; the scream that appears past the boundary is the regenerative loop saturating, exactly as in the model.