The Proving Ground
Change a part, throw, and watch what happens. The physics is the project’s model, ported line by line and checked against it on 52 machines. Every range here is a model value, about 10 to 15% optimistic, and the page sizes no timber: a machine that throws far on screen may not survive in wood.
Loads as multiples of the counterweight’s weight (Mg) and the stone’s weight (mg). The design limits were 5 Mg on a timber hinge and 8 Mg on a steel pin.
A sweep changes one input and leaves the rest as they are. The sling and the prong are not re-tuned and the beam is not re-sized, so the curves show tendencies, not optima. The design study re-optimised every option before comparing it.
The page runs a JavaScript port of the project’s Python model: the equations of motion of beam, counterweight, sling and frame, derived by Lagrange’s method for six layouts, with bearing friction that follows the load, a stone that flies through the air with a drag coefficient of 0.40, and six ways of letting go. Run P9-R01 checked it on 45 machines and run P8-R08 on 52, with the largest range difference under a millionth of a per cent.
What it leaves out: the beam, the frame and the sling are rigid, which is why real throws come out 10 to 15% shorter than the model; the traction crew is fitted to one crew size; and no part is sized for strength, though the loads are reported. A number worth quoting comes from a model run with a run ID, not from this page.
Built with Claude from the project’s model (06_simulator in the project files). The 3D view draws the machine from the same positions the physics gives; proportions of the frame and box are drawn, not modelled.
The simulator guide · The equations · How the model was checked