nx_contactforce_gate.nx

nx_contactforce_gate.nx -- THE CONTACT READBACK IS REAL: it rises with depth, it tells firm tissue from soft, and it is silent when nothing is touching. (softbody SB2 / hapticcare HC1, 2026-08-25.) WHY THESE TEETH AND NOT A NUMBER. The solver carries NO KILOGRAM -- inverse mass is a uniform q12 -- so an absolute newton is not derivable and st_contact_force does not claim one. That makes a tooth of the form "force == X" impossible to write honestly. What a haptic loop actually consumes is COMPARISON: press harder and it must read harder, press firm tissue and it must read harder than soft at the SAME depth. Both are unit-INDEPENDENT, both are exactly what the absent-unit still supports, and both fail loudly for a readback that is a constant, a counter, or noise. THE PROBE RECIPE IS BORROWED, NOT INVENTED. nx_softtissue_gate already measured that teleporting the probe to full depth puts its centre behind the surface and projects far-side vertices BACKWARDS through the chest wall, inverting elements -- an artefact of instantaneous insertion, not of the solver. So the probe approaches from OUTSIDE (z_touch = surface projection + radius) and ramps in over CF_RAMP substeps, the same way that gate does it. Re-deriving the approach would have re-earned their bug. license_tier: ORIGINAL No hw writes (Rule 26). expect_exit: 0