04 MUSCLE

MUSCLE ARCHITECTURE — BIOINSPIRED SPEED

Speed in TCP actuators isn't just a materials problem — it's a systems architecture problem. By mimicking biological motor unit recruitment and elastic energy storage, we push performance beyond single-fiber thermal limits.

THE CORE TRICK — ELASTIC PRE-LOADING

Fine fascicles run continuously at low PWM, keeping elastic tendons under pre-tension at all times. When coarse fascicles fire, the spring is already wound. The tendon releases stored energy simultaneously with muscle contraction → net speed greater than either system alone. Free speed — no extra actuation.

CONTROL ARCHITECTURE
FINE FASCICLES (always on, low PWM)
     maintains pre-tension on elastic tendons
     positional stability + fast small corrections
     analog continuous signal from ESP32

COARSE FASCICLES (burst recruited)
     fires into already-loaded spring system
     tendon releases simultaneously with contraction
     burst PWM pulses from MOSFET driver

OIL-FILLED SLEEVES
     faster cooling → faster reset for next contraction

CONTROL LAYER (ESP32)
     MOSFET per fascicle wired physically
     grouped per muscle in software by default
     split to per-fascicle when needed — wiring already supports it
FULL MUSCLE HIERARCHY
LVL 1
FIBER
Nylon + nichrome co-insert, sewing thread bound, TCP coiled
LVL 2
PAIR
2 fibers twisted opposite direction — 2-ply, self-stabilizing, no return spring needed
LVL 3
FASCICLE
2–3 pairs + Kevlar strain limiters in 4mm PET sleeve (oil-filled)
LVL 4
MUSCLE
Fine fascicles (pre-tension) + coarse fascicles (power) — controlled per muscle via MOSFET
LVL 5
ARM
Muscle groups with elastic tendons, Kevlar ligaments, 3D-printed TPU skeleton
SPEED REALITY
PHASETIMINGNOTES
Coarse contraction0.5–2 secondsNichrome Joule heating — active
Coarse relaxation (dry)3–8 secondsNo oil — air cooling only
Human muscle~100msChemical, not thermal — fundamentally different
Reflex arc~20msSpinal, bypasses brain entirely
TCP suits deliberate, controlled motion — not fast reflexes. Speed limit moves from "how fast can nylon cool" to "how fast can you sense and recruit" — a much more solvable problem in software.