This paper revisits the global asymptotic tracking of robot manipulators with position measurements only. A desired compensation adaptive law plus a saturated feedback term coupled to a new dynamic nonlinear filter is designed to produce global asymptotic tracking while compensating for parametric uncertainty and requiring link position measurements only. A new Lyapunov-based stability argument is utilized to prove the global asymptotic tracking. Simulations are included to illustrate the effectiveness of the proposed approach.