Rotary positioning system, which usually uses direct drive scheme, is often used as motion simulator, radar or satellite antenna, etc. Direct drive is that the torque motor drives the load directly without any transmission like gears, belt or ball-screw. This cancels the gap and simplifies the machine. However, in motion simulator, changing loads endues it with uncertainty. The LQG (Linear Quadratic Gaussian) control law which is widely used in aeronautics is based on a state representation to control. This sort of control law integrates a Kalman filter that estimates the state space vector of the system, and an optimized state feedback depending on one quadratic criterion. The LTR (Loop Transfer Recovery) tuning is also used and provides an excellent robustness/performance compromise. The LQG-LTR control law is proposed for a single axis motion simulator, which is a DDR rotary position servo system with uncertain inertia load, in this paper. Mathematical model is built, based on which some analyses, which focus on the uncertain load, are made. Moreover, the LQG-LTR control law is designed and analyzed. How to tune the parameters is discussed. The results of simulation and experiment show that, LQG-LTR improves the robustness and dynamic performance significantly.