This paper identifies a new mode called "critical voltage mode" in emerging distribution systems. The small-signal stability analysis indicates that load voltage dynamics significantly influence the damping of the newly identified mode. This mode has frequency of oscillation between the electromechanical and subsynchronous oscillation of power systems. A novel voltage controller is designed to damp this voltage mode of the system. The controller is designed using the linear quadratic Gaussian (LQG) method with norm-bounded uncertainty. The approach considered in this paper is to find the smallest upper bound on the H1 norm of the uncertain system and to design an optimal controller based on this bound. The design method requires the solution of a linear matrix inequality. The performance of the designed controller is demonstrated on a distribution test system for different types of induction motors. Simulation results show that the proposed controller with a careful uncertainty modeling has significant performance to improve the voltage profile of the distributed generation system compared to the conventional excitation controller and standard LQG controller.