In this article, we study the design of observer-based robust linear feedback controllers. The uncertainty, which can enterA, and either theB orC matrices, is assumed to satisfy certain matching conditions. Lyapunov techniques are used to establish sufficient conditions for stability, given an uncertainty bound. In particular, sufficient conditions are obtained that, if met, result in stabilizing controllers regardless of the size of the uncertainty entering the system matrix, as long as the standard constraints on the uncertainty entering the input or output matrices are met. As with the case of more general forms of uncertainty, the resulting observers often have high gains. To study performance, the problem of disturbance rejection is considered. Sufficient conditions are presented for obtaining control laws that stabilize the closed loop system, regardless of the size of the uncertainty entering the system matrix, while simultaneously guaranteeing arbitrarily small infinity norm for the transfer function from the plant disturbances to the outputs.