In this paper the design and the implementation of a self-tuning PID controller is presented. The proposed methodology uses a least squares method with covariance resetting and a pole placement design for the calculation of controller parameters. The proposed design is implemented on stable and unstable systems simulated in a COMDYNA GP-6 analog computer, and a comparison of the performance of the self-tuning controller and a classic PID controller is presented in terms of root mean square error. The obtained results show that controllers of adaptive parameters have a better response than controllers of fixed parameters with respect to parameter variations that may affect the conditions of the system.