An algorithm for a model-order reduction method based on impulse/step response error minimization is presented. The denominator of the reduced model is selected by either the Routh method or the retention of the desired poles of the higher order system. The numerator of the reduced model is obtained by minimizing the impulse/step response error along with a steady-state constraint. It is shown that this minimization problem finally leads to a problem of solving linear algebraic equations to obtain the numerator of the model.