RST-controller design by pole placement has been introduced by several authors about fifteen years ago. However, in most of these works, the elimination of steady-state errors resulting from low-frequency inputs has been dealt with only for disturbance inputs, not for reference inputs. As to the latter input, a DC unity gain has been usually taken into account, thus insuring a zero input-output error for stepwise reference inputs, but not for signals of order higher than zero. In this paper, a new design method of RST-controllers is derived, which takes this error criterion into account by means of a secondary diophantine equation, in addition to the usual one.