Filtered digital information representative of the power of an output signal from an amplifier is generated, and control information from a setpoint representative of a desired power for the output signal is generated from the desired response time for the establishment of the power and from a model to be followed for the establishment. In addition, a gain control and a bias current control are established for each initial variable-gain amplifier from the filtered digital information, from the control information and from a polynomial digital regulation of the RST type whose polynomial coefficients are determined from input parameters comprising characteristics of each variable-gain amplifier and from characteristics of the filter generating the filtered digital information.