The object of this method of design is the determination of the smoothing weighting function wf. This approach makes it possible to base the design on minimum total error (signal deformation plus noise) evaluated at every or at any characteristic important point of the experimental signal. In a previous analysis classes of weighting function were defined. In the present paper the optimum wf of each class is calculated. These are universal functions and in the design only the width of the weighting function (the averaging interval) must be determined. The optimum width is calculated as a function ofaprioriinformation on the signal and the noise. Also, the precision with which this information is required is calculated. Examples of design of experimental signal smoothing are developed.