In this paper is presented a new approximation function, based on a least-squares function, which minimizes the loss at the passband. With the proposed function are maximized the loss at the stopband. This optimization produces imaginary zeros, which makes it closer to the ideal function of a brick-wall filter. Comparisons are made with other classical approximation functions used to implement analog and digital IIR filters obtained by bilinear transformation, which demonstrate the advantages of the proposed function. Tables are presented that simplify the determination of these functions.