This thesis deals with how mathematical functions are affected by different kinds of approximations. The aim is to implement a new type of AFC (Automatic Frequency Control) algorithm in a 16-bit fixed point DSP (Digital Signal Processor). Mathematical functions in the algorithm are to be approximated with either Taylorpolynomials or Chebychevapproximation. The approximations should reduce the clock cycles required in the DSP. The algorithm was implementable and the Taylorapproximation turned out to be the optimum approximation. The thesis resulted in a reduction from 28 cycles to four in the DSP without any significant loss in performance. Confidentiality protects the thesis so the results are only showed as relative to alternative methods.