This paper is focused on a cost effective real time implementation of a digital frequency detector. A case-study is presented. It is shown why it is not possible to achieve the requested results with a FIR or a DFT approach. Autoregressive Spectral Estimation techniques are analyzed in order to spot the reasons behind the choice of the “Covariance Least Mean Squares Forward” algorithm. Furthermore practical problems, such as those related to fixed-point approximation, are highlighted in order to reduce memory usage and MIPS load.