In today’s industry, whether it is run-to-failure, preventive, or predictive, mainte-nance is one of the major expenses in the production process. Ball bearings areone of the most vital elements in machinery and maintenance cost for replacementof those elements with interrupting the production is one of the most expensive.Establishing predictive maintenance for those rolling bearings by detecting the pos-sible defects and monitoring the current condition will enable the industry to use themaximum life span of those mechanical devices and reduce the cost of maintenanceconsiderably.Within the ongoing project of condition monitoring by using vibration analysis atRubico AB, this thesis work aims to understand and implement a new algorithmstep-by-step, first as off-line processing, and then on a fixed-point digital signalprocessor to analyze the measured data from industry. Different approaches formaximizing the performance of the algorithm are tested, compared; and the resultsfrom both off-line floating point precision and fixed-point implementation are eval-uated.By running the method on different data sets from industry, it has been shownthat the patented algorithm manages to detect the defects on the inner or outerrace of the ball bearings without a priori knowledge about the measurement objectand environment. The concept for implementation on a fixed-point digital signalprocessor is also proven.