Microprocessor implementation of a time variant floating mean counting algorithm

It is usually assumed that counting algorithms are implemented by acquiring an unknown number of pulses in a fixed period of time. Acquiring measurements at a fixed time interval produces rate meter estimates at a known frequency, however it is possible to acquire an insufficient number of pulses in a sample window to produce accurate rate estimates. Acquiring a fixed number of pulses in a variable time period produces statistically accurate rate estimates but, for very low count rates, may have an unacceptable time response. This paper will discuss the derivation of a time restricted (bounded) fixed pulse floating mean algorithm from a time based floating mean algorithm. This algorithm minimizes rate estimate errors while producing data in an acceptable time period. Implementation of this algorithm on the Motorola 68HC11 and Intel 805X series embedded controllers will be discussed as well as implementation error sources.