The RMS value of a signal is measured by averaging its squared value over an infinite duration, and similarly a power is measured by averaging the product of two signals (voltage and current). These computations of square, product and average are very easy when the signal has been digitised, but this digitisation carries three distinct sources of errors: truncation, sampling and quantification. These three error sources are studied independently of each others, as well as methods for reducing them, especially the use of a weighting window. With this “windowing”, a statistical study of the simultaneous effect of these three error sources has been made by simulation with a very large number of parameters values, and formulas are presented, which link the standard deviation to the resolution and the number of samples.

Key words