A signal which varies over time is monitored to determine an alarm condition, where the sample stream of digital values from an A/D converter is divided in to equal length pieces and a Fourier Transform (FT) algorithm is used to transform each piece of the stream into a three dimensional dataset including frequency domain amplitude, frequency and time. A Frequency Envelope is calculated by taking the maxima over the time dimension for a period of time, leaving a two dimensional frequency domain amplitude vs frequency dataset which is compared with new data arriving to determine the alarm condition for each element of the Frequency Envelope either by applying a constant delta additively or multiplicatively or by using a “leaky bucket” algorithm.