“Introduction:

The knowledge and understanding of the FFT process by vibration specialists has greatly expanded since I gave my first talk on the FFT at the 1980 Annual Meeting in New Orleans. At that time, data acquisition times, windows and the relationship between the AID sample rate and the Filla, of the 400 line spectrum were not widely understood. Today, anyone certified as a Level 1 Vibration Specialist has a pretty good idea of how the FFT vibration analyzer works. In fact he probably uses one daily in his machinery condition monitoring program.

In spite of this improved understanding of the FFT process, there are still some areas where the unsuspecting can get in serious trouble. In some monitoring programs the number of averages to take for accelerometer data seems to be a constant of nature. We always take 8 averages for example. The number of averages should be determined after a review of the character of the data. For extremely noisy data the combination of too few averages coupled with very tight change specs allowed may lead to an unnecessarily high number of alarms.

The FFT process should, strictly speaking, be used only on stationary data. In other words, the data must not be varying in its frequency or its statistical characteristics. Additionally, it was intended to only treat harmonic data where each signal is precisely a harmonic of the center of the first bin.

If these rules were strictly followed, then the FFT would not be the popular processing tool it has become. One must realize however that when one uses the process in non stationary data situations one has to be very careful in setting up the analyzer or serious errors can occur.  The FFT has even been used to process transient data. The idea that the peak hold feature can produce the amplitude portion of a Bode plot during startups or shutdowns can result in getting any data one wants for the response at the critical speed depending on how we set up the analyzer. Some of these situations will be analyzed in the paper.

"The Number Of Averages Is A Constant Of Nature:

The statement that “when looking at accelerometer data from a new machine, a 400 line spectrum and 8 averages is a reasonable first choice”, has been interpreted in some plants as “use a 400 line spectrum and 8 averages for accelerometer data”.”