It’s often useful to understand time-based data through spectrum analysis.  It’s a familiar notion for anyone who likes music.  Stereo equipment sometimes features a display that shows the frequency content of music as is plays—high frequencies, mid-frequencies and bass—depicted with bar graphs.  In scientific applications, frequency analysis is useful for identifying substances or in medical imaging.  A sample under test is excited by some illumination, electrical or magnetic means.  The emissions from the sample are collected and analyzed to obtain a frequency spectrum.  The spectrum is used to identify the nature of the sample.

Conversion of time-domain data to frequency-domain data is made by a procedure called a “Fourier Transform.”   It is named for a 17th century French mathematician, Joseph Fourier, who described a method for converting the range of a mathematical function into a set of component sine waves.  A Fourier transform can be performed by hand or by a digital computer.  The analysis is relatively slow, though—too slow for a display on your stereo. 

In 1965, J.W. Colley and John Tukey published a paper that described a Fast Fourier Transform (FFT) algorithm that was dramatically faster than the Discrete Fourier Transform described by Fourier’s work.  The availability of the FFT made spectral analysis practical for medicine and material science, and it spawned a branch of supercomputing endeavors directed at performing FFTs quickly. 

The Nicolet Instrument Corporation, Madison Wisconsin, was one of the companies that built hardware to process FFTs.  The Nicolet 1280 (early 1980s) is a general purpose 24-bit minicomputer that contains a dedicated FFT board.  It runs an interactive operating system called Nicos.  It is purported to have had a FORTRAN compiler.  The Nicolet 1280 at the Rhode Island Computer Museum was once part of a General Electric nuclear magnetic resonance machine that used FFTs to produce MRI imaging when the industry was relatively young.