What is Signal Processing?

In many areas of engineering and science, an attempt must be made to understand a process by measuring its properties and analysing the signal obtained. Experimentation is usually carried out using one or more transducers followed by digital data acquisition and computer analysis. Although modern equipment can carry out a rapid and potentially very sophisticated analysis, it possesses a number of limitations which can have a detrimental effect on the quality of computed results. Unless these limitations are recognised, significant errors can be introduced. For most applications the data acquisition provides a "time-history" signal representing the output from some measuring instrument such as an accelerometer, microphone, strain gauge or temperature sensor. The process of data acquisition and analysis is commonly termed "signal processing".

The availability of fast multi-channel, analysis systems, together with commercial pressures to produce results, can often exacerbate problems. Vital information is often missed or misinterpreted and the cause of errors are not investigated. Even the simpler single or two-channel analyser can lull the novice operator into a false sense of security. With the press of a few buttons, such systems are capable of producing graphical display and hard copy output. However, the analyser default settings may not be optimised for the particular analysis. Given the same input signal, a second operator may produce a different output. The low cost PC and add-on data acquisition card has produced an expansion in the "build it yourself" inexpensive system. Often a lack of understanding of the specialised commercial product and its capabilities gives the impression that such systems are unrealistically expensive. The novice may embark upon a disastrous cost cutting exercise.

Signal processing procedures may identify a resonant frequency, a mode shape or an out of balance force, but it is the engineer who must decide if these characteristics actually exist. Digital signal processing may be applied to any type of data ranging from engineering dynamics and electronics to bio-medical applications. The data is often obtained from a phenomenon varying in time or space, and the information may be recorded in one or more dimensions. The short course offered by the author emphasised the practical issues of the subject and relied heavily upon examples taken from the author’s experience in noise and vibration. It should be emphasised, however, that the techniques are appropriate to many other applications.