Averaging Techniques for Complex Signals in Automotive Applications 2001-26-0052
When performing measurement and analysis in automotive applications it is common that the collected signals contain both a deterministic and a random component. It is of utmost importance to be able to find the parameter of interest without being too influenced by the other signals, commonly referred to as the noise contribution. It is also common to assume stationarity to the measurement requirement needed for the proper time and frequency domain analysis later applied. Frequency domain averaging using power spectra is possible to use when separating random components from deterministic, given that the stationarity requirement is fulfilled. At least the signal should be ergodic to make this averaging technique work properly. This requirement is rarely fulfilled since a large amount of the measurements are performed with speed or rotations per minute (rpm) as an important operating condition. Thus, it is very difficult to distinguish between the noise and the wanted signal when there are components being both deterministic and random and they may also vary with time. Common examples are when analyzing car compartment sound and the engine components might interfere with the components from the gear box or wheels. Each of them are tonal, or at least more tonal like, and are coupled to the rpm of the engine. On top of this there are measurement noise, background noise and other signals that might interfere with the wanted signal, like resonances in the car body structure, excited by the rpm.
A classical method to handle these situations has been to use an rpm trigger signal that allows time domain averaging for signal separation. There are many challenges associated with the reliability of this reference signal, and a contamination of this signal will quickly degrade the signal separation possibility. A novel method, Frequency Domain Synchronous Averaging, FDSA, has been developed to overcome many of these difficulties. The FDSA method separates the different signals without the classical reference signal limitations. This paper describes, through examples, the classical methods and their limitations and the advantages of using the FDSA technique.