Application of time-frequency analysis for optimum non-equidistant sampling of automotive signals captured at knock

The optimization of modern car engines includes knock control. Knocking combustions are characterized by high frequency resonances of the burnt gas in the combustion chamber. Knock detection is usually based on the estimated resonance energies of pressure or vibration signals. Unfortunately, the resonance frequencies vary with the crank angle. Optimum estimation of resonance energies can therefore be achieved by time varying filtering. Alternatively, optimized non-equidistant sampling leads to resonances with approximately constant frequencies in the discrete time domain. This paper addresses the second approach. A parametric model for the frequency modulation is used and the paper describes a method based on Wigner-Ville analysis to obtain parameter estimates for optimum non-equidistant sampling. Real data experiments demonstrate the usefulness of this approach