Abstract: A method of constructing wavelets by orthogonal filter banks is introduced, and the orthogonal filter banks must be regularity for constructing wavelets. The wavelets constructed respectively by FIR (finite impulse response) filter banks and IIR (infinite impulse response) filter banks are compared, and then the de-noise results are analyzed respectively by seven order Butterworth wavelet (based on IIR filter banks) and Daubechies wavelet (based on FIR filter banks). The results show that the effects by using low order IIR filter banks and high order FIR filter banks are similar. Finally, the technique developed is applied to the diagnoses of fault within an internal combustion engine. It proves that, as an embranchment of wavelet application, Butterworth wavelet can be well applied in digital signal processing and fault diagnosis.