The force identification can be mathematically viewed as the mapping from the observed responses to external forces through a matrix filled with system Markov parameters, which makes it difficult or even impossible for long time duration. A potentially efficient solution is to sequentially perform the identification processing. This paper presents a parametric study on the sequential deconvolution input reconstruction (SDR) method, which was proposed by Bernal. The behavior of the SDR method due to the effects of window parameters, noise levels and sensor configurations is investigated. In addition, a new normalized standard deviation of the reconstruction error over time is derived to cover the effect of only independent noise entries. The sinusoidal and band-limited white noise excitations are tested to be identified with good accuracy even with 10% noise. The simulation results yield various conclusions that may be helpful to engineering practitioners.