This paper studies an online nonparametric identification method that is based on a time-frequency representation of signals using Kalman filter. The method utilizes swept excitation signal by synchronizing the Kalman filter to the frequency of the excitation signal generator and updating the filter gains on a sample-by-sample basis. Four different closed loop identification configuration are studied and experimentally compared by considering direct and indirect identification approaches. Moreover, this paper studies loop transfer function estimation from the closed loop controlled servomechanism and discusses the possibility to use the result for loop diagnostics. The experimental results illustrate the effectiveness of the studied online identification method to estimate nonparametric frequency response of the closed loop-controlled servomechanism at a selected band of frequencies.