Within the framework of system identification, the input / output measured sequences have to be synchronized in order to obtain a good model performance; the impact of the synchronization becomes larger under two conditions, firstly when measurements of the input and output data are in different time instants; secondly when the bandwidth of the measured signals is close to the Nyquist bandwidth of the measuring instrument, as in current trends where systems are pushed towards large bandwidth usage, as in multicarrier operation.
This paper proposes a combined approach in order to introduce the synchronization problem into the model identification procedure, where the overall goal is to produce a device/system model that performs better minimizing the error criterion.
This approach is verified with measurements over a radio frequency power amplifier, In Fig 1. the normalized mean square error (NMSE) is plotted against the synchronization delay (∆), for data that was already synchronized (using a state of art technique with sub-sampling search ), as can be understood if this method it is the best then there should not be any other minima expect ∆=0, however an improvement (in NMSE) is found.
Third Biennial Conference on Advanced RF Measurement Technologies (RFMTC11). Gävle, Sweden. October 4-5 2011