Extraction of RF transceiver system parameters and impairments through detailed analytical modeling combined with a genetic algorithm approach

ABSTRACT To meet stringent market demands, manufacturers must produce Radio Frequency (RF) transceivers that provide wireless communication between electronic components used in consumer products at extremely low cost. Semiconductor manufacturers are in a steady race to increase integration levels through advanced system-on-chip (SoC) technology. The testing costs of these devices tend to increase with higher integration levels. As the integration levels increase and the devices get faster, the need for high-calibre low cost test equipment become highly dominant. However testing the overall system becomes harder and more expensive. Traditionally, the transceiver system is tested in two steps utilizing high-calibre RF instrumentation and mixed-signal testers, with separate measurement setups for transmitter and receiver paths. Impairments in the RF front-end, such as the I/Q gain and phase imbalance and nonlinearity, severely affect the performance of the device. The transceiver needs to be characterized in terms of these impairments in order to guarantee good performance and specification requirements. The motivation factor for this thesis is to come up with a low cost and computationally simple extraction technique of these impairments. In the proposed extraction technique, the mapping between transmitter input signals and receiver output signals are used to extract the impairment and nonlinearity parameters. This is done with the help of detailed mathematical modeling of the transceiver. While the overall behavior is nonlinear, both linear and nonlinear models to be used under different test setups are developed. A two step extraction technique has been proposed in this work. The extraction of system parameters is performed by using the mathematical model developed along with a genetic algorithm implemented in MATLAB. The technique yields good extraction results with reasonable error. It uses simple mathematical operation which makes the extraction fast and computationally simple when compared to other existing techniques such as traditional two step dedicated approach, Nonlinear Solver (NLS) approach, etc. It employs frequency domain analysis of low frequency input and output signals, over cumbersome time domain computations. Thus a test method, including detailed behavioral modeling of the transceiver, appropriate test signal design along with a simple algorithm for extraction is presented.