Behavioral modeling of RF power amplifiers based on SVM

Abstract

RF power amplifiers (PA) are a major source of nonlinearity in a communication system. Accurate behavioral models are indispensable for PA linearization. To describe nonlinear characteristics of power amplifiers, a support vector machine (SVM) based modeling method is presented. The kernel approach and duality theory are employed to train the PA model. Simulation results show that the proposed model provides more accurate prediction of PA output signal compared with classic neural network models.

Support Vector Machine–Based Model for 2.5–5.2 GHz CMOS Power Amplifier

“…Furthermore, it constructs the optimal segmentation hyperplane in the feature space based on the structural risk minimization theory, making the learner achieve global optimization. Therefore, SVM has been widely used for behavior modeling and optimizing PAs [ 15 ].…”

“…Modeling the specification degradation behavior of an RF power amplifier belongs to the nonlinear regression problem formulated as follows [ 15 ]: given the data set {( x 1 , y 1 ), ( x 2 , y 2 ), …, ( x i , y i ), …, ( x l , y l )}, where x i ∈ R n , y i ∈ R , i = 1, 2, 3, …, l , the relationship y = f ( x ) between the inputs and outputs is obtained to predict the output value y corresponding to any input x .…”

“…The linear regression problem is considered first, then the nonlinear regression problem is discussed based on the linear regression problem. Finally, the linear fit function can be expressed as [ 15 ] where w is the weight factor and b is the threshold value.…”

“…Assuming that the given data set can be fitted without error using Equation (1) while satisfying the requirement of accuracy ε , we have [ 15 ] …”

“…It minimizes the actual risk by seeking structural risk minimization and pursues the optimal result under the limited information condition [ 13 ]. There are two main applications of support vector machines, one is classification, and the other is regression [ 15 ]. Moreover, the modeling of power amplifiers belongs to nonlinear regression.…”