COMPLEXITY REDUCTION OF MODEL OPERATIONS IN GENERALIZED MEMORY POLYNOMIAL FOR DIGITAL PREDISTORTION

Digital Predistortion (DPD) has been broadly implemented in Power Amplifier (PA) Linearization, to counter the PA non -linearity effects, which introduce additional operational costs to various PA applications such as base stations, mobile phones, and laptops. The core performance contributor of a DPD system is on its ability to accurately model the PA to acquire an inversed PA model that is used for compensating the input signals before feeding them to the PA. However, the improvement of PA modelling accuracy in DPD usually comes with a cost of increased computational requirements and additional challenges in implementations. In this paper, the popular Generalized Memory Polynomial (GMP) DPD algorithm, is optimized using the Binomial Reduction method to reduce the model operations complexity but maintaining linearization performance. The performance metrics include Normalized Mean Square Error (NMSE), where the pre -distorted PA output is measured against the ideal PA output to acquire magnitude of error in PA linearization. The NMSE measurement is applied on both original and treated algorithm, where they will both be compared. Close to 0 values indicates almost no differences among respective error magnitudes, concluding both algorithms have matching linearization performances. To measure model operations complexity, the number of multiplication operations required for each the original and treated algorithm is calculated, and then compared, where a lower number indicates fewer number of multiplication operations required, indicating lower model operations complexity. Model operations complexity is reduced 38%, with the treated GMP lagging 0.88 dB in NMSE. The difference in linearization performance is close to zero and acceptable, outweighed by the benefits observed in reduction of model operations complexity. The observed advantage would be impactful to almost all Memory Polynomial (MP) based DPD implementations in PAs, especially when PAs are increasing in importance in today's ever connecting world.