Compacting Look-Up-Table Models for All-Digital RF-Transmitters

摘要

Look-up table (LUT) models have been demonstrated to work well for predicting waveforms in circuits that show memory induced, nonlinear behavior. Although, for circuits with deep memory, the models become impractically complex. This work addresses that issue by purposing a method to reduce the memory demand of such models. It is accomplished by grouping model entries according to their dependency on prior symbols within the model's memory depth and combining nodes if this dependency is found negligible.