Parallel Frequency-Domain Simulation of Hyperspectral Waveforms in Nonlinear Power Amplifiers With Memory

We present a parallelization framework prototype for efficient physics-based computer simulation of hyperspectral time-dependent waveforms (i.e., wideband with a large number of frequency components) in nonlinear power amplifiers with memory. It relies on an adaptive algorithm for signal splitting and splicing in the time domain and uses a well-established pseudospectral multifrequency large-signal code, CHRISTINE, as its underlying simulation engine. Included in the model, and calculated from first-principles, are memory effects, such as dispersion and wave reflections. We validate our approach on a specific class of hyperspectral waveforms and study the effect of modifying a set of critical preprocessing parameters on the fidelity and the performance characteristics of the simulation.