A GPU-Accelerated Harmonic Balance Method for Nonlinear Radio-Frequency Circuit Simulation

In this paper, we propose a new parallel harmonic balance (HB) method to accelerate the periodic steady-state (PSS) simulation for nonlinear radio-frequency (RF) circuits. The new method parallelizes construction of the nonlinear Jacobian matrix, which is the most time-consuming part, using graphical processing unit (GPU) platform to boost the simulation performance. Experimental results from several industrial cases show that the proposed parallel HB method leads to 3 x speedup compared to the traditional HB method while still keeping the similar accuracy.