Relaxation-based circuit simulation using dynamic states of circuits

To exploit exactness adaptation in relaxation-based circuit simulation is discussed in this paper. The basic concept is to use appropriate algorithms to simulate subcircuits of different exactness. We dynamically classify subcircuits to three classes of various exactnesses, and use a specific simulation method for each class while in simulating. The proposed algorithm is called exactness-adapted relaxation Newton (EARN). A sub-algorithm called linear relaxation Newton (LRN) that is especially capable of solving strongly coupled subcircuits is firstly proposed, too. All proposed methods have been implemented, and tested by experimental circuits to justify their correctness.