Emulating Diode Circuits with Differentiable Wave Digital Filters

Wave Digital Filters and neural networks are two popular solutions for circuit modelling. This paper describes the development of a Differentiable Wave Digital Filters library. Diode clipper circuits were constructed. A dataset was collected from the circuits and, with the library, was used to train a real-time deployable model. The trained model has higher accuracy and similar computation time when compared to traditional white-box models.