IMS Deep Learning Enabled Generalized Synthesis of Multi-Port Electromagnetic Structures and Circuits for mmWave Power Amplifiers

This work reports an artificial intelligence (AI) assisted method for implementation of multiport electromagnetic (EM) structures. Utilization of AI in mm-wave circuit design is rapidly gaining attraction. While many previous works relied on surrogate modelling of a template, non-intuitive design spaces that provide exponentially larger degrees of freedom can uncover solutions that are not limited to the given template. Although it can be challenging to train accurate predictors for such design spaces, this issue can be tackled with a correct modelling approach. Within this context, we demonstrate the extension of previously laid out design framework to multiport EM structure synthesis [1]. This aims to facilitate the application of AI assisted methods for more comprehensive scenarios such as multiport matching networks, power dividers, diplexers and so on. To demonstrate the flexibility of this method, we show examples for different design goals. Lastly, we report a compact mm-wave amplifier with 3 dB gain bandwidth of 23.6–37.3 GHz, implemented with three port asymmetric power dividers and combiners that emerged through the AI-enabled method.