Deep Learning Enabled Design of RF/mmWave IC and Antennas

Design of RF and mm Wave circuits and systems is a tedious process that requires an expert team to balance out many requirements and trade-offs. This requires numerous sub-tasks to be assigned to different engineers, who then design based on their knowledge, intuition and experience. On the other hand, we are observing a paradigm shift in many other fields, thanks to artificial intelligence (AI). It is not only possible to automate many tasks, but AI can come up with solutions on demand. In this regard, there is ample reason to investigate AI assisted methods for design and automation of RF and mm Wave circuits and systems. This paper introduces an algorithmic synthesis approach that relies on deep convolutional neural networks (CNN) for the modelling of template-free electromagnetic (EM) structures. It is worth noting that once a CNN model is trained, synthesis time is measured within minutes. Moreover, this model can be repeatedly used for different design targets, such as antennas, matching networks, and filters. These points are exemplified with synthesis and measurement results.