Predicting the EMI Induced Offset of a Differential Amplifier Stage using a Neural Network Model

In this paper we present a concept for predicting offset changes on a differential amplifier stage that is exposed to electromagnetic interference (EMI) on its inputs. We do this by using methods that are commonly used in the field of artificial intelligence (AI). To be more precise we develop a regression model based on a neural network topology. In the course of this we first create independent training and test data sets from simulations. The training data is then used to train prediction models, that are different in their structure and complexity. The test data is used to validate these models and to choose the best fitting model. Finally, we show that the model predictions match the real labels well, both for test data within and outside of the training data range, i.e. for higher frequencies than we trained for. Furthermore we provide the code as well as the data needed for the fitting algorithm, that was implemented by using the Tensorflow Python library. This work can be understood as a proof of concept, that can be applied to more complex regression problems to predict EMI induced offset changes.