USB 3.0 IBIS-AMI Model Construction Using Measurement and Neural Network

The input/ output modes are essential for high-speed signal integrity analysis and channel simulation. This work aims to develop a method for generating an IBIS-AMI model for USB 3.0 using measurement data. Instead of requiring a specially designed motherboard with test points for specific measurements, this method uses measurement data obtained from an assembled motherboard. The only available data for measurement in this case is the output voltage waveform from the USB 3.0 port on the motherboard. To address this, a novel approach is proposed to extract all the required parameters for the IBIS-AMI model from a single available measurement using a neural network. The neural network is trained with a set of IBIS-AMI models, each containing parameters with varying values, and a series of voltage waveforms generated from channel simulations with these IBIS-AMI models. Once trained, the neural network can generate the IBIS-AMI model using just one measured output voltage waveform. This constructed model has no limitations related to the output channel and can be applied to different output channels for analysis, making it a versatile tool for high-speed signal integrity evaluation.