A new approach to deriving packaging system statistical eye diagram based on parallel non-linear transient simulations using multiple short signal bit patterns

The eye diagram simulation is very important for checking the signal quality of channels in electronic packaging systems. Most existing simulation tools for creating eye diagram are based on convolution using a single-step response of the channel and require that the I/O devices are modeled as linear time-invariant circuits. However, I/O devices do contain nonlinear elements and some applications have strongly nonlinear I/O devices. We need an alternative eye diagram simulation methodology to simulate these non-linear applications. In this paper, we propose a new approach to deriving the packaging system statistical eye diagram based on nonlinear transient simulations. This is achieved by taking advantage of the ready availability of multi-core networked computers and breaking the bit patterns into many short signal bit patterns and running these simulations in parallel. This can be done with any transient circuit simulator. This approach gives a practical way to overcome the time required to run the more compute intensive non-linear transient simulations. A major advantage of this approach is the capability to include the nonlinear effect of the actual I/O devices. The statistical eye diagram and the BERT scan eye diagram by the proposed approach includes the non-ideal channel characteristics including inter-symbol interference (ISI), the crosstalk from nearby aggressor channels, the random jitter, and duty cycle distortion. The proposed approach is illustrated with the simulation an actual channel from an IBM server. The statistical eye diagram is compared between this approach and the commonly used linear convolution methods.