Fast ISI Characterization of Passive Channels Using Extreme Value Distribution

Accurate prediction of I/O system performance including random and deterministic jitter requires a new simulation framework beyond traditional SPICE simulation. Recently, statistical-based simulation tools, such as StatEyereg and LinkLab, have gained a special interest. These statistical simulation tools compute system bit error rates (BER) based on the probability distribution functions (PDFs) of inter-symbolic interference (ISI) noise. The ISI distribution is calculated by convolving the ISI cursors from the single bit response (SBR) of passive channels [4]. Although this approach produces fast and accurate results, it is limited to differential systems with linear drivers [1]. For more general applications, such as nonlinear single-end signaling drivers as well as the support for coding schemes, transient simulation is desirable but it is also limited by the large simulation time. In this paper, we propose a new approach based on the theory of extreme value distribution (EVD) to speed up the ISI distribution calculation. A regression test of 13 backplane channels is performed to test the accuracy and robustness of the proposed method.