In this thesis, a novel statistical eye-diagram estimation method incorporating bipolar N-tap decision feedback equalizer (DFE) effects is proposed for high-speed serial links. Modern high-speed SerDes systems commonly employ bipolar decision stat...
In this thesis, a novel statistical eye-diagram estimation method incorporating bipolar N-tap decision feedback equalizer (DFE) effects is proposed for high-speed serial links. Modern high-speed SerDes systems commonly employ bipolar decision states of [1, -1] in DFE operation. However, accurately incorporating bipolar DFE behavior into statistical eye-diagram estimation remains challenging for conventional single-bit response (SBR)-based approaches due to their limited capability in modeling nonlinear feedback effects. To address this limitation, the proposed method extends a double-edge response (DER)-based statistical framework by directly incorporating bipolar N-tap DFE behavior. Unlike conventional SBR-based approaches, the DER-based framework preserves both rising and falling edge characteristics, enabling accurate modeling of bipolar DFE feedback effects. The proposed framework identifies the main-cursor (MC) position from the crossing point of rising and falling edge responses and sequentially applies DFE feedback effects to transient and steady-state responses. Based on the resulting DFE-incorporated DERs, probability density functions (PDFs) are formulated and recursively convolved to derive statistical output responses (SORs), from which statistical and BER eye-diagrams are efficiently estimated. The proposed methodology is validated using PCIe Gen 5.0 channels under various DFE configurations, including without DFE, 1-tap DFE, 3-tap DFE, and 20-tap DFE cases. The estimated eye-diagrams show strong agreement with conventional transient simulations, maintaining relative errors within approximately 5% for eye-opening metrics. In addition, BER bathtub characteristics closely match transient simulation results. Compared with conventional transient simulations, the proposed method achieves more than 228× computational acceleration while preserving high analysis accuracy. Furthermore, the proposed framework enables efficient exploration of the DFE coefficient space by rapidly evaluating eye-opening characteristics under various tap conditions. This capability facilitates the identification of eye-opening-sensitive tap regions and supports efficient DFE coefficient optimization without extensive transient simulations. Consequently, the proposed framework provides an efficient and accurate methodology for BER-oriented signal integrity evaluation and DFE optimization in next-generation high-speed serial links.