Signal integrity and power integrity analysis for high-speed PCB systems

Abstract

To achieve high-speed printed circuit board (PCB), the issues in signal integrity (SI) and power integrity (PI) have become unprecedentedly challenging. Motivated by the benefits from ultra-fast clock speeds and shrinking PCBs, SI and PI engineers are actively searching for a more accurate approach to extract circuit parameters of the increasingly small-and-arbitrary geometry interconnects on PCBs, and a more reliable way to measure voltage of two close traces without contact. Although transmission line modelling has been widely regarded as a benchmark for circuit modelling in electronic industries, it becomes gradually less effective in simulating smaller objects with arbitrary geometries. A promising alternative is the partial element equivalent circuit (PEEC), which has attracted substantial interest to overcome a wide variety of challenges. One of the most urgent challenges is, the increase in miniaturization and digital speed in PCBs amplifies the uncertainty effect in the electrical interconnects on PCBs, which makes it an exigent issue to accurately predict PCBs’ performance. To address the above challenge, in this thesis I unveiled a geometry-related stochastic PEEC model to enable uncertainty quantification (UQ) of small and arbitrary shape interconnects in high-speed PCBs for SI and PI. Specifically, the deterministic PEEC model is integrated with an efficient statistical technique—polynomial chaos expansion (PCE), which maps the stochastic processes into polynomial bases and significantly reduces the complexity of the model, leading to a considerable speed-up compared to standard Monte Carlo simulation without loss in accuracy.To fully meet the commitment of PI in high-speed PCBs, one requires measuring the real voltage in the fabricated PCBs, instead of merely relying on simulation results. Non-intrusive voltage detection is a powerful method for PI diagnosis in high-speed links and multiple signal PCB designs. However, due to the dispersive nature of signal delivery passages and interference from neighbourhood signals, it is very difficult to measure the exact voltage of two adjacent traces without contact. The second contribution of this thesis is, a non-contact time-domain voltage measurement method is proposed and introduced to detect voltage signals of two adjacent traces on a PCB. A near-field electric probe is employed and a reconstruction algorithm is developed to recover voltage signals without contact. Through experimental benchmarks, it is demonstrated that the proposed method can successfully recover the common signals, differential signals, and non-correlated signals from adjacent dual traces. In this thesis, geometry-related stochastic PEEC model allows high-speed PCB designers to predict the most crucial geometry-related statistical information and to inspect the dissimilarity between the models and the real world. Foreseeing these results and differences is the first step to create a more robust high-speed PCB system and lower product redesign risk and cost. Meanwhile, the non-contact voltage measurement of two adjacent traces offers a powerful method to diagnose and optimize PI for high-speed and multiple signals PCBs.