Research On Optimization Method Of Board Level Transmission Line In Signal Integrity

With the increase of signal transmission frequency of electronic system equipment,the problem of signal integrity（SI）of transmission line on printed circuit board（PCB）in highspeed equipment has attracted more and more attention.At a higher signal transmission frequency,the conductor on PCB is no longer an ideal conductor,and its inherent resistance,capacitance,conductivity and inductance can’t be ignored,resulting in the distortion of the signal passing through the conductor.However,improper PCB routing design will aggravate the degree of signal distortion,eventually lead to the decline of the quality and efficiency of inter chip communication,and even lead to equipment downtime or interference of other surrounding equipment.The general design method of high-speed transmission line is to establish the transmission line model by using electromagnetic theory and simulation method,simulate and optimize the physical model parameters of the transmission line by using the experience of engineers,and make PCB after obtaining the SI results that meet the requirements.The above process depends on the experience of engineers and is time-consuming,and the results are generally not the most ideal.This thesis uses neural network to model the physical model parameters and SI performance indexes of transmission line,and uses the combination of improved non dominated sorting genetic algorithm and good and bad solution distance method to replace the traditional engineers to optimize the physical model parameters of transmission line based on experience,so as to improve the SI performance of transmission line.The main research contents of this thesis are as follows:（1）The physical model parameters and SI performance index model of transmission line are established.The SI performance index of transmission line can be calculated by the lumped parameter model extracted by electromagnetic simulation,but the lumped parameters are related to frequency,while the W-Element that can characterize the lumped parameters of transmission line is independent of frequency.In contrast,the latter has great generalization ability.Therefore,in this thesis,the neural network is used to fit the relationship between the physical model parameters of the transmission line and W-Element,and the SI performance of the transmission line can be characterized by combining the frequency information.Aiming at the problem that the neural network needs large labeled samples and does not disclose the transmission line data set,this thesis uses electromagnetic simulation to build the data set of transmission line physical model parameters and w-element,and uses a small number of sample points to represent all design variables in the design space,so as to meet the training requirements of neural network and improve the modeling speed.（2）An optimization method of physical model parameters of transmission line is proposed.In order to reduce the dependence on senior engineers,this thesis uses the combination of non-dominated sorting genetic algorithm and good and bad solution distance method to optimize the parameters of transmission line physical model.In order to improve the iterative optimization speed of transmission line,based on the given physical model parameter design space,this thesis adds the constraints of PCB production process and engineers’ design experience conditions in the non-dominated sorting genetic algorithm.The optimization results can be directly applied to PCB transmission line design.（3）Experiments are designed to verify the effectiveness of the parameter optimization method of transmission line model.In this thesis,the above methods are applied to the transmission line design of high-speed memory.The optimized transmission line model parameters are compared with the model parameters obtained by traditional design methods,modeling,simulation and PCB manufacturing experiments,and their SI performance is compared.The simulation and experimental results show that this method can effectively improve the SI performance of transmission line.The method proposed in this thesis can be extended to the modeling and optimization of more complex transmission channels such as non-uniform transmission lines,and is suitable for rapid engineering evaluation in the actual design stage.