Research On Signal Integrity Of Nonlinear Link Based On Worst-Case Analysis

With the continuous development of hardware technology,circuit design is rapidly developing towards miniaturization,integration,high speed,and low power consumption.The increasingly precise devices and corresponding circuit designs continue to amplify small problems on the transmission line,leading to inevitable signal deformation.In order to avoid excessive distortion and distortion caused by issues such as crosstalk,reflection,and jitter during the transmission of high-speed signals,which cannot be properly recognized by the receiving end,the importance of signal integrity analysis has become increasingly important.To measure the reliability of circuit design,worst-case analysis is usually used as a measurement method.On the one hand,different excitation methods may reflect different design defects when the circuit design and electrode material are determined.However,traditional signal integrity testing typically uses fixed frequency square waves and random codes as excitation inputs,which reflect limited signal integrity issues,Usually,more simulation time is required to obtain more complete inter symbol interference information.On the other hand,fast time-domain simulation technology is usually used to obtain the worst-case eye map.As the nonlinear characteristics of high-speed circuits become more apparent,the time cost of high-precision fast multilateral response method in collecting all edge responses increases exponentially.In response to the issue of existing excitation not being able to stimulate more comprehensive signal integrity,this paper proposes an excitation code type that balances eye height and eye width to achieve poorer quality,that is,adding worse inter code interference and inter line crosstalk to the periodic code type that triggers the worst synchronous switching noise.Firstly,study the common influencing factors in signal integrity;Then,based on the impact of different influencing factors on signal distortion,design an excitation code that takes into account the three dimensions of synchronous switching noise,line to line crosstalk,and code to code interference to achieve poorer quality;Next,build a simulation model of the highspeed link to verify the impact of the designed excitation code type on the eye pattern;Finally,compare the method with other excitations and adjust the resonant frequency and code rate of the model to verify the effectiveness and universality of the method;The results indicate that the excitation method proposed in this paper can stimulate more signal integrity issues in a shorter time.This paper proposes a machine learning based eye contour prediction method to address the problem of collecting too many and too long edge responses when using the fast multilateral response method to obtain eye contour.Firstly,an introduction is given to the evaluation of nonlinear influence length and the interception method of tail response in high-speed links;Then,based on the characteristics of the trailing response formed by different code combinations and the composition of the eye map,the influence of the leading bits is divided into linear and nonlinear regions for processing.In the nonlinear region,machine learning models are used to predict the trailing response containing inter code interference,and the predicted trailing response is formed into the contour of the eye map;Finally,the accumulated voltage in the linear region is extended to the eye contour;This method greatly reduces the number and length of responses that need to be addressed in the multilateral response method,enabling eye maps to be obtained in a shorter time frame.