ANN Performance for the Prediction of High-Speed Digital Interconnects over Multiple PCBs

Scharff, Katharina; Schierholz, Christian M.; Yang, Cheng Fu; Schuster, Christian

doi:10.1109/epeps48591.2020.9231490

Cited by 9 publications

(3 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Recently, artificial neural networks (ANN) have been used to predict the S-parameter of high-speed interconnects based on an open SI/PI-Database [1], [3]. However, the data-driven neural networks [3] can lead to negative insertion loss, which violates the passive characteristics that were demonstrated in this paper. Insertion loss is a positive number that represents signal loss from the input power to the output power [4].…”

Section: Introductionmentioning

confidence: 80%

Physics-Enforced Modeling for Insertion Loss of Transmission Lines by Deep Neural Networks

Chen¹,

Tan²

2021

Preprint

View full text Add to dashboard Cite

In this paper, we investigate data-driven parameterized modeling of insertion loss for transmission lines with respect to design parameters. We first show that direct application of neural networks can lead to non-physics models with negative insertion loss. To mitigate this problem, we propose two deep learning solutions. One solution is to add a regulation term, which represents the passive condition, to the final loss function to enforce the negative quantity of insertion loss. In the second method, a third-order polynomial expression is defined first, which ensures positiveness, to approximate the insertion loss, then DeepONet neural network structure, which was proposed recently for function and system modeling, was employed to model the coefficients of polynomials. The resulting neural network is applied to predict the coefficients of the polynomial expression. The experimental results on an open-sourced SI/PI database of a PCB design show that both methods can ensure the positiveness for the insertion loss. Furthermore, both methods can achieve similar prediction results, while the polynomial-based DeepONet method is faster than DeepONet based method in training time.

show abstract

Section: Introductionmentioning

confidence: 80%

Physics-Enforced Modeling for Insertion Loss of Transmission Lines by Deep Neural Networks

Chen¹,

Tan²

2021

Preprint

View full text Add to dashboard Cite

show abstract

“…As input features, the parasitics R, L and C are used, which are changed (by the varying length of interconnects), and a corresponding prediction of delay and power dissipation is made (assuming other parameters to be constant) 2 . The ANN techniques are also used for S-parameter and crosstalk modeling by generating surrogate models [82], [78]. In [82], Latin Hypercube Sampling (LHS) [87] is used for generating the training data, which provides a better distribution of data from every corner of data space, for the prediction of high-speed digital interconnects up to 100 GHz on Printed Circuit Boards (PCBs).…”

Section: Performance Evaluation Of Interconnectsmentioning

confidence: 99%

“…The ANN techniques are also used for S-parameter and crosstalk modeling by generating surrogate models [82], [78]. In [82], Latin Hypercube Sampling (LHS) [87] is used for generating the training data, which provides a better distribution of data from every corner of data space, for the prediction of high-speed digital interconnects up to 100 GHz on Printed Circuit Boards (PCBs). The model was improved by hyperparameter tuning, and the grid search technique was used.…”

Section: Performance Evaluation Of Interconnectsmentioning

confidence: 99%

Machine Learning Techniques for Modeling and Performance Analysis of Interconnects

Tripathi¹,

Vaghasiya²,

Junjariya³

et al. 2021

IEEE Open J. Nanotechnol.

View full text Add to dashboard Cite

Interconnects are essential components of any electronic system. Their design, modeling and optimization are becoming complex and computationally expensive with the evolution of semiconductor technology as the devices of nanometer dimensions are being used. In high-speed applications, system level simulations are needed to ensure the robustness of a system in terms of signal and power quality. The simulations are becoming very expensive because of the large dimensional systems and their full-wave models. Machine learning techniques can be used as computationally efficient alternatives in the design cycle of the interconnects. This paper presents a review of the applications of machine learning techniques for design, optimization and analysis of interconnects in high-speed electronic systems. A holistic discussion is presented, including the basics of interconnects, their impact on the system performance, popular machine learning techniques and their applications related to the interconnects. The performance evaluation, optimization and variability analysis of interconnects are discussed in detail. Future scope and overlook that are presented in the literature are also discussed.

show abstract

SI/PI-Database of PCB-Based Interconnects for Machine Learning Applications

et al. 2021

View full text Add to dashboard Cite

ANN Performance for the Prediction of High-Speed Digital Interconnects over Multiple PCBs

Cited by 9 publications

References 9 publications

Physics-Enforced Modeling for Insertion Loss of Transmission Lines by Deep Neural Networks

Physics-Enforced Modeling for Insertion Loss of Transmission Lines by Deep Neural Networks

Machine Learning Techniques for Modeling and Performance Analysis of Interconnects

SI/PI-Database of PCB-Based Interconnects for Machine Learning Applications

Contact Info

Product

Resources

About