The Signal-Integrity Control Strategy of a TSV Array for a Chiplet-Based System

Area of Science:

• Electrical Engineering
• Computer Engineering
• Materials Science

Background:

• Chiplet-based systems require advanced interconnects like through-silicon via (TSV) arrays.
• Signal integrity is crucial for the performance of high-speed interconnects.
• Optimizing TSV design parameters is complex due to irregular relationships with signal performance.

Purpose of the Study:

• To develop a signal integrity control strategy for TSV arrays in chiplet systems.
• To establish an irregular relationship model between TSV design parameters and signal indexes.
• To optimize TSV design parameters for desired signal performance.

Main Methods:

• Utilized High Frequency Structure Simulation (HFSS) software for TSV array simulations.
• Employed a backpropagation neural network (BP-NN) model to establish parameter-index relationships.
• Applied a particle swarm optimization algorithm with linear decreasing inertia weight (PSO-LDIW) for parameter optimization.

Main Results:

• The BP-NN model accurately predicted signal indexes with relative errors between 0.31% and 5.02% compared to HFSS simulations.
• Optimized TSV design parameters achieved signal index deviations no greater than 5.72% for NEXT, FEXT, and return loss.
• The maximum deviation for insertion loss was 0.0160 dB, demonstrating effective signal integrity control.

Conclusions:

• The developed signal integrity control strategy is feasible for TSV arrays in chiplet systems.
• The combination of BP-NN and PSO-LDIW offers an effective approach for optimizing TSV designs.
• This strategy can significantly improve signal performance in advanced packaging technologies.