Power Delivery Challenges for Optical Interposer in Advanced 3D Packaging

The rapid advancements in artificial intelligence (AI) and machine learning (ML) have resulted in an exponential increase in data processing demands. This surge requires efficient data transfer between compute units and memory units, intensifying the demand for higher bandwidth solutions. Silicon photonics, with its capacity for high-speed data transfer and reduced latency, is set to revolutionize in-package communication, particularly for short-reach applications extending over tens of millimeters.

In this context, the integration of 3D stacking technologies, including xPU (processing units), EIC (electronic integrated circuits), and PIC (photonic integrated circuits), poses significant challenges for the power delivery network (PDN). The primary concern is ensuring reliable and efficient power delivery throughout the densely packed 3D structures. This PhD research will focus on developing comprehensive models to assess the PDN performance for various integration schemes.

The research will entail a thorough analysis of the PDN, examining factors such as the effects of different integration schemes on power integrity, the role of decoupling capacitors in reducing power noise, and the efficacy of power management strategies. By investigating these aspects, the study aims to identify potential bottlenecks and propose innovative solutions to enhance the PDN's robustness and efficiency.

The results of this research will aid in the design of next-generation optical interposers, facilitating seamless integration of silicon photonics in advanced packaging. This will ultimately address the growing demands of AI and ML applications, paving the way for more efficient and powerful computing systems.