#South Korea #Seoul #Hanyang University #PAM-8 #108 Gb/s

Groundbreaking Advancement in High-Speed Communication

In a significant breakthrough, researchers from Hanyang University in South Korea have developed a pioneering multi-path architecture for a PAM-8 receiver frontend system, achieving unprecedented data rates of 108 Gb/s. This innovative technology is set to redefine the landscape of high-speed communication, particularly in vital sectors such as data centers, AI clusters, and networking equipment.

The research, spearheaded by integrated M.S. and Ph.D. student Sangwan Lee alongside Dr. Jaeduk Han, an Associate Professor in electronic engineering, focuses on the critical need for high-voltage and higher-order PAM-8 signals that enable data transmission beyond the 100 Gb/s threshold. The newly proposed system integrates a highly linear receiver frontend with a two-tap feed-forward equalizer (FFE) which is implemented using a 28nm CMOS process. The architecture impressively operates with a total power of just 210.8 mW while maintaining an input range of 1.4 Vppd.

Efficient Architecture Design

The team’s approach utilized a multi-path structure that enhances the linearity-power trade-off, allowing the system to exhibit an extraordinary data transmission rate. This efficiency is achieved by strategically dividing the signal path to handle specific sub-ranges of the total dynamic range, which in turn reduces the load on the ultimate processing stage. Notably, this architecture doubles the linearity without a proportional increase in power consumption — only a 20% rise in required energy.

Moreover, to tackle the challenges posed by channel loss in high-speed signal processing, the researchers developed a separated FFE path. Traditionally, FFE directly manages high-voltage signals, resulting in significant compression of the signal, which affects the overall quality and reliability of data transmission. By decoupling the FFE from the main signal path, the system can compensate for channel loss effectively, leveraging a smaller, attenuated signal to derive compensation values. This innovative separation allows for high-quality signal management even with large input signals.

Implications for Future Technologies

The implications of this work are profound. The technology is primed for immediate adoption within vital infrastructures, particularly for next-generation high-speed data communication systems. It bears significant relevance for data centers and AI clusters, where speeding up communication between servers can drastically enhance the training of large-scale AI models and expedite the processing of enormous datasets. Additionally, it lays the foundation for advanced networking equipment vital for future iterations of Ethernet systems, such as 800G and 1.6T.

Looking to the future, this research is likely to enhance computational speeds in supercomputers utilized for advanced scientific research and simulations. In the next five to ten years, it is expected to be integral in developing robust applications across various fields, such as Artificial Intelligence and augmented/virtual reality, promising advancements in real-time translation, autonomous driving applications, and sophisticated medical diagnostics.

The versatility of this technology also paves the way for sustainable growth in data centers, helping to meet the escalating global demands for data while maintaining an energy-efficient approach.

Dr. Jaeduk Han concludes that the robust infrastructure provided by their innovative design will not only facilitate upcoming breakthroughs in AI and virtual reality but will also aid in achieving a more environmentally sustainable operational model for future technology investments.

For those interested in delving into the specifics, the research details are presented in the paper titled "A 108Gb/s PAM-8 CTLE+FFE Receiver Frontend with 1.4Vppd Input Range in 28nm," showcased at the IEEE Asian Solid-State Circuits Conference (A-SSCC) on November 4. This event is recognized as one of Asia's leading semiconductor conferences, further demonstrating the global significance of the team's findings.