Omni Design Technologies, a provider of Wideband Signal Processing solutions, announced advancements in the development of its 200G-class Co-Packaged Optics (CPO) analog front-end (AFE) IP using an advanced 3nm process.
The development includes a high-swing linear transmit (TX) driver and a low-noise receive (RX) transimpedance amplifier (TIA). The CPO AFE IP targets up to 224Gb/s PAM4 optical I/O to address power and bandwidth challenges in moving data at 200G-class speeds for AI scale-up and scale-out connectivity in data centers.
Increasing bandwidth needs in AI infrastructure are leading to the placement of optical I/O closer to switch and compute packages to achieve higher density and lower energy per bit. Omni Design’s CPO AFE products, built on the 3nm process, provide the electrical-to-silicon photonics interface for high-speed optical links.
Kush Gulati, President and CEO of Omni Design Technologies, stated that as AI training and inference fabrics expand, the industry must address power and bandwidth at interfaces. He described Co-Packaged Optics as a key architectural shift, with the analog front-end contributing to system-level performance and efficiency. The development draws on the company’s analog and mixed-signal expertise to enable optical I/O closer to compute and switching silicon, improving bandwidth density and reducing power.
These advancements expand Omni Design’s portfolio beyond ultra-high-speed data conversion, supporting both optical and electrical interconnects for AI data centers.
Key features include:
224Gb/s PAM4 support for high throughput to meet bandwidth demands of AI scale-up and scale-out fabrics.
Programmable transfer functions for robust link performance.
3nm process for maximum integration density and low power consumption.
