On-the-fly Data Pre-processors
Harnessing silicon photonics for optical data processing, enabling on-the-fly computations and utilizing the ultra-high bandwidth of photonic circuits is one of our goals. The aim is to shift the computational challenge of massively parallel matrix operations into a communications domain problem. This represents a substantial benefit in large detector systems by reducing the quantities of data while moving along the data flow towards storage. Edge computing/sensor applications with intensive matrix multiplication operations and stringent energy/latency constraints require efficient computation of dot products. The architecture leverages 3D heterogeneous integration with driver and compute/memory chips using vertical interconnects.
Objectives
Processing data while moving: Photonic systems allow data to be processed as it is being transmitted, leveraging the high bandwidth and low latency of optical communication. This capability is useful for real-time data processing applications, such as in particle physics experiments and high-speed imaging.
Energy efficiency and integration: Photonic systems offer excellent energy efficiency for certain operations, which is critical in applications with stringent power constraints. Integrating photonic and electronic components in a single platform further enhances overall system performance and scalability.
Deliverables
Design and fabrication of photonic perceptron components: Develop and test photonic components optimized for on-the-fly data processing and high bandwidth.
Integration of photonic and electronic systems: Implement a 3D-integrated control and processing architecture that combines photonic and electronic elements to enhance computational efficiency.
Development of photonic ADC architecture: Create and validate a photonic ADC system for high-speed, energy-efficient digital signal conversion and processing.