Optical Multiply-Add Circuit Using Delay-Coupled Signal Combining
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing optical computing systems, such as spatial and on-chip optical computing systems, face inefficiencies and accuracy issues due to thermal crosstalk and limited operations, particularly in implementing neural network operations like convolution, multiply-add, and pooling operations.
Innovation Solution
An optical computing apparatus comprising a linear operation module, delay modules, a coupler, and optional filtering and nonlinear operation modules, which modulate and combine optical signals to perform multiply-add, pooling, and convolution operations without using heating electrodes, enhancing efficiency and accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If on-chip optical computing systems use heating electrodes to adjust phases of optical signals, then phase adjustment capability is achieved, but thermal crosstalk occurs which reduces accuracy and efficiency
Solution Approach 1:
The patent replaces the thermal field-based phase adjustment (heating electrodes) with an electrical field-based phase adjustment mechanism. The optical modulator uses electrical signals to directly modulate the phase of optical signals, eliminating the need for thermal heating and thus avoiding thermal crosstalk. This substitution of physical field (from thermal to electrical) resolves the contradiction by achieving phase adjustment without the harmful thermal effects.
2Device complexity
If on-chip optical computing systems use limited MZIs to implement convolution operations, then integration is achieved, but multiple multiply-add iterations are required which reduces efficiency
Solution Approach 1:
The patent designs the optical computing apparatus to perform multiple neural network operations (multiply-add, convolution, pooling) using the same integrated optical modulator and coupler structure. The optical modulator can be configured to perform different operations by changing the electrical signal input, making the system multi-functional. This eliminates the need for separate dedicated hardware for each operation type, thereby improving efficiency without increasing physical complexity.
Solution Approach 2:
The patent enables continuous optical signal processing through the integrated optical path where optical signals continuously flow through the optical modulator and coupler without interruption. The system processes optical signals in a continuous manner rather than requiring discrete iterative steps, maintaining the continuous nature of optical computation and improving processing efficiency.
3Adaptability or versatility
If spatial optical computing systems use 4F optical system to implement convolution operations, then convolution capability is achieved, but both speed and volume are not improved
Solution Approach 1:
The patent transitions from the spatial domain 4F optical system to an integrated on-chip optical path, effectively changing the dimensional approach from bulk optical components to planar integrated circuits. This dimensionality change allows the system to maintain convolution capability while dramatically reducing the physical volume and improving operation speed through compact integration and direct optical path coupling.
Data Source
AI summary
An optical computing apparatus and system and a computing method are provided. The optical computing apparatus includes a linear operation module, a first delay module, and a coupler. The linear operation module can modulate, based on received electrical signals, optical signals input to the linear operation module; the first delay module may adjust a delay of optical signals output by the linear operation module; and after the first delay module adjusts the delay of the optical signals output by the linear operation module, the coupler may combine a plurality of groups of optical signals successively output by the linear operation module, to output one group of optical signals used to indicate a computing result that is obtained after a multiply-add operation is performed on one group of data and weights.