Artificial-intelligence-based design optimization of low-noise transimpedance amplifiers for optical detection

Area of Science:

• Electrical Engineering
• Optical Instrumentation
• Artificial Intelligence

Background:

• Transimpedance amplifiers are crucial for electro-optical systems, particularly in optical ultrasound detection.
• Optimizing these amplifiers is key to improving system performance and reducing noise.

Purpose of the Study:

• To present an AI-based design optimization of transimpedance amplifiers for optical detection using a Genetic Algorithm (GA).
• To explore the influence of GA parameters on optimization performance.
• To compare AI-based optimization with Monte Carlo (MC) methods and experimental validation.

Main Methods:

• Utilized a Genetic Algorithm (GA) for the design optimization of photodetectors.
• Investigated the impact of GA parameters such as population size, generations, and mutation rate.
• Compared GA optimization results against Monte Carlo (MC) optimization and systematic search benchmarks.
• Experimentally validated the AI-based optimized photodetector.

Main Results:

• AI-based optimization with specific GA parameters (population size 1000, 10 generations, 5% mutation) required only 104 evaluations, achieving results within 0.12% of the maximum merit.
• MC optimization needed significantly more evaluations (3.4 × 10^5) for comparable statistical results.
• Demonstrated power-law scaling of performance with initial population size for both GA (exponent 1.2) and MC (exponent 0.88).
• Experimental validation confirmed the AI-based optimization's accuracy.

Conclusions:

• AI-based design optimization using GA is a highly efficient and promising method for developing low-noise transimpedance amplifiers.
• This approach offers superior performance and reduced computational cost compared to traditional methods like MC optimization.
• The validated AI method is applicable to various photodetector designs, including those for ultrasound detection, general-purpose use, and quantum processing research.