Optical systems for digit-serial computation

Area of Science:

• Computer Science
• Optical Engineering
• Digital Systems

Background:

• Traditional digital systems face limitations in speed and precision for complex computations.
• Optical computing offers potential for high-speed parallel processing.
• Digit-serial (on-line) arithmetic enables concurrent operations, but requires efficient implementation.

Purpose of the Study:

• To present novel optical systems and algorithms for digit-serial computations.
• To achieve high accuracy and processing speed without analog-to-digital converters.
• To enable parallel calculations and variable precision through new arithmetic approaches.

Main Methods:

• Introduction of modified signed-digit number representation for parallelism.
• Development of new arithmetic rules for addition, subtraction, multiplication, and division.
• Design of architectures utilizing optical bistable devices and optical interconnects.

Main Results:

• High-accuracy digit-serial computations achieved without analog-to-digital converters.
• Parallel processing enabled through carry-free addition and concurrent operations.
• Variable precision demonstrated with problem- and step-invariant algorithms.

Conclusions:

• The proposed optical systems and algorithms offer a significant advancement in digit-serial computation.
• Novel architectures facilitate efficient parallel processing and high-speed arithmetic operations.
• This approach provides a foundation for future high-performance optical computing systems.