High-Frame-Rate Low-Noise Global Shutter CMOS Image Sensor for High-Speed Machine Vision

Area of Science:

• Electronics
• Computer Vision
• Sensor Technology

Background:

• Machine vision applications demand high-resolution, high-frame-rate image sensors.
• Existing sensors often face limitations in speed, noise, or power consumption for demanding tasks.

Purpose of the Study:

• To present a novel global shutter CMOS image sensor optimized for high-speed machine vision.
• To detail the sensor's performance characteristics, including frame rate, resolution, noise, and power consumption.

Main Methods:

• Developed a 5 µm voltage-domain global shutter pixel with dual-gain capability.
• Fabricated the sensor using a 65 nm, four-metal Back-Side Illumination (BSI) process.
• Implemented 16 CML data ports for high-speed video streaming.

Main Results:

• Achieved UHD resolution (3840 × 2160) with frame rates up to 1141 FPS (12-bit) and 1694 FPS (8-bit).
• Measured low read noise: 3.04 e⁻ (global shutter) and 2.15 e⁻ (rolling shutter).
• Total power consumption is 5.5 W with an aggregate throughput of 119 Gbps.

Conclusions:

• The ForzaFAST581 sensor meets the demanding requirements of high-speed machine vision.
• Its combination of high frame rate, low noise, and power efficiency makes it suitable for advanced applications.
• Selectable bit depths allow for application-specific frame rate optimization.