A Sensitive Dynamic and Active Pixel Vision Sensor for Color or Neural Imaging Applications.

Area of Science:

• Neuromorphic Engineering
• Computer Vision
• Sensor Technology

Background:

• Event cameras offer high dynamic range (DR), low latency, and reduced data rates, but often lack sensitivity for detecting subtle visual contrast.
• Existing event cameras struggle with detecting small visual contrast, limiting their application in sensitive areas like biological imaging.

Purpose of the Study:

• To present the SDAVIS192, a novel vision sensor with enhanced sensitivity for temporal contrast detection.
• To introduce a new methodology for characterizing event camera detection thresholds using signal-to-noise ratio (SNR).
• To explore the potential of the SDAVIS192 for applications such as calcium imaging.

Main Methods:

• Developed and fabricated the SDAVIS192 sensor using a 180-nm Towerjazz CIS process.
• Implemented an in-pixel preamplification stage to boost temporal contrast sensitivity.
• Established a characterization methodology to measure DVS event detection thresholds and fixed pattern noise with SNR considerations.

Main Results:

• The SDAVIS192 achieves high sensitivity with temporal contrast thresholds as low as 1% (OFF events) and 3.5% (ON events).
• Despite reduced effective intrascene DR, automated operating region control enables up to 110-dB DR for OFF events.
• Characterization at 30-dB SNR shows DVS temporal contrast threshold fixed pattern noise between 0.3%-0.8%.

Conclusions:

• The SDAVIS192 demonstrates significantly improved sensitivity for temporal contrast detection compared to previous sensors.
• The developed characterization methodology provides a robust way to assess DVS performance.
• The sensor's high sensitivity makes it a promising candidate for demanding applications like calcium imaging in neuroscience.