Key points are not available for this paper at this time.
Emerging machine vision applications require efficient detection of both dynamic events and static grayscale information within visual scenes. Current dynamic vision and active pixel sensors (DAVIS) technology integrates event-driven vision sensors and active pixel sensors within single pixels. However, the complex multi-component pixel architecture, typically requiring 15-50 transistors, limits integration density, increases power consumption, and complicates clock synchronization. Here, a charge-coupled phototransistor is presented that uses dual photosensitive capacitors to provide gate voltage to a single transistor channel, enabling simultaneous capture of dynamic and static information, surpassing existing DAVIS technology. Under illumination, both top and bottom gates generate photogenerated electrons through a charge-coupling effect; electrons in the top gate are blocked by a thick dielectric layer, producing a stable current change for static grayscale detection, while electrons in the bottom gate tunnel through a thin dielectric layer, creating transient current spikes for dynamic event detection. This device demonstrates a dynamic range of 120 dB and a response time of 15 µs, comparable to traditional DAVIS pixels, while significantly reducing power consumption to 10 pW and overcoming clock synchronization issues. This charge-coupled phototransistor paves the way for the development of high-performance, low-power, and highly integrated machine vision technology.
Feng et al. (Mon,) studied this question.