Proc. SPIE 11051, 32nd International Congress on High-Speed Imaging and Photonics, 110510B (28 January 2019); doi: 10.1117/12.2524492

Event: International Conference on High-Speed Imaging and Photonics 2018, Enschede, The Netherlands


Over 100 Million Frames per Second High Speed Global Shutter CMOS Image Sensor

by R. Kuroda*, M. Suzuki, S. Sugawa
1 Graduate School of Engineering, Tohoku University, Sendai, Japan
* Corresponding Author []


This paper presents advancement of ultra-high speed (UHS) global shutter CMOS image sensor technology exceeding 100M frames per second (fps). The development of key technologies toward the next generation UHS global shutter CMOS image sensor are overviewed, that includes high density analog memory integration, pixel-wise memory array architecture, and burst correlated double sampling (CDS) operation. By introducing the newly developed signal readout scheme with minimized pixel pulse transitions, a frame period of photo-electrons transit time is achieved. The fabricated chip prototyping a 3D stacked structure achieved 100Mfps with 80 record length and 125Mfps with 40 record length under room temperature without any cooling systems.

Time resolution of image sensors is one of the fundamental performances of image sensors. Figure 1 summarizes relationship of performance categories with regards to speed and time resolution of image sensors. For high speed image sensors utilized in high speed video cameras, the time resolution is determined by the frame period(1-6). On the contrary for image sensors utilized in high speed shutter cameras or lock-in detection cameras such as for time of flight range imaging, fluorescence lifetime imaging and etc(7-8), the time resolution is limited by shutter period or pixel modulation period, which can be shorter than the frame period. Both types of sensors have been extensively researched and their time resolution is continuously improving by continuous technological breakthroughs.

Keywords: Burst CMOS image sensor; 3D stacking; analog memory; burst correlated double sampling