AIStorm’s AI-in-sensor integrated solutions present designers with an ultra-low-power fully integrated chip for AI image workloads, enabled using AIStorm’s patented charge-domain-based processing. This analogue architecture offers faster speed, lower power consumption and noise, and smaller size than competing technologies while overcoming the bandwidth limitations of other attempts at analogue AI computing.
Thanks to direct coupling between on-chip sensors and processing, slow and power-hungry digitisation is eliminated, enabling performance that is equivalent to >1 GHz digital processing architectures while using 200x less power.
The AIS-C100A ‘Mantis’ is a 96 x 96-pixel grayscale image sensor with a programmable AI processor running at <80 µW for always-on operation (waiting). The chip also features GPIOs, an LED driver at up to 40 mA, power management, and more in a 6.4 x 6.4-mm OLGA package. MantisConnect combines this with a Cortex-M3 processor and Crypto Secure subsystem in addition to ultra-low-power Wi-Fi and Bluetooth connectivity (approx. 1 mW, always on).
Key AI use cases include occupancy detection, face detection with sex and age differentiation, intelligent wake-up, human and animal detection, intrusion detection, and access point wake-up.
If you do not require AI processing, AIS-C100C ‘Cheetah’ is a 20 x 44-px ultra-high-speed imager for up to 10 kFPS operation, and it is available in a 4.6 x 5.5-mm OLGA package. This chip features many of the mentioned peripherals included with Mantis and supports use cases that include gaming mouses, barcode reading, high-speed inspection and assembly, high-speed tracking and sports.
Learn more about Mantis and Cheetah in these product briefs, and if you like the look of this technology for your next commercial project, fill out the form below, and ipXchange will facilitate evaluation.