An ultra-low power CMOS vision sensor, embedding spatio-temporal contrast estimation and binarisation, has been demonstrated through novel auto-exposure pixel architecture with improved contrast sensitivity compared with previous implementation. The processing involves three neighboring pixels and uses passive charge-transfer mechanism, with no dc power consumption. Contrast is computed with 3% sensitivity on images with more than 100dB dynamic-range. Asynchronous data readout takes less than 100μs, enabling high data-rate operation, up to 8000 fps. Efficient imager raster-scan skips all the empty rows, shrinking the sensor readout time down to 10μs, which dynamically adapts to the amount of data to be dispatched. Data compression and delivering is implemented through addressdriven technique. The square pixel pitch is 26μm with 25% fill factor. The sensors consumes 30μW at 30fps and 3.3 V, with a typical 70% output data compression.
A 30uW 100dB Contrast Vision Sensor with Sync-Async Readout and Data Compression
Massari, Nicola;De Nicola, Marco;Gottardi, Massimo
2010-01-01
Abstract
An ultra-low power CMOS vision sensor, embedding spatio-temporal contrast estimation and binarisation, has been demonstrated through novel auto-exposure pixel architecture with improved contrast sensitivity compared with previous implementation. The processing involves three neighboring pixels and uses passive charge-transfer mechanism, with no dc power consumption. Contrast is computed with 3% sensitivity on images with more than 100dB dynamic-range. Asynchronous data readout takes less than 100μs, enabling high data-rate operation, up to 8000 fps. Efficient imager raster-scan skips all the empty rows, shrinking the sensor readout time down to 10μs, which dynamically adapts to the amount of data to be dispatched. Data compression and delivering is implemented through addressdriven technique. The square pixel pitch is 26μm with 25% fill factor. The sensors consumes 30μW at 30fps and 3.3 V, with a typical 70% output data compression.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.