Sony Stacked Field Of View Sensor for Event Detection
Japan, Tokyo-Sony Corporation (hereafter: Sony) and Prophesee have announced that they have jointly developed a stacked field of view sensor for event detection, which has a smaller 4.86μm pixel size and a higher 124dB (Or higher) HDR performance.
The new sensor can asynchronously detect the brightness change of each pixel, and output only the coordinate and time data of the detected pixel, thereby achieving efficient, high-speed, low-latency data output. Combined with Sony’s unique stacked CMOS image sensor technology, the sensor has a small size and low power consumption, but has extremely high resolution and still performs well in low-light environments.
Combined with Prophesee’s Metavision event detection visual sensing technology, The sensor can achieve fast pixel response, high time resolution and high throughput data readout. It is suitable for a variety of machine vision applications, such as detecting fast-moving objects in a variety of environments and conditions.
At the International Solid-State Circuits Conference (ISSCC) in San Francisco, USA on February 16, 2020, Sony announced the new sensor and its performance results. Its specific characteristics and specifications are as follows:
Realized the industry’s smaller 4.86μm pixel size event-based stacked field-of-view sensor with small size and high resolution
The pixel chip (top) and the logic chip (bottom) are combined with a signal processing circuit to detect changes in brightness based on asynchronous incremental modulation methods. Each pixel of two separate chips is electrically connected in a stacked configuration using a Cu-Cu connection (*1). In addition to the industry’s smaller 4.86μm pixel size, the sensor uses high-resolution 40nm logic technology to achieve high-density integration, with a 1 / 2-inch, 1280×720 HD resolution.
The industry’s highest 124dB HDR performance (or higher) through the high light entrance rate (*2)
By placing only the backlight pixels and a part of the N-type MOS transistor on the pixel chip (top), the light entrance rate is increased to 77%, thereby achieving the industry’s highest 124dB HDR performance (or higher). Sony’s high sensitivity / low noise technology accumulated over the years during the development of CMOS image sensors enables event detection to be performed in low light conditions (40mlx).
High time resolution, high output event data readout
Frame-based sensors output the entire image at fixed intervals based on the frame rate, while event-based sensors use row selection arbiter circuits *3 to asynchronously select pixel data. By adding time information with a precision of 1 μs to the pixel address where the brightness changes, event data with high time resolution can be ensured. In addition, by effectively compressing the event data, that is, the brightness change polarity, time and x / y coordinate information of each event, a high output event rate of 1.066 Geps (4) has been achieved.
(1) A technology that provides electrical continuity through a connected Cu (copper) pad when stacking a back-illuminated CMOS image sensor part (top chip) and logic circuit (bottom chip). Compared with TSV wiring, the connection is achieved by penetrating electrodes around the pixel area. Compared with this method, this method has greater freedom in design, improves productivity, reduces size and improves performance. Sony announced the technology at the International Electronic Equipment Conference (IEDM) in San Francisco in December 2016. (2) The ratio of the aperture (excluding the light-shielded portion) of the light incident surface of each pixel.
(3) A circuit that determines priorities in the Y-axis direction in response to a request for multiple pixels whose brightness changes. (4) Number of events per second.