IN Brief:
- ST has introduced monochrome and RGB global-shutter sensors for always-on embedded vision.
- The devices combine compact die size, low-power operation, event wake-up, and MCU-friendly interfaces.
- Small edge devices are adding more local visual awareness without relying on continuous high-power image streaming.
STMicroelectronics has introduced two ultralow-power global-shutter image sensors designed to bring always-on vision into compact, battery-powered, and energy-harvesting devices.
The VD55G4 monochrome sensor and VD65G4 RGB colour sensor extend ST’s BrightSense portfolio and are now available to early adopters. The devices target wearables, AR, VR, XR headsets, smart home appliances, IoT nodes, medical devices, laptops, and compact embedded vision systems.
The sensors are built around a 0.56MP, 804 x 704 pixel array and use a 2.16µm global-shutter pixel. Each die has a 2.73mm x 2.16mm footprint, supporting use in size-constrained designs where lens stack, board area, battery capacity, and thermal headroom are limited.
Output options include RAW8 and RAW10, with MIPI CSI-2, SPI, I3C, I²C, and I3C control interfaces available. The interface mix allows the sensors to be connected to both microcontroller-based systems and more capable application processors, depending on frame-rate, processing, and power requirements.
The architecture is built around low-power visual awareness rather than continuous high-power image streaming. An embedded auto wake-up function allows the sensors to remain in an ultralow-power sleepy mode and trigger the host processor when an event is detected. Operating power scales with frame rate and enabled features, with the product page listing less than 2mW for sleepy operation and up to 35mW at 60fps.
The devices include on-chip image processing functions such as autoexposure, noise reduction, defect correction, multi-exposure support, frame statistics output, background subtraction, crop functions, and GPIOs. Those features reduce the amount of low-level image handling required from the host MCU or processor, helping conserve system power and simplify software development.
Always-on vision is increasingly being designed into products that cannot support the energy budget of a conventional camera subsystem. Smart glasses, medical patches, low-power monitoring nodes, industrial identification systems, and compact autonomous devices all need context from the surrounding scene, but continuous image streaming can overwhelm battery capacity or energy-harvesting limits.
Event-driven image sensing changes the system partition. The sensor can monitor for scene changes, generate frame statistics, manage exposure, and wake a host processor only when required. That allows more visual intelligence to be placed in small edge devices without forcing every product towards a high-power application processor architecture.
Product development still depends on careful optical and system design. Lens selection, illumination, sensor placement, interrupt handling, wake-up latency, and false-trigger control will determine how the parts behave outside controlled evaluation conditions. Privacy and local processing choices will also influence system architecture in medical, smart-home, and wearable products.
ST is supporting the sensors with development boards for platforms including STM32 and Raspberry Pi, along with camera modules, evaluation software, drivers, and a software development kit. As embedded vision moves into smaller edge devices, evaluation hardware and software support are becoming central to sensor adoption.
