Artificial intelligence (AI) is fundamentally transforming the capabilities of machine vision and imaging applications. Edge AI, a branch of AI, enables localized decision-making at the edge, playing a crucial role in applications such as 5G networks and the Internet of Things (IoT) in new infrastructure. By deploying and implementing edge AI in devices, TI helps customers in China and globally accelerate the adoption of this key technology across various industries, driving the practical application of AI and contributing to the development of China's new infrastructure. Edge AI enhances the accuracy and sensitivity of real-time analysis of high-resolution imaging data, enabling rapid detection of patterns or anomalies. These capabilities are supporting robot-assisted surgery, 3D imaging-based safety and monitoring platforms, and vision-based industrial robots in factory automation environments.
With advancements in sensor technology and the realization of ultra-high-definition or 4K video image sensors using miniature modular packaging, this shift towards edge AI processing has become possible.
Applications such as robot-assisted endoscopy platforms and machine vision cameras require the transmission of high-resolution image data from sensors mounted on extremely small probe tips connected to video acquisition and analysis systems via ultra-fine cables. Control information from the video acquisition system simultaneously flows back to the probe tip, providing a method for controlling the probe tip's position through tilt and zoom controls.
Figure 1 illustrates vision-based control systems for industrial mobile robots, which require extremely low latency to achieve real-time acquisition and analysis of high-precision video data. The process of transmitting control signals via a reverse path to adjust camera position also requires low latency. The number of cables and conductors, conductor size, and power dissipation at the sensor end can be significant limiting factors for space-constrained applications such as endoscopes and machine vision.
Many existing high-speed interface technologies can help achieve reliable transmission of high-resolution video data, but vision-based control systems also have drawbacks. For example, standard technologies such as Ethernet introduce additional latency due to protocol-related overhead. Since Ethernet physical layer devices cannot directly connect to the sensor's local video interface, these devices require additional wiring and components (such as crystal oscillators). For instance, a 4MP, 30fps high-resolution imager can generate approximately 3.2Gbps of video data. A single gigabit-based Ethernet link (such as 1000-BaseT) cannot provide sufficient throughput to carry this uncompressed high-resolution data, which introduces artifacts into the image stream and can potentially lead to errors in machine vision-based video processing.
Dedicated serializer/deserializer (SerDes) technologies (such as the V3Link TSER953 serializer and the TDES954 and TDES960 deserializers) work together to transmit high-resolution video, control signals, and power simultaneously over a single ultra-fine wire. These devices facilitate the establishment of links between sensors and processors to aggregate clock, uncompressed video, control, power, and general-purpose input/output signals, as shown in Figure 2.
In this configuration, signals and power are transmitted through a forward channel that runs from a serializer in the sensor module to a deserializer or deserializer hub in the sensor fusion analysis system. This channel also transmits control signals and power.
The V3Link deserializer also provides an embedded clock for all connected serializers, enabling video synchronization across multiple sensors and supporting video stitching, image blending, 3D reconstruction for stereo vision, and depth sensing. Using the frame synchronization signal generated internally by the TDES960, you can synchronize multiple cameras with an accuracy of 600ns, enabling multiple time-triggered machine vision applications. Master clock synchronization is achieved by extracting the reference clock from the reverse channel, eliminating synchronization errors caused by the relative drift of different oscillators acting as clocks for multiple imagers.
Adaptive equalizer technology for reducing signal loss and power consumption
In addition to supporting single-wire transmission of video data, control signals, and power, V3Link devices also include adaptive equalizer technology that compensates for up to 21dB of loss at 2.1GHz, enabling the use of extremely thin 28 to 32 American Wire Gauge (AWG) cables. The higher the AWG number, the thinner the cable and the higher the signal loss.
Thinner cables offer greater flexibility, enabling applications where sensors can be positioned in confined, space-constrained environments, such as those using endoscopes. The ability to transmit both power and control signals on a single thin cable also minimizes the number of conductors required.
With a typical power consumption of 250mW at the sensor end, the V3Link serializer consumes very little power, enabling the integration of the sensor and serializer into a very compact area. It consumes no power and generates no heat, thus requiring no additional space. Thanks to V3Link's proprietary master clock synchronization technology, no crystal oscillator or any other oscillator is needed at the sensor end, further reducing cost and overall space requirements.
Conclusion
From medical imaging applications to machine vision cameras, edge AI is driving the demand for real-time video acquisition, transmission, and analysis. V3Link SerDes ICs help engineers meet these needs while reducing cabling, power consumption, and overall system cost. V3Link devices offer versatile link technologies, making them ideal for most applications requiring real-time acquisition, transmission, and analysis of high-resolution video data. These devices support various cabling configurations (coaxial cable, unshielded twisted pair, and shielded twisted pair) and multiple clock modes (such as synchronous and asynchronous).
