Lattice Blog


Simplify Multi-Camera Designs with CrossLink

Simplify Multi-Camera Designs with CrossLink
Posted 03/14/2017 by Tom Watzka

Posted in

CSI-2 cameras are a hot commodity. Driven by new regulations requiring backup cameras on new cars and continual improvements in camera-based Advanced Driver Assistance System (ADAS) functions, the global vehicle camera market is expected to grow at a CAGR of 11 percent between 2015 and 2020. Growth in the camera market is also driven by the rapidly escalating use of drones, which market research firm Radiant Insights expects to jump from $2B in 2016 to over $21B by 2022. At the same time, developers of Virtual Reality (VR) systems, 360° cameras and smartphones are adding more high performance cameras to their systems.

The development of new video bridging technologies has been the key to the proliferation of these new sensors for high volume designs. Lattice’s CrossLink device, introduced less than a year ago, offers developers the design flexibility and fast time-to-market benefits of an FPGA, along with optimized power and efficiency. CrossLink delivers the most flexible, highest bandwidth, lowest power and smallest footprint solutions for several high-growth market segments.

The implementation and popularity of the CrossLink FPGA devices can also be attributed to its IP support for a wide variety of common protocols and interfaces in mobile image sensors and displays. Designers can use these configurable soft core IPs as standardized blocks and spend more time on the unique aspects of their design.

New Use Cases

In February 2017, Lattice expanded its portfolio of CrossLink IP cores with additional modules that open to the door for development of new solutions. An excellent example is the new 4:1 CSI-2 to CSI-2 mux/merge IP core. Designers of augmented reality and VR systems, or those developing depth perception or gesture recognition applications, typically need to connect multiple image sensor interfaces to the application processor (AP). In some cases, the processing latency between image sensors and imaging data may be too long. In other situations, the AP may not feature enough CSI-2 interfaces to support the number of sensor inputs required for the application.

With this IP, designers can merge two of the four image sensors together in left/right format and use a single GPIO pin on the CrossLink device to switch between the two sets of merged image sensors. This data aggregation allows the designer to minimize processing latency between the image sensors and imaging data and, if needed, reduce the need for CSI-2 interfaces on the AP. Data aggregation can be performed by stitching the image sensor frames together in a side-by-side configuration or arbitrating data packets based on a virtual channel.

Four Input to One Output MIPI CSI-2 Camera Aggregator Bridge
Fig. 1: Lattice’s 4:1 CSI-2 to CSI-2 mux/merge IP allows designers to mux between four image sensors and then merge the two chosen video streams into a single MIPI CSI-2 video image output.

The new CrossLink bridge is completely programmable, supports any input encoding, a variable number of data lanes (1, 2, 4) and MIPI CSI-2 inputs and outputs at up to 6 Gbps. The output data rate is 2x the input data rate. The video interface bridging device is available as free IP in Lattice Diamond®.

Designers using this IP can implement stereovision solutions not only in the XY plane, but in the 3D plane as well. And by allowing designers to vary which channels to send to the processor to implement stereo depth, the new IP allows them to vary the distance and depth of the image.

Whether designers are building augmented reality systems, drones or automotive subsystems, demand is clearly growing for more complex video systems that can support 360° views, depth perception and gesture recognition. New video bridging technologies like Lattice’s CrossLink solutions offer the performance and design flexibility designers need to construct these multi-camera solutions while minimizing overall cost, power and footprint.