IN Brief:
- Rev 2.0 reduces the PolarFire Ethernet Sensor Bridge footprint by 60% and supports four cameras.
- MIPI CSI-2 sensor streams are carried to NVIDIA edge platforms over dual 10Gb Ethernet.
- FMC expansion and optical latency measurement extend the board beyond basic camera-interface conversion.
Microchip Technology has released Revision 2.0 of its PolarFire FPGA Ethernet Sensor Bridge, reducing the board footprint by 60% while doubling support to four MIPI CSI-2 cameras.
The production-ready board is designed around NVIDIA Holoscan Sensor Bridge technology and connects high-speed cameras and sensors to NVIDIA Jetson AGX Orin and IGX Orin or Thor platforms over dual 10Gb Ethernet. Microchip has also moved the revised hardware to USB-C power and says it is offered at a lower price than the first generation.
The PolarFire FPGA receives MIPI CSI-2 data, converts the incoming sensor stream into packets, and transmits it through a 10G MAC over Ethernet. Rev 2.0 is pre-programmed for four MIPI CSI-2 cameras, giving development teams a defined multi-camera architecture rather than requiring separate interface hardware for each sensor.
Microchip has also added onboard optical latency-measurement circuitry. Used with NVIDIA’s Latency Display Analysis Tool, the hardware is intended to measure the path from sensor capture through the AI pipeline, providing a system-level timing check rather than relying only on latency figures reported by individual software or processing stages.
Shakeel Peera, vice president of Microchip’s FPGA business unit, said: “Developers want to spend their time building high-value edge AI applications, not stitching together proprietary sensor interfaces. With low power PolarFire FPGA technology at its core, this second-generation Ethernet sensor bridge delivers a power-efficient, secure foundation in a significantly reduced form factor to help teams move faster from development to deployment in edge AI systems.”
The board includes an FPGA Mezzanine Card connector alongside MIPI CSI-2, I²C, UART, GPIO, and a standard PMOD connection. Microchip says the architecture can be extended to interfaces including SLVS-EC 2.0, 12G-SDI, HDMI, and DisplayPort without redesigning the core platform.
Those expansion options should be separated from the configuration delivered today. Microchip’s current product documentation states that Rev 2.0 comes pre-programmed for four MIPI CSI-2 cameras; other protocols rely on the FPGA and expansion architecture rather than being presented as identical out-of-box configurations. That distinction is useful for engineers evaluating how much development work sits between the reference board and a production system.
Using Ethernet as the sensor transport can reduce dependence on short point-to-point camera links around the edge processor. Cameras can be placed farther from the compute platform, while the FPGA handles protocol conversion and the network carries the sensor stream into the NVIDIA system. The approach also creates a common transport layer when several sensors have to be aggregated.
Bandwidth is only part of the design problem in robotics, industrial automation, and medical imaging. The useful inference result also depends on how predictably frames move from exposure through transport and processing. The onboard optical measurement circuit is therefore more than an accessory: it gives teams a way to test the timing behaviour of the complete pipeline under representative conditions.
The development package includes cameras, cables, schematics, a reference design, and design collateral, with RTL development supported through Microchip’s Libero SoC Design Suite. Microchip timing and power-management devices are also integrated into the platform, including the MCP16701 PMIC, which the company has validated for PolarFire FPGA use.
Integration with NVIDIA’s Holoscan SDK gives the board a defined software environment for sensor-heavy edge applications. Microchip identifies medical systems, industrial equipment, and humanoid robotics among the target uses, all of which place pressure on board area, power consumption, sensor count, and measurable latency.
Rev 2.0 is available directly from Microchip and through authorised distributors. The smaller footprint and doubled camera count are straightforward hardware improvements, but the broader engineering proposition is the consolidation of camera inputs, FPGA conversion, Ethernet transport, latency measurement, and development resources on one reference platform that can be evaluated before those functions are committed to a custom design.
