Arrow Electronics simplifies AI development

The plug-and-play board, with Arduino-compatible header, simplifies building IoT edge devices leveraging the programmable logic, RISC-V with Altera’s Nios V, DSP accelerators, and hard IP blocks.

Easy to use, and containing an Arrow USB Programmer2 on-board, the CYC5000 is ready to use for an AI camera application with a neural network running on the device. The board is optimised for low power consumption, and can be used for learning, proof of concept, or full production in customised versions.

The Altera Cyclone V FPGA, as the main feature of the CYC5000, has 25,000 logic elements, 25 DSP blocks, 50 18 x 18 multipliers, and FPGA memory blocks on-chip. The board also contains 64MBit SDRAM and 64MBit Flash, a USB interface, and runs from a single 5V supply.

In addition to the Arduino header, a CRUVI connector eases adding high-speed signal adapters for extra interfaces and applications. The AI camera demo uses an image sensor connected via the CRUVI adapter and runs a convolutional neural network, image filtering, MIPI in, and HDMI out, on the Cyclone V. Implemented in less of half of the FPGA’s available logic elements, the algorithm can analyse 10k samples per second, at only a few hundred milliwatts, highlighting opportunities for edge AI with a green footprint.