Edge AI Nexus-Based FPGA: Promwad Kicks Off Designs Based on Latest Lattice CrosslinkU-NXs
By Andrei Smirnov
Head of the Adaptive Computer Systems Unit at Promwad
In September, our partner Lattice announced the release of its new CrossLinkU-NX FPGA family with integrated USB interface, which was marketed as Edge AI Nexus-based FPGA. Our clients will be among the first to have access to this advanced FPGA platform for video processing, interface bridges and multiplexers.
The new CrossLinkU-NX FPGA family is based on the Lattice Nexus platform and delivers best-in-class image processing performance while reducing power consumption by up to 75% compared to similar FPGAs.
Within its compact form factor, CrossLinkU-NX features an integrated USB interface with accelerated data transfer rates of up to 5 Gb/s, facilitating the development of applications for a wide range of industries. Also, CrossLinkU-NX has an FPGA fabric optimised for AI inferencing. With this feature, the new FPGAs can be used in network video recorders (NVR), video cassette recorders (VCR), transcoding systems, video editing systems, industrial video systems, machine vision systems, and USB 3.0 cameras.
Lattice proposes to utilise the features of the new platform for a wide variety of uses:
Here are more important features of the CrossLink-NX-33U for embedded vision applications features:
- 33K logic cells, 3.7 Mb of embedded memory and 64 DSP blocks for AI and image signal processing;
- higher processing performance compared to CrossLink-NX-33;
- support for image sensor processing form factor: available in 84 pin WLCSP and 104 pin FCCSP;
- optimised for integration with image sensors;
- sleep mode operation and fast booting recovery (less that 15 ms);
- high-speed interface and low total operating power USB 3.0 Interface with PHY;
- backward-compatible with USB 2.0.
Partnering with Lattice allows our engineering team to be among the first to have access to documentation for the latest technologies and create advanced and cost-effective solutions with the new Lattice CrosslinkU-NX FPGA family:
- industrial сommunication protocols IP cores / EtherCAT, CAN, others;
- digital camera interfaces, MIPI-to-USB, MIPI-to-Ethernet;
- optoelectronic conversion boards;
- video encoding accelerators;
- video interface converters, HDMI-to-LVDS, SDI-to-HDMI, etc.
Lattice-based designs represent a significant part of our portfolio of FPGA-based designs, and below, we summarise some of the projects.
Case Study #1. Crosslink-NX FPGA-Based Firmware for Serial and GPI Fiber Optic Transceivers
For Lynx Technik AG, a German company developing products for professional A/V and broadcasting, we developed firmware to migrate their FPGA-based hardware from Intel to Lattice. The migration was essential due to supply constraints, which jeopardised our customer's contract to deliver tens of thousands of devices.
In this project, we developed firmware that converted high-speed optical data to a low-speed parallel set of such interfaces as RS232/RS422 and GPIO in both directions. Also, we implemented the bridge function using the Crosslink-NX FPGA.
This is what the firmware architecture looked like:
Crosslink-NX FPGA-based solution software architecture
Thanks to our engineering support, the client successfully migrated their solutions to the new platform and secured their supply contract. Our partnership with Lattice allowed us to utilise the required products and start production on time.
Case Study #2. Automotive Video Surveillance System Migration to Lattice Crosslink-NX Platform
Now, we are upgrading the video surveillance solution for a European supplier working in the automotive industry. Our client's solution is based on components that will soon be discontinued, and our task is to migrate their hardware to Crosslink-NX FPGA, develop ASPICE CL2-level software and add new features to the solution.
Lattice Nexus FPGA Applications
CrossLinkU-NX Evaluation Board. Source: latticesemi.com
The Nexus FPGA family by Lattice is used for several of the most common development scenarios:
- Data acquisition and preprocessing from multi-sensor systems such as motion sensors, temperature sensors, fuel metres, cameras, and displays. AI-powered hardware accelerators for image stitching, colour conversion, and neural network processing have been developed on Lattice chips. These low-power and high-performance chips provide cascaded data processing, which reduces main processor utilisation in real-time safety-critical applications.
- Design of bridges to communicate between different interfaces or buses in an electronics system. Bridges act as a layer that carries data from one interface to another and enables device communication. For example, Mazda used a low-power Lattice FPGA to develop an interface bridging solution for safety-critical Advanced Driver Assistance System (ADAS) applications in their crossover vehicles.
- Multiplexers design to receive video streams from multiple cameras and transmit them to an embedded processor in bird-eye or surround-view applications.
- Functional safety certified software development: Lattice FPGAs are designed for fault-tolerant operation and protected by AES256 bitstream encryption and ECC256 bitstream authentication.
- Infotainment display bridging and processing: Lattice FPGAs provide low latency for image processing and help deploy safe and reliable displays and sensors with human-machine interface (HMI).
The CrossLinkU-NX FPGA boosts connectivity, simplifies design projects, improves data transfer, offers a compact form factor and low power consumption. Its advantages make the platform suitable for the development of multiplexers, bridges, and functional safety solutions.
Promwad and Lattice partnership gives us access to all tech documentation and priority support to develop the best solutions for our customers. Contact us to discuss your tasks and requirements, and we will advise you on how they can be addressed with the latest Lattice technology!