Adapter Board Development for Automotive Radar

Client

A company specializing in AI and automotive perception.

Challenge

The project goal is to receive data from the sensor board and transfer it to a PC. In parallel, the received data on FPGA must be organized in the defined order to be transferred to a PC over 10G Ethernet.

Our customer has developed an RF radar prototype using a sensor IC by Analog Devices. The radar sends a radio signal and captures its reflections back.

Our engineering task is to develop the adapter board which connects to Xilinx ZCU102 EVB and the customer’s radar board via FMC.

The adapter board transfers signals between the sensor board and ZCU102: 900 MBits/s per LVDS data lane (16 lanes), clocks, and management signals.

Solution

1. Hardware Design

Our hardware engineers have developed the adapter board which connects the customer’s sensors board and ZCU102 EVB. Also, the adapter provides power supply and voltage conversion for the sensor board.

documentation development;
PCB design;
library development;
PCB tracing;
simulations and tests.

2. Software Development

Our FPGA engineers have developed FPGA firmware and Cortex-R5 bare-metal application. The FPGA firmware performs the following processes:

receiving LVDS data from the sensor board,
parsing and checking the consistency of the received data,
forming the data to customer-defined packets and transferring the packets over 10G Ethernet using RTP and UDP protocols.

The Cortex-R5 application performs configuring and management of the FPGA logic and sensors.

The application has the following layers:

FPGA IP cores drivers,
sensor’s driver,
HAL,
business logic.

Also, the application provides a CLI via UART for a user to configure and manage various parameters of the FPGA IP cores and sensors.

Python scripts have been written for automated testing, managing, and configuring the entire system via UART using CLI and customer’s utility purposes.

Besides, Promwad engineers have made a path for configuring sensors, receiving FTP packets, and low-level control.

In addition, they have implemented:

input path;
serial and parallel bus reception;
data processing: parsing, merging;
debugging;
bring-up.

Business Value

The customer received a complex solution for computer vision, automotive perception, and other niches. The solution easily deploys on custom hardware platforms, meeting the exact requirements as a customer’s platform. An example of possible hardware for such a solution is SoM with Xilinx FPGA.

Project highlights: