Labs / Board Reviews & Technical Insights

Avnet ZUBoard 1CG Development Board: A New Choice for Deep Learning

Author: Zhuang Xiaoman · Source: Freak Embedded on WeChat, published 2024-08-13 · License: CC BY 4.0

Contents

Abstract

This article introduces the features, architecture, and hardware units of the Avnet ZUBoard 1CG development board, gives a basic introduction to connecting to the board through the Ethernet interface and the serial port, and demonstrates its functions with two application examples.

Main Text

The Avnet ZUBoard 1CG is an FPGA development board based on the RISC-V architecture:

Avnet ZUBoard 1CG Development Board: A New Choice for Deep Learning
Avnet ZUBoard 1CG Development Board: A New Choice for Deep Learning

Its main features include:

  1. Based on AMD Zynq UltraScale+ architecture:

    1. The ZUBoard 1CG gives engineers flexibility and versatility to experiment with and learn the AMD Zynq UltraScale+ architecture.

  2. Powerful processing performance:

    1. The board carries 81K programmable logic cells, a dual-core Arm Cortex-A53, and a dual-core Arm Cortex-R5F processor.

    2. It is equipped with caches and on-chip memory, providing strong computing performance for applications.

  3. Rich storage and interfaces:

    1. 1GB LPDDR4 memory and a 256Mb QSPI Flash or microSD card slot provide flexible storage options.

    2. 10/100/1000 Ethernet, a USB 2.0 host interface, and a JTAG/UART interface give developers a variety of connection options.

  4. Complete power and clock management:

    1. The board can be powered through the USB-C port and is equipped with a Microchip controller and a TDK power module.

    2. A Microchip oscillator and an ECS crystal provide stable clock sources for the core devices.

  5. Rich peripherals and user interaction:

    1. Slide switches, buttons, LEDs, and so on allow users to interact with the board.

    2. The onboard temperature and pressure sensors provide environmental data.

  6. High expandability:

    1. Through the Samtec connectors and the Click Board interface, the ZU+ PS GTR transceivers, PS MIO, and PL I/O pins can be accessed, supporting high-speed expansion.

  7. Development environment support:

    1. Board definition files supported by Vivado and a PetaLinux BSP allow developers to get started quickly.

    2. It is suitable for various application scenarios such as bare-metal, Linux, and Vitis AI accelerators.

Overall, the ZUBoard is a development board based on the open RISC-V architecture, with flexibility, expandability, and rich peripheral support, widely used in embedded and IoT fields. It provides developers with a powerful hardware platform and rich development support.

1. Application Areas

In the market, the ZUBoard 1CG is positioned for high-performance embedded applications, and is especially suitable for application scenarios that require strong processing power and flexibility, such as machine learning, image processing, edge computing, and IoT.

Avnet ZUBoard 1CG Development Board: A New Choice for Deep Learning

2. Development Board Hardware Units

2.1 Processor

The large chip in the center is the "heart" of the board, the AMD Zynq UltraScale+ MPSoC, which integrates a dual-core A53 APU and a dual-core R5 RPU. It handles most of the computing tasks.

Avnet ZUBoard 1CG Development Board: A New Choice for Deep Learning
Avnet ZUBoard 1CG Development Board: A New Choice for Deep Learning

2.2 Memory and Storage

The chips next to it include the DDR4 memory and the QSPI flash, which are used to store data and the operating system.

Avnet ZUBoard 1CG Development Board: A New Choice for Deep Learning

2.3 Peripheral Interfaces

Avnet ZUBoard 1CG Development Board: A New Choice for Deep Learning

2.4 User Buttons and Switches

Avnet ZUBoard 1CG Development Board: A New Choice for Deep Learning

3. Demo Section

Next we will show you how to get started with the ZUBoard in 2 minutes. Two demos are prepared here: the first is gesture digit recognition for image processing, and the second is a gesture-controlled software robot(Both examples come from Hackster)Before running the demos, we first set up the environment and connect to the development board over Ethernet.

3.1 Environment Setup

First, connect the SD card to the computer with a card reader:

Avnet ZUBoard 1CG Development Board: A New Choice for Deep Learning

Format the SD card:

Avnet ZUBoard 1CG Development Board: A New Choice for Deep Learning

Write the image file (the image file can be downloaded from the Baidu Netdisk link in the references section):

Avnet ZUBoard 1CG Development Board: A New Choice for Deep Learning

3.2 Connecting to the Development Board over Ethernet

Connect the cables to the computer; the main power input is 15V:

Avnet ZUBoard 1CG Development Board: A New Choice for Deep Learning

After power-on, the onboard LED lights up:

Avnet ZUBoard 1CG Development Board: A New Choice for Deep Learning

Press the SW7 power button; the white light turns on and the board starts booting:

Avnet ZUBoard 1CG Development Board: A New Choice for Deep Learning

The blue light turns on, indicating that the board has booted successfully:

Avnet ZUBoard 1CG Development Board: A New Choice for Deep Learning

Download the MobaXterm software (the download URL is given at the end of the article):

Avnet ZUBoard 1CG Development Board: A New Choice for Deep Learning

Select the USB-connected serial port:

Avnet ZUBoard 1CG Development Board: A New Choice for Deep Learning

Set the baud rate to 115200:

Avnet ZUBoard 1CG Development Board: A New Choice for Deep Learning

Turn off flow control:

Avnet ZUBoard 1CG Development Board: A New Choice for Deep Learning

Useifconfigto view the network IP address:

Avnet ZUBoard 1CG Development Board: A New Choice for Deep Learning

Configure the SSH service to enter the internal system automatically. After logging in, we will be inside the system and can begin programming and configuration work. Copy the network IP address into the remote host box:

Avnet ZUBoard 1CG Development Board: A New Choice for Deep Learning

Select Sprcify username and enter root:

Avnet ZUBoard 1CG Development Board: A New Choice for Deep Learning
Avnet ZUBoard 1CG Development Board: A New Choice for Deep Learning

The following page will then pop up automatically; we have now successfully connected to the development board over Ethernet:

Avnet ZUBoard 1CG Development Board: A New Choice for Deep Learning

3.3 ASL Gesture Digit Recognition Demo

  1. Switch to the project directory:Use thecd asl_classification_vitis_aicommand to switch to the project directory

  2. Run the real-time gesture digit recognition script:Execute thepython3 asl_classify_live.py --model=./model_mobilenetv2/B{#}/asl_classifier.xmodelcommand to run the real-time gesture digit recognition script

Next, we can recognize gesture digits!

Avnet ZUBoard 1CG Development Board: A New Choice for Deep Learning

3.4 Gesture-Controlled Software Robot Demo

Follow these steps:

  1. Confirm the application is active: Execute the command xmutil listapps, and confirmavnet-{platform}-benchmarkwhether the application is active. The{platform}here should be replaced according to your hardware platform (ZUBoard or Ultra96-V2)

  2. QueryDPUarchitecture: Use the command xdputil query | grep DPU to query the DPU architecture loaded in the programmable logic (PL). Note down the architecture information; it should be B2304 for the Ultra96-V2 platform and B512 for the ZUBoard platform

  3. Create the workspace directory:mkdir -p webinar_ws/src

  4. Switch to the workspace directory:cd webinar_ws

  5. Set upROSenvironment: Execute the command source /usr/bin/ros_setup.sh to set the ROS environment variables

  6. CompileROSworkspace: Use the command colcon build to compile the packages in the ROS2 workspace.

  7. Set up the local environment: Execute the command source ./install/local_setup.sh to set up the environment for the compiled packages.

  8. List executables: Use the command ros2 pkg executables | grep py_vision to listpy_visionthe executables in the package.

  9. Launch the ROS2 demo: Execute the command ros2 launch py_vision webinar_demo_launch.py to launchpy_visionthe demo in the package. This demo may be the main program for controlling a robot with sign language recognition.

As you can see, the related demo has started:

Avnet ZUBoard 1CG Development Board: A New Choice for Deep Learning
Avnet ZUBoard 1CG Development Board: A New Choice for Deep Learning
Avnet ZUBoard 1CG Development Board: A New Choice for Deep Learning

4. User Experience

The ZUBoard 1CG is designed to be an ideal platform for exploring bare-metal programs, Linux, or the Vitis AI accelerator. Its powerful processing capability and rich interface resources allow developers to easily implement complex embedded applications. The temperature and pressure sensors integrated on the board add environmental data acquisition to the system, which broadens its range of applications. In addition, through its rich expansion interfaces, the ZUBoard 1CG can be conveniently connected to various peripherals, which further improves its flexibility and expandability.

References

Avnet ZUBoard 1CG Development Board: A New Choice for Deep Learning