Qualcomm RB3gen2 Reference Images

Overview 

The Qualcomm RB3 Gen2 development kit (also known as the Dragonwing RB3 Gen 2) is a compact, high-performance IoT and edge-AI development kit built on the Qualcomm QCS6490 SoC.

Compute & AI Power Octa-core Kryo 670 CPU, Adreno 643 GPU, 12 TOPS AI capability via integrated Hexagon DSP/NPU.
Multi-camera support Qualcomm Spectra 570L ISP; dual CSI cameras included in the Vision-kit; supports additional stereo, ToF, depth vision sensors.
Connectivity Wi-Fi 6E (802.11ax, 160 MHz, 3.6 Gbps), Bluetooth 5.2, plus USB-C, USB-3, PCIe Gen-3.
Modular & expandable 96Boards-compliant mezzanine connectors (MIPI CSI/DSI, PCIe, GPIO, I²C, SPI, UART).
Storage & memory 6GB LPDDR4x RAM, 128 GB UFS onboard, micro-SD and NVMe via PCIe.
Use-cases Tailored for power-efficient on-device AI in robotics, vision systems, drones, security cameras, inventory management, SLAM, path planning and surveillance.

Variants 

There are two variants of the RB2gen2: the core kit and the vision kit. The vision kit includes two cameras and mounting bracket.

There is also a lite version of the RB3gen2, but this kit is a different product which has a different SoC. The SoC on the RB3gen2 lite kit is currently not supported by the reference images.

Vendor 

Vendor: Qualcomm
Reference Hardware: RB3gen2
SoC: QCS6490
Architecture: ARM64
Purchase: The development kit can be purchased from Thundercomm.

Supported Features in Linaro Reference Images 

Feature	Supported?	Notes
Boot to Console	Yes
Ethernet	Yes	Non-redistributable firmware required
USB-A	Yes	Non-redistributable firmware required
HDMI	Yes
Wi-Fi	Untested
Bluetooth	Untested
UART	Yes
SPI	Yes
I2C	Yes
GPIO	Yes

Download 

Prebuilt Image: Currently we do not distribute public binary images for this development kit. Follow the instructions below to build a reference image from source.

Building the Image 

Follow the Quick Start Guide to setup the build environment and clone the source code.

After setting up the build environment, build the image by running:

cd lightspeed
kas build kas-cfgs/qcs6490-rb3gen2-core-kit.yaml

Flashing the Image 

To flash the board, you need a recent version of qdl, an open-source tool to download images to Qualcomm processors.

Build qdl from source (in a separate directory) by running:

sudo apt install git build-essential libxml2 libusb-1.0-0-dev
git clone https://github.com/linux-msm/qdl.git
cd qdl
make

To flash the image to the board, plug the board into the computer using a USB-C cable. Press the QDL button. Power the board using the 12V power cable and press the ON/OFF button for a second.

The board should now be connected to your PC in EDL mode:

$ lsusb
Bus 001 Device 020: ID 05c6:9008 Qualcomm, Inc. Gobi Wireless Modem (QDL mode)

Flash the image to your board by running:

cd build/tmp/deploy/images/qcs6490-rb3gen2-core-kit
qdl prog_firehose_ddr.elf rawprogram*_WIPE_PARTITIONS.xml
qdl prog_firehose_ddr.elf rawprogram0.xml rawprogram1.xml rawprogram2.xml rawprogram3.xml rawprogram4.xml rawprogram5.xml patch*.xml

Remove the USB-C cable and power from the board. Boot the board by applying power and pressing the ON/OFF button for a second.