Getting started with Coralboard

Congratulations on getting a new Coralboard Google I/O Edition!

The Coralboard is designed to be highly capable in a headless configuration without a display. In the headless configuration, you will need to connect to the Coralboard from your host machine and program it entirely through a command-line interface (CLI) using a terminal window.

While the default setup is headless, Coralboard does support an optional MIPI display if your project requires a visual interface. For more details on compatible displays, pinouts, and hardware configuration, refer to the User's Guide.

The Google I/O Edition kit includes a few accessories for the Coralboard so you can experiment with on-device AI models. Some assembly is required, so follow these steps to get started.

1. What's in the box?

Synaptics Coralboard 2GB Limited Edition
Sensor HAT accessory board
CSI camera module
CSI flex ribbon cable

You will also need (not included): A Phillips #0 screwdriver and a USB-C cable to connect your machine.

2. Assemble the sensor HAT

Included with the Google I/O edition only, the sensor HAT is an optional accessory that attaches to the Coralboard. The sensor HAT adds a camera for capturing images, two microphones for capturing audio, a piezo buzzer for sound output, and a few LEDs as programmable indicators.

Connect the camera module: First, plug the flex ribbon cable into the camera module. Pull the lock bar, insert the flex cable as shown, and push the lock bar back in to secure the flex.

Connect to the Coralboard: Next, plug the other end of the ribbon cable into the CSI connector on the main Coralboard (located between the two sides of the mikroBUS socket). Pull the lock bar, insert the flex cable as shown, and push the lock bar to secure it.

Mount the camera: Place the camera module on the dedicated position mounts on the HAT board and use the four provided M1.6 screws to secure it.

Attach the HAT: Plug the HAT board down onto the Coralboard. The camera should be facing up (outward) with the flex cable folded neatly under the HAT board as shown.

3. Connect and boot up

Boot Switch Configuration

Coralboard includes a physical switch that determines the location of the software boot image. It can be loaded from the eMMC memory (default) or from an SD card (optional). Ensure the switch is set to the '1' position as shown in the image.

1: eMMC boot (default, this is the OFF position)
ON: SD card boot (optional, not supported out-of-the-box)

Power the Coralboard

Coralboard is powered through a USB-C connector located in the lower right-hand corner of the board. This power may be supplied through a connection to a host machine (where a data connection is desired) or a USB-C power supply.

Status LEDs

Coralboard features three status LEDs to help you monitor operation:

Green LED: Power indicator (always on when powered up)
Red LED: Heartbeat indicator (two short blinks with a delay)
Blue LED: IO indicator (sporadic based on memory access)

4. Open a terminal and log in

In the headless configuration, you will access the operating system using a terminal window using one of two options:

Option 1 - Use the Android ADB Interface (via the host USB-C cable)

On your host machine, download Android SDK Platform-Tools and see details on using the ADB interface

Open a terminal on your host machine and use the following command to open a shell:

adb shell

Option 2 - Use a data cable with the sensor HAT

To connect to the Coralboard using the sensor HAT, plug a USB-C data cable directly from the HAT to your host machine, open your preferred terminal emulator and connect using the designated serial port set to a baud rate of 115200.

Login: root
Password: (none)

See the User Guide for more detailed examples of connecting over serial.

5. Connect to a network (recommended)

For the best experience, we recommend you connect your Coralboard to a network. There are a few different methods but we recommend using Network Sharing through the USB port.

Requirements:

Host machine that supports network sharing

Follow these steps:

With the USB type C cable connected, enable network sharing on your host machine (the method depends on your machine type - Windows, macOS, or Linux supported, see User Guide for details).
Share the network to the "NCM".
On the Coralboard, enable DHCP: udhcpc -i usb0

If you have an m.2 Wi-Fi module then see the User Guide for more details on how to enable it.

6. Try out the AI demos!

Once you have successfully logged into the terminal, and connected to a network, download example applications to start experimenting with on-device AI.

Each demo application is organized into its own subdirectory. Navigate to any of the subdirectories and open its README file for specific execution instructions and details. Available Demo Subdirectories:

gemma_translate – Text translation using Gemma
jellectronica – Interactive audio and sensor demo
Object_detection – Real-time object identification using the camera
Function_calling – LLM tool use and function calling workflows
Image_classification – Image categorization models

For more on each of these applications, head over to the Demos page.

To obtain the examples, connect to a Coralboard terminal and clone the examples to the home directory.

cd /home/root
git clone https://github.com/synaptics-astra-demos/sl2610-examples

Follow the instructions in the top level README.md to get started.

cd sl2610-examples
more README.md

Can't get online?

Network sharing is disabled on some host machines. The object detection and image classification examples support offline dependency install.