Installation
Requirements and installation procedure for the Neuro N6 Arduino core.
The Neuro N6 Arduino core is installed by cloning its repository into the Arduino sketchbook. It is distributed as a git repository rather than through the Arduino Board Manager. Three external tools must be present on the system path. The procedure is the same on Windows, macOS and Linux.
Requirements
| Tool | Purpose | Notes |
|---|---|---|
| Arduino IDE 2.x or Arduino CLI | Build environment | Arduino IDE 1.x is not supported |
| Arm GNU Toolchain | Compiler | GCC 11 minimum. Version 13.3.rel1 is the tested release. Must be on PATH |
| STM32CubeProgrammer | Firmware signing | The signing utility is located automatically |
| Python 3.8 or later | Build scripts and uploader | Requires the pyserial package |
| ST Edge AI Core 4.0 | Neural network compilation | Required only for sketches containing #pragma neuron6 |
Compiler
The Arm GNU Toolchain is downloaded from Arm's developer site and its bin directory added to PATH. A prebuild hook verifies that arm-none-eabi-gcc is present and at least version 11, and stops the build with a message if it is not. The minimum arises from a section keyword in the linker script that older linkers do not accept.
The core does not contain per-operating-system compiler paths; the compiler is always resolved from PATH.
STM32CubeProgrammer
STM32CubeProgrammer supplies the signing tool that produces the _Trusted.bin image the bootloader accepts. The build locates it in the default installation directory for each operating system. The CubeProgrammer graphical interface is not used in normal operation.
Python
Any Python 3.8 or later. The uploader depends on pyserial:
pip install pyserialOn Windows the core invokes Python through a shim, tools/run_python.cmd, that locates a working interpreter when a bare python command resolves to the Microsoft Store stub. This is common when Python was installed through conda.
Board package
The repository is cloned into the sketchbook's hardware directory. The directory names determine the vendor and architecture the IDE reports, so they must be as shown:
<Sketchbook>/hardware/Ohmlab/stm32n6/On Windows the sketchbook is normally Documents\Arduino.
cd Documents/Arduino/hardware/Ohmlab
git clone https://github.com/ohmlab-ltd/stm32n6.gitAfter the IDE is restarted, Neuro N6 appears under Tools > Board. The bundled libraries and examples are available whenever that board is selected.
ST Edge AI Core
Sketches that compile a neural network require ST Edge AI Core, because the model is compiled for the NPU on the host during the build. Camera-only and sensor-only sketches do not require it.
On Windows the build searches C:\ST\STEdgeAI. On other systems, or for a non-default location, the environment variable NEURON6_STEDGEAI is set to the installation directory. The bundled models require version 4.0.
Verification
With the board connected over USB-C, Tools > Board > Neuro N6 is selected and the board's port chosen under Tools > Port. The example File > Examples > NeuroN6Examples > 01.Basics > Blink is uploaded. The first build compiles the entire core and takes one to two minutes. A blinking green LED indicates a working installation. First Upload describes the upload process and recovery of an unresponsive board.
Updating
The installation is a git checkout and is updated with git pull in the stm32n6 directory. The IDE is restarted afterwards so that it rescans the libraries. If a newly added bundled library is not detected, the IDE cache directory (%LOCALAPPDATA%\Arduino15\cache on Windows) is cleared once.