Creative Hardware

An Open-Hardware Board for Driving Four Brushless Motors at Once

BeagleBoard.org's QuadESC pairs four independent three-phase motor drivers with a neural-network-accelerated microcontroller, with full KiCad designs and firmware published for anyone building kinetic or robotic work.

Kinetic art, robotics, and anything with moving parts driven by brushless DC (BLDC) motors has historically meant either an off-the-shelf ESC per motor or a custom driver board designed from scratch. BeagleBoard.org’s new BeagleConnect QuadESC sits in between: a single 60 x 48mm open-hardware board that drives four BLDC motors independently, with every design file published for anyone to build, modify, or fork.

The control side runs on the BeagleConnect Torque module, built around Texas Instruments’ AM13E23019 microcontroller — a 200MHz Arm Cortex-M33 core with 512KB flash and 128KB SRAM, plus a floating-point and trigonometric math unit and a TinyEngine neural-network accelerator on board. That last piece matters for anyone wanting to run learned control policies (rather than purely classical PID loops) directly on the device, without offloading to a separate compute board. CAN connectivity comes via a TCAN844 transceiver for linking multiple boards or talking to other CAN-based robotics hardware.

The motor side uses four TI DRV8323SRTAR gate drivers feeding twelve TI CSD88599Q5DC half-bridge power devices — three phases per channel, four channels total — with per-channel current, phase-voltage, and temperature sensing plus overall battery-voltage monitoring. Power input runs 3S–6S (9–25.2V), and the board’s current design targets are 40A continuous and 70A burst per motor, though BeagleBoard.org’s own documentation is careful to flag these as design targets pending thermal validation, not an established rating — a detail worth taking at face value rather than treating as a spec sheet guarantee.

What “open” actually means here: editable KiCad PCB designs, firmware source, an application programming guide, and an XDS110 debug/programming guide are all published directly through the project’s GitHub repository and documentation site, alongside a schematic PDF. One caveat surfaced in early coverage: the AI-assisted diagnostics and advanced motor-control features the neural accelerator enables aren’t active out of the box — they depend on firmware a user develops and loads themselves, same as the rest of the open-hardware model Behringer’s recent releases (also covered on this site this week) have leaned on: the hardware and starting code are public, what gets built on top is up to whoever picks it up.