Standard hobby-grade BLDC motor controllers lack high-speed inline current feedback, absolute encoder integration, and robust thermal management. This limits their application in direct-drive robotic joints which require precise torque control and haptic force feedback.
The primary goal of this project was to design, prototype, and build a high-precision, low-cost brushless DC (BLDC) actuator module with integrated closed-loop control. The actuator is intended for direct-drive robotic joints and force-feedback systems that require high torque density, absolute position awareness, and zero-backlash force estimation. By integrating the motor driver, analog sensing paths, and a 32-bit ARM microcontroller on a circular 40mm board matching a NEMA 17 footprint, the project achieves an ultra-compact mechanical profile with clean, high-bandwidth force response.
CONSTRAINT
The actuator control board is designed around a modular, low-noise architecture divided into three key subsystems:
1. Control Unit (Logic Stage): Utilizes an STM32G474RET6 MCU operating at 170MHz. High-speed SPI lines poll the AS5048A magnetic absolute encoder at 5 kHz, while the complementary PWM timer (TIM1) operates at 25 kHz.
2. Inverter Power Stage: A 3-phase half-bridge utilizing six CSD18532Q5B N-channel power MOSFETs driven by a DRV8305 gate driver. The gate driver is controlled via SPI to dynamically adjust gate charging current and limit EMI switching noise.
3. Current Feedback (Ana
TRADE-OFF
1. Inline vs. Low-Side Shunt Sensing: Low-side sensing is easier to route but only measures current when the low-side MOSFET is active. I chose inline sensing, which allows continuous current measurement throughout the entire PWM cycle. The trade-off was that inline sensing requires higher common-mode voltage rejection amplifiers, increasing component cost.
2. 3D-Printed Housing vs. Milled Aluminum: For rapid prototyping, PETG FDM 3D printing was chosen over CNC-milled 6061-T6 aluminum. This reduced prototyping time from 2 weeks to 4 hours. The trade-off is lower thermal conductivity, meaning
OUTCOME
The project successfully met its design goals:
- Hardware Validation: The 4-layer PCB was fabricated and assembled successfully. Power rails remain stable (+24V, +5V, and +3.3V) with less than 25mV ripple during full-load transitions.
- Firmware Control: Field Oriented Control (FOC) was established with a loop update rate of 25 kHz. Phase current transitions are clean, and torque ripple is below 3%.
- Force-Feedback: Direct-drive torque control was calibrated using the Python GUI, enabling active compliance and real-time haptic feedback interaction.
Reflecting on this build, the FOC actuator project was a success in demonstrating high-bandwidth torque control, but highlighted several key lessons for version 2.0:
1. Thermal Management Limitations: 3D printing the enclosure in PETG was excellent for rapid prototyping, but it severely bottle-necked the motor's performance. Under continuous stall or high-torque loads, the PETG housing absorbed heat without dissipating it, leading to thermal throttling. In the next revision, I will transition to a CNC-milled 6061-T6 aluminum body which will act as a structural heat sink.
2. Analog Noise Mitigation: Routing the kelvin connections for current shunts adjacent to high dV/dt phase switching nodes introduced significant ground bounce noise. Resolving this required manual board modification to add extra bypass capacitors. For future layouts, I will utilize a dedicated analog ground island and optimize current sense trace routing with symmetrical differential pairs.
3. Component Consolidation: The discrete gate driver and MOSFET layout consumed substantial board space. Moving forward, I would explore using an integrated power stage (smart power module) to shrink the PCB footprint and simplify current loop routing.
