LunaDrive: Pushing a Flat BLDC Motor to a 96 V Bus with GaN FETs

How fast a dynamic robot moves is ultimately capped by the rotational speed limit of its actuators, not by its control algorithms. High-power flat BLDC motors have driven this generation of dynamic robots forward, but most commercially available servo drives assume a bus voltage of 48 V or lower, and that ceiling applies directly to instantaneous maximum speed. What dynamic motion needs — jumping, fast leg swing, sudden acceleration — is a large burst of energy in a short window.

What blocks the route is not inertia in circuit design but a physical trade-off in the power devices: voltage rating and current capacity squeeze against each other in a semiconductor, so conventional Si MOSFET drivers struggle to achieve high voltage and high current at once. GaN FETs are widely seen as the way out, yet compact drivers that can be mounted on the rear side of a flat BLDC motor remain scarce — and if the drive electronics cannot fit inside the motor itself, the form-factor advantage of a flat motor is largely cancelled.

Core Innovation

• A GaN FET driver built for high-power flat BLDC motors. The target is a drive circuit whose shape suits the tight space behind a flat motor rather than an external box bolted nearby.
• Driving beyond the rated voltage on purpose. Instead of staying inside the datasheet envelope, the work deliberately drives the motor past its rated operating range to raise instantaneous speed, and studies delay compensation in that high-speed region. Once the electrical frequency climbs into the kilohertz range, delays that were negligible in the control loop begin to erode commutation and current-loop accuracy, which makes delay compensation a necessity rather than an optimization.
• Demonstrating on a battery pack rather than a bench supply. The high-speed load-lifting experiment runs from a 24S (100 V) LiPo battery, so what is being validated is operation under onboard power rather than a limit set by a laboratory DC source.

Results

• 96 V bus operation with 30 A continuous current when a heat sink is attached.
• 80 A peak current and a maximum electrical frequency of 3110 Hz.
• A high-speed load-lifting experiment powered by a 24S (100 V) LiPo battery, offered as evidence of applicability to dynamic robot operation.

Limitations

No efficiency figures are reported, and efficiency is the dividing line between “works on the bench” and “fits on the robot”: at 96 V and tens of amperes, drive losses and heat rejection decide whether the design can run continuously. Thermal data covers only the heat-sink case; behavior without a sink, and the actual thermal coupling when mounted behind a flat motor, are unstated — yet that rear-mount form factor is the whole point. The electrical frequency ceiling of 3110 Hz converts to a mechanical speed only through the pole count, which the abstract omits, so joint output speed cannot be inferred. Delay compensation is described as studied rather than quantified against the rated range. The abstract also says nothing about electromagnetic compatibility, protection logic under bus transients, device cost or supply, or long-duration reliability and lifetime data — precisely what an industrial buyer needs.

Industry Implications

Once a robot’s speed ceiling is set by bus voltage, power electronics become as much of a bottleneck as the algorithms, and this constraint has been badly underweighted. Humanoid and legged teams spend their attention on control and structure, then quietly accept the capability limit of the 48 V ecosystem when they choose drives. The signal here is that GaN devices plus compact drive circuits may shift that ceiling up a notch, and that shift propagates through the whole chain: battery cell count and bus voltage class, thermal design, and the high-voltage safety and compliance burden that follows. Crossing the conventional 60 V low-voltage safety boundary changes insulation, creepage, protection, and certification costs, and those hidden costs are what actually drive a platform’s voltage-class decision. For actuator suppliers there is a second implication, about form factor: putting the drive behind the motor means a given joint volume can deliver higher instantaneous speed, which changes dynamic performance more directly than raising the motor constant.

---