Engineering portfolio · personal projects

Michael Douglas

Electrical engineer (B.S., University of Maryland). I design, build and test hardware end to end: brushless motors and motor controllers, a custom quadcopter with its own flight firmware, and the rigs to test them safely.

01

Custom brushless outrunner motor

12-slot / 14-pole · SolidWorks · 2026
Wound 12-tooth laminated stator, CAD render
Prototype 2 stator: laminated electrical steel, 12 wound teeth.
Problem

Understand, then eventually replace, the off-the-shelf motors on my quadcopter.

Approach

Built an air-core PLA prototype first to prove the winding and test method, then redesigned with an iron core, changing one variable so the gain is measurable. Lamination went through five DXF revisions (V1–V5).

Result

Prototype 1 spins on a commercial BLHeli ESC, with back-EMF captured on a scope. Prototype 2 design is frozen and released as a machine-shop quote pack: four toleranced part drawings plus the lamination drawing.

Magnetic designWinding layoutGD&T / tolerancingDFMOscilloscope
Rotor bell CAD
Rotor bell with vented top plate.
Section through magnet and back iron
Magnet bonded to a steel back-iron ring.
Stator lamination drawing
Lamination drawing (V5) for etch or wire-EDM.

Machining drawings

Rotor bell drawing
Rotor bell
Base flange drawing
Base / hub: bearing seats in one setup
Shaft drawing
Shaft: h6 journals, circlip groove
Back iron drawing
Back-iron ring
02

Sentinel quadcopter

~300 g · C/C++ firmware · 2026
Sentinel full CAD assembly
Designed: full assembly with sensors and electronics modeled to pin level.
Sentinel built quadcopter
Built: 3D-printed frame, hand-wired flight electronics.
Problem

Build a small quadcopter from the subsystem level up, with my own flight-control firmware on an 8-bit microcontroller (32 KB flash).

Approach

Wrote C/C++ flight control: 200 Hz attitude, altitude and optical-flow position loops. Iterated the frame in CAD (arm V0 → V1_3, legs V1 → V3), split every part to fit a 7 × 7 in print bed, and checked arm stress with FEA. Logged 14 channels of telemetry at 25 Hz on every flight. Designed its custom brushless motor controller (section 03).

Result

Stable 18 s position holds at 300 mm. Hold error dropped from 132 to 69 mm RMS across firmware versions, and every flight anomaly was root-caused from the logs.

Position hold track chart, flight 1 versus flight 11
Hold track at 300 mm: v19 vs v20 firmware.
Bar chart of hold error over 14 flights
Hold error across 14 flights; anomalies in red.
Motor command and roll during flight 10
Flight 10: logs showed a lost motor connection, not a tuning fault.
ItemResult
Thrust budget~130 g per motor, ~510 g total; thrust-to-weight ≈ 1.7
Arm FEAPeak ~5.8 MPa at 3 N per arm (≈2.3× max motor thrust)
Yaw biasTraced to one motor producing ~3.6× the reaction torque of the others
SensorsIMU, time-of-flight altimeter (I²C), optical flow (SPI)
03

Custom brushless motor controller

4-channel PCB · bring-up · 2026
Assembled motor controller board
Assembled 4-channel board.
Bench demo with four motors
Bench demo: one controller driving four motors.
Problem

Off-the-shelf motor controllers are a black box. I wanted one I understand down to the comparator, for Sentinel and future aircraft.

Approach

Simulated each block in LTspice, laid out a 4-layer PCB, had it fabricated and assembled, then flashed open-source AM32 firmware to all four channels. When motors lost sync on first spin, I ruled out firmware and power-stage causes in order, then simulated the back-EMF sense network.

Result

Found a filter mismatch putting ~11 µs of false timing error after every PWM edge. Removing one part fixed it, and the board now drives motors on the bench. Next revision matches the sense filters on all channels.

Motor controller block diagram
Block diagram, one of four channels.
04

Test equipment

Gimbal rig · thrust stand · 2026
Built PVC gimbal rig
Built: PVC frame, rings printed in bolted sections.
Slide-in axis lock
Slide-in lock on each axis.
Sentinel mounted in the gimbal rig
Sentinel at the center of rotation.
Problem

Tune and verify each control loop without risking the aircraft or myself.

Approach

Designed an orbital 3-axis tuning rig with roll, roll-pitch and roll-pitch-yaw configurations (each lockable, so one loop is tuned at a time), plus a linear rig. Built a thrust stand with a load cell (HX711), INA226 voltage/current sensing and an IR RPM sensor, alongside a watt meter.

Result

Measured the thrust budget and motor imbalance that set the flight-controller tuning, and moved first spin-ups off the bench and into a constrained rig.

Gimbal rig CAD in three configurations
Gimbal rig CAD (SolidWorks).