This is part two of the fixed-wing project, where a pile of parts becomes a real electrical system. If you have not read part one yet, start there: Fixed-Wing Foamboard Trainer Project.
I ran an experiment. Not on code. Not on architecture. On hardware. I wanted to see whether the AI workflow I use in software development - spec-first thinking, tight feedback loops, interactive debugging, and rapid iteration - would hold up in a physical build.
No deep pre-reading. No tutorial binge before touching the parts. Just curiosity, components, and AI as an active technical partner. The goal was to build and understand a 4-channel fixed-wing power and control system from scratch.
The Rules of the Experiment
- I would not pre-study the topic.
- I would not binge YouTube tutorials before starting.
- I would rely almost entirely on AI to plan, explain, wire, debug, and troubleshoot in real time.
I wanted to test whether the workflow I already trust in software could transfer directly to physical systems. It turns out: it absolutely can.
System Drawings
If you are building in a similar spot, these diagrams are meant to make the architecture legible at a glance: how power flows, how control signals flow, and what the ESC arming path looks like.
Power Architecture (Battery to 5V Rail)
Control Signal Flow (Stick to Motion)
ESC Arming Checklist (Fast Debug Path)
Step 1: Shopping by Conversation
Instead of searching for a pre-approved beginner kit, I described the destination: eventually build my own drone, but start with fixed-wing fundamentals.
We converged on an A2212 1000KV motor, 40A ESC with 5V/3A BEC, 3S LiPo, 9x4.7 props, MG90S servos, RadioMaster Pocket, ER5A receiver, servo tester, extensions, prop adapters, and foamboard linkage hardware.
- A2212 1000KV brushless motor
- 40A ESC with 5V/3A BEC
- 3S 1000mAh LiPo
- 9x4.7 props
- MG90S servos
- RadioMaster Pocket + ER5A ELRS PWM receiver
- Servo tester, servo extensions, prop adapters, and linkage hardware
I did not choose these from comparison charts.
The useful part was not just the shopping list. It was the why behind each decision: What does 3S mean? Why PWM instead of CRSF? Will this ESC power servos safely? The explanation layer changed everything.
Step 2: Unboxing to First Electrical Life
First bring-up chain: battery to ESC to servo tester, ESC to motor, and BEC rail to servos. The motor beeped. The servos moved. First electrical life.
Right after that came the fear: did I just overvolt something? That was the first big AI-mentor moment, and the BEC concept clicked in a way it never would have from passive watching: battery voltage comes in high, ESC regulates a safe 5V rail out for receiver and servos.
What I Learned (Before I Realized I Was Learning)
1) What 2S and 3S really mean
Not just a number on a listing: series cell count, nominal voltage, fully charged voltage, and why each downstream component has a very different tolerance envelope.
Learn more: LiPo battery fundamentals and lithium-ion voltage ranges (Battery University).
2) Why swapping any two motor wires reverses direction
Brushless motors are 3-phase systems. Swap any two leads and you reverse phase order, which reverses rotation. Elegant and immediately testable.
Learn more: Brushless DC motor fundamentals.
3) How ESC arming actually works
Minimum throttle at startup, trim awareness, channel direction, and endpoint calibration all matter. It felt like debugging a startup handshake protocol in the real world.
4) Channels, PWM, and signal flow
The control path is now clear: stick input to radio, radio over ELRS, receiver output as PWM, PWM to ESC and servos, then motion in the airframe.
Learn more: ELRS PWM output behavior and EdgeTX input/mix/output pipeline.
5) The BEC revelation
The ESC is not only a motor controller. It is also the power architecture anchor for the aircraft electronics.
Learn more: Electronic speed control (ESC) and battery eliminator circuit (BEC).
Where AI Surprised Me
The biggest surprise was not generic correctness, it was interactive troubleshooting depth. When I said, "the motor rapidly beeps," I did not get a canned response. I got a sequence: arming logic, throttle checks, channel mapping, reverse verification, endpoint calibration, and then clarifying questions to isolate fault state.
It felt like pairing with a senior engineer, except the domain was hardware.
The Meta Experiment
In software, my loop is define, ask, test, adjust, repeat. This experiment showed the same loop works in electrical systems, RF links, motor control, and mechanical linkages.
AI did not replace understanding. It accelerated understanding through active, contextual dialogue while the system was alive on the bench.
What This Means and What Is Next
This was not just an RC project. It was a hypothesis test: can AI-guided workflows extend beyond code into physical debugging, voltage safety, wiring mistakes, and system bring-up? Yes.
What's Next
- Foamboard airframe build
- Aileron and linkage geometry tuning
- CG experiments and trim iteration
- Possible flaperons and later stabilization
- Eventually autonomy
Only after it worked did I watch tutorials. The difference was huge: I was no longer consuming passively. I was validating and deepening a model I had already built with my own hands.
Final reflection: electric motors, control channels, initialization sequences, and failure modes are just another kind of distributed system. The tools are different. The thinking is the same.