Supernova 3220: Bedroom motor smashes 580kph - narrow, hot, lethal efficiency
TL;DR: Chris Rosser designed the AOS Supernova 3220, picked a 32mm stator from RCM Power, tuned the rotor with Bayesian optimisation, and had RC inpower machine and wind the parts. Luke and Mike pre-cooled motors with dry ice and hit 580 km/h in Dubai.
The quick background: why anyone cares
Luke and Mike Bell set a new world drone speed record on 22 June 2025. They hit 580 km/h with Peregrine, powered by the Supernova 3220.
What problem the team had to solve
Their previous setup peaked at 480 km/h. Existing T-motor and APC gear had no headroom. They needed more torque, more power, and slimmer drag profile.
Core design insight
Once you optimise a motor’s magnetic geometry, that geometry stays optimal across props and voltages. You change KV to match props and cells, not the magnet layout.
Stator selection — the trade-off you’ll learn to hate
Big stator teeth carry flux. Big coil area lowers resistance. Those two want the same space. Chris picked an RCM Power 32 mm stator as the best compromise for torque and packing.
Aerodynamics constrained the choice
Luke and Mike demanded a narrow motor. Increasing diameter adds drag and costs top speed. So the team stuck to circa‑30 mm diameter while packing power inwards.
Rotor optimisation — no simple rules, just maths
Rotor geometry, magnet tiles and the flux ring form a high‑dimensional problem. Chris used Bayesian optimisation to hunt the sweet spot without burning compute on every random variant.
Bayesian optimisation — cheap guesses, fewer sims
Bayes builds a model of performance and uncertainty. It picks the next most promising geometry to simulate. That saves time over blind search in a dozen dimensions.
Simulation workflow — 2D first, 3D for the edges
They started with 2D radial transient sims for quick screening. Final tuning used 3D transient sims to capture end‑effects and rotor overhang gains.
Manufacturing handoff — RC Inpower does the dirty work
Design in CAD means nothing if parts arrive sloppy. RC inpower stamped stators, machined bells, sourced magnets, and wound coils to tight tolerances.
Materials and tight tolerances
They used 0.15mm non‑oriented silicon steel for stators. Magnets were 52‑SH grade for high temperature stability. Flux ring alloy stayed confidential.
Assembly details that matter
RC inpower built custom fixtures to install magnet tiles and keep minimal gaps. They use aerospace adhesives and precision CNC for bell and base to 0.01mm.
Windings, insulation, and balancing
They wind with automated 4S machines, use high‑temp enamel (claimed 250°C military grade), and run resistance, inductance, and insulation tests. Dynamic balancing hits 1–3 mg accuracy.
Dubai: perfect and hostile for speed
Hot air is thin air. That lowers aerodynamic drag and helps batteries deliver higher power. But heat wrecks motor magnets and ESCs fast.
How they beat overheating — dry ice
Motors were already hot past 500 km/h. In Dubai, magnets risked demagnetising mid‑run. Luke and Mike pre‑cooled motors with dry ice to −78°C for extra seconds of full throttle.
The record and the data
On the final validated run Luke briefly saw 579 km/h on OSD. GPS log later confirmed an instantaneous top speed of 580 km/h. Guinness certified it.
Next target: 624 km/h and the headaches ahead
They want 624 km/h to beat every all‑electric vehicle. Current issues: magnet cracking, extreme RPM when you run a 6S motor at 12S voltages, and brutal thermal shock.
Possible fixes and future motors
Reinforce magnets, rethink flux ring materials, and design larger Supernova variants. Expect new motors, stiffer rotors, and more exotic alloys in coming months.
Where to read more
Watch Luke and Mike’s interview here: https://www.youtube.com/watch?v=ZJ1bGOlHzGI. See detailed Supernova 3220 specs here: https://www.aos-rc.com/designs/aos-supernova-3220-925kv
Related reading on motor choices and magnet tech from the archive:
How to choose motors: https://www.unmannedtechshop.co.uk/2015/11/07/how-to-choose-the-best-motor-for-your-drone/
About Supernova motors and class claims: https://www.unmannedtechshop.co.uk/2023/08/14/revolutionizing-fpv-motors-aos-supernova-2207-unveiled/
Magnet manufacturing and choices: Sintered vs. Bonded: The Definitive Magnet Showdown for Drone Motor Domination?
Inside Peregrine’s record chase: Inside the Record-Shattering Peregrine FPV Drone's Quest for Unthinkable Speeds
FAQ
Q: Why reuse the same magnetic design for different props?
Because once magnetic geometry is optimised, you tune KV in the windings, not the magnets. That keeps flux paths ideal while matching RPM and torque.
Q: How did dry ice help without cracking magnets?
Dry ice bought a few full throttle seconds before magnets heated. But it created extreme thermal cycling. That increases cracking risk over many runs.
Q: Can hobbyists duplicate this at home?
Not safely. This project pushes materials and voltages far beyond typical FPV builds. You need precision manufacturing, specialised magnets, and proper testing rigs.
Q: What broke first during tests?
Magnets showed cracking and partial demagnetisation. Also ESCs and bearings ran hot under sustained full throttle at extreme RPMs.
Takeaway box
Short, sharp nuggets for skimmers.
- Supernova 3220 proves careful magnetic design scales across props and voltages.
- Pick a stator for the best flux‑to‑winding tradeoff — RCinPower 32 mm worked here.
- Bayesian optimisation gets better rotors with fewer heavy sims.
- Thermal management beats brute force — dry ice won a record, but stressed magnets.
- Expect reinforced Supernovas and more exotic materials on the next attempt.
This article was based from the video How I designed a 580kph world-record breaking motor in my bedroom