The Toroidal Propeller Test That Defied Expectations

The FPV drone community has been circling toroidal propellers for years now, treating them like some kind of aerodynamic curiosity—visually striking, theoretically interesting, practically untested. A YouTuber known as FPV Geek decided to actually find out if they work, which required building something that didn't exist: a reliable way to test tiny drone props.

What started as a simple question turned into a three-month engineering project. The first version of his test bench worked, technically, but relied on manually controlling throttle by hand. Try comparing thrust measurements when your input method is "push this stick smoothly and hope for the best." The curves never matched between runs. Science demands repeatability. His setup delivered chaos.

The Precision Problem

Version two needed automation—full computer control, identical test scenarios every time, no human variability. Getting there meant solving problems that don't appear in build guides. The creator initially assumed he could control the electronic speed controller (ESC) directly from his app, the way a flight controller does. That assumption cost him time. He eventually got motor control working, but through a workaround he doesn't fully detail in the video. The specifics matter less than the principle: engineering is rarely a straight line from problem to solution.

The hardware choices reveal how testing small things requires different thinking. His first test bench used a 5-kilogram load cell because he had one lying around. Whoop motors produce thrust measured in grams, not kilograms. A load cell works through strain gauges—apply weight, the metal bends microscopically, resistance changes. When you're only using one or two percent of the measurement range, that tiny signal drowns in vibration and motor noise. Switching to a 100-gram load cell transformed the data quality.

"Because we're only producing a few grams of thrust, I'm really just using maybe 1 or 2% of the full measurement range," he explains. "The signal becomes extremely small and suddenly vibrations and motor noise become huge in comparison."

He also switched microcontrollers twice. Started with an ESP32 (overkill), moved to an Arduino Nano (too limited), settled on an ESP8266 (just right, plus Wi-Fi for future upgrades). Then he added an OLED display, redesigned the enclosure to fit an emergency stop button that arrived larger than expected, fought with 3D-printed TPU feet that wouldn't cooperate, accidentally broke the display by pressing too hard, and iterated endlessly on details that don't make dramatic video footage but determine whether a tool actually works.

The result: PropBench V2, a clean enclosure housing load cell, flight controller, microcontroller, emergency stop, and custom software that runs three automated tests. Max thrust (spin to full speed three times, average the results). Acceleration (time to reach maximum speed). Efficiency (set a thrust target, measure power required to maintain it). Each test produces a high score for quick comparison.

The Toroidal Test

With the bench finally ready, he could address the original question. He tested two setups: motors and battery from a long-range build against 75mm toroidal props, and tiny brushless motors from his smallest 4K drone build against 35mm toroidals.

The 75mm toroidal props surprised him. They worked. Not just worked—they outperformed standard props in his tests. "The 3D printed toroidal prop even beats the 35mm gem prop in every category," he notes with audible surprise. He'd expected them to explode under the stress of high-RPM whoop motors. They didn't.

The 35mm toroidals failed completely. During max thrust testing, they couldn't reach the 10-gram target he'd set for the efficiency test. He had to lower the threshold to 5 grams just to get results. The output: noise, minimal lift, basic non-functionality. He based the design on standard gem props, calculating pitch and blade size accordingly. Something in the translation to toroidal geometry broke.

"My guess is that there's something wrong with the design here," he admits. The video includes no definitive answer to what went wrong. Scale might matter. Geometry that works at 75mm might not translate linearly to 35mm. Or the design parameters for toroidals require different calculations than standard props. The test bench revealed the failure. It didn't explain it.

This is where scientific method meets YouTube constraints. Understanding why the 35mm props failed would require more testing, more design iterations, probably consultation with aerodynamics expertise the creator doesn't claim to have. That's a different video, or a series, or material for a research paper nobody will write because the stakes are hobby drones, not aircraft certification.

What the Data Shows

The test bench itself proved more interesting than any single prop comparison. Running standard props at different scales illuminated efficiency relationships: lower KV motors with larger props consistently outperformed higher KV motors with smaller props when measuring power consumption for a given thrust target. This isn't news to experienced builders, but having quantified data transforms folk wisdom into measurable reality.

The creator also acknowledges his acceleration test might be flawed—sampling rate issues, maybe needs to run calculations on the microcontroller instead of the app. He weighted it at only 5% of the overall score because he doesn't trust the results. This kind of transparency is rare in tech coverage. Most people hide their methodology's weaknesses. He puts his in the video.

He's releasing everything—STL files, parts list, app code—for free on his Patreon and GitHub. The test bench is printable on a Voron Zero. Anyone can build one, improve it, test things he hasn't tested. The PropBench V2 story isn't finished because he's made it unfinishable by others. This is how open-source hardware should work.

The toroidal props remain a partial mystery. They work at one scale, fail at another, for reasons that demand more investigation. Three months of building a test bench answered one question and generated several more. That's not a disappointment. That's exactly how inquiry proceeds.

—Bob Reynolds