Inside-Out Wind Turbine With a De Laval–Inspired Nozzle

Most wind turbines have externally mounted blades. This one, however, hides them: the entire turbine is contained within a 3D-printed housing shaped like a de Laval nozzle. The duct is wide at the inlet and narrow in the middle, accelerating the air precisely where the blades are located. The housing itself acts as a wind concentrator, with no additional moving parts.

Inside, neodymium magnets rotate with the rotor, while six coils on ferrite rods are fixed around it. No brushes, no sliding contacts - nothing to wear out.

This tabletop version is a concept model. One of the coils outputs raw AC: enough for measurements; another one is wired to a diode bridge that turns its output into DC, though it doesn't power anything yet. But no part of the design is limited to this size: the final step covers what to upgrade - thicker wire and more turns, a smaller magnet-to-coil gap, proper bearings, larger blades - to build a version that actually charges a battery.

Materials:

- Neodymium magnets, N52, 10 mm × 5 mm discs — 24 pcs

- Ferrite rods, 8 × 63 mm (M400NH grade; any AM-radio antenna ferrite rod works) — 12 pcs

- Enameled copper wire (magnet wire), Ø 0.71 mm (~AWG 21) — one 200 g spool (~55 m)

- Fully plastic bearings, 60 mm ID / 78 mm OD / 10 mm wide — 2 pcs (no steel parts for the magnets to grab)

- Rectifier diodes for the bridge (1N4007 or similar) — 4 pcs

- Electrolytic capacitor, 470 µF / 16 V — 1 pc

- 3D-printed housing parts (STL files attached — mine were printed in PLA by a print shop)

- Epoxy glue

- Thick cardboard — several sheets

- Popsicle sticks — a good handful

- 1 long, thick wooden stick (a candy-apple / corn-dog stick, or a piece of dowel)

- 1 cardboard tube from a toilet paper roll

- Empty soda cans — 2 pcs (optional; explained in Step 1)

- Electrical tape — 1 roll

- Reinforced duct tape — 2 rolls

Tools:

- Soldering iron, solder, rosin flux

- Hot glue gun with glue sticks

- Hacksaw

- Utility knife

- Sandpaper

- A measuring instrument: a very sensitive multimeter (millivolt range) or, better, an oscilloscope (used in Step 4)

- A laptop with Blender (free) — in case you need to tweak the models

- 3D printer (or a print service) for the housing parts

A note on scaling: if you resize the models for your printer, keep one dimension unchanged — the cylinder must stay Ø 60 mm. The plastic bearings (60 mm bore) are glued straight onto it, the rotor rides on the bearing, and the front piece's smaller opening is sized to match the same Ø 60 mm. Everything else can be rescaled freely.

The build starts with the 3D-printed parts. I don't own a 3D printer, so I took the files to a local print shop — any printing service can make these, and the parts are small enough to keep it affordable. I made the models in Blender (it's free), so it's worth having a laptop around in case you want to tweak them before printing.

I'm not sure the dimensions are stored in the files themselves, so here they are, just in case:

- Central cylinder: Ø 60 mm, 130 mm long

- White funnel (the nozzle) with its ring: 140 mm across at the widest point, 100 mm tall

The Ø 60 mm is the one dimension to respect: the plastic bearings (60 mm bore) glue directly onto the cylinder, and the rotor rides on them. And don't look for any mounting holes in the models — there aren't any. The ferrite rods get glued to the outside of the cylinder later in the build.

Originally, I printed two identical white funnels, but then I realized the front of the housing needs a different shape — more rounded, closer to a ball. That's the green piece in the photos, and it gave me the most trouble. Full honesty: I got its shape slightly wrong. I corrected it with a hacksaw, and then had to enlarge it in two ways: first, I widened its smaller opening, and then I made the whole piece wider and longer with strips cut from soda cans, held on with reinforced duct tape. That's what the optional soda cans in the supplies list are for — if you model the shape right from the start, you won't need them at all.

So learn from my mistakes and model this piece more elongated and a bit more generous in size from the beginning. I'm not attaching dimensions for it — you'll want to make your own version anyway. Only two things really matter: its smaller opening must match the cylinder diameter (Ø 60 mm), and it should be around 100 mm tall. Taller is fine — honestly, even better.

This is the main part of the build. The cylinder is Ø 60 mm — the same as the bearing bore — so everything in this step is built around the two plastic bearings.

For each side I cut rings out of thick cardboard, sized so that both diameters — outer and inner — are slightly larger than the bearing's outer and inner diameters. My advice: glue the blades and the magnets on first, and only then join everything with the cylinder parts. I did it the other way around and paid for it with a lot of fiddly work.

Each bearing assembly gets 12 magnets, glued in three groups of four with 120° between neighboring groups. Two of the three groups face one pole outward, and the third group faces the opposite pole. The blade shape is not important — slightly angled popsicle sticks are enough. What matters is that everything is glued on solidly; I recommend epoxy for this.

Now the coils. Each bearing gets 6 coils, 60° between neighbors — that's simply 360° divided evenly by six, just as the magnet groups sit at 360°/3 = 120°. All the coils end up combined into a single output, so this is a single-phase generator. All coils on one bearing must be wound in the same direction (it doesn't matter which one). Mine have 100 turns each, and they need to sit as close to the magnets as possible, so wind in several compact layers rather than one long thin one. To keep the gap between the ferrite rods and the magnets minimal, I filled the whole space between the cardboard and the bearing with hot glue and left openings only for the coils — so the glue supports them from the side, not from underneath.

Now the wiring. In the last photo, I did my best to draw the diagram (apologies for my drawing skills). Connect the coils every other one, pairing each upper coil with a lower one — wire everything like this first. Then pick four neighboring coils: from two adjacent ones take their two upper ends, and from the other two adjacent ones take their two lower ends. These are your two output leads. I marked them + and − in my sketch, but strictly speaking, the output stays AC until it passes the diode bridge — so just think of them as the two output wires for now.

P.S. Don't worry if you count only three magnets per row in the photo — it just caught an earlier stage of the build. Each row ends up with four.

I recommend sawing the cylinder so that a narrowed section ends up between the funnels and the coils — cutting off roughly 20–30 mm from each end works well. (And once again: attach the coils first, join the cylinder parts after.)

I also linked the blades of the upper bearing to the lower one - this is where the candy-apple stick finally earns its place in the supplies list. At this scale, there's no way to shape the blades so that the airflow (a hair dryer, in my case) could spin the far bearing on its own, so the stick makes the front rotor drag the rear one along.

Thanks to the cardboard rings from the previous step, joining the cylinder and the bearings went easily, and nothing rubs or blocks the rotation.

The soldering itself is probably one of the easiest steps: I've already shown what connects to what, so all that's left is to make those joints permanent. (follow basic safety rules and don't burn yourself; the iron is hotter than it looks.)

The diode bridge is a bit trickier. I've attached a photo of the schematic. The key to reading it: each diode is drawn as a triangle pressing against a bar, and on the real diode, that bar is the painted stripe on its body. To place a diode correctly, match the stripe on the part with the bar in the drawing - that's the whole trick.

Once you see that pattern, the outputs are easy to find: the point where two stripes meet is the DC plus, and the point where two triangles meet is the DC minus. One more thing - the schematic doesn't show the capacitor. It gets soldered right where the DC output is marked. And here's the trap: on a diode, the stripe marks the cathode, but on an electrolytic capacitor, the stripe marks the minus leg: same stripe, opposite meanings. So the capacitor's striped leg goes to the point where the triangles meet, and its other (plus) leg goes to the point where the stripes meet. Double-check this before spinning anything - an electrolytic soldered backwards can pop.

There are only four photos here - the very first attempt and the final one - but this step took more time than almost anything else. The problem has a name: cogging. The magnets are strongly attracted to the ferrite rods, so the rotor "sticks" in certain positions, and the blades have to be large enough to break it free before it can spin at all. I went through many blade sizes and shapes before the rotor finally overcame that magnetic grip.

As I said back in Step 2, the blades inside the bearings don't matter much - in my design, the main work is done by these external blades.

My final blades are made from short pieces of the same winding wire (they work as a skeleton, so each blade holds its shape and can be bent into the right curve), reinforced duct tape, popsicle sticks, and hot glue — the whole thing is mounted with epoxy. And reinforce the blade bases as much as you can: that's where all the force goes.

I measured two coils: a bare one and the one wired to the diode bridge. All bare coils are identical by design, so one calculation covers them all. Everything below is in mean values, to match what the instruments report.

Inputs: n = 6.5 rev/s; magnet orbit r = 30 mm; magnet stack Ø10 × 20 mm (4 × N52 discs, Br ≈ 1.43 T); gap z = 9 mm; coil N = 100 turns of Ø 0.71 mm wire; ferrite rod Ø 8 mm.

BARE COIL: PREDICTION

B ≈ 0.08 T — the field at the rod tip. This comes from the standard formula for the on-axis field of a cylindrical magnet: plug in the stack size (Ø10 × 20 mm), the N52 strength, and the distance z = 9 mm. At the magnet's face, the same formula gives ≈ 0.7 T — the 9 mm gap alone costs a factor of ~9.

Φ = B × A_rod × 1.5 ≈ 0.08 × 50 mm² × 1.5 ≈ 6 µWb — the flux through the coil. A_rod = πd²/4 ≈ 50 mm² is the rod's cross-section, and the ×1.5 is there because ferrite conducts magnetic field much better than air, so the rod funnels in field lines from a region a bit wider than itself.

Δt ≈ 8 ms — the duration of one magnet pass. The magnets move at v = 2πrn ≈ 1.2 m/s, and the flux rises and falls while a magnet crosses the ~10 mm zone in front of the rod: 10 mm ÷ 1.2 m/s ≈ 8 ms.

U = N × ΔΦ/Δt = 100 × 6 µWb / 8 ms ≈ 0.07 V peak → about 25 mV as a mean reading (the pulses are short and sparse, so the average sits well below the peak).

R ≈ 0.13 Ω — the coil's own resistance: 100 turns around a ~Ø10 mm winding is about 3 m of wire, and Ø 0.71 mm copper has 0.043 Ω/m, so 3 × 0.043 ≈ 0.13 Ω.

I_short = U / R ≈ 0.07 / 0.15 ≈ 0.5 A peak → on the order of 0.15 A mean. This is the short-circuit current — the most the coil can push.

COIL WITH THE BRIDGE: PREDICTION

A silicon diode opens at ~0.6 V, and in a bridge, the current always crosses two of them: threshold = 2 × 0.6 = 1.2 V. Our peak is 0.07 V — seventeen times too small. Prediction: U_DC ≈ 0, I_DC ≈ 0. The diodes never open. Not a wiring mistake - a scale problem, and the final step is about fixing exactly this.

WHAT WE ACTUALLY MEASURED (mean values)

Bare coil — 20.4 mV and 168 mA

THEORY VS REALITY

The bare coil is a clean win for the theory: 20.4 mV measured vs ~25 mV predicted, and 168 mA vs ~150 mA — both within 20%, which is excellent for a cardboard-and-hot-glue build.

The bridge behaved exactly as calculated, too: with the input seventeen times below the diode threshold, its DC output showed practically nothing. That's not a failure — that's the prediction coming true. In this build, the bridge is a demonstration piece: it shows how AC turns into DC in principle, and it starts earning its keep in the scaled-up version, which is what the final step is about.

P.S. There is only one diode bridge in the whole machine, and that's deliberate: I rectified a single channel for demonstration and left the second bearing's coils as raw AC. And sorry, the only photos I have are of the oscilloscope itself, not of the actual measuring session.

Time to close the machine up: the rear funnel goes on, and the whole thing gets a stand.

My stand is nothing fancy:a sheet of cardboard, popsicle sticks, a toilet paper tube, hot glue, and reinforced duct tape. This part is completely optional and entirely up to you: the stand doesn't affect how the generator works, so build whatever holds it steady - a wooden base, a 3D-printed frame, anything. Mine is simply made of what was left on the table.

That's it for the build. One step remains - the tips and lessons I collected along the way.

If I could restart this project, I would change a single thing, and every other improvement falls out of it on its own: do the calculations first, build second. I did it backwards — I fitted the design to the 3D models I already had, so the parts dictated the physics instead of the physics dictating the parts.

Here is what calculating first would have shown. The most expensive number in this whole build is the magnet-to-coil gap: at 9 mm it eats almost nine tenths of the magnetic field — and that gap is not glue or sloppiness, it is literally the wall of the bearing sitting between the magnets and the coils. So the right order is the opposite of mine: first find the thinnest bearing you can get, then design everything around it. A few millimeters shaved off there beats any other upgrade on this page.

The second lesson is scale. A bigger build means bigger blades, and bigger blades mean each rotor stage catches enough air to spin on its own — no wooden stick tying the two stages together, a simpler assembly, less friction.

And if you want to push this design from "physics demonstrator" to "actually charges a battery", the shopping list writes itself: a thin bearing to shrink the gap (every millimeter there buys back a big slice of field), several hundred turns of thinner wire instead of my 100 × 0.71 mm, Schottky diodes like the 1N5819 instead of ordinary silicon (they open at roughly half the voltage), and larger blades in real wind instead of a hair dryer. Shrink the gap and multiply the turns together, and the output grows by a factor of ~20 — from millivolts into volts, past the diode threshold, into charging territory.

A bonus measuring tip: an oscilloscope alone cannot measure current — it only sees voltage. If you need a real current reading, solder a 1 Ω resistor into the loop, measure the voltage across it, and I = U / R gives you the current directly.

That's the whole journey. If you build your own - bigger, calculated first, with a thinner bearing - I would genuinely love to see it in the comments.

Archived version with DOI: https://doi.org/10.5281/zenodo.21449769