← Crank Charger

The circuit, from hand to phone

A hand at 90 rpm, a stepper motor run backwards, eight diodes, a capacitor and a converter: how the crank charger is predicted to put 10 W into a phone, worked through one number at a time. Every figure here comes from the project's design script and its ngspice model, the same numbers as the design document.

The whole path

Your hand turns the ring gear; the planets spin the sun 5.35 times faster; the sun turns the stepper, and the stepper, run backwards, is a generator with two separate coils. Each coil makes alternating current, so each gets its own bridge of four diodes. Both bridges pour into one DC bus, a capacitor keeps it steady, a buck converter brings it down to 5 V, and the phone takes what it wants through a USB meter.

coil A16.1 V pk, 0.9 Ω, 2.5 mHcoil B16.1 V pk, 0.9 Ω, 2.5 mHPLUS, the DC bus (7.3 V)MINUS8 × Schottky (SB560), two bridges1000 µFbuck7 V in, 5 V outQC3.0, 88 %USBmeterphone5 V
Drawn the way the card is built: a PLUS rail along the top, a MINUS rail along the bottom, four legs of two diodes between them, and each coil sitting between its own pair of legs.

1. The stepper is a generator

Turn the shaft and the rotor's magnet sweeps its flux through each coil 50 times per turn. Faraday's law makes the voltage proportional to how fast that flux changes, so the peak voltage grows in a straight line with speed. The one constant that sets it is the same one that links torque to current, and it can be read off the motor's label.

Gear ratio
91 / 17 = 5.35
Motor speed
90 rpm × 5.35 = 482 rpm, ω = 482 × 2π / 60 = 50.5 rad/s
Motor constant
k = Thold / (√2 · Irated) = 1.26 / (1.414 × 2.8) = 0.318 V per rad/s
Coil voltage
E = k · ω = 0.318 × 50.5 = 16.1 V peak per coil
Frequency
f = 50 pole pairs × 482 / 60 = 402 Hz
Inductance in ohms
X = 2π f L = 2π × 402 × 0.0025 = 6.3 Ω, against R = 0.9 Ω
Current ceiling
I = E / √(R² + X²) = 16.1 / 6.4 = 2.5 A; the limit at any speed is k / (p L) = 2.54 A
2026-11-03T22:02:14.052374 image/svg+xml Matplotlib v3.10.8, https://matplotlib.org/ 0 200 400 600 800 1000 motor speed, rpm 0 5 10 15 20 25 30 35 coil voltage, V peak 16.1 V at 482 rpm 0 200 400 600 800 1000 motor speed, rpm 0 1 2 3 current into a short, A peak ceiling k / pL = 2.54 A 2.5 A at 482 rpm
Crank faster and the voltage keeps climbing, but the coil's inductance pushes back just as fast, so the current it can deliver flattens near 2.5 A. Torque follows current, so the torque is pinned too: the only way to more watts is more speed, which is why the gears multiply speed and not torque.

2. Two bridges and a capacitor

Each coil swings between plus and minus 16 V. Four Schottky diodes round it flip every negative half upward, two of them carrying the current at any moment, each taking about half a volt. The two coils are a quarter wave apart, so one is near its peak while the other passes through zero, and the bus is topped up four times per cycle. The capacitor fills on each bump and gives back in the gaps.

After the diodes, nothing drawn
16.1 − 2 × 0.5 = 15.1 V
Bus current at 10 W out
11.4 W / 7 V = 1.6 A
Ripple, hand estimate
I / (4 f C) = 1.6 / (4 × 402 × 0.001) ≈ 1.0 V at most
Ripple, ngspice
0.05 V peak to peak on a 7.3 V bus
Why 7 V and not 15
under load the coil's own 6.4 Ω uses up most of the 16 V; the diodes and the bus hold the coil's ends about 8 V apart
2026-11-03T22:02:14.159355 image/svg+xml Matplotlib v3.10.8, https://matplotlib.org/ −10 0 10 20 volts coil A voltage coil B, a quarter wave later DC bus 7.3 V 0 1 2 3 4 5 time, ms (two cycles of the 402 Hz coil wave) −2 −1 0 1 2 amps coil currents, 1.5 A peak
ngspice at 90 rpm on the handle, the coils loaded so the bus takes the converter's 11.4 W. The current only flows near each coil's peak, when its voltage is higher than the bus plus two diode drops, and the bus barely moves.

3. The converter and the USB port

The buck converter switches the bus on and off fast and smooths the result, keeping about 88 percent of what it takes in. It needs about 7 V to make 5 V, and that fixes everything upstream: the coil has to produce 7 V plus two diode drops before any watt reaches the phone. On the other side, a plain 5 V, 2 A port gives the phone at most 10 W, and the phone decides how much of that it asks for.

Converter in
10 W / 0.88 = 11.4 W
Duty, ideal
5 V / 7 V = 71 percent of the time switched on
Slowest useful crank
coil must give 7 + 2 × 0.5 = 8 V, so ω = 8 / 0.318 = 25 rad/s: about 45 rpm at the handle with nothing drawn, more with the phone on
Port limit
5 V × 2 A = 10 W, reached at about 90 rpm
1 percent of a phone
15 Wh / 100 / 10 W = 0.9 min of cranking
2026-11-03T22:02:14.191164 image/svg+xml Matplotlib v3.10.8, https://matplotlib.org/ 0 20 40 60 80 100 120 140 160 handle speed, rpm 0 5 10 15 20 watts bus below 7 V: the converter cannot start 10 W: a 5 V, 2 A port your arm the phone 90 rpm: 10.0 W for 17.5 W of arm
The design script's prediction against handle speed. The model's first watts appear near 38 rpm, where its coarse simulation grid starts; the hand estimate above puts the useful start a little higher. From 90 rpm on, extra speed only makes your arm work: the port is already full.

4. Where the arm's work goes

Walking backwards from the phone: power out is power in times efficiency, and power is torque times speed. At the motor that is 0.28 N m; at the handle 1.85 N m, about 18.5 N on a 100 mm grip, roughly 2 kg.

2026-11-03T22:02:14.221247 image/svg+xml Matplotlib v3.10.8, https://matplotlib.org/ 0 5 10 15 20 25 watts at 90 rpm your arm motor shaft DC bus phone 17.5 W 14.1 W minus 3.3 W: two gear meshes, 0.9 each (a guess) 11.4 W minus 2.8 W: coil copper and the diodes (ngspice) 10.0 W minus 1.4 W: converter, 88 percent
57 percent of the arm's work reaches the phone. The gear loss is a guess (0.9 per mesh); the coil and diode loss comes from ngspice; iron loss in the stepper is not modelled, so the real numbers will be a little worse.

Generated from the crank charger repository: design.py, sim/alternator_sweep.py and sim/waveforms.py. Motor constants from the NEMA23 label: 1.26 N m holding, 2.8 A, 0.9 Ω, 2.5 mH. The converter's 88 percent and the 0.9 per gear mesh are assumptions.