Why would a transformer turn?
A house needs a few kilowatts at 120/240 volts; the line outside carries 7.2 kilovolts. Today a pole or pad transformer closes that gap with one magnetic circuit soaked in oil. The design is a century old and nearly perfect at its job, and it has three fixed properties: it is a hand-assembled tank of oil, it is built in one piece, and it is electrically invisible to the grid’s spinning machinery.
The Turning Transformer changes all three. The line drives one long, continuous coil that never moves. Iron taps along that coil carry its flux into small station solenoids in staggered rings around a rotor; fixed magnets on the rotor are pushed, and small free-turning magnets are pulled. The rotor spins an ordinary low-voltage alternator, which makes the house supply.
- No oil
- Nothing in the machine needs to be immersed. The high-voltage coil can be cast in resin, and the rotor chamber sealed. Pumped down to a vacuum, the chamber loses its air drag and gains vacuum’s high breakdown strength across its gaps. Filled instead with a dielectric refrigerant, it cools itself: the gas boils off the hot windings and condenses on the shell. Either way there are no gallons of mineral oil to weep down a pole, catch fire or be cleaned out of a yard.
- Three assemblies
- The machine separates into a high-voltage stator module, a rotor cartridge and a low-voltage alternator. Each is made, tested and stocked on its own line, by the processes motors already use: stamped laminations, machine-wound bobbins, sintered magnets, composite tube. The high-voltage module, which carries the real insulation risk, has no moving parts and never needs opening. The rotor cartridge carries every wearing part and can be changed at the curb.
- Inertia in every house
- Its rotor is locked to the line’s frequency, so it behaves like a small synchronous machine. When a power plant trips and frequency starts to fall, millions of these rotors slow together and hand their stored energy back. A grid that is losing its turbines gains a flywheel at every meter.
Transformer shortages make the question urgent: the U.S. Department of Energy reports some transformer prices rising four to nine times in five years, and public utilities report critically low spares (DOE; APPA). But the case for a turning transformer rests on what it can do that an oil tank cannot.
Inside the machine
The whole step-down fits on one bedplate about a metre long. The left-hand housing is the high-voltage module: the line enters through a single porcelain bushing and never leaves it, and its coil, iron taps and station solenoids stay still while the rotor turns inside them. The rotor cartridge is one glass-epoxy shaft carrying both magnet groups, with a ribbed insulating spacer between them. The right-hand housing is the alternator, which makes the house supply. Drag to turn the view; pinch or scroll to come closer.
Cutaway model
Proportions follow the sizing below: a 10 cm rotor, roughly 6 cm of active station length per ring group, and 15–20 cm of insulating spacer. Rotation is slowed for viewing; the machine runs at 3,600 rpm. Red and blue are the fixed push magnets; gold are the free-turning pull magnets in their cradles.
How much shaft does a house need?
The rotor is synchronous: a two-pole rotor locked to 60 hertz turns at 3,600 rpm. For 5 kW from an alternator assumed 93 percent efficient, the shaft must carry about 5.38 kW.
τ = P / ω = 5,376 W ÷ 377 rad/s = 14.3 N·m
F = τ / r = 14.3 N·m ÷ 0.05 m = 285 N of total tangential pull at a 5 cm rotor radius
L = τ / (2πr²σ) = 6.1 cm of active rotor at an air-gap shear stress σ of 15 kPa
The last line sets the scale. Small permanent-magnet machines routinely work their air gap at 10–20 kilopascals of magnetic shear. At that stress, a 10 cm rotor only needs about 6 cm of magnetically active length. The machine is the size of a shop motor, not a pole pig. Its difficulty lies elsewhere: making one phase of alternating current turn it smoothly, and winding a coil that the line voltage cannot break.
One phase has to travel
The distribution line to a house is a single phase. If every station is fed in step, each pulses at the same instant and the field only breathes in and out: a standing wave. Any standing wave is two travelling waves of equal strength turning in opposite directions:
B cos ωt cos θ = ½B cos(θ − ωt) + ½B cos(θ + ωt)
The rotor locks to the forward wave. The backward wave sweeps past it at twice line frequency, adding a 120 Hz torque that swings as large as the useful torque itself. The fix comes from the stations themselves. Each one is a small secondary of the mains coil, and a series capacitor or inductor in its loop shifts its current phase. Splitting the stations into two groups, one shifted +45° and one −45°, gives two currents in quadrature — the trick inside every capacitor-run motor. Reversing a station’s winding adds a free 180°.
A shift of r costs amplitude: the station’s impedance rises by 1/cos r, so its current falls by cos r. That makes a surprising result. Tuning each station to its exact ideal phase is worse than the simple two-group split, because the stations that need nearly 90° of shift go almost dark, and that uneven brightness pattern recreates a backward wave of about a third of the useful wave. The split scheme with 48 stations leaves a 6.6 percent ripple.
Station ring
- Forward wave
- —
- Backward wave
- —
- 120 Hz torque ripple
- —
Each bar is one station’s instantaneous current, drawn outward for one polarity and inward for the other. The arrow is the net field the rotor feels. The trace below plots torque on a synchronously locked rotor over two line cycles, as a multiple of its mean. Wave strengths are fractions of one station’s full current. Spatial harmonics from finite station width are not modelled.
Make the free magnets keep up
The pull side of the machine relies on a magnet that can turn on its own axle. A magnetic dipole m in a field B has energy U = −m·B. If it can swing into line with the field before the field changes much, it is always attracted toward stronger field, whichever way the current flows. A fixed magnet only helps on one half-cycle; a free magnet that keeps up helps on both.
Whether it keeps up is set by its natural frequency, the rate at which the field can swing it:
f0 = (1/2π) √(mB/J) = (1/2π) √(12MB / ρ(3r² + ℓ²))
For a cylinder of NdFeB (M ≈ 955 kA/m, ρ ≈ 7.5 t/m³) as long as it is wide: f0 ≈ 1.48 √(B/0.1 T) / d, with d in metres
Natural frequency falls as size grows, so the magnets have to be small. A 5 mm magnet in a 0.1 tesla station field swings at about 300 Hz, five times line frequency; a 20 mm one manages only 74 Hz.
The simulation turned up a harder problem. In a field that only alternates back and forth along one axis, a free magnet with any bearing friction does not flip with the field. It drifts sideways until it lies perpendicular to the field, where torque and attraction both vanish. That orientation is the stable one under rapid alternation. In a field that rotates, the same magnet turns with it and stays more than 95 percent aligned once f0 is at least twice line frequency. The quadrature phasing that cleans up the rotor’s torque is also what makes the pull magnets work: each one becomes a tiny synchronous rotor of its own.
Free-magnet tracking
- Natural frequency
- —
- Mean alignment
- —
Copper: field direction. Gold: magnet direction. Six cycles are shown after 18 cycles of settling. Mean alignment is the average of m·B as a fraction of perfect tracking; it measures how much of the available pull the magnet delivers. Damping ratio 0.3 stands in for bearing and eddy-current drag. Remanence 1.2 T.
Winding the mains coil
The one component that meets 7.2 kilovolts is the long mains coil. Its turn count follows from the flux it must carry each half-cycle, its wire from the current, and its loss from both.
| Peak flux linkage, √2·V/ω | 27.0 V·s |
|---|---|
| Turns | 4,502 |
| Volts per turn | 1.6 V |
| Line current | 0.84 A |
| Wire diameter | 0.60 mm |
| Wire length and copper mass | 2.25 km · 5.7 kg |
| Resistance and copper loss | 138 Ω · 98 W |
| Extra magnetizing current per millimetre of air gap | 0.19 A |
| Line current and coil loss without compensation, power factor 0.8 | 1.05 A · 153 W |
The last row is why the mains coil and its iron must stay still. Each millimetre of gap in its flux path adds about 0.19 A of magnetizing current, a quarter of the working current, and drags the power factor down. A gap-free stationary tube with iron taps keeps magnetizing current to a laminated core’s few percent. The rotor gap lives only in the station circuits, which run at low voltage on their own secondary windings.
Those circuits still draw reactive current: the stations’ leakage reactance and the rotor gap need magnetizing, and every reactive ampere they demand is reflected up through the mains coil. Capacitors on the station secondaries supply it locally instead. They are the same capacitors that set each station’s phase in the graded-turns scheme, sized now to bring each secondary near unity power factor as well. The mains coil then carries almost only the real current, which is the 0.84 A in the table; without them, at a power factor of 0.8, the coil would carry 1.05 A and lose 153 W instead of 98 at full load. The utility gains too: every house presents a near-unity load to the line.
Compensation removes reactive current, not magnetic loss. The core’s hysteresis and eddy losses are paid on flux, which the capacitors leave unchanged. At a typical 1.2 kW the saving is small, about 6 W, because copper loss is small there. The capacitors also add a resonance: tuned against the stations’ leakage inductance, it must sit well away from the third and fifth harmonics, or it will amplify them.
Steady-state stress is mild: 1.6 volts between adjacent turns. Lightning is not. Equipment on a 15 kV-class line is expected to survive a 95 kV impulse, and a steep surge front piles most of its voltage onto the first few percent of turns. That is the same problem a conventional transformer’s line-end turns face, solved the same ways: reinforced insulation on the entry turns, an electrostatic shield and a surge arrester at the terminal.
Where the energy goes
The energy passes through four conversions instead of one, and their efficiencies multiply. The upper chain is a well-built baseline unit; the lower is the same machine with the five upgrades described in the next section, drawn to the same scale.
A house rarely runs at full load. The average U.S. home draws a little over a kilowatt across the year, so the losses that run whenever the machine is energized matter most: core magnetization, bearing friction and air drag. Those are taken here as 150 W for the baseline machine, about 42 W with all five upgrades, and 35 W for a 10 kVA transformer.
Annual loss
- Turning Transformer
- —
- 10 kVA transformer
- —
- Extra energy and cost each year
- —
- Upgraded Turning Transformer
- —
- Upgraded extra energy and cost each year
- —
Each loss is a fixed part plus a load part that grows with the square of power. Baseline machine: 150 W fixed, 1,051 W total at 5 kW. Upgraded: about 42 W fixed, 424 W total. Transformer: 35 W fixed, 150 W load loss at 10 kW. Using a steady average understates load losses slightly for a real, peaky household load.
At typical use the baseline machine spends roughly 1,400 kWh a year more than a transformer, about $250 at 17 cents. With all five upgrades the gap shrinks to about 240 kWh, roughly $40. That is the price of availability. Against a multi-year wait or a several-fold price rise for a transformer, it can be the better bargain.
Closing the efficiency gap
The 1,051 W budget above splits into parts that behave differently. About 150 W is fixed: core magnetization in three magnetic circuits plus bearing and air drag, paid every hour the unit is energized. The rest is load loss — mostly copper — which grows with the square of output and almost vanishes at a household’s typical kilowatt. Five upgrades attack different parts of that budget.
- Graded station turns
- The two-group split leaves stations up to 45° away from their ideal phase. Instead, give every station its own series capacitor and enough extra turns to restore full current after the shift. Copper loss depends on ampere-turns, not on how they are divided, so the extra turns cost nothing. With shifts capped at 70°, the field becomes a nearly ideal travelling wave: station copper loss falls 18 percent and torque ripple drops from 6.6 to 1.9 percent. The capacitors handle only about 1.3 volt-amperes of reactive power for each watt of station copper loss.
- Generous copper
- Wind at 3 A/mm² instead of 5. About 60 percent more copper cuts every winding’s resistive loss by 40 percent. Small motors are designed tight because copper is expensive; a machine that runs for decades on every house can afford more.
- Amorphous-metal cores
- At 60 Hz, amorphous ribbon loses roughly a fifth as much power to magnetization as ordinary silicon steel. It saturates lower, so the cores grow about 15 percent. This attacks the fixed loss, which is why it is the strongest single change at typical household load.
- Segmented Halbach rotor
- Arranging the rotor magnets so their fields add on the air-gap side and cancel inside gives roughly 15 percent more gap flux. Torque for the same current rises by the same amount, so copper loss falls about 28 percent. Cutting each magnet into insulated segments starves the eddy currents that the residual backward wave drives in them.
- One shared rotor
- Run both machines on one rotor and one pair of grounded end bearings. The rotor need not be metal: a glass-fibre epoxy tube carries 14 N·m easily, and the sintered magnets sit in two separate groups, one under each stator, divided by a ribbed insulating spacer. That spacer keeps the isolation gap on the shaft, needing roughly 15–20 cm of clearance and creepage for a 95 kV impulse. The price is stiffness: glass-epoxy is about an eighth as stiff as steel, so the tube must be wide and short enough to keep its whirl critical speed well above 3,600 rpm.
Upgrade bench
- Full-load efficiency
- —
- Heat at 5 kW
- —
- Fixed loss
- —
- Extra over a 10 kVA transformer each year
- —
Multipliers applied to the baseline components: graded turns, station copper ×0.82 and station stray ×0.55; generous copper, all copper ×0.6; amorphous cores, all iron ×0.2; Halbach rotor, machine copper ×0.72 and stray ×0.4; shared rotor, bearing and air loss ×0.55. They are first-order engineering estimates. The coupling between upgrades — a Halbach rotor needs less iron, more copper needs a larger frame — is not modelled.
With every upgrade, full-load efficiency climbs from 82.6 to about 92 percent, and fixed loss falls from 150 W to about 42 W. At an average 1.2 kW, the yearly penalty against a transformer shrinks from about 1,440 kWh to about 240 kWh — around $40 at 17 cents. The order matters: cores first, because they cut the loss that never stops; copper and magnets next, because they only pay at high load.
Beside a pole-mount transformer
A common pole-mount unit for a handful of homes is rated 25 kVA. Manufacturers list such units at 370 to 570 pounds, about 170 to 260 kg, holding around 18 gallons of oil in a tank roughly 56 cm across and 94 cm tall (NPC Electric; RESA Power). The Turning Transformer is sized for one 5 kW house.
Per kilowatt the turning unit is heavier: about 19 kg/kW against 7 to 10 for the 25 kVA tank. The difference is structural. A transformer has one magnetic circuit; this machine has three, because the mains coil, the station motor and the alternator each need their own iron. The mains circuit cannot shrink: carrying 27 volt-seconds at 7.2 kV takes the same flux, and so the same iron, in any design. The other two are 5 kW rotating machines, and those shrink with air-gap stress and diameter. Active length falls with the square of rotor radius, so a 7 cm rotor needs half the length of a 5 cm one; a Halbach rotor raises the stress; one shared rotor drops a bearing set. A wider, shorter rotor also stores more energy, which is the next section’s subject.
What the scale buys back is the material. The tank’s weight is welded steel and oil; the turning unit’s is laminations, copper and magnets that motor factories already make by the million, in pieces small enough to lift by hand.
How each one fails
Equipment on a pole is judged as much by how it dies as by how it lives. The two machines fail very differently, and audibly.
The pole-mount tank
- Lightning or a squirrel at the bushing. A blue-white flash, a bang heard several streets away, and the fuse cutout swings open and hangs from the crossarm. Every house on the unit goes dark at once.
- Slow overload. Nothing to see or hear. Summer after summer the paper insulation cooks a little in hot oil, until one evening it fails without warning.
- Internal fault. Arcing gas pressurises the tank. Usually the relief valve spits oil; occasionally the tank ruptures and the oil burns on the pole.
- Leaks. A dark stain creeping down the pole, and a cleanup crew in the yard.
The Turning Transformer
- Bearing wear. Weeks of warning: a whine that rises in pitch, then a growl that carries to the kitchen. A vibration sensor reports it long before that. The rotor cartridge is swapped; the high-voltage module stays where it is.
- Overload or a house fault. The rotor slips a pole. The steady 120 Hz hum breaks into a rhythmic brrap-brrap as it falls out of step, the lights pulse in time with it, and protection opens within a second or two. Then the house goes quiet. The line sees only the bounded pull-out current, never a bolted fault.
- A seized free magnet. One cradle sticks. A new, slightly sour buzz, a few more watts of heat, and a little more torque ripple. Nothing goes dark.
- Loss of the line. The rotor coasts down like a jet engine spooling off, a falling whine that lasts a minute or more. Momentary dips of a few cycles never reach the house at all.
- Insulation failure across the spacer. The one that matters: line voltage reaching the house side. A grounded screen between the magnet groups and a ground-fault trip must make it end with a pop, a smell of hot epoxy and an open line fuse.
The tank usually fails suddenly and takes every house it serves with it. The turning machine fails gradually, one house at a time, and mostly by announcing itself.
Ten million small flywheels
A grid’s first defence against losing a power plant is inertia: the kinetic energy of every machine turning in step with the line. When generation suddenly falls short, frequency drops at a rate set by the power lost and the energy stored,
df/dt = ΔP · f₀ / (2Ek)
and the slower it falls, the more time governors and reserves have to respond. Inverter-based solar, wind and batteries store no rotational energy, so as they replace turbines, inertia falls. Texas’s grid operator, ERCOT, sets a critical inertia level of 100 GW·s, below which a double plant trip could outrun its frequency response; its recorded lows have been in the 115 to 130 GW·s range (ESIG; PUCT).
Every Turning Transformer is a synchronous machine. Its rotor alone, at 0.05 kg·m², stores 3.6 kJ at 3,600 rpm. A 22 kg steel disc 30 cm across adds 0.25 kg·m², bringing the unit to 21 kJ, with a rim speed of only 57 m/s. In a sealed vacuum chamber that disc costs almost nothing to keep spinning. Texas has about 10.75 million households (Census). Equipped this way, they would hold about 230 GW·s: more than the entire grid’s inertia on its leanest nights.
Grid inertia
- Flywheel disc, 4 cm steel
- —
- Energy stored per home
- —
- Inertia added to ERCOT
- —
- Initial frequency fall after losing 2.75 GW
- —
Base case: 10.75 million homes and a lean-night system inertia of 129 GW·s. The 2.75 GW loss stands for ERCOT’s two largest units tripping together. The rate is the swing-equation value at the first instant, before any governor or reserve responds.
The effect is fast and unconditional. A rotor locked to the line cannot help slowing when the line slows, and as it slows it carries part of its own house’s load, so the grid sees demand fall at the instant it needs it to. No telemetry, controller or market signal is involved. It is the same physics that makes a turbine hall steady, spread across ten million kitchens.
It has costs that must be engineered. Millions of synchronous machines can swing against each other and against the grid, so their damping matters as much as their inertia; the free magnets and station circuits must act as damper windings. After an outage they cannot all restart at once, or the returning line would meet a wall of starting current. And each unit can only give back what its pull-out torque allows: about twice rated power, for as long as its rotor has speed to spend.
What the shaft gives back
- Frequency without electronics
- A synchronous rotor turns at exactly line speed, so the alternator makes exactly 60 Hz with no inverter, controller or firmware.
- A wall in both directions
- A conventional transformer shares one core between the line and the house, so whatever couples through it passes either way. Here the only path is shaft torque, which carries slow, smooth power and little else. The harmonic currents of chargers, heat-pump drives and switching supplies reach the line as a trace of torque ripple, smoothed by the rotor. Noise and surges induced in the house wiring, which is a sprawling antenna, stop at the alternator. Line switching transients, flicker and harmonics never reach the house, which gets a fresh sine wave from its own machine.
- A steadier load
- Electronic loads draw constant power: when voltage sags they pull harder, which pushes a weak feeder further down. Behind a rotor, the house’s swings are buffered by stored energy, and the line sees a synchronous machine that naturally resists change instead of amplifying it.
- Where the wall leaks
- A few picofarads across the insulating spacer still pass the steepest edges, such as a lightning front; a grounded screen between the magnet groups drains that path. The frame, bearings and neutral must be arranged so they never bridge the gap. The line side still meets the line, so the high-voltage module keeps its surge arrester and reinforced entry turns.
- Built-in fault limit
- A short in the house can only demand the rotor’s pull-out torque. Beyond it, the rotor slips out of step and the line sees a bounded current instead of a bolted fault.
- Ride-through
- The bare rotor’s 3.6 kJ, released over a 5 percent speed sag, covers about 64 ms of full-load power, roughly four line cycles of a voltage dip. With the 0.25 kg·m² flywheel the same sag covers nearly 0.4 seconds.
What remains to prove
The figures above are first-order models, chosen to show where the machine works and where it strains. Three questions decide it.
- Starting. A synchronous rotor cannot pull itself up to speed from a 60 Hz field. It needs damper bars, as in a line-start permanent-magnet motor, or a small pony motor on the low-voltage side. Free magnets must also survive the slip while it starts.
- Station coupling. Iron taps that draw flux from one long coil into many small secondaries interact with each other. A magnetic finite-element model of a short section should confirm that each station receives its share and that the two phase groups stay near 90° apart under load.
- Voltage regulation. A permanent-magnet alternator’s voltage droops as load rises. A wound alternator field, or switched secondary taps, holds 120/240 V across the house’s load range.
A first prototype can test the whole idea at safe voltage: a 240 V mains coil, the same station geometry, and a 1 kW alternator. Torque ripple, free-magnet alignment and loss can be measured there and scaled. Only the mains coil changes on the way to 7.2 kV.