The brushless motor

Inside the GO-M8010-6, behind the gearbox and in front of the driver board, sits the part that makes all the torque: a ring of copper coils facing a ring of magnets. This chapter explains how that arrangement turns, lets you drive one in a live model, and then takes the real motor apart, stator and rotor, drawn to scale.

1.1 What the module asks of its motor

Unitree rates the module, not the motor, so the motor's duty has to be worked back through the gearbox. The ratings are: peak torque 23.7 N·m at the output, maximum output speed 30 rad/s at 24 V, maximum current 40 A, bus voltage 12 to 30 V, reduction 6.33, all in a housing 96.5 mm across and 42.3 mm deep weighing about 530 g.^[1]

Dividing the torque by the ratio, the rotor must produce about 3.7 N·m at the moment of peak demand. Multiplying the speed by the ratio, it must spin up to about 190 rad/s, which is 1,800 revolutions per minute. With 21 pole pairs (section 1.2 explains the term) that is an electrical frequency of about 635 Hz, the rate at which the driver must reverse the current in every coil; it sets how thin the steel sheets must be, which part two will take up. The current limit of 40 A and the bus of 24 V bound the copper: the coils must carry that current for a moment without the enamel burning, and their resistance must be low enough that 24 V can push the current against the motor's own back-voltage at full speed.

The envelope is the harder constraint. The gearbox takes the front of the housing, the board takes the back, and the motor gets the ring in between, roughly 86 mm in outer diameter by 12 mm long by our estimate from the photographs.^[2] That shape, a thin wide ring, is why the motor is an outrunner with many poles: torque in a permanent-magnet motor scales with the square of the air-gap radius for a given length, so the design pushes the magnets as far out as the housing allows. The heat has one way out: through the stator into the aluminium hub and housing, with no fan. Unitree says the coils are dipped in varnish partly to improve that path.^[1]

1.2 How a brushless motor turns

Start with the vocabulary, because the rest of the chapter uses it without further explanation.

Stator

The part that does not move. A ring of steel teeth, separated by slots, each tooth carrying a coil of enamelled copper wire. The steel is a stack of thin sheets (laminations) rather than a solid block, for a reason we get to in part two.

Rotor

The part that turns. A ring of permanent magnets, north and south alternating, on a steel ring called the back iron that closes the magnetic circuit behind them. In an outrunner the magnet ring is outside the stator and spins around it; in an inner-rotor (or inrunner) motor the magnets are inside the stator bore.

Poles and pole pairs

Each magnet is one pole. With 2p magnets there are p pole pairs. The GO-M8010-6 has 42 magnets, so p = 21.

Air gap

The thin cylindrical space, well under a millimetre, between the tooth tips and the magnet faces. All the torque is made across it.

Phase

The coils are wired into three groups, A, B and C, fed by three currents. Which tooth belongs to which phase follows a fixed rule from the slot and pole counts, called the star of slots; the model below applies that rule and colours the teeth accordingly.

Electrical angle

One pole pair is one electrical cycle. Turning the rotor by one pair of magnets (360°/p mechanical) brings the field back to the same state, so the electronics count angles in electrical degrees = mechanical degrees × p. For 21 pole pairs, one mechanical turn is 21 electrical cycles.

The mechanism in three sentences

A current in a coil wrapped on a steel tooth makes that tooth a magnet, north or south depending on the direction of the current. Feed the three phases with currents that rise and fall in sequence, and the pattern of north and south teeth moves around the stator: a rotating field. The permanent magnets on the rotor are pulled toward the opposite teeth and follow the pattern, so the rotor turns at the speed of the pattern, and the torque that pulls it is largest when the magnets lag the field by a quarter of an electrical cycle, 90° electrical.

Notice which part moves. In the GO-M8010-6 the coils are on the inside and stay still, and the magnets are on the outside and turn: it is an outrunner. The shaft is attached to the magnet bell, not to the coils.

That last point is the whole secret of a brushless drive. In a brushed motor, a mechanical switch (the commutator) keeps the field 90° ahead of the rotor automatically. In a brushless motor nothing mechanical does it, so the driver board measures the rotor angle with an encoder, adds 90°, and sets the three currents to put the field there. The torque is then

where τ is the torque in newton-metres, I the amplitude of the phase currents in amperes, δ the torque angle between the rotor's magnet axis and the peak of the stator field in electrical degrees, and K_t the torque constant in N·m/A, a number that belongs to the motor: it collects the magnet strength, the number of turns on each coil, the gap radius and the stack length. With δ held at 90°, torque is simply K_t times current, which is why a joint driven this way can estimate its own torque from the current it measures. For this motor, a teardown measured K_t ≈ 0.26 N·m per ampere at the rotor, before the gearbox.^[2]

See it move

The model below draws a stator and a rotor with the real tooth and magnet counts, lights each tooth with the current in its coil, extracts the resulting field, and lets the magnets chase it. Start with the textbook motor of 3 coils and 2 poles, where one lap of the field is one lap of the rotor, and then switch to the GO-M8010-6's 36 teeth and 42 magnets: the same physics, folded 21 times, so the rotor turns 1/21 of a turn per lap of the field. The full version of this model, with a speed loop and timing errors, is a separate note.

Three things to notice. First, in the drawing the black arrow (the field) stays a quarter cycle ahead of the blue arrow (the rotor) whatever the load, because the controller re-aims from the measured angle every frame; raise the load and the rotor slows, but the angle holds. Second, switch to six-step drive: the field now jumps between six fixed positions, the torque angle swings between 60° and 120°, and the torque trace ripples by about 14 %. That ripple is why joint motors are driven with sinusoidal currents from field-oriented control and not with the cheap six-step method of a cooling fan. Third, compare the 12-teeth motor with the 36-teeth one: the colour pattern of the 36 is the pattern of the 12 repeated three times. The GO-M8010-6 stator is three copies of the smallest useful building block.

Why 36 teeth and 42 magnets

Two integers fix more of a motor's character than any other choice. The ratio q = Q / (3 · 2p), the number of slots per pole per phase, is 36 / 126 = 2/7 for this motor. A fraction below one means a fractional-slot concentrated winding: one coil per tooth, the coil spanning a little less than one magnet, the pattern of a drone motor rather than an industrial servo. Two consequences matter here.

The winding factor k_w says what fraction of the copper's effort becomes torque: a coil that does not sit exactly under a north and a south pole at the same time loses a little. For 36N42P the star of slots gives k_w = 0.933, recomputed for this page, not read from a table. The cogging count says how many times per turn the unpowered rotor clicks from one preferred position to the next: the least common multiple of 36 and 42, which is 252. Many shallow clicks rather than few deep ones, and the rotor turns smoothly by hand, which matters in a joint whose gearbox is too short to hide it. For comparison, the other classic family is 18 slots and 16 poles, reported for the knee motor of Unitree's G1 humanoid: q = 3/8, k_w = 0.945, 144 cogging cycles per turn. The model above lets you run either.

1.3 The motor, part by part

Now the real object. Unitree publishes the module's envelope and ratings and a few construction details; a 2025 teardown by Simplexity photographed every subassembly and measured the motor on the bench.^[2] From those two sources the drawings below are built. The outer dimensions and every count are from the sources; the internal radii are our estimates from the photographs and are labelled as such.

1.4 What we could not verify

• No source gives the stack length, the magnet grade, the wire gauge or number of turns, or the mass of the stator or rotor. The lamination diameters are the catalogue values for this motor class, not measured on a Unitree part; the magnet size and air gap are borrowed from an open-source build on the same stator; the bell and hub are drawn from photographs.
• The air gap, which the drawings show as a thin line, has not been measured by anyone.
• Electrical values are from one unit, measured once, by one team.
• The 18-slot, 16-pole figure quoted for the G1 knee motor comes from a single secondary report and is used here only as a comparison.

Sources

1. Unitree, GO-M8010-6 Motor User Manual V1.0 (2022) and product page unitree.com/go1/motor. Ratings, 6.33 ratio, stator steel, coating, impregnation, winding.
2. Simplexity Product Development, Unitree Go2 Motor Teardown (July 2025). Photographs, counts, delta winding, electrical measurements.

Part two of this chapter: the decisions behind the design, and how the stator is made. Part three: the rotor, assembly and test.