Reaction Wheel

Turning without pushing on anything

To turn a car you push on the road; to turn a boat you push on the water; to turn an aircraft you push on the air. Every familiar way of changing direction works by shoving against something outside the vehicle. A spacecraft in orbit has nothing to push against. There is no road, no water, no air worth the name — just the vehicle and the vacuum. So how does the Hubble Space Telescope swing from one galaxy to another and then hold its aim steady enough to photograph a star for ten hours?

The answer is the reaction wheel: a heavy flywheel spun by an electric motor whose body is bolted firmly to the spacecraft. Speed the wheel up clockwise and an equal-and-opposite reaction torque acts on the spacecraft, rotating it counter-clockwise. The satellite turns by pushing against its own spinning wheel — and because nothing leaves the vehicle, it costs no propellant, only the electricity the solar panels produce for free. This is attitude control by momentum exchange, and it is the quiet workhorse behind nearly every pointing instrument in space.

The principle is conservation of angular momentum. A spacecraft drifting in orbit with no external torque on it has a fixed total angular momentum. That total can be redistributed internally — between a wheel and the body — but it cannot be created or destroyed. Spin the wheel up and the body must spin the other way so the books balance. The same physics lets a figure skater spin faster by pulling in her arms, and lets a cat right itself in mid-air without anything to push against.

The physics in one equation

The whole device reduces to Newton's rotational second law. The torque a wheel produces equals the rate of change of its angular momentum:

τ = dH_wheel/dt = I_wheel · dω_wheel/dt

Here I_wheel is the wheel's moment of inertia (kg·m²) and ω_wheel is its spin rate (rad/s). Because total angular momentum is conserved in the closed spacecraft-plus-wheel system, the torque on the body is exactly the negative of the torque on the wheel:

I_body · dω_body/dt  =  − I_wheel · dω_wheel/dt

So the body's angular acceleration is the wheel's acceleration scaled down by the (large) inertia ratio. A worked slew makes this concrete. Suppose a 500 kg imaging satellite has a body moment of inertia of I_body = 200 kg·m² about the slew axis, and a reaction wheel with I_wheel = 0.05 kg·m². We want to rotate the satellite 30° (0.524 rad) and we are willing to take 60 seconds, accelerating for the first half and decelerating for the second.

Slew angle:        θ = 0.524 rad over t = 60 s (accel 30 s, decel 30 s)
Body accel:        α_body = 4θ / t²  = 4·0.524 / 3600 = 5.8e-4 rad/s²
Body peak rate:    ω_body,max = α_body · (t/2) = 5.8e-4 · 30 = 0.0175 rad/s (≈1°/s)
Required torque:   τ = I_body · α_body = 200 · 5.8e-4 = 0.116 N·m
Wheel accel:       dω_wheel/dt = τ / I_wheel = 0.116 / 0.05 = 2.32 rad/s²
Wheel speed gain:  Δω_wheel = 2.32 · 30 = 70 rad/s ≈ 665 rpm per slew

A modest 0.12 N·m wheel torque turns the whole half-tonne satellite a degree per second, and each 30° slew costs only about 665 rpm of wheel speed — well within a 6,000 rpm budget. Note the asymmetry the inertia ratio buys you: the wheel had to change speed by 70 rad/s to move the body by only 0.0175 rad/s. The wheel does the "running"; the spacecraft does the slow, precise "turning." That gearing-down of motion is exactly why a small, fast wheel can deliver fine, slow, smooth pointing.

Storage, not propulsion — and why that bites back

It is tempting to think of a reaction wheel as a fuel-free thruster, but that mental model is wrong and dangerous. A thruster removes momentum from the vehicle permanently by throwing mass overboard. A reaction wheel stores momentum inside the vehicle. Every bit of momentum the wheel absorbs is still on board, sitting in the spinning rotor, waiting.

That distinction is the source of the actuator's defining weakness: saturation. Real satellites are never in a perfectly torque-free environment. A handful of tiny external torques act continuously:

• Solar radiation pressure. Sunlight carries momentum; on a satellite with asymmetric solar panels it produces a steady torque of order 10⁻⁵ N·m. Tiny, but it never stops.
• Aerodynamic drag. In low Earth orbit (below ~600 km) there is enough residual atmosphere to push on a satellite's leading face, offset from its centre of mass.
• Gravity-gradient torque. The near side of an elongated satellite feels slightly more gravity than the far side, twisting it toward a "long axis down" attitude.
• Magnetic torque. Any residual magnetic dipole in the spacecraft's wiring and structure interacts with Earth's magnetic field.

To hold a fixed attitude against a steady disturbance torque, the wheels must continuously absorb the incoming momentum by spinning ever faster. A constant 2×10⁻⁵ N·m disturbance, for example, adds about 1.7 N·m·s of stored momentum per day. A wheel holding 1 N·m·s at 6,000 rpm will run out of headroom in well under a day. When a wheel reaches its maximum rated speed it saturates: it can absorb no more momentum, and any further disturbance starts to turn the spacecraft.

The fix is desaturation (also called momentum dumping): apply a genuine external torque to dump the stored momentum so the wheels can spin back down to a low bias speed. There are two standard tools, and the choice tells you a lot about a mission:

• Magnetic torque rods. Electromagnets that push against Earth's magnetic field. Free (just electricity), but weak and only usable near a magnetised planet. Hubble and most LEO satellites desaturate this way, typically once or twice per orbit.
• Reaction-control thrusters. Small chemical or cold-gas jets that work anywhere, including deep space, but burn finite propellant. Interplanetary probes have no magnetic field to push on, so they spend hydrazine on desaturation — and that fuel budget often sets the mission lifetime.

Three wheels, four wheels, and a famous failure

One wheel controls one axis. Full three-axis control of a satellite therefore needs a minimum of three wheels, one each for roll, pitch, and yaw. Almost every operational spacecraft, though, carries a fourth wheel skewed off the body axes in a pyramid (tetrahedral) arrangement. The fourth wheel is pure redundancy: lose any one of the four and the remaining three can still be blended to produce torque about all three axes, so a single motor or bearing failure does not end the mission.

How much that fourth wheel matters was written across the sky by NASA's Kepler space telescope. Kepler hunted for exoplanets by staring at 150,000 stars and watching for the 0.01 percent dip in brightness when a planet transited — a measurement that demanded sub-arcsecond pointing held for years. It launched in 2009 with four wheels. One failed in July 2012; a second failed in May 2013, dropping the count below the three needed for precise three-axis control. The prime mission was over. Rather than give up, engineers devised the ingenious K2 mission: they oriented the telescope so that solar radiation pressure pushed symmetrically along its body, using sunlight itself as a "virtual third wheel" to balance the axis the dead wheels could no longer hold. K2 ran four more years and found hundreds of additional planets — a vivid lesson in both the criticality of wheel redundancy and the cleverness available when it runs out.

Inside the can

A flight reaction wheel is a sealed unit, often the size of a hatbox, containing a few essential parts:

• Rotor (the flywheel). A balanced metal disc, usually steel or a steel rim on an aluminium hub, sized so its moment of inertia gives the required momentum at the rated speed. Most of the wheel's mass sits at the rim, where it contributes most to inertia.
• Brushless DC motor. Spins the rotor in either direction, with no brushes to wear out or shed debris in vacuum. Closed-loop speed control resolves to a fraction of an rpm.
• Precision bearings. The lifetime-limiting part. Spinning at thousands of rpm for a decade or more in vacuum, the bearings rely on tiny, carefully metered amounts of lubricant. Lubricant migration and bearing wear are the most common cause of on-orbit wheel failure.
• Tachometer and drive electronics. Measure wheel speed and command motor current so the attitude-control computer gets exactly the torque it asks for.

Representative hardware spans a huge range. A 3U CubeSat might fly a wheel the size of a coin storing ~1 milli-N·m·s and massing 30 grams. A large observatory like Hubble uses wheels storing on the order of hundreds of N·m·s, each massing tens of kilograms. The same equation, H = I·ω, scales across five orders of magnitude.

The dark side: jitter

The very smoothness that makes reaction wheels good for pointing has a sharp edge. A spinning rotor is never perfectly balanced — there is always a residual static imbalance (centre of mass off the spin axis) and dynamic imbalance (the inertia axis tilted from the spin axis). These inject a small periodic force and torque, the wheel jitter, at the spin frequency and its harmonics. At 6,000 rpm the fundamental is 100 Hz, and the harmonics march up into the hundreds of Hz.

For most satellites this micro-vibration is irrelevant. For an arc-second-class telescope it can be catastrophic, smearing a long exposure. If a wheel harmonic happens to coincide with a structural resonance of the spacecraft or its instrument, the vibration is amplified and the image blurs. Mitigations form a layered defence: balance the rotors to extreme precision; choose operational speed ranges that keep all harmonics off known structural modes; mount the wheels on passive vibration isolators (soft elastomer or flexure mounts that roll off above a few Hz); and, on the most demanding instruments, add active vibration cancellation. The James Webb Space Telescope, for instance, isolates its reaction wheels to keep jitter well below its few-milliarcsecond pointing budget.

Reaction wheels versus the alternatives

A reaction wheel is one of several attitude actuators, each with a distinct niche. The table contrasts the four common choices for changing or holding a spacecraft's orientation.

Property	Reaction wheel	Control moment gyro (CMG)	Magnetic torquer	RCS thruster
Working principle	Change wheel spin rate	Tilt a constant-speed flywheel	Push on planet's B-field	Eject reaction mass
Consumes propellant	No	No	No	Yes
Torque per unit mass	Low–moderate	Very high (10–1000×)	Very low	High, in pulses
Smoothness / precision	Excellent (continuous)	Good, but has singularities	Coarse	Poor (discrete pulses)
Saturates?	Yes — needs dumping	Yes — needs dumping	No	No (limited by fuel)
Works in deep space	Yes	Yes	No (needs a field)	Yes
Typical user	Telescopes, imaging sats, CubeSats	ISS, large agile sats	Desaturation, small LEO sats	Desaturation, large slews, deep space

The pattern is clear. Reaction wheels win on smoothness, precision, and propellant economy, which is why they dominate pointing instruments. CMGs win when you need huge torque to slew a big or agile vehicle quickly — the International Space Station holds attitude with four double-gimbal CMGs storing ~4,760 N·m·s each, dwarfing what reaction wheels could supply at the same mass. Magnetic torquers and thrusters are the desaturation partners that keep the momentum-storage devices from saturating. A real spacecraft usually carries a combination: wheels for fine control, torquers or thrusters to dump momentum.

Where reaction wheels fly

• Space telescopes. Hubble (four wheels, ~7 milliarcsecond stability), Kepler, TESS, and the James Webb Space Telescope all slew and hold on reaction wheels closed around fine star sensors.
• Earth-imaging satellites. Commercial constellations (Planet's Doves, Maxar's WorldView, Airbus Pléiades) use agile reaction-wheel systems to snap a target, then quickly re-aim at the next — sometimes imaging dozens of sites per orbit.
• CubeSats and smallsats. Miniature reaction wheels, often sold as integrated three- or four-wheel modules, brought precise pointing within reach of university and startup missions on tight budgets.
• Interplanetary probes. Dawn, New Horizons, and many others carry reaction wheels for fine pointing of cameras and antennas, with thrusters reserved for desaturation since there is no magnetic field to push on far from Earth.
• Communications satellites. Geostationary comsats commonly run a momentum-bias wheel for passive gyroscopic stiffness plus reaction wheels for precise antenna pointing toward the ground.

Failure modes and trade-offs

• Bearing wear. The dominant lifetime limiter. Years at thousands of rpm in vacuum on a microscopic lubricant film eventually degrade the bearings, causing rising friction (drag torque), increased current draw, and finally seizure. Kepler's two wheel failures were bearing-related. Cures: better lubrication, run-in procedures, redundant fourth wheel.
• Saturation drift. If desaturation is missed or a torquer fails, the wheels spin up to their limit and the satellite loses attitude — sometimes tumbling into safe mode. Cures: robust momentum-management software, generous wheel speed margin.
• Jitter and resonance. Rotor imbalance couples into the structure and blurs imagery if a harmonic hits a structural mode. Cures: precision balancing, speed-range exclusion zones, isolation mounts, active cancellation.
• Zero-crossing friction (stiction). When a reaction wheel is commanded through zero rpm, static friction in the bearings causes a brief torque discontinuity that disturbs fine pointing. Cures: bias the wheels away from zero, or use a momentum-bias arrangement so no wheel ever crosses zero.
• Electronics or motor failure. A driver or winding fault disables a wheel entirely. Cure: the redundant fourth wheel and graceful three-wheel reconfiguration.

The overarching trade-off is the same one that defines the device: a reaction wheel buys you exquisite, fuel-free, continuous pointing, but only by storing momentum it must eventually be helped to give back. Design the attitude-control subsystem well and that bargain is almost free; design it badly and the spacecraft tumbles the first time the wheels fill up.