A Thompson Coupling is a constant-velocity (CV) shaft coupling that transmits rotation between two misaligned shafts at exactly equal input and output speeds, with no second-order velocity ripple. It pairs two Hooke (universal) joints with a passive intermediate control yoke that mechanically forces the joints to bisect the input-output angle on every revolution. The design solves the speed-ripple problem inherent to single Cardan joints, which makes it suited to high-torque drivelines where vibration and gear chatter would otherwise destroy bearings — marine sterndrives, industrial test rigs, and helicopter tail-rotor shafts being the named applications.
Thompson Coupling (CV) Interactive Calculator
Vary shaft angle and input speed to compare single Hooke-joint velocity ripple with ideal Thompson coupling cancellation.
Equation Used
The single Hooke joint comparison uses the standard velocity-ratio extrema: maximum speed ratio is 1/cos(beta) and minimum speed ratio is cos(beta). The Thompson coupling places two Hooke joints around a controlled bisecting yoke, so the second joint cancels the first joint's second-order speed ripple and the ideal output speed equals the input speed.
- Ideal Thompson control yoke holds the bisecting plane exactly.
- Single Hooke comparison uses one joint at the full working angle beta.
- Bearing play, shaft compliance, backlash, and torque wind-up are neglected.
- Angle range is limited to the article's practical Thompson range of about 25 deg.
Operating Principle of the Thompson Coupling (CV)
A single Hooke joint — the cross-and-yoke universal joint you see on every truck driveshaft — has a problem. When the input shaft turns at constant speed and the joint operates at an angle, the output speed oscillates twice per revolution. At a 20° working angle the output swings roughly ±6.4% above and below the input, which shows up as torsional vibration, gear noise, and bearing fatigue further down the driveline. The Thompson Coupling cancels that ripple by pairing two Hooke joints back-to-back and adding a passive control yoke that geometrically forces the intermediate shaft to sit on the bisecting plane between the input and output axes. When the bisecting condition holds, the velocity error of the first joint is exactly cancelled by the second, and the output shaft tracks the input to within parts-per-million homokinetic accuracy.
What makes the Thompson different from a conventional double Cardan is the control mechanism. A double Cardan relies on a centring ball or a sliding yoke that approximates the bisecting plane — close enough for low-angle automotive driveshafts but not exact, especially under load. Hugh Thompson's design uses four spherical bearings on a control yoke that mechanically constrain the intermediate yoke positions, so the bisecting condition holds at any angle up to roughly 25° and at any torque level the bearings can carry. If the control-yoke pivot bushings wear or the spherical bearings develop radial play above about 0.05 mm, the bisecting plane drifts under load and second-order velocity ripple reappears — usually first detectable as a 2× rotation-frequency vibration at the output flange. That is the failure mode you watch for.
The geometry is unforgiving on machining. The two Hooke joint trunnion centres must be coplanar with the control-yoke pivots to within about 0.025 mm, otherwise the homokinetic condition is only approximated. Get that wrong and you have an expensive double Cardan, not a Thompson Coupling.
Key Components
- Input Yoke: The driving fork that transfers torque from the input shaft into the first Hooke cross. Trunnion bores typically held to H7 fit with the cross bearings, runout under 0.02 mm at the flange face. This is the sole torque entry point and sees the full applied torque without speed averaging.
- First Hooke Cross (Spider): A four-trunnion cross with needle-roller bearing cups carrying the input-to-intermediate angular displacement. Each trunnion is ground to within 0.005 mm diameter tolerance to maintain bearing preload. Operating angles up to 25° are typical; beyond that, needle bearing life drops sharply due to oscillation amplitude.
- Intermediate Yoke Pair: Two yokes joined by a short shaft section that carries motion from the first cross to the second. This element experiences the speed ripple internally — its instantaneous angular velocity oscillates at 2× shaft frequency — but the ripple is invisible at the output because the second joint cancels it.
- Control Yoke: The defining feature of the Thompson Coupling. A passive linkage with four spherical bearings that geometrically constrains the intermediate yokes so they always lie on the bisecting plane between input and output axes. Bearing radial play must stay under 0.05 mm or homokinetic accuracy degrades.
- Second Hooke Cross: Mirrors the first cross. Receives the rippling motion from the intermediate yoke pair and converts it to constant-velocity output because the control yoke has positioned it precisely on the bisecting plane. Identical bearing tolerances to the first cross.
- Output Yoke: Delivers constant-velocity rotation to the driven shaft. Output flange runout under 0.02 mm. With the control yoke working correctly, output angular position tracks input to within a few arc-seconds at any working angle within the rated range.
Real-World Applications of the Thompson Coupling (CV)
The Thompson Coupling sits in a narrow but valuable application space — drivelines that need constant-velocity transmission at angles too large for a Rzeppa or tripod CV joint, at torque levels that would destroy a plunge-style automotive CV, and in environments where double-Cardan approximation error is unacceptable. You see it where speed ripple causes structural resonance, where the driveline runs at a permanent fixed angle, or where the torque is too high for a homokinetic joint of the ball-and-cage variety.
- Marine: Volvo Penta IPS-style sterndrive applications and Thompson Couplings Ltd installations on commercial workboats, where the engine sits on a soft mount and the propshaft runs at a permanent 8–15° angle to the engine output flange.
- Aerospace: Helicopter tail-rotor driveshafts on platforms where the tail boom flexes under flight load and a single Cardan would inject 2P vibration into the airframe — Thompson-type CV couplings have been evaluated as alternatives to flexible disc couplings on Sikorsky and Bell research platforms.
- Industrial test equipment: Engine and gearbox dynamometer drivelines where the unit under test moves on a soft mount and the dyno shaft must remain at constant velocity to avoid corrupting torque-ripple measurements at the load cell.
- Heavy machinery: Articulated mining trucks and underground LHD loaders where the driveline crosses the articulation joint at angles up to 20° and conventional double Cardans need frequent rebuild due to phasing wear.
- Defence vehicles: Tracked-vehicle final drives and turret drives where torque levels exceed Rzeppa joint capacity and the driveline must maintain constant velocity to avoid stabilised-platform jitter.
- Renewable energy: Tidal turbine power take-off shafts where the rotor nacelle yaws relative to the seabed-mounted generator and constant velocity is required to avoid generator-side torque pulsations.
The Formula Behind the Thompson Coupling (CV)
The single equation that justifies the Thompson Coupling's existence is the velocity ratio of a single Hooke joint as a function of operating angle. This is what the Thompson cancels. Computing it tells you how much speed ripple you would have without the cancellation — and therefore how much vibration energy the control yoke is actually removing from your driveline. At low angles below about 5° the ripple is under 0.4% and you can usually live with a single Cardan. At nominal mid-range angles of 15–20° the ripple climbs into the several-percent range and structural resonance becomes a real risk. Above 25° you are off the map for any Hooke-based design and need a different topology entirely.
Variables
| Symbol | Meaning | Unit (SI) | Unit (Imperial) |
|---|---|---|---|
| ωin | Input shaft angular velocity | rad/s | RPM |
| ωout | Output shaft angular velocity (single Hooke joint, before cancellation) | rad/s | RPM |
| β | Operating angle between input and intermediate shaft axes | degrees or radians | degrees |
| θ | Instantaneous input shaft rotation angle | degrees or radians | degrees |
| Δωpeak | Peak-to-peak velocity ripple amplitude — what the Thompson cancels | % of input speed | % of input speed |
Worked Example: Thompson Coupling (CV) in a hydraulic dynamometer driveline
You are commissioning a 750 kW hydraulic dynamometer driveline at an engine test cell. The engine output flange sits on rubber isolation mounts that allow up to 18° of static angular offset to the dyno input shaft, and the test programme demands torque-ripple measurements clean enough to resolve a 1% combustion variation. Input speed is 3000 RPM. You need to know how much speed ripple a single Cardan would inject — and therefore why the Thompson Coupling is on the bill of materials.
Given
- ωin = 3000 RPM
- βnom = 18 degrees
- βlow = 5 degrees
- βhigh = 25 degrees
Solution
Step 1 — for a single Hooke joint, the peak-to-peak velocity ripple as a fraction of input speed reduces to:
Step 2 — at the nominal 18° operating angle, evaluate the ripple amplitude:
That is a 300 RPM peak-to-peak swing on a 3000 RPM input — twice per revolution, at 100 Hz. On a dyno bedplate that frequency is right in the band where structural modes live, and the torque cell will see it as a phantom 100 Hz combustion artefact that completely masks the 1% combustion variation you are trying to measure.
Step 3 — at the low end of the range, β = 5°, the ripple drops sharply:
Below 5° you can usually run a single Cardan and live with it — most light truck propshafts operate here. At the high end of the design range, β = 25°:
Nearly 20% peak-to-peak ripple. At 3000 RPM that is a 590 RPM swing twice per revolution and the torsional acceleration alone will fatigue-crack a stub shaft inside a few hundred hours. The Thompson Coupling cancels all of this by holding the intermediate yoke on the exact bisecting plane.
Result
At the nominal 18° operating angle, a single Hooke joint would inject roughly 10. 0% peak-to-peak speed ripple at 100 Hz into your dyno driveline — completely unacceptable for 1%-resolution combustion measurements. The Thompson Coupling reduces this to a few arc-seconds of residual position error, effectively zero. Across the operating range, ripple climbs from 0.76% at 5° to 10% at 18° to nearly 20% at 25°, so the engineering payoff scales nonlinearly with angle — the Thompson earns its cost above about 10°, becomes mandatory above 15°, and is the only Hooke-based option above 20°. If after installation you still measure a 2× rotation-frequency vibration at the output, the most likely causes are: (1) control-yoke spherical bearing radial play exceeding 0.05 mm letting the bisecting plane drift under load, (2) phasing error between the two Hooke crosses where the trunnions are not aligned within 0.025 mm coplanarity, or (3) intermediate shaft length that does not match the design nominal — even a 1 mm length error throws off the bisecting geometry at high angles.
When to Use a Thompson Coupling (CV) and When Not To
The Thompson Coupling is not a default choice. It is more expensive and more complex than a single Cardan, heavier than a Rzeppa CV, and overkill for most automotive applications. You pick it when ripple cancellation must be exact, torque is high, and operating angle exceeds what a ball-type CV can handle. Here is how it stacks against the realistic alternatives.
| Property | Thompson Coupling | Double Cardan Joint | Rzeppa CV Joint |
|---|---|---|---|
| Maximum operating angle | ~25° | ~30° (but only approximately CV near zero) | ~47° |
| Speed ripple at 18° | ~0% (true homokinetic) | ~1–3% residual | ~0% (true homokinetic) |
| Torque capacity (typical) | High — limited by Hooke cross bearings, scales to thousands of Nm | High — same Hooke cross bearings | Moderate — limited by ball-cage contact stress |
| Relative cost | Highest — precision control yoke and matched bearings | Moderate | Low — mass-produced automotive part |
| Complexity (part count) | High — 4 spherical bearings plus 2 crosses plus control linkage | Moderate — 2 crosses plus centring ball | Low — single ball-and-cage assembly |
| Lifespan at rated load | 10,000+ hours typical industrial duty | 5,000–8,000 hours, sensitive to phasing wear | 3,000–5,000 hours under fixed angle |
| Best application fit | High-torque fixed-angle CV drives, dynos, marine, defence | Truck and 4x4 driveshafts at moderate angle | Automotive halfshafts with plunge motion |
Frequently Asked Questions About Thompson Coupling (CV)
The double Cardan is only constant-velocity at the centring ball — the ball mechanically forces the two yokes to a roughly equal angle, but only at that single point and only under low load. Under torque, the ball loads asymmetrically and the bisecting plane drifts by a fraction of a degree, which reintroduces a small second-harmonic ripple of typically 1–3% at 18°.
The Thompson uses four spherical bearings on a passive control yoke that geometrically constrain the bisecting plane regardless of torque. The constraint is geometric, not frictional — that is the difference. Truck driveshaft catalogues call the double Cardan a CV joint because at near-zero angle it behaves like one, not because it is homokinetic at the angles where it actually lives.
Once you have ruled out control-yoke play and trunnion coplanarity, the next suspect is unequal yoke phasing on assembly. The two Hooke crosses must be installed with their trunnion planes oriented correctly relative to the control yoke — get one cross 90° out and you have built a velocity-error doubler instead of a canceller. This is a surprisingly common assembly error because the cross looks symmetric on the bench.
The second suspect is intermediate shaft torsional resonance. The Thompson removes velocity ripple at the output, but the intermediate shaft itself still oscillates internally at 2× frequency. If that frequency hits a torsional natural mode of the intermediate shaft, you can excite a resonance that radiates structurally even though the output is kinematically clean. Check whether your 2× excitation frequency lines up with the intermediate shaft's first torsional mode.
A phased Z-driveline — two single Cardans with the intermediate shaft yokes in-phase and the input/output shafts parallel — does cancel ripple geometrically, and it is much cheaper than a Thompson. The catch is that it only cancels when the two operating angles are exactly equal. In a fixed installation that is achievable; in any application where the relative geometry changes (vehicle suspension travel, marine engine mount deflection, articulated machinery), the angles diverge under load and the cancellation breaks down.
Pick the phased Z when geometry is rigid and well-defined. Pick the Thompson when the geometry can flex, when you cannot guarantee parallel input and output, or when the cost of intermittent ripple — fatigue, instrumentation noise, NVH complaints — exceeds the price delta. As a rough rule, if your installation can hold input/output parallelism within 0.5° under all load conditions, a phased Z works; if not, the Thompson earns its keep.
The cancellation argument relies on the two Hooke joints' velocity-error functions being exactly 180° out of phase. That phase relationship is set by the geometric position of the trunnion centres relative to the control-yoke pivot plane. If the four trunnion centres are not coplanar with the control-yoke pivots, the second joint's error function is shifted slightly in phase from the first joint's, and they no longer null perfectly.
At 0.025 mm coplanarity error you get a few arc-seconds of residual position error — invisible in practice. At 0.1 mm, the residual climbs to a noticeable fraction of a percent of speed ripple, and the whole point of specifying a Thompson over a double Cardan disappears. This is why the machining is the expensive part — not the bearings, not the materials, but holding that coplanarity through final assembly.
The 25° limit is not arbitrary kinematic — the homokinetic condition still holds geometrically up to roughly 35–40°. The limit is the needle-roller bearing inside each Hooke cross trunnion. At high operating angles each needle oscillates over a larger arc twice per revolution, and above about 25° the oscillation amplitude exceeds what needle bearings tolerate before false-brinelling and pitting begin.
If your application genuinely needs 30°+ at constant velocity, you are out of Hooke-based topology entirely and should be looking at a Rzeppa or a tripod CV joint sized for the torque. Pushing a Thompson to 30° will work for tens of hours but the cross bearings will be scrap before the end of a normal duty cycle.
It cancels the kinematic speed ripple completely, and because torque ripple in a Hooke joint is the inertial reaction to the speed ripple (T = J × dω/dt of whatever rotates downstream), removing the speed ripple removes the inertia-driven torque ripple along with it. That is the whole reason it shows up on dynamometer drivelines.
What it does not cancel is friction-driven torque variation inside the joint itself — the small second-harmonic component caused by needle bearing rolling resistance varying with trunnion angular position. This is typically two orders of magnitude smaller than the kinematic component and only matters in extremely high-resolution torque measurements. For a 1% combustion-variation dyno measurement you will not see it; for a sub-0.01% precision metrology drive you might.
References & Further Reading
- Wikipedia contributors. Constant-velocity joint. Wikipedia
Building or designing a mechanism like this?
Explore the precision-engineered motion control hardware used by mechanical engineers, makers, and product designers.
