Mechanical Engineering
Hooke Joint Velocity Fluctuation
The cyclic speed error every U-joint produces — and how pairs cancel it
A Hooke (Cardan) joint transmits torque between angled shafts but introduces a sinusoidal velocity error twice per revolution. Pairs of joints in phase cancel it perfectly.
- Error frequency2 × shaft RPM
- Peak ratio1 / cos(β)
- Trough ratiocos(β)
- β = 20° error~6% peak
- β = 30° error~15% peak
- AnalysedHooke, 1676
Interactive visualization
Press play, or step through manually. The visualization is yours to drive — try it before reading on.
Watch the 60-second explainer
A condensed visual walkthrough — narrated, captioned, under a minute.
The exact formula
For a single Hooke joint with input shaft angle θ₁ (measured around its own axis), input angular velocity ω₁, and shaft-angle misalignment β between input and output axes, the output shaft's instantaneous angular velocity is:
ω₂ / ω₁ = cos(β) / (1 − sin²(β) · cos²(θ₁))
Two extrema occur every half revolution of the input. When cos²(θ₁) = 0 (θ₁ = 90° or 270°), the denominator becomes 1 and ω₂/ω₁ = cos(β) — the minimum. When cos²(θ₁) = 1 (θ₁ = 0° or 180°), the denominator becomes 1 − sin²(β) = cos²(β) and ω₂/ω₁ = 1/cos(β) — the maximum.
The peak-to-peak fluctuation is therefore 1/cos(β) − cos(β) = (1 − cos²β)/cos β = sin²β / cos β. This grows quickly with β: at 10° about 1.5% peak, at 20° about 6%, at 30° about 15%, at 45° an unworkable 50%.
Why two cycles per revolution
The Hooke joint's spider is a four-armed cross. Two opposite arms pivot in the input yoke; the other two opposite arms pivot in the output yoke. Each pivot is a hinge that allows the joint to bend.
As the input yoke rotates, the spider's input-yoke pair of arms sweep a circle in the input shaft's transverse plane. The other pair — locked into the output yoke — must sweep a circle in the output shaft's transverse plane. The two planes are tilted relative to each other by β. The projection from one plane to the other is what introduces the velocity error: when the spider's "long" axis lies along the line of nodes (intersection of the two planes), the geometry maps differently than when it lies perpendicular.
The four-fold symmetry of the cross means the same projection mismatch repeats every quarter revolution of the spider — but the spider rotates at the average of the two shaft speeds, so one full rotation of the spider corresponds to one full rotation of the input shaft, producing two error cycles per shaft revolution.
Worked example: β = 20°
A rear-wheel-drive truck driveshaft sees about 20° of angular misalignment at full unloaded suspension droop (the rear axle hangs low; the transmission output is fixed by the chassis). Let's quantify what a single Hooke joint at 20° does to a driveshaft turning at 3000 rpm.
- β = 20°, so cos β ≈ 0.9397, sin β ≈ 0.342.
- Peak ratio = 1/0.9397 ≈ 1.0642.
- Trough ratio = 0.9397.
- Peak-to-peak = 0.1245, or roughly ±6.2% about the mean.
- Input speed: 3000 rpm = 50 rev/s; error frequency: 100 Hz.
- If input angular velocity is 314 rad/s, output oscillates between 295 rad/s and 334 rad/s at 100 Hz.
The angular acceleration is the time derivative of ω₂: peak value ≈ ω₁² · sin²β / cos β ≈ 314² × 0.117 / 0.94 ≈ 12,300 rad/s². On a driveshaft inertia of 0.02 kg·m², that's an oscillating torque of 246 N·m at 100 Hz — easily detectable as cabin vibration.
Cancellation with two joints
Two Hooke joints in series can cancel the velocity error if three conditions are met:
- Equal joint angles. Joint A operates at angle β₁; joint B at angle β₂. Cancellation requires β₁ = β₂.
- Intermediate shaft yoke phasing. The intermediate shaft has a yoke at each end. These yokes must lie in the same plane (or, by some conventions, rotated 90° — depends on which arm is considered "lead"). Get this wrong and the second joint adds error rather than subtracting.
- Parallel input and output axes. When β₁ = β₂, the input and output shafts are parallel (offset, but parallel). This is the "Z-configuration"; the alternative "W-configuration" has the two joints angled the same way and the input-output axes intersect at the centre.
Most truck and RWD car driveshafts are Z-configuration: the gearbox output, the centre bearing, and the differential input are arranged in a shallow Z. The two joint angles are equal in magnitude, the intermediate yokes are phased correctly, and the output velocity equals the input velocity exactly — at every instant, not just on average.
Common installation errors
- Phasing wrong by 90°. The yokes are mounted with their pins crossed instead of aligned, so the second joint amplifies the first joint's error to almost 12% peak rather than zero. Symptom: violent driveline shudder. Cure: pull the driveshaft, re-clock the slip-yoke spline so the front and rear U-joint pivots align in the same plane.
- Unequal joint angles. The transmission tail-shaft is angled down 5° and the differential nose is angled up 8°. The two joints' velocity errors don't cancel because they're different magnitudes. Symptom: residual 2-per-rev vibration at highway speed. Cure: shim the differential mount until the angles match.
- Tail-shaft droop. The driveshaft is straight when the truck is unloaded but bends at large angles when fully loaded — and the angle at the front joint may differ from the angle at the rear. Some setups use a hinged-Z driveshaft layout to keep the angles matched across the load range.
- Wrong U-joint orientation. Modern U-joints have a directional installation — the grease nipple should point toward the driver to allow service access. Some mechanics install backward, and while the joint works mechanically, lubrication intervals get missed.
Hooke joints vs CV joints
| Single Hooke | Two Hookes (paired) | Rzeppa CV | Tripod CV | Double-cardan | Disc / flex coupling | |
|---|---|---|---|---|---|---|
| Velocity error | 2ω at all angles | 0% when phased | 0% | 0% (small axial pulse 3ω) | 0% | Damped, not zero |
| Max angle | 30° (1 joint) | 15–20° per joint | 47° | 26° | 30° | 5° |
| Plunge | Slip yoke external | Slip yoke external | Limited (plunging Rzeppa) | ±25 mm | None | Slight |
| Cost | $ | $$ (2 joints + intermediate) | $$ | $ | $$ | $ |
| Common use | Industrial PTO, low-angle shafts | RWD/truck driveshafts | FWD outboard halfshafts | FWD inboard halfshafts | 4WD transfer-case shafts | Pump couplings |
| Service | Grease nipple, easy | Two nipples | Boot + grease pack | Boot + grease pack | Greaseable nipple | None |
When you must pair joints
Rules of thumb from drivetrain practice:
- β < 3°: single Hooke joint is fine; error is below 0.2%.
- 3° < β < 8°: single Hooke joint acceptable for tractor PTOs and slow shafts; risky for high-speed automotive.
- 8° < β < 15°: use two Hookes in phase; check angles match within 1°.
- 15° < β < 30°: double-cardan joint or CV joint; the centring yoke of the double-cardan does the phasing automatically.
- β > 30°: CV joint mandatory; Hooke pairs can't be packaged with intermediate shafts in cars at this angle range.
Frequently asked questions
What is the formula for Cardan error?
If a Hooke joint operates at angle β between the two shaft axes, and the input shaft turns at constant angular velocity ω₁, then the output angular velocity ω₂ follows ω₂ / ω₁ = cos(β) / (1 − sin²(β)·cos²(θ)) where θ is the input shaft's angular position. The ratio oscillates between cos(β) (twice per revolution) and 1/cos(β) (twice per revolution), with peak-to-peak fluctuation of (1 − cos²β)/cos β. At β = 20° this is about 6% peak; at 30° about 15%; at 45° it climbs to 50%, dangerously large.
Why does it oscillate twice per revolution?
Picture the spider's two pivot pairs. When the input yoke is horizontal, the spider pivots are in the bisecting plane and the output yoke aligns with it. Quarter-turn later the geometry has rotated so the spider's other axis lies in the bisecting plane, and the projection from input rotation to output rotation has changed. The cycle repeats every half input revolution because of the four-fold symmetry of the cross, hence twice per revolution. The mathematical version is the cos(2θ) term that emerges when you solve the kinematics.
How do two joints cancel the error?
Two Hooke joints in series — input shaft → joint A → intermediate shaft → joint B → output shaft — produce cancellation if three conditions hold: (1) the input/output shaft axes are parallel to each other (or, equivalently, the two joint angles are equal in magnitude); (2) the intermediate shaft's two yokes are in the same plane (phased 0° between them is wrong — they must be aligned with each other or 90° depending on convention); and (3) the joint angles are equal. When all three hold, joint B introduces exactly the opposite velocity error to joint A, leaving the output shaft with constant velocity equal to the input shaft's. Most truck driveshafts approximate this; getting the angles unequal is the most common installation error and reintroduces vibration.
Why is it called Hooke's joint?
Robert Hooke described and analysed the four-pivot cross-coupling in his 1676 Treatise. Gerolamo Cardano had sketched a similar device in 1545, but Hooke gave the rigorous mathematical treatment, including the velocity-fluctuation formula. In English the device is usually called the Hooke joint or universal joint; in continental Europe, the Cardan joint or Kreuzgelenk. Engineers use 'U-joint' colloquially. All three names refer to the same geometry.
At what angle does Cardan error become a problem?
Below about 3°, the peak velocity error is under 0.2% and inconsequential. From 3–10° it grows to 1.5% and starts to excite mild torsional vibration in long shafts. Beyond 15° (peak error ~4%) you must use two joints in phase or switch to a constant-velocity joint. At 30° the error is 15% peak-to-peak and a single joint is unusable. Truck driveshafts are designed for 3–8° at each joint; cars at full suspension travel might see 10°. CV joints at front-wheel steered axles handle up to 47° because they have zero velocity error by geometry.
Can the velocity error be useful?
Surprisingly yes — some mechanisms exploit Cardan error rather than cancelling it. The Newall reducer uses a single Hooke joint at large angle to convert constant input rotation into a deliberately fluctuating output, which then drives a mechanical filter or pulser. Old artillery sighting systems used the same trick. Most engineering applications, though, treat the fluctuation as pure parasitic loss and design it out with paired joints or CV couplings.