A servo motor coupled to a ball screw. The axis had 0.1 degrees of lost motion at the motor shaft. The customer thought the servo was missing steps. It wasn’t. The coupling was a jaw-type with 1 degree of backlash. At the ball screw, that 1 degree at the motor (with a 2:1 belt) becomes 0.5 degrees at the screw. That’s 0.087 mm of linear lost motion on a 10 mm lead screw. Enough to miss a position. This is how I select couplings.
The three coupling types
| Type | Backlash | Torsional stiffness | Misalignment | Use for |
|---|---|---|---|---|
| Jaw (elastomer) | 0.5-1° | Medium | Good (parallel, angular) | General motion, pumps, conveyors |
| Bellows | 0 (zero-backlash) | High | Poor (small angular only) | Servo, stepper, precision |
| Disk | 0 | Very high | Good (angular, parallel) | High-speed, precision |
| Oldham | 0.5° | Medium | Very good (parallel) | High parallel misalignment |
| Rigid | 0 | Very high | None (must be perfect alignment) | Only when perfectly aligned |
Why backlash kills servo accuracy
A servo motor is a closed-loop system. It commands a position, the encoder tells the drive where the motor is. But if the coupling has backlash, the motor can turn 1 degree before the screw moves. The drive thinks it’s at position X. The screw is at position X – 1 degree. The part misses. For a 10 mm lead screw, 1 degree at the motor is 10/360 = 0.028 mm at the screw. That’s small. But with a 2:1 belt and a jaw coupling, it’s 0.028 × 2 = 0.056 mm. Add the belt’s own backlash (0.5 degrees at the pulley), and you’re at 0.1 mm. That’s the lost motion the customer saw.
What I changed
1. Switched to a bellows coupling. The jaw coupling (1 degree backlash) became a bellows coupling (zero backlash). The bellows is a thin-walled metal tube that flexes slightly to accommodate angular misalignment (0.5 degrees max). It has no play. The motor turns, the screw turns. No lost motion. The position accuracy improved from 0.1 mm to 0.01 mm.
2. Checked alignment. A bellows coupling doesn’t tolerate parallel misalignment. I dial-indicated the motor and screw. The parallel offset was 0.05 mm. The bellows coupling tolerates 0.02 mm max. I shimmed the motor mount to bring it under 0.02 mm. The coupling doesn’t bind.
3. Sized for torque. The bellows coupling must transmit the peak torque without wind-up. The motor’s peak torque is 7.2 N·m (3x rated). The coupling’s rated torque is 10 N·m. That’s fine. But torsional wind-up (how much the coupling twists under load) matters. A bellows coupling’s torsional stiffness is about 5000 N·mm/degree. At 7.2 N·m = 7200 N·mm, the twist is 7200/5000 = 1.44 degrees. That’s not backlash — it’s elastic wind-up. It’s repeatable (the coupling twists the same way every time), so the servo’s encoder on the motor shaft sees it as a position error. The drive compensates. But if the wind-up is too much, the axis feels “springy.” For high precision, I use a disk coupling (stiffer, 20,000 N·mm/degree). The twist drops to 0.36 degrees.
The misalignment budget
Every machine has some misalignment. The question is how much, and which coupling tolerates it. I measure it with a dial indicator: mount the indicator on the motor shaft, read the runout at the screw shaft. Parallel misalignment (the shafts are offset) and angular misalignment (the shafts are at an angle). For jaw couplings, I tolerate 0.1 mm parallel and 1 degree angular. For bellows, 0.02 mm parallel and 0.5 degrees angular. For disk, 0.05 mm parallel and 1 degree angular. If the measured misalignment exceeds the coupling’s tolerance, I either realign the shafts or use a more forgiving coupling.
The coupling I use for servo axes: bellows or disk (zero backlash), sized for peak torque, with torsional stiffness high enough that wind-up is under 1 degree. The jaw coupling is fine for conveyors and pumps, not for precision positioning. The axis that missed by 0.1 mm had a jaw coupling where a bellows should have been. Zero-backlash isn’t a luxury — it’s how the closed loop stays closed.
