A rotary index table that stalled on the last 5 degrees of every index. The table was driven by a 1:40 worm gearbox with a 400W servo. The customer thought the servo was undersized. They had sized it from the catalog: 1:40 ratio, so output torque is 40x motor torque. That math is right in one direction and dangerously wrong in the other.

The catalog torque rating for a gearbox is the output torque at a given input speed. At 1:40, a 400W servo (1.27 Nm rated) gives 50 Nm at the output, minus efficiency. Worm gear efficiency at 40:1 is roughly 70-80%, so the useful output is about 35-40 Nm. That’s the continuous number. The table stalled because the worm gear’s efficiency collapses as speed drops — at very low speed, a 40:1 worm can be under 50% efficient, and at standstill the self-locking torque dominates. The output torque at low speed is nowhere near the catalog number.

What the catalog doesn’t say

Gearbox catalogs list output torque at a reference speed, usually 1450 rpm input (or 1500). At that speed the gearbox is at its most efficient. At lower speeds the efficiency drops, and for worm gears it drops fast. A 1:40 worm that delivers 38 Nm at 1450 rpm input might deliver only 18 Nm at 300 rpm input. The customer’s index cycle decelerated the table over the last 5 degrees — the input speed dropped to a crawl, the worm’s efficiency collapsed, and the table stalled with the servo at full current.

The servo wasn’t undersized. It was running at the wrong point on the gearbox curve. Doubling the servo size would have added cost and inertia without fixing the real issue, because the efficiency collapse is a property of the worm pair, not the motor.

What actually fixed it

Three options, we took the second.

1. Switch from worm to helical-bevel. A helical-bevel gearbox keeps 94-97% efficiency down to very low speed. The same 400W servo with a 1:40 helical-bevel delivers about 48 Nm at the output, flat across the speed range. The table never stalls. The tradeoff: helical-bevel doesn’t self-lock. The table needs a brake or a holding device to stop under load.

2. Add a second-stage reduction. We kept the worm but changed the drive to a belt stage: a 2:1 belt reduction between the servo and the worm input. The worm now runs at a higher input speed for the same table speed, keeping it in its efficient range. The belt adds 2x torque at the worm input, which more than covers the efficiency loss. The belt is cheap, quiet, and the table now indexes without stalling.

3. Change the index profile. The controller’s motion profile was trapezoidal with a hard deceleration at the end. A modified sine profile decelerates more gradually, keeping the input speed higher through the final approach. That’s a software change. It helped but didn’t fully solve it on the last degree.

The sizing rule

For worm gearboxes, size for output torque at the actual running speed, not the catalog reference speed. If your application spends time below 25% of rated input speed, multiply the catalog output torque by the worm efficiency at that speed (ask the manufacturer for the curve, or assume 60% for ratios over 30:1 at low speed). If the application reverses or indexes precisely, consider that worm backlash (1-3 degrees at 40:1) may be too much anyway.

Worm gear output torque collapses at low input speed, and the catalog doesn’t tell you. The stalling table wasn’t an undersized servo — it was a gearbox running in its dead zone. Add a belt stage, switch to helical-bevel, or soften the deceleration. Recalculate at your real speed before upsizing the motor.