Engineering data

Harmonic drive efficiency: why the datasheet number isn't what you'll get

"85% efficient" is a true statement about exactly one operating point: rated torque, moderate speed, warm grease. Your machine spends most of its life somewhere else on the map. Here's the rest of the map.

Strain wave gears run 75–90% efficient at rated load, genuinely good for a 100:1 single stage. But engineers get burned by that number, because unlike a ratio or a tooth count, efficiency is not a property of the gearbox. It's a property of the gearbox at an operating point, and the operating points where machines actually live — light load, low speed, cold mornings — are where the datasheet number quietly falls apart.

Where the power goes

Three loss mechanisms, all speed- and temperature-dependent:

The shape of the map

Because the last two losses are roughly fixed at a given speed, efficiency depends heavily on how much useful torque you're pushing through against them:

Operating pointWhat happensConsequence
At rated torque, warmFixed losses are small next to the loadThe catalogue figure: 75–90%
At 10–20% of rated torqueFixed losses dominateEfficiency can fall below 50%
Cold start (0 °C and below)Grease viscosity multiplies dragStarting torque several times the warm value
High input speed, sustainedHysteresis + churn heat the flexsplineThermal limit before torque limit

None of this is a defect; it's the physics of any preloaded, grease-filled precision mesh. The mistake is budgeting motor torque, battery capacity, or thermal headroom from the single rated-point number.

The three torque numbers nobody reads

Buried below the headline specs, every honest strain wave datasheet carries three small torque values that answer real design questions:

Decision that follows directly: a vertical axis holding a load against gravity needs a brake, full stop. Power loss plus a back-drivable reducer means the load comes down. Several of our harmonic servo joint frames are catalogued with and without an integrated brake for exactly this reason: pick the brake variant for gravity axes.

The thermal loop

Efficiency and temperature chase each other: losses heat the flexspline, heat thins the grease (briefly helping), then sustained heat ages the grease and raises friction (permanently hurting: the wear clock from the failure-modes article ticks faster). For high-duty-cycle axes, check the steady-state housing temperature against the operating window under the real ambient, enclosure, and mounting conduction. A joint bolted to a big aluminium arm sheds heat; the same joint in an insulated shroud doesn't.

The checklist

Before you commit a motor and battery to a strain wave axis:

If you'd rather hand us the duty cycle and ambient conditions and get back a sized, quoted part number, that's the same-working-day service; it's what the RFQ form is for.

Related: How to size a harmonic drive · Reading the datasheet: backlash, lost motion, stiffness.

Budget from real numbers.

Send the duty cycle, the ambient, and the battery you have to live with. We'll size the joint against all of it — usually the same working day.

Talk to an engineer