Harmonic Drive FAQ: Mounting, Catalog Specs, Speed Reducers, Disc Brake Components & Bevel Gear Alternatives

If you've landed here because you're searching for a harmonic drive, a best harmonic drive mount, or even the phrase 'what uses a bevel gear,' you're probably sizing a motion system and getting lost in spec sheets. This is the FAQ I'd send to a customer before I review their first order.

1. What is a harmonic drive, and why would I use it as a speed reducer?

A harmonic drive is a compact speed reducer built around three components: a wave generator, a flexspline, and a circular spline. The wave generator elliptically deforms the flexspline; the teeth engage in a small zone; and the output moves a fraction of a turn per input revolution. Single-stage ratios typically fall in the 30:1 to 160:1 range.

What makes it useful is torque density and near-zero backlash. In a robot joint or a telescope mount, you want high reduction without adding a heavy multi-stage gear train. The trade-off, from my quality seat, is that the flexspline is a compliant component. It has a limited fatigue life, and its stiffness is lower than a large planetary gearbox. I'd choose a harmonic drive when compactness and low backlash matter more than surviving direct shock loads.

2. What is the best harmonic drive mount for a servo actuator?

The best mount is the one that keeps the wave generator aligned with the load. Not the most expensive one. Not the one with the biggest flange. The one that holds concentricity.

The field failure in March 2023 changed how I think about mount alignment. One actuator came back with a worn flexspline after roughly 800 hours. The gearbox was not defective. The mounting face on the customer's adapter plate had a flatness error around 0.02 mm on a 120 mm flange. That was enough to load one shoulder of the flexspline unevenly.

For a standard servo actuator, I specify a machined adapter plate, not a hand-finished aluminum plate, and I check the pilot bore concentricity before assembly. Why does flatness matter? Because a harmonic drive's preload is set inside the unit. Once you bolt it to a surface that pulls the housing out of shape, you've changed the preload. Use a cross-pattern torque sequence, and if the load has high moment, support it with an outboard bearing instead of using the output flange as the only bearing support.

3. What should I look for in a harmonic drive catalog?

Before you compare torque numbers, look at the conditions behind the torque numbers. I review 200+ unique spec sheets every year, and the biggest trap is a catalog that lists rated torque but does not say at which input speed and duty cycle.

Things I read before anything else:

• Rated torque at the rated input speed
• Peak torque and its allowed duration
• Maximum allowable radial and axial load on the output
• Torsional stiffness, especially for precision indexing
• Backlash class: standard vs. low backlash
• Moment of inertia at the input

I only believed the catalog-suffix lesson after approving a first article with the right ratio but the wrong output flange. That $22,000 redo came from a one-letter part number difference. Now every contract includes the full part number and the mounting face drawing. If you see a harmonic drive catalog with 'zero backlash' printed without a measurement condition, be suspicious. Backlash is measured in arc minutes under a defined torque, not in absolute terms.

4. How does a harmonic drive compare with other speed reducers?

A harmonic drive is a speed reducer, but it is not the only low-backlash option. A planetary gearbox uses multiple planet gears around a sun gear. It can handle higher shock loads and is generally stiffer in larger sizes. The downside is that reaching a 100:1 single-stage ratio is hard; you usually need multiple stages, and backlash goes up with each stage.

A cycloidal drive has rolling elements and a lobed cam. It can survive heavy impact loads and offers very low backlash, but it tends to be larger and heavier for the same reduction. If you ask me, the decision is not about which technology is universally better. It's about where the load comes from and how much mounting space you have.

To be fair, a large planetary can beat a harmonic drive on stiffness. But for a semiconductor wafer-handling arm or a telescope mount, you don't want a 3-stage planetary with 6 arc min of cumulative backlash. You want one compact stage with near-zero lost motion.

5. What do disc brake components have to do with a harmonic drive?

More than you'd think, if you're buying an integrated harmonic drive motor. Many of these actuators ship with a fail-safe holding brake. That brake usually includes a brake disc, an armature, friction lining, and a spring set. When power is removed, springs press the armature against the disc to hold the load. When power is applied, the brake releases.

During assembly reviews, I check the air gap between the armature and the brake disc. A larger air gap means a slower release and lower holding force. I also make sure the friction lining isn't contaminated with grease from the gearbox side. Grease on a brake disc is a classic root cause for load drift at power-off.

After specifying a brake version, I kept second-guessing, wondering if a separate brake would be easier to replace. But the integrated unit has become easier to verify because the brake supplier's inspection report comes with the actuator. If you buy generic disc brake components from a third party, you may save money on one part and lose it on mating tolerances. Not ideal. Prefer the manufacturer's listed brake kit or a documented equivalent.

6. What uses a bevel gear, and when would I choose it instead?

A bevel gear is used where the input and output shafts need to change direction, typically 90 degrees. Classic examples include vehicle differentials, right-angle power takeoffs, and some manual machine tools. The teeth are cut on conical blanks, which lets them mesh while the shaft axes intersect.

If your application requires a right-angle drive, a bevel gear is often the correct starting point. But a bevel gear alone does not provide high reduction. It also tends to generate axial thrust loads, and its backlash can be difficult to control unless you use ground gears and adjust shims.

The better architecture is often a bevel gear set followed by a harmonic drive. The bevel stage changes the direction; the harmonic drive provides the reduction and low backlash. So, what uses a bevel gear? Plenty. What uses a harmonic drive? Anything that needs a compact, low-backlash speed reducer on the same axis as the motor.

7. How do I pick the right harmonic drive ratio for a motion axis?

Ratio selection starts with the speed and torque at the application, not the maximum motor speed. A motor that runs at 3,000 rpm should not be geared to a load that needs 10 rpm by selecting a 50:1 drive if the reflected inertia ratio ends up too high. I look at inertia matching first: reflected load inertia divided by motor inertia. People debate the exact number, but a ratio above 10 can make tuning more difficult.

For a precision indexing table, I check the torque at the load, including acceleration torque and holding torque. Then I multiply by a safety factor for load variation. Is a catalog peak torque number useful? Yes, for short acceleration events. But you want continuous rated torque to cover the cycle-average torque. If the average torque is over 80% of rated torque, I usually recommend one frame size larger. That is not a universal rule, and I don't push it if the duty cycle is low. For an actuator that moves every few seconds, temperature rise will drive the selection more than peak torque.

8. How do I check backlash and lost motion before I approve a drive?

Per AGMA, backlash is the amount by which the width of a gear tooth space exceeds the thickness of an engaging tooth. That definition matters, but it does not tell you how to measure the drive as an assembly.

In our verification protocol, we lock the output flange, apply a reversing torque, and measure angular displacement with a rotary encoder. The result changes with the torque level. A drive that shows 0.5 arc min at 10% rated torque can show 1.2 arc min at 100% torque. So I always ask for the measurement conditions.

Granted, that's the kind of testing not every customer can do. At minimum, ask the manufacturer for backlash measurement criteria: which torque, which direction, and which output position. If a supplier says 'zero backlash' but cannot describe how it was measured, I'd treat the claim as marketing.

The other number I watch is lost motion after torque reversal, because it includes flexspline windup and bearing deflection, not just tooth clearance. For a precision positioning stage, lost motion matters more than static backlash. I'd rather take a drive with 0.8 arc min measured lost motion than one with 'zero backlash' and no test data.