A Procurement Checklist for Harmonic Drive Components: 6 Things I Check Before Every Order

If you're sourcing a harmonic-drive reducer or the individual harmonic drive components inside one, this checklist is for you. I've managed procurement for a 40-person automation integration company for the past six years. That means I've tracked roughly $180,000 in motion control spending—including every harmonic drive order we've placed since 2021. I'm the person who reads the fine print and keeps the TCO spreadsheet.

In that time, I've built a six-step checklist. Some steps are obvious if you've been buying gearboxes for years. One step, I learned the hard way after a $400 mistake and a delayed project.

Who Should Use This Checklist

Use this if you're an engineer or procurement person who needs to order harmonic drive components for a robotics joint, a semiconductor equipment stage, or a telescope mount. It's not a design tutorial. It's a checklist to run before you hit send on the purchase order. It helps especially if you've ever received exactly the part number you ordered—but not the right system.

The 6 Checks I Run on Every Harmonic Drive Order

1. Confirm exactly which harmonic drive components you're buying. From the outside, a harmonic drive looks like a single gearbox. The reality is that a typical reducer is a matched set: a wave generator (input), a flexspline, and a circular spline. Some vendors sell them as a kit; others sell the components separately. People assume all three are included in every harmonic drive quote. They're not. During a Q2 2024 vendor switch, we almost paid a separate price for the flexspline because I assumed the quote included the full set. It didn't. The invoice came in $1,700 above the original quote.

2. Calculate your real operating efficiency—don't just default to 70%. The number you'll see in most catalog material is harmonic drive efficiency 70%. That's a starting point, not a universal number. Real efficiency depends on reduction ratio, input speed, output torque, and lubrication. At low input speeds—say, a servo motor turning at 50 rpm for a telescope mount—the efficiency can be much lower than 70% because friction takes a bigger share. In an audit of our 2023 spending, we found two gearboxes that were oversized because the engineers used the 70% figure at a point where the actual efficiency was closer to 55%. If you're sizing a motor, use 70% as a first pass, then ask the vendor for the efficiency curve at your operating speed. I still kick myself for not doing that on a 2019 order. The motor we chose was undersized by about 30% at our operating point, and we had to rework the whole axis after the first build.

3. Know your input motor: servo or 3 phase induction motor. This sounds basic, but I've had vendors ask me 'what's a servo motor?' during a technical call—and I've also seen engineers assume a standard 3 phase induction motor can do the same low-speed, high-torque job as a servo. The short answer: a servo motor is a closed-loop motor with a feedback device (encoder or resolver) and a drive that commands position, speed, and torque continuously. A 3 phase induction motor is fixed-speed and open-loop. Both can drive a harmonic-drive reducer, but each changes the sizing. A servo motor can produce torque at low speed, so the harmonic drive's stiffness and near-zero backlash matter most. A 3 phase induction motor runs near constant speed, so the reduction ratio and efficiency assumptions matter more.

4. Don't forget the flexible coupling between the motor and the wave generator. This is the part I'd bet you've forgotten on at least one BOM. A flexible coupling—bellows style is common—absorbs misalignment between the motor shaft and the harmonic drive input. The input side can spin at 3,000+ rpm, and even a few thousandths of misalignment gets transmitted as vibration and bending load into the wave generator. Skip the coupling, or under-spec it, and you'll shorten the life of the harmonic drive. I learned this when I skipped the final review on a rushed order because 'it's basically the same as last time.' It wasn't. The coupling was missing from the BOM. $400 and a week of delay later, the project shipped. So glad I did a full checklist on the next order.

5. Compare total cost of ownership, not unit price. I've compared quotes from eight vendors over three months for one frame size. Vendor A's price was lower, but the quote didn't include freight or the test report. Vendor B was 11% higher, but included the test report, a spare flexspline, and a one-year warranty with a guaranteed lead time. When I calculated total cost across the project and likely downtime, Vendor B was cheaper. Vendors won't tell you this, but the first quote is often not the final price. Once you've proven you're a reliable customer, there's usually room to negotiate on price or lead time. We implemented a three-vendor quote policy in 2023 as a result. That policy has saved us roughly 8% annually.

6. Verify the zero backlash claim with actual test data. Every harmonic drive vendor says 'zero backlash.' What they mean is near-zero backlash within the rated life. Backlash can increase with wear, and 'lost motion' is usually the more honest metric. Ask for the test report from the actual unit if you can. Over the past 6 years, I've seen a 25:1 unit with measured lost motion better than 0.5 arc-min, and another 'zero backlash' unit with 3 arc-min of slop right out of the box. The 3 arc-min unit came from a supplier we no longer use. Put the acceptance criteria in the PO: 'Backlash / lost motion shall be verified and reported at the specified torque before shipment.' That one sentence is worth more than any warranty. If I remember correctly, the part number was in the 40-series size, but don't quote me on that.

Notes and Common Mistakes

If you take one thing from this, take this: harmonic drive components are a system, not a pile of parts. The wave generator, flexspline, and circular spline are designed to work together. Changing one component without understanding the effect is how projects get delayed.

• Mistake 1: Mixing flexspline and circular spline from different manufacturers. The tooth profiles need to match. Some people think any flexspline of the same size will work with any circular spline. It won't.
• Mistake 2: Assuming 70% efficiency applies at your operating point. It's an average rule-of-thumb, not a guarantee. Get the efficiency curve.
• Mistake 3: Missing the flexible coupling until assembly. It's a small line item but a big source of delay. Add it to the PO and to the lead time.
• Mistake 4: Letting a low unit price override a weak test report. If the vendor won't put backlash or lost motion in writing, move on.

On the efficiency point, I'll be honest: I don't trust any single efficiency number, including 70%, until I've seen the curve. A good vendor will send it. If they won't, that tells you something.

Efficiency is also a procurement process. If you can automate your quote comparison and approval workflow, do it. We moved from a manual email chain to a shared cost-review sheet and cut our quote-to-order time from 5 days to 2. That's the kind of efficiency that shows up in your bottom line.