2026-08-10 · Jane Smith
Harmonic Drive Checklist: 7 Steps I Use Before Every Install (After $2,800 in Mistakes)
-
Who This Checklist Is For
-
The 7-Point Checklist
-
Step 1: Verify the Ratio and Backlash on the Actual Part
-
Step 2: Check the Efficiency Curve at the Speed You'll Actually Use
-
Step 3: Listen for the "Bad Timing Belt Sound"
-
Step 4: Confirm the Wave Generator and Flexspline Are a Matched Set
-
Step 5: Use a Rotating Torque Sensor for the Run Test
-
Step 6: Check the Linear Bearing Size Before Designing Around It
-
Step 7: Record a Backlash Baseline, Then Test It Again Later
-
Step 1: Verify the Ratio and Backlash on the Actual Part
-
What's Changed in Harmonic Drive Tech (and What Hasn't)
-
Common Mistakes When Working with Harmonic Drives
I've been handling motion control orders for seven years. In that time, I've made—and documented—12 mistakes that cost roughly $3,200. The worst ones weren't the big, dramatic failures. They were the "it looked right on the screen" ones.
This checklist is the one I wish someone had given me before I first built a harmonic-drive gearbox into a robot joint. If you're about to install or troubleshoot a harmonic drive reducer, gearbox, or integrated actuator, use it. It's seven steps, most take under an hour, and Step 4 is the one almost everyone skips.
Who This Checklist Is For
If you design OEM equipment, build robot arms, design telescope mounts, or maintain semiconductor tools, this applies. Harmonic drives are great because they are compact and have near-zero backlash. But they're also unusual: the wave generator/flexspline interaction means small installation errors show up as noise, torque ripple, or early failure.
Use this list when you're bench-testing a new unit, before you torque the housing bolts, or when a machine that ran fine starts making a noise that seems to come from the motor area.
The 7-Point Checklist
Step 1: Verify the Ratio and Backlash on the Actual Part
Don't trust the model number. Don't trust the packing list. Open the box, read the plate, and if there's a test certificate, check the number against the unit.
I made the classic spec error in my first year: I ordered a 50:1 unit, but because I only wrote "50" on the PO instead of the full model code, the vendor shipped a 30:1. It looked identical from the outside. I didn't catch it until I spun the input one full turn and the output moved 12 degrees instead of 7.2. That mistake cost me a $600 redo plus a week of waiting.
Checkpoint: before mounting, turn the input exactly one revolution and measure the output angle. Compare it to the datasheet. If the ratio is right, then check backlash at the output with a dial indicator. "Near zero" should feel like less than half a degree, depending on the size and ratio.
Step 2: Check the Efficiency Curve at the Speed You'll Actually Use
The number "harmonic drive efficiency 70%" gets thrown around as if it's universal. It's not. I tested a 50:1 unit in 2024 that ran at 74% efficiency at 1,500 rpm, but at 500 rpm it dropped to about 63%. The same gearbox, different speed, different result.
Why this matters: if you're sizing a motor for continuous duty, the difference between 70% and 63% changes your heat estimate. Ask the manufacturer for the published efficiency curve. If they give you one number and no curve, treat it as marketing.
If you've got a unit that's already installed, you can check this roughly with a motor that has known current draw and a torque sensor, but that's a bigger project. For now, just get the curve before you mount.
Step 3: Listen for the "Bad Timing Belt Sound"
A whine, a click, or a rhythmic "chirp" can sound exactly like a bad timing belt sound. I've chased that noise more than once. Sometimes it was a bad timing belt. But the more frustrating cases were harmonic drives where the belt was fine and the real problem was a failing wave generator bearing or a flexspline that had developed a crack.
Quick test: with the motor unpowered, rotate the output by hand and feel for rough spots. Then run the drive at low speed and put a stethoscope (or a long screwdriver against your ear) on the housing near the wave generator. If the noise repeats every output revolution, it's downstream of the belt.
If you hear a bad timing belt sound at a new installation, check belt tension first. Then check the flexspline. It's a 20-minute check that can save you from ordering an unnecessary belt or pulley replacement.
Step 4: Confirm the Wave Generator and Flexspline Are a Matched Set
This is the step that most people skip, because most people assume that if the parts fit, they're compatible. In a harmonic drive, that's not a safe assumption.
In September 2022, I ordered a replacement flexspline for a 40:1 unit. I kept the existing wave generator because it looked fine. It fit. For about twenty minutes. Then the drive started making a noise I now recognize as teeth skipping. The output was jerky, and the flexspline was scored beyond saving. The manufacturer eventually told me the two parts were from different production runs and the tooth geometry had been revised. I didn't know there was a revision because I hadn't checked the harmonic drive news for that series.
If it fits but it's not a matched pair, it's not installed.
So: if you're replacing anything, ask the supplier "Is this a matched set with the wave generator?" If they hedge, make them find out. And write the part numbers down together.
Step 5: Use a Rotating Torque Sensor for the Run Test
My gut once told me a gearbox was "lumpy" even though the spreadsheets looked good. I ignored it. Later, a bench test with a rotating torque sensor showed a 12% torque ripple spike every output revolution. The flexspline was damaged.
You don't need an expensive dyno for this. If you have access to a rotating torque sensor, put it between the motor and gearbox, record torque while running at constant speed, and look for repeatable spikes. Static "drag torque" tests won't catch this. The drive can feel fine when you turn it by hand and still have a problem under rotation.
If you're integrating a motor with a harmonic drive, this step is worth the extra hour. It's cheaper than a field failure.
Step 6: Check the Linear Bearing Size Before Designing Around It
Here's the random one. If you've ever googled "what size is lm8luu linear bearing", the short answer is 8mm bore, 15mm OD, and about 35mm long, depending on the manufacturer. That "or about" is the reason this is on the list: the length of an LM8LUU can vary slightly, and if you design a housing around the catalog value instead of the actual part, you'll have problems.
I once built a harmonic drive linear stage and used an LM8LUU from one supplier. The carriage was designed to the standard length. The bearing I received was 1.5mm longer than the drawing. It fit into the carriage housing, but after about 20 cycles the rail bound because the bearing jammed against the end stop. The design looked correct. The part was correct. The size was not.
Step 7: Record a Backlash Baseline, Then Test It Again Later
Harmonic drives have near-zero backlash when new. Over time, flexspline wear makes that number larger. That's normal. What's not normal is if it jumps suddenly.
Before you put the unit into service, measure and record the lost motion at the output with a dial indicator. Lock the input, apply a small torque both ways, and read the deflection. Write it on the test sheet. If you don't have a baseline, you won't be able to tell whether a "backlash problem" six months later is a real issue or just normal break-in.
I skipped this on my first unit. When the customer asked if backlash had changed at a later service, I had no answer. That didn't end well.
What's Changed in Harmonic Drive Tech (and What Hasn't)
One piece of harmonic drive news worth knowing in 2025: integrated actuators are everywhere. More companies sell a motor, wave generator, flexspline, and encoder as one package. That saves space and removes the motor-gearbox coupling step. But it also makes it harder to test components separately. If you're using an integrated unit, run the same checks, especially Step 5, because the housing hides the interface.
The fundamentals haven't changed. The wave generator still stretches the flexspline, and precision still depends on matched geometry. The execution has changed, though. What was best practice in 2020 may not apply in 2025. That's not a bad thing; it just means your checklist needs to evolve.
Common Mistakes When Working with Harmonic Drives
Three things I repeat to myself before every install:
- "Budget" is not a spec. I saved $150 on a no-name unit and paid $1,200 to replace it after a field failure.
- "Standard" means nothing until written down. Grease type, input shaft tolerance, and flexspline material all vary by vendor.
- Noise is a clue, not a diagnosis. A sound that resembles bad timing belt sound could be a belt, a bearing, or the gear mesh itself.
This checklist isn't fancy. It's the result of mistakes, redo costs, and one very long phone call. If you're installing a harmonic drive for the first time, follow these steps. If you're an old hand, maybe there's one step here you haven't been doing. I know there was for me.