2026-09-16 · Elena Markovic
The Hidden Cost of 'Probably Good Enough': Harmonic Drives, Backlash, and Last-Minute Motion Control
About once a month, I get a call that starts with a simple question: “What size VFD do I need for a 5 hp motor?” My answer usually surprises them. That’s not really the question you should be asking.
I coordinate rush applications for a precision drive manufacturer. Most of my job is triage when a machine is down, a prototype is late, or a deadline is about to be missed. In the last eleven years, I’ve handled more than 400 emergency orders: broken harmonic drives, wrong gearboxes, misapplied motors, controllers that wouldn’t talk to each other. The pattern is consistent. Someone buys a component based on power rating and price, not on mechanical behavior.
Not because the people are careless. Because they’re under time pressure. They ask the fastest question, get a fast answer, and then the machine tells them the truth: the motor has enough power, but the drive train doesn’t behave the way the control loop needs it to.
The Surface Problem: You Think You Need a Part
Last week, a customer needed to replace a gear motor in a packaging line. The gear motor had failed after a jam. The maintenance engineer ordered a replacement with the same frame size, same ratio, and similar torque. That was the fastest path back to production, so I understood the decision.
The axis ran again. But it chattered on every reversal. The servo system became unstable. The motor started overheating. The line was still down.
The diagnosis took thirty minutes. The replacement gearbox had backlash. Not much — maybe a couple of arcminutes. But a dead zone in a gearbox is not just a small error. It’s a control problem waiting for a mechanical failure. The servo loop turns the motor, the load doesn’t move, then it moves suddenly. That is the definition of a limit cycle.
That’s the gear motor trap. Gear motors are convenient. Motor and gearbox in one housing, ready to mount, easy to order. But when you are building a positioning system, you can’t select one by torque and ratio alone. You have to control gearbox backlash. Otherwise, you are buying a mechanical problem that will be blamed on the electronics.
What the ZWO AM5 and AM7 Really Show
The same confusion shows up in a different place: telescope mounts. The ZWO AM5 harmonic drive mount became popular because it offered astrophotographers a small, lightweight platform that could carry a surprising amount of payload. The ZWO AM7 harmonic drive mount price is higher, and for a reason.
Why is a harmonic drive inside those mounts? Because a mount has to track the sky for many minutes or hours, and it can’t lurch when the drive direction reverses. In an ordinary gearbox, backlash creates a small movement of the final axis. At long focal lengths, that looks like a jump. In astrophotography, a jump is not cosmetic. It ruins exposures.
A harmonic drive doesn’t remove that dead zone entirely, but it gets close enough that a servo loop can handle it. The principle is unusual: a wave generator deforms a thin flexspline so that teeth engage the circular spline at two opposite points. That creates a kinematic coupling with very low backlash and many teeth in contact. You also get a large reduction ratio in a single stage.
When people look at a premium mount and ask why it costs so much, they’re comparing it to a parts list. But what’s inside a mount like the AM7 is not a pile of parts. It’s an integrated drive train: harmonic drive gearing, servo motor, encoder, and control electronics. The hardware is only part of the cost. The tuning, testing, and support are hidden in that price.
That brings me to the real question I ask on every rush project. What is this part actually doing inside the servo loop?
The Deeper Problem: Backlash, Inertia, and the Wrong Question
When a drive train oscillates under load, or when a motor runs hot, there are two mechanical suspects: backlash and inertia mismatch. Both get ignored when people quickly replace parts.
Backlash Is Not a Minor Spec
Backlash is the lost motion between input and output when torque reverses. In a standard gearbox, the teeth need clearance for lubrication and thermal expansion. But that clearance means the output shaft can move a little before the teeth re-engage. The motor turns. The load doesn’t move. Then it jumps.
Servo drives cannot fix backlash. A servo drive sees a nonlinearity — a dead zone. Turn the gain up, and the axis oscillates. Turn it down, and the axis feels sluggish. There is no PID setting that creates stiffness where the mechanical path has free motion.
The numbers matter more than they seem. One degree of backlash at the output of a 500 mm arm creates roughly 8.7 mm of motion at the tip. One arcminute creates about 0.15 mm. That is why “small” gearbox backlash is not small when you are positioning a tool, a telescope, or a robot.
That is why designers choose low-backlash components like harmonic drives for critical axes. But let’s be precise: “zero backlash for infinite cycles” does not exist. Every mechanical system wears. Per FTC advertising guidelines, product performance claims have to be truthful and substantiated. If a vendor promises zero backlash forever, ask to see the test data. A responsible spec sheet gives you a limit, such as less than 1 arcmin for a new unit.
Inertia Changes Everything
The VFD sizing question is really an inertia question. I often get calls that start with: “What size VFD do I need for a 5 hp motor?” It’s not a dumb question. But it is incomplete.
A VFD has to deliver motor current, not just horsepower. A typical three-phase 5 hp motor at 230 V draws about 15 A at full load; at 460 V, about 7.5 A. A 5 hp VFD can be fine for a pump running at constant speed. For a load that starts under heavy torque, accelerates frequently, or runs across a wide speed range, the next size up is often the right call. Not because the motor needs more power, but because the VFD needs to handle current peaks and heat.
If you are asking that same question for a positioning axis, the real issue is somewhere else. A VFD is a speed controller for an induction motor. It is not a servo amplifier. For true positioning with low backlash, you need a motor with feedback and a controller that closes the position loop. A harmonic drive alone does not turn an induction motor into a servo motor.
Then there is load inertia. The reflected inertia seen by the motor drops by the square of the gearbox ratio. With a 100:1 gearbox, a load inertia of 1 kg·m² is reflected as 0.0001 kg·m². That is why a small motor on a high-ratio harmonic drive can control a massive telescope arm. The ratio does more than multiply torque; it transforms the load into something the servo loop can manage.
Mini Linear Actuators: Same Math, Different Axis
The same logic applies to linear motion. Mini linear actuators are popular for good reasons: they are compact, simple, and relatively inexpensive. They are also a common source of position error.
A mini linear actuator usually has a lead screw. The nut has some clearance. Under load reversal, that clearance becomes a dead zone, just like gear backlash. If you are moving a cover or adjusting a guard, that’s acceptable. If you are positioning a sensor and expecting micron-level repeatability, the screw clearance will be the limit.
The answer is not always a harmonic drive. It could be a preloaded ball screw, a linear servo motor, or a mini linear actuator with a low-backlash screw. But you have to check which spec you are actually buying: the ability to move a load, or the ability to hold a position. They are not the same.
The Real Cost Is Delay, Not Parts
This is where my work gets intense. After years of rush jobs, I can tell you that when backlash fails an acceptance test, the part isn’t the most expensive thing. The consequence is: downtime, service calls, missed events, penalty clauses.
In March 2024, a semiconductor customer called 36 hours before a factory acceptance test. A harmonic drive in a wafer handling arm had developed backlash after a crash. Normal lead time for a replacement was four weeks. The contract had a $50,000 penalty clause for late acceptance.
We had a drive with the same ratio in stock, but with a different input flange. Our applications team machined an adapter, verified the concentricity, and shipped the drive express the next morning. The rush fee? The customer barely noticed it. The penalty would have been impossible to ignore.
I’ve also seen the opposite. We lost an order a few years ago because we tried to save on freight. A customer needed a flexspline quickly. Shipping by standard Priority Mail seemed reasonable. It arrived one day late. The customer reallocated the work and never ordered from us again. The shipping method didn’t save money; it cost us a relationship. Since then, our internal policy is simple: never put a critical component on a non-guaranteed delivery option if a guaranteed option exists. USPS explains the difference between Priority Mail and Priority Mail Express on usps.com. That guarantee is worth far more than the extra postage.
You Are Buying Certainty, Not Just Hardware
When I process rush orders, I’m not selling parts. I’m selling certainty. Schedule certainty: the machine will run Monday. Acceptance certainty: the test will happen Friday.
Some people resent rush fees. I see them differently. A rush fee reflects disrupted plans, unpredictable dependencies, and risk taken by the supplier. People think rush orders cost more because they are harder. In reality, they cost more because they are less predictable. And unpredictability is expensive.
That causation matters. It’s not that expensive components are better because they cost more. It’s that a supplier who delivers a consistent, well-specified system can defend a price. If you buy the cheap option that will “probably” arrive on time, you become the risk holder. Sometimes that works out. Sometimes it doesn’t.
The Short Version: What to Do Next
The good news is that avoiding these problems doesn’t require a master’s degree in mechanical engineering. It requires starting with the behavior you need, not with the part you want to order.
- Define the positioning requirement in numbers. How much arcminute or millimeter error can the axis tolerate? If it’s below roughly 0.1 degrees, a standard gear motor probably will not get you there.
- Calculate load inertia, not just load torque. Choose the reduction ratio so the reflected inertia fits the motor. Harmonic drives are strong in this area because they offer high ratios in one compact stage.
- Size the electronics by current and duty cycle, not by nameplate horsepower. The question “what size VFD for 5 hp motor” should lead you to load torque, starting torque, and duty cycle. For positioning, use a servo amplifier, not just a VFD.
- Decide whether the device is producing motion or holding position. A mini linear actuator is great for pushing and pulling; it is less great for precision positioning unless you verify its reversal loss.
- Add supply chain buffer. If a part is critical, get a guaranteed delivery date and a backup plan. Accepting “maybe next week” is a decision, and it’s usually a bad one.
The answer is not always a harmonic drive. A gear motor might be the right choice for a conveyor. A mini linear actuator may be perfectly adequate for a vent or a simple adjustment mechanism. Harmonic drives have their own limitations: they need proper lubrication, alignment, and protection against overloads. But when you have an axis that must reverse under load and hold position, backlash is the boundary. A harmonic drive makes that boundary usable.
So next time you search for the ZWO AM7 harmonic drive mount price, or ask what VFD you need for a 5 hp motor, stop and ask a better question first. What is this axis supposed to do? Is this a speed control problem or a position control problem? How much backlash can the application truly tolerate? And what will a failure cost if the part arrives too late?
In the hundreds of emergency orders I’ve handled, the best projects started with a condition, not a component. The part came after. That is why certainty is worth the premium.