A Quiet Failure on Test Station 4

On June 11, 2024, I was standing in front of Test Station 4, watching position data drift. The first label applicator had failed its 72-hour endurance run the day before. By Wednesday, six out of fifty production units had the same symptom: their print heads were slowly losing position.

I’m a quality compliance manager at an OEM that builds automated packaging lines. For the past four years, I’ve reviewed roughly 200 machines per year—electrical safety, software logic, motion logs. This was going to be a long week.

From the outside, the axis looked healthy. No hot motors, no fault codes, no loose parts. The machine simply placed labels a little further to the right every few hundred cycles. Quiet failures are the ones that scare me, because they don’t stop production until a customer notices.

The Technician Asked the Right Question

While we were pulling the cover off one of the units, a technician looked at me and asked, “What’s a stepper motor? I thought this was a servo axis.”

It was a fair question. If we had asked it a year earlier, we would have saved ourselves a lot of trouble.

Here’s the short definition: a stepper motor is a brushless, synchronous electric motor that moves in fixed increments. It has no brushes or commutator. Instead, stepper motor drivers send current through stator windings in a sequence, and the rotor follows the changing magnetic field. Each pulse moves the shaft by one step—usually 1.8 degrees, or 200 steps per revolution. Microstepping makes the motion smoother, but the idea is the same.

The critical detail is that a standard stepper motor does not report its own position. The controller counts pulses and assumes the shaft moved. That’s an open-loop system. If the load torque exceeds the motor’s available torque for even one step, the rotor stalls. The controller doesn’t know. It keeps counting pulses, and the position error silently carries forward until the machine is re-homed.

Closed-loop stepper systems with encoders exist, and they fix part of this problem. But when people ask “what’s a stepper motor?” they usually mean the classic open-loop version: pulse in, step out, no feedback. That’s what we had installed.

Why the Spreadsheet Looked Fine

The original decision wasn’t made by careless engineers. It was made in a design review with torque curves and a cost target. The motion was straightforward: move the print head along a belt, hold still while the label is printed, return. A stepper motor was the lowest-cost way to meet the required positioning at the prototype’s speed.

The spreadsheet didn’t include the things that vary in the field: cable carrier drag, belt stiffness at different temperatures, and the extra inertia of a full label roll. On the warm test floor, the torque margin looked healthy. On a cooler production line, with more drag in the cable track, the demand crossed the motor’s torque curve. The motor stalled.

Stepper torque also falls as speed increases, so the margin shrinks exactly when acceleration demand is highest. The motor wasn’t defective. The load was less predictable than the model assumed.

Here’s the part I don’t like telling: I was in that review and I approved the design. I asked whether we should add an encoder. The answer was that the driver would flag a fault. That answer was only half true. Basic stepper motor drivers amplify step and direction commands; they don’t know where the rotor is.

I’m not an electrical engineer, so I won’t pretend I tuned anything myself. From a quality perspective, the mistake was approving an assumption instead of testing the failure mode. That’s on me as much as anyone.

Looking in the Maxon DC Motor Catalog PDF

After the failure, the engineering manager asked me to help evaluate alternatives. We started with our normal drive supplier because we knew the documentation and the warranty process. On my desktop, I keep a copy of the maxon DC motor catalog PDF for exactly these moments.

The catalog is more than a parts list. It shows continuous torque, speed/torque curves, and winding options for each family. It also makes a useful distinction between brushed DC motors, steppers, and brushless DC motors. We needed to stop losing position under changing load, and we needed more speed headroom for future revisions.

A hybrid closed-loop stepper with an external encoder probably would have solved the stalling problem. But the envelope was tight, and the thermal budget in the control cabinet was already stretched. A maxon brushless DC motor from the EC family gave us much better continuous torque in the same diameter, with lower heat loss. The motor itself, however, cannot position anything by itself. It needs feedback and a controller.

So we paired the motor with an encoder and specified a servo drive for maxon motor applications—in maxon’s range, that meant an EPOS4 positioning controller. The EPOS4 handles three jobs: it commutates the motor, controls torque or speed, and closes the position loop using the encoder. That closed loop was the actual fix.

It also changed the failure mode. If the axis can’t reach the commanded position, the machine now stops and tells us, instead of quietly producing mislabeled cartons.

Validation, Because a Catalog Is Not a Test Plan

I’ve learned not to sign off on a motor change because the datasheet looks good. We put an old stepper machine and the new brushless machine side by side, then ran them through 20,000 cycles. Every 1,000 cycles, we increased the drag on the cable track to simulate a dirty or cold production line.

The stepper began to show small position errors after the fifth increase. The brushless motor held its commanded position. When we pushed it beyond its continuous torque rating, the controller stopped the axis cleanly rather than losing count. Case temperature under the same cycle dropped from 62 °C to 39 °C. Those are numbers from our test logs, not from marketing material.

If your application is well-behaved, a stepper is fine. But the point of quality testing is to find out what happens when it isn’t.

The Real Price Tag

The upgrade added about $145 in material cost per unit. On fifty machines, that’s roughly $7,250. The six failed units had already cost more than $18,000 by the time we counted rework hours, replacement parts, freight, and two field service visits. The payback was not hard to explain to management.

We converted all fifty units, not just the six that failed. As of Q4 2024, our field and final-test logs show no repeat of this failure mode. I wish I could erase the delay, but I’m glad we caught it before the machines reached customers.

What I Tell Customers Now

Stepper motors are not bad motors. There are thousands of stepper-driven axes running reliably for years. They work best when the load is known and constant, speeds are moderate, and a missed step wouldn’t cause damage. If you choose a stepper for those conditions, it can be the right engineering call.

But my one-sentence answer to “what’s a stepper motor?” is this: it’s a brushless motor that moves in discrete increments, and in its standard form, it doesn’t tell you if it actually moved. That’s not a design flaw. It’s a design assumption. You should make that assumption deliberately.

If the load varies, speeds are high, or losing position would hurt people or product, then a brushless DC motor with an encoder and a servo drive is usually the better architecture.

Now, whenever I review a motion design, I ask one question before signing off: “If this motor misses a step, how long will it take the machine to know?” If the answer is “never,” we are accepting that risk on purpose. I’d rather spend ten minutes explaining the tradeoff than explaining a recall.