The part number that looked right

I've been handling motor and drive orders for eight years. In that time, I've personally made—and documented—14 significant mistakes that cost roughly $32,000 in wasted budget. This one is the reason I keep a checklist. I keep a folder called 'post-mortems' with a note for each one. This one has a photo of a worn commutator and a service report I still reread before motor selections.

In March 2021, our team was quoting a small rotary indexing station for a packaging line. The customer wanted a 24 V motor, a gearhead, an encoder, and enough torque to rotate a 3-kg tray in 90-degree increments. They didn't ask for high speed or huge margin. The duty cycle seemed light: two seconds of motion, then a 10-second pause while a part was placed. The customer had burned through two other suppliers before coming to us, so the quote mattered more than the sticker price. We wanted to look like the easy team to work with.

My first instinct was to compare ac servo motors and a closed-loop stepper motor controller. Both felt like overkill for the load, and the target price was tight. So I opened the maxon-motor catalog and picked a 25-mm maxon brushed dc motor with a 64:1 planetary gearhead. The datasheet showed more than enough torque, and the encoder made the positioning scheme straightforward.

That should have been the end. It wasn't. The order was for one prototype followed by nine production units. The first unit was supposed to be the easy one.

What the datasheet didn't tell me

The prototype ran beautifully. We cycled it 1,000 times in the shop with no load, then 200 cycles with the tray. Repeatability stayed inside 0.1 degrees, and the motor barely got warm. We shipped it, and for about two weeks everything looked fine.

Then the field service report came back: Intermittent stall at position 4, grinding noise during index.

My first thought was the gearbox. I fell into a 'how ball bearing made' search hole that afternoon—cold heading, flash grinding, heat treating, lapping, then sorting by ISO 3290 grade. I learned that bearing noise can be caused by a contaminated race or a marginal preload. But the motor bearings were fine. The gearhead was fine. The problem was the motor itself.

Bearing manufacturing reference: high-precision balls are cold-headed from wire, ground, heat-treated, and lapped to a tolerance often under 1 micron. Final sorting is done by ISO 3290 grade. Source: ISO 3290-1 and Machinery's Handbook, accessed January 2025.

Here's what I'd missed: the application didn't just move in 90-degree steps. It also held the tray under power during those 10-second pauses. In a brushed dc motor, continuous stall current at the same rotor position means the brushes dwell on the same commutator bars. Over time, that concentrates heat and brush wear in one small zone.

The maxon brushed dc motor I'd chosen was designed for intermittent motion, not for holding torque with the armature locked for extended periods. The datasheet included a torque-speed curve and a life note—I just hadn't connected it to my duty cycle. I was too focused on peak torque and encoder resolution.

The most frustrating part: the motor passed every static test. You'd think a torque-speed curve would be enough, but duty cycle changes everything. I had chosen the motor because it was a well-built Swiss product. It was. The failure was my application engineering, not the hardware.

What made it worse was that I had considered the alternatives. An ac servo with a holding brake would have been far more robust. A stepper motor controller with reduced-current idle would have held the position without overheating. Both were in the maybe column. I chose the brushed motor because it was smaller and lower-cost, and I ignored the one feature that mattered most. It wasn't a lazy choice; it was a numbers-first choice. The torque and speed curves checked out. My mistake was stopping at the curves.

When I spec'd the ac servo motors, the smallest unit I could find was a 100 W servo with a separate drive and brake. The customer's control cabinet didn't have room for another 240 V driver. The stepper motor controller option required a 48 V supply and a better PLC output card. Both were solvable, but they weren't as simple as the 24 V brushed motor on the spreadsheet. Simplicity on paper doesn't mean simplicity in the field.

Part of me wanted to blame the datasheet. Another part knew I hadn't read it carefully enough. That's when the field failure changed how I think about motor selection.

I had mixed feelings about the rework. On one hand, it was a genuine design mistake. On the other, the replacement used a more complex architecture, and that complexity paid off. That experience forced me to separate 'looks right' from 'is right.'

The rework

The replacement was not a bigger maxon motor; it was a NEMA 17 closed-loop stepper motor controller package with a 20:1 planetary gearbox. The stepper's idle current reduction meant it could hold the tray without cooking the windings, and the controller's encoder input gave us the position confirmation we needed. It wasn't as compact, but it was the right tool.

The driver was set to 70% current during the 10-second hold, just enough to keep position with the brake off. The first production run completed 20,000 cycles without a stall. I'm not going to pretend the original maxon brushed dc motor could never handle this duty; the issue was that I hadn't verified it for that specific mode. 'Probably fine' is how bad purchases happen. The swap also required a new mounting bracket and a slightly bigger enclosure. The customer approved the change without a redesign fee, which tells you how much they valued reliability over size.

The cost of my mistake: $1,380 for the replacement and freight, plus about 2.5 days of field service labor, including a Saturday drive to the customer site at triple overtime. Total roughly $2,900, depending on what I count. The customer's deadline slipped by a week, and my credibility took a hit that no purchase order can show. The accounting system shows the first two numbers; the trust cost doesn't show up anywhere.

The checklist that caught the next 47 mistakes

A few months later, I started scheduling a weekly check of maxon motor news today—product alerts, datasheet revisions, end-of-life notices. That's how I learned about a brushless EC version in the same frame size with a longer life rating. It would have been a better fit, but the release was announced after our project shipped. The lesson stuck: a part number is a decision, not a default.

After that project, I wrote a 12-point motor-selection checklist. It's not clever. It's just the things I forgot:

  • Duty cycle: Is the motor moving, holding, or both? If holding, how much current at standstill?
  • Motor family: brushed, brushless, stepper, or ac servo motors? Match it to the motion profile, not just the torque number.
  • Continuous versus peak torque, with the gearhead efficiency included.
  • Life expectations: brush wear, bearing load, ambient temperature.
  • Check maxon motor news today and product change notices before locking the BOM.

The checklist starts with questions, not part numbers. I used to start with a motor and work backward. Now I start with the motion profile and work forward. It's printed and taped to my monitor—not as a decoration, but as part of the design review template. Engineers sometimes treat checklists as busywork. This one is different because it's built around the mistakes that actually show up in the field, not the ones you learn in a classroom.

The list has caught 47 potential errors in the past 18 months. Some were caught before ordering; a few during design reviews. The math is simple: 5 minutes of verification beats 5 days of correction. I'm not arguing that every failure is avoidable. But most of the near-misses I log are the same pattern: a specification was available, and I didn't connect it to the application until it was too late.