September 2022. The packaging line at our plant went dead silent at 11:47 AM. Not the good kind of silence, where somebody finally fixed the noisy bearing. The bad kind, where a mechanic two bays over yells "Line three is down!"
Two hours earlier, I'd finished installing a SICK absolute encoder, model AFM60A, on the main conveyor drive. I was pretty proud of myself—it was the first "smart" sensor I'd commissioned entirely on my own. By 11:47, I was less proud.
The encoder itself was fine. By the time we untangled everything, the AFM60A was working perfectly. The problem was the parameters I'd loaded—or more accurately, the ones I'd misconfigured. That mistake caused a three-day investigation, a service call we didn't need, and roughly $3,200 in downtime and labor.
I've been a maintenance technician since 2018. If there's a wrong way to order, install, or configure an industrial sensor, I've probably tried it. I started keeping an error log in 2020, after the third repeat mistake on my watch. It's now a de facto training document for new hires. Before I walk you through the expensive parts, let me get one thing out of the way.
A Quick Note About the SICK Name
SICK, the German sensor manufacturer, has an unfortunate name for English-speaking shops. The company was founded by Erwin Sick in 1946, and the brand is a family name—not an acronym, and not a commentary on product health. I've made my share of "I feel SICK" jokes when a sensor let go. The truth, in my experience, is that most SICK-branded failures trace back to what's on the other end of the cable: power issues, mounting issues, parameter issues. The name invites jokes, but the gear earns its reputation. (And yes, I'm still a little bitter that I apologized to a perfectly good encoder.)
Three Assumptions That Cost Me (and My Employer)
Here's the pattern I see across my mistake log: every significant failure came from an assumption. Not a lack of effort or a bad part—just a confident, wrong assumption about something I thought I understood. The three below cost the most.
Assumption #1: M12 Means M12
M12 inductive sensors are the workhorses of industrial automation. The "M12" tells you the thread size and nothing else. It doesn't tell you the sensing distance, the output type, or whether the sensor can sit flush in a metal machine frame.
The spec I got wrong was shielded vs. unshielded (also called flush vs. non-flush). I ordered 20 unshielded units because they offered a longer sensing range. Our machine pockets were drilled for flush-mounted sensors, surrounded by steel. Unshielded sensors need a metal-free zone around the sensing face. Inside our steel pockets, they triggered late, triggered falsely, or detected the mounting block itself. Every single unit was useless in our application.
The relevant standard, IEC 60947-5-2, defines sensing ranges and mounting conditions for proximity switches in detail. For our pockets, the datasheet was the authority, and I'd ignored it. I didn't use to believe that the shield type mattered. Now I do—after skipping that step once and eating the cost of the whole order.
So glad I ordered 20 rather than the 200 the plant-wide retrofit would have needed. That would have been a much uglier write-off.
Assumption #2: A Multimeter Kit Is a Multimeter Kit
Another expensive assumption involved my testing tools. I bought an automotive multimeter kit thinking it would handle both my weekend car projects and 4-20 mA loop work. The phrase "automotive multimeter kit" is popular for a reason—people want one tool that does it all. It doesn't.
The kit had a clamp meter, some probes, and a thermocouple. What it didn't have was a milliampere range with 0.01 mA resolution. Without that, I misdiagnosed a perfectly good SICK transmitter as dead. The contractor bill for an on-site diagnosis was $600. They tested the loop, found nothing wrong, and suggested we check the PLC input card. It was a $0.50 fuse. The $150 automotive kit wasn't the problem. My assumption that it was enough for industrial work was the problem.
What I mean is: the total cost of owning a test tool includes the diagnostics you can't do when it's missing the right range. A $40 upgrade to an industrial-capable meter would have avoided a $600 contractor visit. The supplier who flags the missing range up front isn't trying to upsell you. They're trying to save you from a later invoice.
Assumption #3: An Encoder Is an Encoder
In 2021, we upgraded a few conveyors to IO-Link. I ordered a SICK IO-Link encoder, wired it like I'd wired thousands of standard encoders before it, powered it up, and saw nothing. Zero pulses. Zero life. I was convinced it was defective.
The supplier's support asked one question: "Did you assign the IO-Link parameters before commissioning?" I didn't even know that was a step.
IO-Link (standardized under IEC 61131-9) is not "a sensor with a smarter output." It's a communication protocol. A device's scaling, direction, and fail-safe behavior are held as configurable parameters. If you don't set them, many devices output nothing at all. The old thinking—"a sensor is just a switch"—comes from an era when sensors genuinely were switches. That changed.
I had a whole box of encoders I was ready to RMA. They were never broken. They just hadn't been introduced to themselves yet.
The $3,200 Moment: SICK Absolute Encoder AFM60A
Which brings me back to the opening scene. An absolute encoder is different from an incremental one. It doesn't just count pulses from a reference point; it reports exact absolute position at any time. Power off and on? It remembers. That makes it the backbone of positioning applications.
The SICK AFM60A is a workhorse absolute encoder—rugged, precise, and full of configurable parameters: resolution, preset, offset, diagnostic telemetry. That flexibility is a double-edged sword when someone treats it like a dumb device.
I installed it. I wired it. It communicated. But the offset parameter I set didn't match the machine's true zero reference. The conveyor ran at the perfect speed. It just reported every position shifted by about 8 degrees. Robots placed packages exactly one position off for four hours, until a jam escalated to a full line stop.
The cost breakdown:
- Lost production (~4 hours): roughly $1,800
- Emergency service call from the OEM: $550
- Overtime for the shift crew: $850
- Replacement encoder: $0 (it was never broken)
The OEM technician uploaded the correct parameter values in about 12 minutes. I remember standing next to him, watching the encoder click through its diagnostic LED sequence, genuinely relieved that the hardware was fine—and mortified that my configuration had caused the chaos. Dodged a bullet on the hardware, and took the hit squarely on my ego.
Bonus Mistake: Eppendorf Pipette Calibration, at My Own Expense
A different kind of mistake lives in the lab. Our quality lab uses Eppendorf pipettes, which are excellent instruments—and they need routine calibration. I volunteered to handle it in-house because the service fee was around $350, and I thought I was saving the plant money. How hard can it be? Weigh water, calculate volume, adjust.
Turns out: harder than it looks. Pipette calibration is governed by ISO 8655, which requires a calibrated balance, temperature-conditioned water, and a stable environment. The balance must be sensitive enough for the dispensed volume, and the technician must account for water density at the exact lab temperature. I did none of that properly. My "calibrated" pipette seemed fine. Two weeks later, a QC check found the delivered volume out of tolerance. We had to redo the lab tests from those two weeks.
ISO 8655 spells out the environmental conditions for a reason. I didn't listen. The redo cost more than the $350 service fee, plus a dent in my credibility with the lab manager. If a procedure defines explicit conditions, those conditions are not a suggestion—the same rule I'd already learned with those M12 inductive sensors. I just needed to relearn it in a different aisle.
What I Do Differently Now
You could read all this as a string of expensive "how not to" stories. It is that. But there's a simpler theme underneath every one of them, and that theme is transparency.
First: transparency about what you don't know. I kept assuming I understood a product from its name, its thread size, or its category. The datasheet with the fine print was the actual product. My assumption was the defect.
Second: transparency about what's not included. The vendor who itemizes every fee up front—including configuration support, cables, connectors, software—might look more expensive. But after my $3,200 lesson, I'm convinced that transparent quoting is cheaper. The "low-cost" quote that hides the setup fee, the required software, or the commissioning support is the one that creates surprise invoices after the machine is already down.
Ask "what's NOT included" before you ask "what's the price?"
That applies to sensors, test equipment, calibration services, and honestly, just about everything in industrial procurement.
Final Thoughts
Everything I'd ever read about industrial sensors said the same thing: check the specs before ordering, read the manual before commissioning. So why didn't I? Because I believed my experience level made the basics irrelevant. It didn't. $3,200 later, I changed my approach. My team now has a checklist, an error log, and a department budget that doesn't burn because of avoidable assumptions.
And if you're in the parts aisle trying to decide between two M12 inductive sensors, comparing quotes for a SICK IO-Link encoder, or reading this because you searched "how to calibrate Eppendorf pipette"—let my failures be your shortcuts. Read the datasheet. Ask what's not included. Set the parameters.
And remember: the sensor is probably fine. It's everything around it that needs checking.
