The Encoder That Passed Every Spec and Still Shut Down the Line
September 2022, around 2 p.m. Our case-packing line stopped mid-cycle. The replacement absolute encoder was a SICK unit, same family as the original, and every line on the datasheet checked out: 4096 PPR, SSI interface, 24V, IP67. Bolted it on, ran the line, and it lost position after the first index. Operator reset it. Two hours later, it lost position again.
Seven hours of downtime. A $4,300 reorder. And the worst part — nothing was actually broken.
I'm an instrumentation and maintenance engineer handling sensor and test-equipment sourcing for 9 years. I've personally made — and documented — four significant procurement mistakes totaling roughly $27,000 in wasted budget, scrapped material, and lost production. This is what I learned from them, and the checklist our team now uses.
The Real Problem Hides Behind the Spec Sheet
Here's the thing nobody tells you when you're new: the spec sheet tells you what a device can do. It doesn't tell you whether it will do what you need. That gap is where most of the money goes.
1. "Programmable" Doesn't Mean "Pre-Configured"
A sick programmable encoder — the AFS60 and AFM60 lines, for example — can be configured for resolution, output direction, zero position, and interface type. That flexibility is genuinely useful. But when I ordered a rush replacement, I assumed "programmable" meant the supplier would ship it matched to our drive.
It didn't. The unit arrived at its default 1024 PPR. The encoder worked flawlessly — at the wrong resolution. The drive had no idea what to do with the signal and dropped the position every cycle.
Everyone told me to verify the as-shipped configuration before installation. I only believed it after ignoring it once and eating a $4,300 reorder plus a full shift of downtime.
The same trap exists on absolute encoder SICK replacements in general. The catalog makes it look like a simple part-number lookup, but three things have to match at once: mechanical interface (solid shaft vs hollow shaft, flange type), electrical interface (SSI, BiSS-C, or IO-Link), and resolution or multiturn range. Get the first two right and the third wrong, and you get a silent failure — the kind that doesn't throw an alarm until the line is already down.
We didn't have a formal spec-review process for replacement parts back then. Cost us when that encoder went in without anyone checking the shipped configuration against the drive parameters. That was the third time a "drop-in" replacement wasn't, so I finally built a pre-install verification sheet. Should have done it after the first one.
2. The Hygrometer That Was "Working Fine"
In 2021 we installed a relative humidity sensor in our cleanroom dry-box. It read 42% RH. It stayed at 42% RH for three months. Everyone was happy.
It wasn't. The sensor was mounted right next to the air inlet, where airflow was highest and most stable. The actual chamber humidity was swinging between 28% and 55%, and our process ran out of spec the whole time. A roughly $180 hygrometer gave us about $6,000 in scrapped material because nobody explained two basic rules to me:
- Sensors go where the environment represents the process, not where it's convenient to run a cable.
- Response time matters. A slow hygrometer averages out spikes, so it looks stable when it isn't.
I'm fairly skeptical now of any environmental reading that looks too steady. Real processes have noise. A flat line usually means the sensor is either lying or in the wrong place.
3. The 62 MAX IR Thermometer, and Why "Point and Shoot" Is a Trap
I bought a Fluke 62 MAX IR thermometer in 2019 to check bearing temperatures during walkarounds. It's a genuinely decent little unit — 10:1 distance-to-spot ratio, -30 to 500°C range, covers most routine maintenance checks.
Here's what I didn't understand: at 30 cm, the 62 MAX measures a spot roughly 3 cm across. At 90 cm, it averages a 9 cm spot. If you're checking a 2 cm bearing housing from a meter away, you're averaging the bearing against the surrounding ambient — and you'll get "normal" readings right up until the bearing fails.
I got the reading. I got it wrong. A bearing seized in February 2019, and that's how I learned that an IR thermometer isn't a magic wand — it's a spot-size calculation you have to do in your head before you trust the number.
4. Fluke vs FLIR Thermal Cameras: The Comparison I Got Wrong
I spent two weeks in 2022 deciding between a Fluke and a FLIR thermal camera. I read every Fluke vs FLIR thermal cameras comparison I could find online. Most of them focus on resolution — 160×120 vs 192×144, and so on — which is exactly the wrong place to start for maintenance work.
I bought the one with more pixels. Wrong metric.
Pixels matter, but NETD — noise equivalent temperature difference — is what actually tells you whether you can see a 2°C anomaly on a warm motor. A high-resolution camera with mediocre sensitivity will hide the small deltas you're trying to catch. A lower-resolution camera with good sensitivity won't. I also ignored ergonomics — menu structure, battery life, whether the thing is usable in gloves at 2 a.m. — because I was comparing spec tables instead of using the tools.
When I finally compared the two cameras side by side on the same motor — same load, same ambient — I understood why the cheaper one produced the more useful thermal image. It wasn't resolution. It was sensitivity.
This is where I became a believer that output quality directly shapes how clients perceive your work. When I hand a thermal image to a client to justify a maintenance recommendation, a blurry or ambiguous photo undercuts the entire report. Clients don't judge your engineering by your spreadsheet — they judge it by what you put in front of them. Saving $600 on a camera and losing a $15,000 service contract is a trade I won't make again.
What This Actually Costs
Adding it up, because the total tells the real story:
- Wrong-configuration encoder, 2022: $4,300 reorder plus 7 hours of stopped line time
- Cleanroom hygrometer misplacement, 2021: roughly $6,000 in scrapped material
- Bearing failure missed with an IR thermometer, 2019: $890 repair plus a 3-day production delay
- Thermal camera decision, 2022: $600 "saved," roughly $15,000 service contract lost
Roughly $27,000 in avoidable costs across four purchases. Not one of them involved a defective product. Every single one involved a product that met its datasheet and missed my actual application.
The Checklist I Wish I'd Had in 2017
I won't pretend this is exhaustive. It's what our team uses now, and it's short on purpose:
- Write the exact configuration into the PO. For any programmable or configurable encoder, specify resolution, direction, zero position, and interface. Don't assume "programmable" means "matched to my drive."
- Match all three interfaces on absolute encoders: mechanical, electrical, resolution. An absolute encoder that's "almost right" is just a wrong part with a datasheet attached.
- Mount environmental sensors where the process happens, not where wiring is easy. Check response time against your process cycle before you trust a steady reading.
- Do the D:S math before you trust an IR thermometer. If the spot size at your working distance is bigger than the target, you're measuring the room, not the part.
- Compare thermal cameras on NETD, not just pixels — and handle one before buying. Spec sheets don't tell you whether you can actually use the thing in the field.
That's it. The underlying lesson is simple: a spec sheet tells you what a device can do. It never tells you whether it will do what you need. Closing that gap is the whole job.
I've been on the wrong side of it four times. The checklist is cheaper.