I’m the person people call when a CMM suddenly “measures wrong.” For eleven years, I’ve specified, approved, and maintained the measurement equipment for a mid-sized precision manufacturing company—so I’ve earned a fair number of those calls. The uncomfortable part is that I’ve also personally made and documented fourteen significant measurement mistakes in that time. They total roughly $26,000 in wasted budget, and that number doesn’t include the trust I lost each time I blamed a good instrument for my own bad decision.
In September 2022, three of those moments happened in a single week. That week became the basis for our whole measurement checklist.
The week three instruments “failed”
On Tuesday, the night shift measured a milled bracket and found a critical datum 0.0024 inch from nominal. The day shift measured the same bracket a few hours later and got 0.0002 inch. Their conclusion: the Renishaw probe was drifting.
On Wednesday, maintenance said a pump’s overload kept opening even though their logged current looked stable at 216 A—well below the protection setting. Their conclusion: the clamp meter was no good.
On Thursday, a customer challenged the insulation-resistance report for a motor we had rebuilt. Our tester said 83 MΩ; after the motor sat in the customer’s workshop for two days, their megger indicated 117 MΩ. Their conclusion was predictable: our tester was faulty.
Three failures in five days. Three teams pointing at three instruments. In every case, the instrument was right.
The deeper cause wasn’t something I could calibrate
The CMM program was fighting physics
The probe was fine. The stylus was fine. The problem was the inspection routine. A programmer had rewritten the night-shift path to collect more points while moving all three CMM axes at speed. With a long stylus, rapid acceleration turns machine inertia into measurement error. The day shift still ran the older, slower path, which is why the same machine, same part, and same probe produced such different results.
A calibration certificate doesn’t catch that, because calibration measures static performance under ideal conditions. The instrument wasn’t malfunctioning; the task I gave it was unrealistic. That realization is also what eventually pushed us toward the Renishaw REVO 5-axis CMM system. I used to think of five-axis scanning as “the fast option.” The more important difference is dynamic: the head’s two rotary axes move the stylus while the big CMM axes accelerate far less. We can collect many points without injecting inertia-related error into the reading. The probe wasn’t the problem. The motion was.
The clamp meter was telling the truth, but not the whole truth
The pump load was fed by a variable-frequency drive, so the current waveform was not a clean sine wave. An average-responding clamp meter can sit at 216 A while the actual waveform still contains the peaks that trip an overload. That’s not a broken meter; it’s a limitation I should have factored in when I approved the original purchase.
The Fluke 337 clamp meter that solved the case had true-RMS measurement and a dedicated inrush capture mode. On the first cold start, it logged a 640 A spike that the overload was responding to. The older meter wasn’t lying. It was simply the wrong class of measurement for the waveform.
Megger vs insulation tester is a category mistake
Thursday’s insulation dispute taught me the third and perhaps biggest lesson: environment. In electrical testing, people type “megger vs insulation tester” as if it were a real face-off between two competitors. To me, that phrasing means the underlying question still isn’t clear. Megger is a brand name that became a generic term for an insulation resistance tester. What you should compare is test voltage, measurement range, and whether the unit can log readings over time. But even a great insulation tester only gives you a snapshot of one moment.
Insulation resistance is not a constant. Temperature, humidity, and moisture move it around from day to day. A motor parked on a damp wooden skid for a week can lose a big chunk of its insulation resistance. The MR60 moisture meter pro we keep at the receiving dock—we use it to check inbound crates and skids—read 31% moisture content in that skid. Healthy dry wood is usually below 16%. The motor had been absorbing moisture for days. When the customer moved it into a dry workshop for two days, the readings rose. Neither tester was lying. The storage process was.
The price of blaming the instrument
That week cost about $10,000 in direct costs: a calibration service call and overtime to re-inspect parts, half a day of downtime during the pump diagnosis, and an extra engineering visit to defend a report that was actually correct. The indirect cost was larger. Every time I told a team “the instrument is bad, send it back,” I stopped them from looking at the real cause. Calibration would pass, the tool would return, and the same problem would come back a month later.
The most expensive sentence I ever say is still: “It’s the instrument; let’s recalibrate.” It sounds responsible. It’s often just a way to avoid admitting the measurement process was wrong.
What I do now (short version)
I stopped starting buying decisions with a brand or a category. Every purchase now begins with the condition the instrument will face. Is the current waveform distorted? Is the stylus going to move fast along a curved surface? Could moisture change the quantity I’m measuring? The answers decide the class of tool, and only then do I talk about brands and models.
For dimensional work, we now standardize on Renishaw products for our critical CMM applications—not because of the name, but because the REVO architecture removes a specific class of dynamic errors that used to show up in our part data. For electrical troubleshooting, we make sure maintenance crews have access to true-RMS clamp meters with inrush capture, not just the cheapest basic current clamp in the catalog.
I also stopped treating data logging as optional. The REVO system’s point-cloud output and the trend logs from our insulation testers mean we can compare today’s measurement with last month’s under the same conditions. That single change caught more problems than any annual recalibration. It also made us faster: when we stopped retyping readings and reacting to false alarms, inspection turnaround dropped by roughly 40%. That is the kind of efficiency I should have cared about years ago.
I’m not saying manual prechecks are worthless. For a one-off job, they still have a place. But if you own the process, automated data capture removes the boring, repetitive steps that create transcription errors—and it gives you a record you can actually defend when a customer questions your numbers.
One caveat before you quote me too hard: I’m writing this in early 2025, and measurement equipment and software change quickly. Verify current specifications against your actual process before making a decision.
Next time an instrument “lies,” get past the calibration sticker and ask: what did I ask this tool to measure, and in what conditions? In all three cases from September 2022, the honest answer was embarrassing—but it was also far cheaper than replacing a good instrument.
