The Calibration Shortcut That Caused a 3-Day Shutdown

It was 2:47 PM on a Thursday when the panel alarm went off. Tank B-204 was reading 42% on the Endress+Hauser level transmitter. The overflow port was already pouring product onto the floor, which was a strange thing to see when the screen claimed we had almost 60% of headroom left.

That was my first real lesson in calibration—or rather, in what happens when someone "adjusts" a transmitter without checking the reference first. I'm a process engineer with 11 years in chemical manufacturing. I've made enough calibration mistakes to write a short book about them. Now I maintain our team's checklist so nobody else repeats my errors.

The Surface Problem: "The Reading Doesn't Match"

Here's how the story usually goes: an operator notices that the instrument reading on the control system doesn't match the sight glass. Or the local gauge. Or the dipper reading from yesterday's shift. At that point, someone inevitably suggests "adjusting" the transmitter to make it agree with the "real" value.

That's the surface problem—and it's almost never the real problem.

What most people don't realize is that the sight glass or local gauge is often the wrong reference. The sight glass on B-204 was clogged with polymer residue. The "real" value that the operator insisted on was actually a stuck float indicator. Once you adjust a calibrated transmitter based on an unverified reference, you've broken the measurement for both the reference and the instrument. The transmitter wasn't lying. We made it lie.

Deep Cause #1: Calibration vs. Verification Confusion

"Calibrating" a level transmitter isn't the same as shifting the displayed value to match something else. Most engineers know the difference when asked. But when you're in a hurry—or rather, when the production manager is standing behind you asking, "How long is this going to take?"—the line between "trimming" and "calibrating" gets dangerously blurry.

The proper Endress+Hauser level transmitter calibration procedure, per the operating manual, involves:

• Establishing a reference standard that's traceable to a national or certified reference
• Performing a zero trim and a full-scale (span) trim under controlled conditions
• Verifying at mid-range and quarter points, not just at 0% and 100%
• Documenting as-found and as-left values

We'd done the first two steps. But we skipped the mid-range verification because we were already late for a product changeover. The instrument was calibrated at the endpoints—and the nonlinearity in the middle of the range didn't show itself until we were dumping product onto the floor.

Deep Cause #2: The Wrong Reference (My Turbidity Disaster)

In September 2022, I made an even worse mistake with an Endress+Hauser turbidity sensor on our wastewater treatment line. We have a permit limit on effluent turbidity, and readings seemed off. A lab grab sample came back at 8.2 NTU on a rainy Tuesday. Our sensor showed 6.5. So we "calibrated" the sensor to read 8.2.

Turns out—and this is the part I still kick myself over—the grab sample had been sitting on the lab bench for three hours before analysis. Solids had settled. The lab value was influenced by stratification, and I used it as a truth reference. The sensor was then off in the opposite direction for two weeks before anyone noticed. That's a permit violation we got lucky on, not a near-miss we earned.

What I know now: turbidity sensors are sensitive to the technique used to verify them. The stored calibration curve is meant to be checked against stable, well-mixed standards—not against grab samples that can stratify during transport and waiting time. The manual says this. I didn't read it carefully enough.

Deep Cause #3: Environmental Effects on Load Cells

The third incident was in January 2024, in our filling hall. Load cells on three filling lines were drifting. The operator's solution was to "zero out" the buildup every morning to get the displayed weight back to zero. Each morning, he'd make a zero correction, and the line would run fine for a few hours.

But here's the insidious part: zeroing the scale to compensate for material buildup doesn't restore accuracy—it shifts the underlying mechanical zero. If you keep compensating for what is actually a physical buildup, you compress the calibrated range of the load cell. Over a month, we distorted the effective measurement range by about 14%. We discovered it when a customer complained that drums were underfilled. That was a $3,200 order, straight to rework, plus the credibility hit with a customer we'd served for years.

I should add that even good load cells drift with temperature changes, vibration, and mechanical stress. They are not "set and forget" devices. The preventive approach—which we now follow—is to inspect and clean the load cell platform before verification, not assume zero is zero.

What This Pattern Actually Costs

Let me put the numbers together from these three incidents:

• B-204 overflow: $12,000 in wasted raw material, 3 days of production downtime, and a lot of uncomfortable conversations with the area manager.
• Turbidity sensor mishap: two weeks of unreliable compliance data, elevated risk of a permit violation, and a strained relationship with our environmental reporting contact.
• Load cell distortion: $3,200 order redone at our cost, 1-week delay for the customer, plus the hard cost of re-inspecting dozens of other drums before release.

Total: roughly $15,000 in direct costs, plus delays, reputational damage, and the not-easy-to-quantify hit to my own credibility within the plant. And that's the stuff we even caught. I suspect there are more near-misses we never noticed.

Prevention Over Cure: The 12-Point Pre-Calibration Checklist

After the third mistake, I created our team's standard pre-calibration checklist. It's deliberately paranoid, and it's the cheapest insurance we own. Here's the core of it:

• Verify the reference standard is within its own calibration date.
• Confirm the process is stable before starting (temperature, pressure, flow; no tank mixing if it's a level check).
• Use the manufacturer's documented procedure—for us, the Endress+Hauser operating manual for each transmitter, whether it's the level transmitter calibration procedure or the turbidity sensor handbook.
• Check for physical buildup, corrosion, or damage on the sensing element before doing anything.
• Never adjust a transmitter based on a local gauge or a grab sample unless that reference has been independently verified.
• Perform at least a 4-point verification: 0%, 25%, 75%, 100%—not just endpoints.
• Record all readings in the calibration log, not on a scrap note.
• Track drift across calibration cycles; if a transmitter drifts noticeably two times in a row, flag it for repair.
• Clean and inspect load cells before any zero or span adjustment.
• For turbidity sensors, use stable, well-mixed calibration standards, never a grab sample.
• When calibrating our Mettler Toledo pH meters—and we use several handheld units—always use two buffers that bracket the expected range, not a single buffer. (That's a lesson from a separate incident I'd rather not re-live.)
• After calibration, run a final loop check to confirm the signal reaches the control system correctly.

Since we rolled this out in Q1 2024, we've caught 47 potential errors—expired reference standards, buildup on sensing elements, a loose connection that was moments away from failing. A five-minute verification in the plant has replaced what used to become five days of correction.

Oh, and another small addition that's been useful: I keep a thermal imaging camera for iOS in my toolbox. It doesn't replace any of the verification steps above, but it's remarkably good at spotting hot spots on motor connections and instrument enclosures—those are often early signs of an electrical problem that will eventually affect the measurement loop anyway.

The Bottom Line

Calibration is not a one-time event. It's a relationship between the instrument, the process, and the reference standard—and each one needs to be checked and respected. If you take anything from this, take this: five minutes of verifying the reference before you adjust anything is infinitely cheaper than the alternatives.

I still kick myself for B-204. If I'd checked the sight glass before letting someone tweak the transmitter, I'd have noticed it was clogged. The transmitter was telling the truth the entire time. I just made it lie.