Test & Measurement
A Sensor Dies at 2 AM. Here's How I Get the Plant Running by Noon.
-
The short answer: three checks before you approve anything
-
Why you should trust this triage order
-
Case 1: pressure readings drifting 36 hours before an audit
-
Case 2: oil flow measurement, and why we didn't buy a Coriolis meter
-
Case 3: the thermal camera question
-
Case 4: the emergency that never happened
-
What the quote doesn't say
-
When this approach breaks down
The short answer: three checks before you approve anything
When a sensor fails and the line is down, you have roughly one shift to make three decisions correctly. Get any of them wrong, and what looked like a four-hour fix turns into a three-day outage.
Here's the order I use, every time:
First, confirm what actually failed. Roughly a third of the "dead sensor" calls I've handled in the past two years weren't sensor failures at all. They were wiring, contamination, a drifting setpoint, or a PLC input card that had been flaky for months. Replacing a healthy sensor costs you the part, the labor, and the four hours you didn't have.
Second, pick a proven part with real local availability. Not "available on the website." Not "ships in 3-5 business days." On the shelf, within driving distance, today. In my experience, the difference between a same-day fix and a missed deadline is almost never the technology—it's whether someone physically has the part.
Third, get the total number in writing before the PO goes out. Base price, expedite fee, freight, calibration certificate, connector, mounting bracket, tech time. The quote that looks 20% cheaper on the line item frequently isn't cheaper on the invoice.
That's the whole method. The rest of this is what it looks like in practice, including the parts where I got it wrong.
Why you should trust this triage order
I run field service escalations for an industrial automation distributor. Nine years, somewhere north of 300 emergency instrumentation calls, mostly mid-size plants in automotive and food processing. Same-day turnarounds for OEMs and integrators who call me because their customer is already standing at the machine.
In March 2024, a Tier 2 automotive supplier called at 4:40 PM on a Thursday. Their leak-test station was reading pressure drift, and they had a customer audit at 8 AM Friday. Normal lead time on the replacement we eventually used was five days. We found one at a sister distributor 90 minutes away, paid $340 in expedite and courier fees on a $610 sensor, and had it installed by 10:15 that night. The audit passed. If we'd ordered the "cheaper" one online, the audit would have happened with a station flagged non-conforming.
I've also blown it. In 2022, I shipped a replacement flow meter to a food plant without confirming the calibration certificate was included. It wasn't. The meter arrived Friday afternoon, worked fine, and sat unused until Monday because their QA team wouldn't sign off on an uncalibrated instrument. Nobody's fault but mine. That call is the reason I now ask about documentation before I ask about price.
Case 1: pressure readings drifting 36 hours before an audit
The call above was a classic. A pressure sensor on a leak-test fixture was reading about 0.04 bar high, consistently, across three shifts. Not dead—just wrong enough to fail parts that should have passed.
We swapped in an ifm PN2070 pressure sensor, SC1 variant. The SC1 version ships with the M12 connector already on it, which matters more than it sounds like it should at 9 PM. If you order the bare unit and discover you need a connector kit, you've just added a day.
Three things made that swap work:
- The PN2070 has a display, so the maintenance tech could verify zero and span on the bench before installing it. No laptop, no config software, no waiting on me to remote in.
- It speaks IO-Link, so the PLC could read the process value and the diagnostic flags without rewiring the analog input. That's the difference between a 20-minute swap and a 2-hour one.
- We had the part. That's it. That's the whole advantage.
Honestly, I'm not sure why so many plants keep a spare for their most critical photoelectric sensors but not for pressure transmitters on test fixtures. My best guess is that pressure sensors feel permanent—nobody plans for them to drift. They do drift. Keep one on the shelf.
Case 2: oil flow measurement, and why we didn't buy a Coriolis meter
Different plant, different problem. A hydraulic test bench needed flow measurement on the return line, roughly 4 to 40 liters per minute of hydraulic oil, and the customer wanted to know within a couple of percent what the flow actually was. They'd been using a turbine meter that had been rebuilt twice and was drifting again.
The "correct" answer, if you ask a metrology engineer, is a Coriolis flow meter. Direct mass measurement, no moving parts, excellent accuracy, handles viscosity changes without recalibration. It's also roughly five to ten times the price of a volumetric meter based on quotes we collected in Q3 2024, it's heavy, it needs straight pipe runs upstream and downstream, and it adds pressure drop to a return line that didn't have much margin.
I went back and forth on this for about a week. Coriolis gave us certainty. The alternative gave us a workable number at a fifth of the cost with a footprint that fit the existing plumbing.
We went with an ifm oil flow meter sized to the line. Not because Coriolis is wrong—it isn't, and for custody transfer or high-value blending I'd specify it without hesitation—but because this bench needed a trend, not a fiscal measurement. The customer wanted to know if a pump was degrading over weeks, not to bill anybody by the kilogram.
Here's something vendors won't always tell you: the accuracy spec on the datasheet assumes a specific installation. Straight runs, no elbows right at the inlet, proper grounding. If your pipe run looks like spaghetti, a tighter spec on paper doesn't buy you a tighter number in the plant. We rerouted about 40 centimeters of pipe and got a reading that tracked within a couple of percent against a portable reference. Good enough for the application, which is the only standard that matters.
Case 3: the thermal camera question
Somebody always asks about a thermal camera for iPhone during these calls, usually right after a motor starts running hot and nobody can find the infrared gun.
Two things to know. There is no thermal sensor built into an iPhone—you're buying an attachment that plugs into the Lightning or USB-C port, or connects over Wi-Fi. And the resolution on the sub-$500 units is low, typically well under 100x100 thermal pixels, which is fine for spotting a hot bearing and useless for reading a precise temperature on a small component.
For triage, that's genuinely useful. In November 2024, a tech used a phone attachment to find a failing contactor on a control panel that had been intermittently tripping for two weeks. Thirty seconds of scanning, one hot lug, problem solved. The measurement wasn't accurate to the degree. It didn't need to be.
Where people get burned is treating a phone attachment as a replacement for a real thermal imager. It isn't. If you need to document temperature rise for a warranty claim or an insurance report, you need the actual instrument with a calibration certificate. The phone attachment tells you where to look. It doesn't tell you what to write down.
Case 4: the emergency that never happened
Not every instrumentation crisis is instrumentation. A pharma lab I work with had a pipette calibration fail on a Monday morning, which meant a batch release slipped by two days. Root cause: residue buildup in the pipette barrel from months of use without disassembly.
If you want the short version of how to clean an Eppendorf pipette properly: check the manual for your specific model first, because the allowed cleaning agents vary. Then, in general terms—separate the lower part from the handle, wipe the exterior with 70% ethanol or isopropanol, clean the disassembled lower section with a mild detergent solution or the agent your manual specifies, rinse with distilled water, and let everything air dry completely before reassembly. Don't soak the whole pipette. Don't use aggressive solvents. And recalibrate after any full disassembly, because cleaning can shift the adjustment.
The lab's fix cost them about $180 in technician time and two days of schedule. A quarterly 15-minute cleaning routine would have cost them nothing. I've seen the same pattern with sensors—the plants that keep spares and clean their optics don't call me. The ones that call me at 2 AM are usually the ones who were "too busy" to do the boring maintenance.
What the quote doesn't say
Here's my standing opinion, and it applies whether you're buying a $600 pressure sensor or a $9,000 Coriolis meter: the number you see on the first quote is almost never the number you pay.
I've learned to ask "what's not included" before I ask "what's the price." The distributors who list every fee upfront—expedite, freight, calibration, connector, restocking—usually come out cheaper in the end than the ones quoting a low number with a shrug. Not always. Usually.
Real talk: when you're in an emergency, you're not in a strong negotiating position, and some suppliers know it. They'll quote the part and mention the rush fee after you've committed. Ask about it before. Every time.
When this approach breaks down
This triage order works for standard industrial instrumentation in non-hazardous areas. It does not work everywhere.
If you're in an ATEX or Class I Div 1 area, you can't swap in whatever's on the shelf—you need the certified variant, and that usually means a lead time. If you're doing custody transfer or regulated metering, a Coriolis or positive-displacement meter is the right answer regardless of cost, and "good enough for trending" is not an acceptable argument. If the root cause is a process problem—foaming, cavitation, particulate—replacing the sensor just means you'll be replacing it again in six weeks.
And my experience is based on roughly 300 calls, mostly mid-size plants in automotive and food. If you're running a refinery, a semiconductor fab, or a cleanroom pharma line, your constraints are different and my shortcuts may not survive contact with your QA department. Take the triage order. Leave the specific part numbers if they don't fit.