"Defective. Intermittent signal loss."
That was the note attached to the returned Balluff radar sensor. The yellow RMA tag told me the customer had shipped it back after two weeks of unreliable level readings in a bulk hopper.
I pulled it out of the box and gave it the standard once-over. I review every returned unit that comes through our quality lab—a couple hundred a year, more or less—and the first two checks are always visual. Housing clean. Threads intact. No signs of impact or moisture. On a bad day, you find the smoking gun in a hairline crack or a bent mounting thread. This one passed both.
Then I put it on the bench and did what I always do: powered it up, pointed it at a reference target at a known distance, and read the output. Spot on. I let it run for two hours. No dropouts. Twenty-four hours later, still rock solid. (Which, honestly, made me more suspicious than a clean failure would have. Intermittent faults never show up when you're looking for them. That's the rule.)
People ask me what quality inspection looks like in practice. If you've ever worked in commercial print, it's the same idea. According to the Pantone Color Matching System guidelines, brand-critical colors should be matched within a Delta E of 2 or below—beyond that, trained observers notice the difference. Print production calls for 300 DPI minimum at final size; at 72 DPI, you see the pixelation immediately. Standards exist for a reason: you can't reproduce what you can't measure consistently. I don't guess at a sensor's behavior—I test it against a reference and wait for the problem to show up. This sensor didn't have one.
So I had two options. Sign off on the return and process a replacement, or go see the installation myself. In my earlier days, I'd have taken the first option without a second thought. Ask me now, and I'll tell you that's exactly how you end up replacing a perfectly good product with another perfectly good product and calling it quality control. So I did something I don't always have time for: I called the customer and asked if I could visit the site.
The Installation
The customer was a mid-sized packaging plant. They use the radar sensor for bulk level measurement in a hopper; if the sensor gives an incorrect reading, the line either overfills or stops. The maintenance lead walked me out to the line and pointed up at the sensor, mounted about twelve feet off the floor.
I saw the problem in about three seconds.
Someone had extended the mounting bracket with a piece of steel angle to work around a new conduit run. The sensor wasn't pointing down into the hopper—it was angled across it, and a steel support beam sat right in the detection field. The radar sensor was doing exactly what it was designed to do. It was measuring the beam. Not the fill level. That's it. Period.
The maintenance lead nodded in that way people nod when they're embarrassed. "We had to move it to make room for the conduit," he said. "Didn't think it would matter."
It took us twenty minutes to fix the mount and re-aim the sensor. The reading settled to the correct level immediately. The same unit that had been shipped back as defective went back into service that afternoon.
Why does this matter? Because that incident wasn't the first time I'd seen a good product blamed for something it never did. In my opinion, most "sensor failures" fall into one of three buckets: misapplication, improper installation, or misreading. Actual hardware defects are the minority, but nobody sees it that way on the factory floor. First instinct: blame the instrument.
The Plant Walk: A Crash Course in Measurement Literacy
The maintenance lead asked me to stop by his office on the way back to grab a screwdriver. On his belt was a digital caliper with the battery cover held on by electrical tape. It wasn't one of ours—a hardware brand—but the issue was universal. He pressed on the taped cover to make the display come on. "It works," he said. "Just annoying." (I didn't say it out loud, but tape over a caliper battery cover is exactly the kind of thing that causes intermittent readings. Loose battery contacts make the unit flicker and reset mid-measurement. The replacement cover costs a few dollars. You shouldn't be holding a precision measurement tool together with tape.)
On the way to the hopper, we passed a conveyor line with a Balluff encoder on the drive end. "That one's been giving us erratic counts," he said. I glanced at the cable routing—the encoder cable ran parallel to the motor's power cable for about ten feet, no separation. Classic EMI setup. A power cable running alongside a signal line can induce ghost pulses on an encoder output. The encoder itself was probably fine. The cable routing wasn't. I mentioned it, and he wrote it down in a greasy notebook. (Not that I have hard data on how often EMI explains erratic encoder counts, but from my experience, it's more often than people assume.)
Then we stopped at a hydraulic press. He nodded at the pressure gauge on the line—a 1009 pressure gauge, from the look of it. "This one's been reading low. Thinking about replacing it." I suggested he check it against a reference gauge before ordering anything. He found one, teed it in, and we watched both gauges. Sure enough, once the line reached operating temperature, they read the same. The 1009 gauge wasn't broken—it was reading a cold line, that's all. Same story, twice in one morning: good instrument, misread situation.
Then he asked me something completely out of nowhere: "Do you know how to read a Sensus water meter? The night shift keeps writing down the wrong numbers, and we can't figure out why."
As it happens, I knew exactly what he was talking about. The newer Sensus digital meters—the iPERL series, for example—don't just show the reading and stop. The LCD cycles through several screens: diagnostic codes, test codes, and then the actual consumption display. If you look at it at the wrong moment, you'll copy a number that looks like a reading but isn't. You have to wait for the screen with the correct unit label, or you'll write down the wrong data. The night shift at that plant had been recording a diagnostic code for weeks without knowing it.
Standing in that plant, I finally put all the pieces together. It wasn't the radar sensor. It wasn't the 1009 gauge. It wasn't the caliper. It wasn't the Sensus water meter. Every one of those "broken" instruments was doing its job. The common thread was a lack of basic measurement literacy—not because the crew was careless, but because nobody had ever shown them how to read what the instruments were telling them.
The Lesson That Stuck
I only believed that after ignoring it once. Early in my quality career, a customer returned a sensor, I signed off on a replacement without visiting the site, and three weeks later, they returned the replacement with the same complaint. Same installation mistake caused both "failures." That one cost us an $800 replacement and a difficult conversation about why we hadn't fixed the problem the first time. I've never made that assumption since.
So we did something different this time. Instead of just closing the return ticket, we put together a short training session for the plant team: how to mount a radar sensor and aim it properly, how to verify a gauge before replacing it, how to route encoder cables away from motor lines, and how to identify the right screen on a Sensus meter before writing down a reading. It wasn't a sales presentation. It was customer education.
Even after we scheduled the session, I kept second-guessing. What if nobody shows up? What if they nod politely and forget everything by the end of the week? The two weeks between scheduling and delivering were stressful. I hit send on the calendar invite and immediately thought: did I just volunteer us for something we can't measure?
The session itself went fine. The follow-up is what convinced me. The maintenance lead called a month later—not to complain, but to tell me that the night shift had read the Sensus meter correctly every day for four weeks straight. He'd also replaced the caliper battery cover. And the radar sensor? Still running. Not a single dropout. (The encoder started counting reliably again too, once they moved the cable away from the motor feed.)
I don't have hard data on how many returns are false alarms, but based on the returns I've personally tested over the years, my sense is that more than a quarter fall into the "product was fine, installation or reading was wrong" category. That's not a criticism of customers. If anything, it's a criticism of the industry for not doing better at education. We make more than 30 categories of industrial sensors, but all of that technology is useless if the person on the floor can't interpret what it's telling them.
An informed customer is the best customer. I know that sounds like a slogan, but in my experience, it's a practical fact. Customers who understand what they're measuring ask better questions, catch their own errors sooner, and call us only when they're actually stuck. They don't waste their time, and they don't waste ours. A product is only as good as its installation and its interpretation.
In print production, standards like the 300 DPI rule and Delta E tolerance keep everyone honest. In sensing, it's the same discipline applied to a different world: calibration references, installation guides, and the willingness to ask one more question before you blame the instrument. A radar sensor pointed at a steel beam doesn't measure your hopper. A 1009 gauge on a cold line doesn't tell you anything useful. A caliper with a taped battery cover isn't a measurement tool—it's a guess. And a Sensus water meter that nobody knows how to read is just a screen with numbers on it.
Measurements are only as good as the person reading them. And that, honestly, is the part of my job I never expected to care about the most.
