I'm a quality manager at a controls integration company. I review every pump skid before it ships—roughly 200 units a year. In March 2023, I was standing in front of a packaged chilled water skid that kept throwing a false low-level alarm. The HMI logged 'tank level sensor fault' for about twelve seconds, then it cleared. Classic intermittent. And my first instinct was to blame the transmitter.
We use Dwyer pressure sensors/transmitters on most of our pump packages because the loop calibration and terminal layout are consistent. The alarm was coming from the Dwyer level transmitter in the condensate tank. It was sending 11.2 mA at what should have been a stable 50% level when I checked it with a 115 multimeter. From the outside, it looked like an electronic failure. The reality was cable damage.
The Setup
The skid was part of a $117,000 order for a lab building's HVAC cooling loop. It had passed hydrostatic testing, and we were in the 72-hour run-in phase. The level signal dropped out twice during the first night. The night operator swapped in a spare Dwyer level transmitter from stock—a $210 part. By mid-morning, the same fault appeared.
That should have been my first clue: two transmitters, same wiring, same fault. But I still assumed the sensor was the problem. I knew I should have megger-tested the cable before touching anything, but I thought, 'a brand-new skid? What are the odds?' The odds caught up with me.
The Turning Point
I grabbed a C2 thermal imaging camera and scanned the junction box on the side of the skid. There was a warm spot where the level transmitter cable entered the gland. Not hot—maybe two degrees above ambient. But it was the only thermal anomaly in the box. I opened the enclosure and checked the terminals with a 115 multimeter. The multimeter itself has a NIST-traceable calibration certificate, so I trusted the reading.
Loop voltage was 24.1 VDC and steady. The transmitter output was 11.2 mA, which corresponds to 50% level on a 4–20 mA loop. The sensor was fine. It was reading exactly what it should. Let me rephrase that: the sensor was measuring exactly what it saw. The problem was what the cable was telling it.
How to Use a Megger Insulation Tester
I pulled out the megger insulation tester. This is the tool I should have used before ordering that $210 spare.
- Lock out the circuit and verify it's de-energized. Per NFPA 70E, this is non-negotiable.
- Disconnect both ends of the cable so you're testing the cable, not the transmitter or PLC input.
- Set the test voltage to 500 V for a 24 VDC signal cable. Don't use 1,000 V unless the cable insulation is rated for it.
- Connect the black lead to ground, and the red lead to one conductor. Hold the test button for 60 seconds and record the reading.
- Repeat for each conductor, and always compare conductor-to-conductor if the cable has a shield.
This is where the mask came off. One conductor read 0.4 MΩ. The other conductor read over 99 MΩ. On a 24 V control circuit, I want to see 10 MΩ or better; our internal rule is even simpler—if it's below 10 MΩ, I don't accept it. That 0.4 MΩ reading told me moisture had gotten into that one conductor. The cable jacket was fine to the naked eye, but the gland grommet was cracked, and condensate had wicked down inside.
We replaced the 50-foot cable, replaced the gland grommet, and re-terminated both ends. The original Dwyer level transmitter went back in—or actually the spare stayed in, and the original went back to spare stock. Either way, no more fault. The 72-hour run finished clean.
The Lesson
The cable was $34. The transmitter was $210. The diagnostic time, the night call, and the delayed shipment cost us roughly $1,900. I'm not 100% sure of the exact accounting, but it was enough to hurt.
People assume intermittent readings mean the instrument is bad. What they don't see is the wiring, the termination, the environment. I've learned never to assume 'new skid' means 'new cable.' On a brand-new build, the cable is exactly where mistakes happen: a pulled gland, a nicked jacket, a stripped screw. The instrument didn't fail. The installation did.
This is also a total cost of ownership story. Total cost of ownership—TCO—is just as important on the plant floor as it is in the purchasing office. The cheapest way to install an instrument isn't just the price of the instrument. It's the price of installing it correctly, testing the loop, and troubleshooting when something goes wrong. A $34 cable and a $10 grommet cost us a full day because I skipped the verification step.
In quality work, the test equipment—a decent 115 multimeter, a C2 thermal imaging camera, a megger insulation tester, and the discipline to use them—is what actually keeps the total cost low. If you're dealing with a noisy or intermittent level signal, do yourself a favor: before you order a replacement transmitter, use a megger insulation tester on the cable. I know it's easy to blame the sensor. Most of the time, the sensor is just the messenger.
