A $47,000 Power Quality Lesson: Fluke 43B, a 4 Channel Oscilloscope, and an Absolute Encoder Profibus Problem

It was a Tuesday morning in March 2024, and I had a pallet of 120 drive panels staged for final release. Our customer's line was waiting. We were already late, and the pressure to sign off was pretty intense. The vendor had sent a test report for every panel, and all the numbers were green. I remember reading 'absolute encoder Profibus communication — within specification' and thinking, 'Okay, we are good to ship.'

Honestly, that sentence is exactly what I want to see as a quality inspector. It is also the sentence that almost cost us the relationship with that customer.

The Fault

We did not ship all 120. We shipped three, because that's what the customer had room for on the line. Within an hour, the PLC started dropping Profibus nodes. The encoders would read position fine for a while, then lose a few counts. The line would fault. The customer's controls engineer reset it, and it ran for another twenty minutes, then faulted again.

At first, everyone blamed everyone else. The field tech said the customer's PLC program was bad. The customer's electrician said our panels had a ground loop. The encoder manufacturer's helpline said the most common cause of intermittent position loss on an absolute encoder Profibus link is a bad cable or a missing terminating resistor. We checked the PROFIBUS cable against the IEC 61158 guideline: termination, baud rate, cable type. All matched. The STP jacket was fine. The baud rate matched.

Then we replaced two encoders anyway, because it was faster than arguing. The fault never went away.

The Argument

On Thursday of that week, I flew to the plant. I do not normally do field visits, but the customer was upset, and I was the one who had approved that vendor test report. I brought a Fluke 43B power quality analyzer because it was sitting on the shelf, and honestly because a colleague swore by it. That was a bit of luck. It turned out to be exactly the right tool.

The customer's maintenance manager said something I still repeat: 'We need to know why this keeps happening, not who is responsible.' (Should mention: that manager was the one who decided to stop swapping parts and start taking measurements. He deserves most of the credit.)

What the Multimeter Did Not Show

I started with a multimeter. I am a Fluke fan, and if you ask me what the best Fluke multimeter for automotive or industrial troubleshooting is, I will usually say the 87V. The low-pass filter is great around variable-frequency drives, and the leads take a beating. But a multimeter only gives you a snapshot. I measured the 24V supply at an encoder, saw 24.1V, and thought, 'This is fine.'

That was the trap. 'Fine' at idle is not the same as 'fine' under load.

What the 43B Showed

We set the Fluke 43B power quality analyzer on the incoming 115V control power feed. It is basically a handheld scope with logging, so it sat there and watched. In about twenty minutes, it caught the problem: the line voltage dipped below 98V for three or four cycles every time a big variable-frequency drive on another line started. The dip was not long, but it was deep enough to make the absolute encoder Profibus interface lose its place.

Then we dug out a 4 channel oscilloscope (a Fluke 190 Series scope we keep in the lab, though I want to say a portable one is better in the field—do not quote me on model numbers). We put one channel on the Profibus data line and another channel on the AC power input. Whenever the power line dipped, the data line spiked. The correlation was obvious. It was like watching a weather radar and seeing the storm exactly when the alarm went off.

So the encoder was not bad. The Profibus cable was not bad. The PLC was not bad. The 115V feed was bad, and the encoder was just the canary.

About Thermal Cameras and Walls

One of the electricians asked me, 'Can thermal cameras see through walls? FLIR's own FAQ says no.' He had an FLIR unit on the truck, and he was hoping to scan the junction box behind the panel without opening it. A thermal camera sees surface temperatures, not what is behind sheet metal. But the reason he asked was smart: he had already noticed that the junction box wall was warm.

The thermal camera did not see through the wall. It saw the heat radiating from the wall. And that heat came from an undersized neutral connection in the feeder junction box. Every time the big VFD pulled current, that connection produced enough heat to warm the box, and enough voltage drop to sag the control power. We found a charred-looking neutral splice in a box that had passed the original wiring inspection. (Oh, and the 'bad' encoders? Both passed perfectly when we returned them.)

The Real Fix

The repair itself was not that complicated: a new neutral splice, a dedicated control transformer tap for the encoder power supply, and a torque check on every termination in that junction box. The expensive part was the disruption. Two days of line downtime. A hotel bill for me and the field tech. Freight for a replacement panel, even though we did not end up needing it. And the reputational hit that comes when a customer's production line stops because your panel's power feed was marginal.

In my head, that adds up to about $47,000, though I might be misremembering the exact number. What I do remember is that the measurement that caught it took 20 minutes, and the measurement that should have caught it would have taken maybe 15 minutes before shipping.

What I Do Now

I only truly believed in power-quality testing after ignoring it and paying for it. Now every panel with a PLC, an encoder, or any kind of fieldbus gets a short soak test before I approve it. We put the Fluke 43B power quality analyzer on the incoming feed and leave it there for a while. If the voltage dips under load, we fix the feed before we ship the panel. According to Fluke's published spec sheet, the 43B is rated CAT III 600V, which is the right category for this kind of panel work.

People ask why we do not just rely on component vendors' test reports. That is the surface illusion. A test report proves the component worked in a test lab under ideal conditions. It does not prove it will work on your customer's line next to a 50HP drive starting every 45 seconds. The question everyone asks is, 'Is this absolute encoder Profibus compatible?' The better question is, 'Is the power going to be stable enough for the encoder to keep its position when the rest of the factory starts drawing current?'

That lesson was expensive. But the checklist that came from it is cheap. Five minutes of verification beats five days of correction. Every single time.