Tuesday, 7:41 AM
It was a Tuesday in Q1 2024. The email arrived at 7:41 AM, before I’d finished my coffee. Our incoming inspection team had flagged a batch of components from a supplier we’d worked with for years. The subject line: “TDK transformer rejects — DCR out of spec.”
I’m a quality/compliance lead at an electronics components distributor. I review every product before it reaches the customer—roughly 200 unique SKUs per year. In 2024, I rejected about 6% of first deliveries for spec deviations. That number doesn’t sound dramatic, except when you’re the one signing the return documents.
The parts in question were part of a larger order for a medical device customer. They were building a wearable blood pressure monitor. The design included a TDK InvenSense C300 blood pressure sensor for the sensing side and a small TDK transformer for the isolated power stage. We’d ordered both together: 5,000 C300 sensors and 5,000 transformers.
That morning, the transformer looked like the problem. The DC resistance was reading 3.14 Ω against a spec of 2.8 Ω max. Our internal tolerance was 10% over the datasheet maximum, so the accept limit was 3.08 Ω. We were above it. I told my team to quarantine the lot and prepared the rejection notice.
Look, I’m not someone who enjoys rejecting parts. But I also know the cost of letting a questionable batch through. One bad transformer in a medical device is worse than three delayed shipments. So I pulled the trigger. We notified the customer, the distributor, and the manufacturer. Then the vendor asked a simple question that made my stomach drop:
“Can you describe the exact test setup you used?”
I couldn’t. Not fully. That’s when I realized the story was about to change.
The C300 and the transformer
Let me back up. The C300 sensor was new to us. It wasn’t our first blood pressure sensor, but it was the first TDK InvenSense part we’d handled in that category. We’d used TDK inductors, EMI filters, and transformers for years. The C300 was a different animal: a sensor with an integrated signal-conditioning chip, and it needed careful handling.
We weren’t testing the sensor itself that morning. The transformer was the failure. Or so we thought.
Here’s what we had done: our technician took a handheld multimeter, set it to ohms, and measured across the transformer’s primary winding. The reading was 3.14 Ω against a spec of 2.8 Ω max.
That 0.34 Ω difference doesn’t sound like much. In a medical power supply, it can matter. More importantly, it was above our documented accept limit. We rejected the lot.
The vendor’s request for the test setup sat in my inbox for an hour. I had two options: stand firm or dig deeper. I chose to dig. And the more I dug, the more I didn’t like what I found.
First, nobody had zeroed the test leads before measuring. The multimeter’s own lead resistance was about 0.3 Ω. We were measuring the transformer plus the leads, not the transformer alone. That alone could account for most of the false failure.
Second, the multimeter was one we’d purchased for an automotive line, not for precision component testing. It was a solid unit for checking alternator output and parasitic drain. But the best multimeter for automotive diagnostics isn’t automatically the best multimeter for milliohm-level transformer measurements. I’d forgotten that distinction.
Here’s the thing: test equipment selection matters. A high-quality automotive multimeter might have a CAT III rating and a bright screen, but if it lacks low-resistance mode or manual lead null, you’ll chase ghosts.
Second-guessing the data
We re-ran the test the next morning with a four-wire micro-ohmmeter. I remember standing over the bench, watching the reading settle: 2.84 Ω. Within tolerance. Then we tested eleven more samples from the same lot. All of them were inside our acceptance limit.
Not ideal, but workable. We had a false positive. The “bad” TDK transformers were actually fine. Our test setup was the problem.
I have mixed feelings about the rejection decision. On one hand, our protocol caught something that looked wrong, and we stopped before shipping suspect parts. On the other, we wasted three days, made the vendor jump through audit hoops, and damaged our credibility with the medical customer. The lesson wasn’t “don’t reject.” The lesson was “verify before you blame.”
What I mean is that rejecting a component isn’t just about protecting the customer—it’s also about protecting the supplier relationship, because if you reject a good part for the wrong reason, you lose credibility; and once you lose credibility, every future negotiation gets harder.
As for the TDK InvenSense C300 blood pressure sensor, that turned out to be a different story entirely. In the same batch, we had two C300 samples with slightly higher offset voltage than expected. We initially suspected a soldering issue. Then we noticed our test fixture had a bad ground connection. When we fixed that and compared the sensor output to a calibrated reference, the readings were normal.
Honestly, I’m not sure why the bad ground affected one unit more than the other. My best guess is that the fixture’s contact resistance shifted the reference input, but I’d love to hear from someone with more sensor test experience. All I know is that the sensor itself was innocent. The fixture was the culprit.
There’s something satisfying about finding the real root cause. After the late nights and cross-checks, the payoff was seeing the C300 output match the reference after we corrected the fixture. That’s the feeling that keeps a quality person going.
The actual culprit
So what do we blame? Not the transformer. Not the sensor. The actual culprit was our verification workflow. It was too dependent on someone grabbing a multimeter that wasn’t meant for that measurement.
After that incident, I changed our incoming inspection procedure. The first step isn’t measuring the part. It’s measuring the test setup.
- Check the test lead resistance before each session.
- Use lead-null or zero-offset compensation when the DUT is under 10 Ω.
- For low DC resistance, use a four-wire method or a bench meter designed for it.
- Let the parts reach room temperature before measuring.
- Record the test setup in the same report as the measurement.
The 12-point checklist I created after my third mistake has saved us an estimated $8,000 in potential rework. That’s not a guess I make lightly—it’s from the five false rejections we caught before they turned into vendor disputes in the past year.
What I’d do differently
If you’ve ever had a project delayed by a “bad batch” that turned out to be a test fixture issue, you know the feeling. You also know how much easier it is to prevent it than to clean it up.
What I’d do differently:
- Verify the instrument before verifying the part.
- Write down the exact probing points before the measurement, not after.
- If a rejection would surprise the supplier, treat that as a red flag—not proof, but a reason to pause.
- For automotive electronics, choose a multimeter based on the measurement you need, not the vehicle you’re working on. The best multimeter for automotive diagnostics has the features you need, like true RMS, low-impedance voltage detection, and a quality probe set.
We also added a step for sensors like the C300: before testing, inspect the fixture contacts and ground reference. A tiny change in contact resistance can look like a failing component. That wasn’t in our old protocol.
Our medical customer still uses the TDK InvenSense C300 blood pressure sensor in their monitor design. The TDK transformers in that same lot went into production. No field failures, so far. But I’m careful when I say that—I don’t guarantee components for life, and neither should anyone else in this industry.
As of January 2025, FTC business guidance on advertising (https://www.ftc.gov/business-guidance/advertising-marketing) says claims must be truthful, not misleading, and substantiated. That applies to us as a distributor, to the sensor manufacturer’s documentation, and to the customer who ultimately puts the product on the market. I don’t call a sensor “medical grade” unless the manufacturer’s datasheet and regulatory filing support it.
So, bottom line: prevention over cure. Five minutes of verifying your test setup beats five days of explaining a rejection. Trust the data, but trust your test equipment first. And if a TDK transformer and a TDK InvenSense C300 blood pressure sensor suddenly look too far out of spec to be true—check the probes before you send that email.
