If you’ve ever had a prototype pass every bench test only to fail spectacularly in the first production run, you know the sinking feeling I’m talking about. It usually starts with a Friday afternoon call: “Hey, something’s off with the boards. They’re… fuzzy. Maybe 5% under spec.”
This is the story of how I learned that not all components are created equal—and why a seemingly small difference in consistency became a $22,000 lesson.
The Setup, and the Assumption
In early 2023, I was managing quality for a mid-size medical device OEM. We were developing a new portable diagnostic tool—think a handheld reader for blood assays. The power management subcircuit was critical, and our lead engineer spec’d a Vishay current-sense resistor (the WSL series, if you’re keeping score) for current sensing.
Why Vishay? Honestly? Because the datasheet looked great, and the lead time was two weeks better than the next competitor. That was pretty much the extent of the decision process. And at the time, I didn’t question it. The design team had used Vishay in older products, and everything had been fine. So on paper, it was fine.
We ordered 5,000 units for the first engineering build. The resistors arrived, looked good, passed incoming inspection (resistance, power rating—the usual). Prototypes went together in a week. First bench test: perfect. Second test: perfect. Third test: still perfect. We were on track.
Then came the temperature chamber. That’s where things got… interesting.
The 57% Failure That Changed My Mind
The spec called for the current-sense circuit to maintain ±1% accuracy from -10°C to +60°C—fairly standard for medical handhelds. Our engineer, confident after the bench tests, sent ten prototype boards to the chamber for an overnight soak. The next morning, six out of ten boards were reporting measurements that drifted by 4-7% at the temperature extremes. That’s a hard fail for a diagnostic device. The remaining four boards? They were rock solid—within 0.8%.
At first, we blamed the PCB assembly. Maybe a cold solder joint? A grounding issue? We swapped boards, reran tests, and got the same split result. It wasn’t the layout.
Then our designer, who apparently had some experience with passive components, suggested we check the resistor lot code. We pulled the 5,000 resistors and found three different date codes mixed in the same reel. Three. And the resistors that failed? They were all from the earliest date code—the one that had been in inventory for 18 months.
Here’s where I made my first real mistake. I called our distributor. “We’re getting mixed lots,” I said. “Can you ship us a fresh, single-lot reel for the next batch?”
The answer was polite, but firm: “We can’t guarantee single-lot on that part. It’s a high-volume commodity resistor. Most customers don’t care.”
And they were right—mostly. Most customers didn’t care. But for a medical device, the temperature coefficient (TCR) drift across different manufacturing runs can vary by as much as 25-50 ppm/°C. That’s enough to push a ±1% circuit outside spec if you get the worst-case combination of lots.
So now I had a problem. I needed a resistor that could deliver consistent TCR across multiple production runs, but I didn’t have the leverage (or the volume) to demand single-lot shipments from a commodity line. And the alternative? Changing the design.
The Hail Mary (and Why It Almost Worked)
I asked our engineer: “Is there a version of Vishay’s resistor with tighter drift? Maybe a ‘professional’ or ‘automotive’ grade?”
He came back with an alternative: the Vishay WSLP series—a power metal strip resistor, same basic package, but with a TCR of ±75 ppm/°C max vs. the ±150 ppm/°C of the standard part. The cost was higher—about 40% more per unit—but still within our BOM budget. And more importantly, Vishay typically manufactured these in smaller, controlled lots, making single-lot shipments more feasible.
I placed a test order for 2,000 units, specifically requesting a single date code. The distributor came back: “Confirmed. Single lot. Lead time is the same.”
We re-ran the temperature chamber test with the new parts. Eight boards this time. All eight passed. Drift was consistent across all units—0.5-0.7% across the entire temperature range. The fix worked.
So, happy ending, right? Not quite. We had already built, and partially scrapped, the first 50 prototype boards. The rework and retesting cost us about $22,000 in engineering hours and materials. That’s the part I still think about.
What bothers me isn’t the money—it’s that the failure was entirely avoidable. The datasheet didn’t lie. The distributor didn’t lie. But our process didn’t ask the right question. We had assumed that because a part number existed, all units of that part number would behave identically. They don’t. They’re made at different times, on different lines, with different materials. The only way to guarantee consistency is to control the lot.
What I Actually Learned (Not the Obvious Stuff)
Conventional wisdom says: choose reputable brands and you’ll be fine. In practice, I’ve found that brand alone isn’t enough—it’s about how you specify the part within that brand’s portfolio. Every component line has a range of options: military, automotive, industrial, commercial. The commercial version might look identical on paper, but the process control applied to the automotive or precision line is often far tighter. That’s not a secret, but it’s rarely taught to design engineers.
So here’s what I did differently after that project:
- For any circuit that cares about temperature or drift (and that’s most of them), I now include a line item in the spec requiring a single manufacturing date code and a lot acceptance test (LAT) on the first 100 units. This adds about 2-3 days to procurement but costs essentially nothing compared to rework.
- I stopped trusting “standard” product grades for anything safety-critical. The cost delta is usually smaller than you think—5-40%—and is almost always cheaper than a field failure. The WSLP part I switched to was 40% more expensive per piece, but on a $1,500 board, that’s a rounding error.
- I now have a rule of thumb: if the component cost is less than 5% of the total BOM value, don’t economize on the part. Cheap out on the housing, the label, the packaging—not the thing that makes the circuit work.
Oh, and one more thing. When I later ran a blind side-by-side test with 20 engineers—first the original mixed-lot resistors, then the single-lot WSLPs—I asked them to identify which board was “more reliable.” Without knowing the difference, 17 out of 20 picked the single-lot board based solely on the consistency of the power-up behavior. Perception matters, but in this case, the perception was backed by data.
The Bottom Line (If There Is One)
Looking back, I don’t regret choosing Vishay. Their parts are solid. What I regret is treating a component brand as a guarantee of manufacturing consistency. They’re not the same thing. A brand can have the best datasheet in the world, but if your distributor ships you three different lots in one bag, you’re just rolling the dice.
If you’re designing something that has to work in a temperature range, or for a medical device, or for any application where drift matters, ask your supplier one question that I never used to ask: “Can you guarantee single-lot supply for the entire production run, and what’s the cost difference?”
The answer might surprise you. It might cost a little more upfront, but I can tell you from experience: it costs a whole lot less than a 22,000-dollar redo and a missed launch date.