October 2022. 11:47 PM. I'm in the lab with a prototype blood pressure monitor that just spent 12 hours in a cold storage shed. The display reads 72.1 kg. The reference scale says 67.8. I press zero, wait for the device to warm up, and watch the number climb to 76.3 kg. Four kilograms of phantom weight, appearing from nowhere.
It wasn't one bad unit. It was every unit with the load cell I'd approved.
Let me explain how I got here, because if you've ever picked a sensor for a product, you might recognize the trap.
How I Ended Up Choosing the Wrong Load Cell
I'm a product engineer at a connected health device company. For the last six years, I've been the person who selects sensors — load cells, thermistors, pressure sensors, the whole menu. I've personally made (and documented) a dozen or so significant mistakes in that time, totaling more wasted budget than I like to admit. This one takes the trophy for the most avoidable.
A quick disclaimer: my experience is based on roughly 20 sensor selection cycles across three product lines, mostly medical and consumer health. If you're designing for aerospace or automotive, some of this won't translate. Those worlds operate by different rules. But for the rest of us, I'm pretty confident the pattern holds.
The product was a "platinum" edition blood pressure monitor — a premium unit for clinics and serious home users, with a seat-integrated scale that measured weight and blood pressure in one sitting. Calling it "platinum" kind of locked us into a certain level of accuracy. Which, in hindsight, was a commitment we should have taken more seriously from day one.
I spec'ed the Vishay 3310 from the start. It's a single-point load cell (designed to sit under a scale platform and handle off-center loads), anodized aluminum, 5 to 50 kg capacity. Boring in the best way. It has built-in temperature compensation, OIML approval for legal-for-trade weighing, and it comes from Vishay Israel's load cell operation — the line that traces back to Tedea-Huntleigh, a company that's been making load cells since the 1970s.
Then procurement found a cheaper alternative. 22% lower unit cost. "Same specs," they said. I went back and forth for two weeks. The trusted supplier with the boringly perfect 3310, or the new vendor with real savings. On paper, the cheap part looked acceptable. My gut said "don't."
I still kick myself for overruling my gut.
The Cheap Load Cells Looked Fine. That Was the Problem.
We ran the alternative parts through our standard bench tests. At 25°C, they passed. Linearity was correct. Hysteresis was acceptable. Repeatability was surprisingly good. I told my test engineer, Victor, that maybe this would work out. (Should mention: Victor has a personal rule that every medical device we build has to survive a night in a cold storage room, because "one day a patient will keep this in a garage in Minnesota." Victor is usually right.)
The bench tests went so well that we assembled 30 prototypes and shipped them to friendly clinics for beta testing. For two weeks, nothing. Then a beta tester called: the integrated scale was showing patients 4 to 7 kg heavier than their actual weight, usually in the afternoon when the exam room heated up.
That's when Victor put one of the prototypes into the cold shed. And that's how I ended up in the lab at 11:47 PM, watching a medical device invent four kilograms from nothing.
I pulled the vendor's datasheet, looking for answers. It listed "temperature effect on zero: 0.02% of rated output per °C." On a 50 kg cell, that's 10 grams per °C. But "typical," it said. Not "maximum." The real-world parts drifted far beyond that number, and the vendor's spec was, to put it generously, aspirational.
This is where I learned a distinction that sounds pedantic and isn't: typical values describe the parts the vendor is proud of. Maximum values describe what you can rely on across all units. Vishay gives you maximums. This vendor gave us typicals, and the gap between the two was the gap between our budget and our credibility.
A four-kilogram error in a bathroom scale is annoying. In a "platinum" blood pressure monitor, where patient weight can influence readings and clinical decisions, it's a liability. And per FTC advertising guidelines (ftc.gov), a claim like "platinum-grade accuracy" has to be substantiated with evidence. A calibration sheet from a vendor that only tested parts at 25°C was not evidence we wanted to show anyone.
So we recalled all 30 prototypes, scrapped the cheap load cells, and started over.
Going Back to Vishay, After the Damage
The 3310 isn't the most exciting component in the world. That's precisely the point. Its datasheet is brutally honest, with maximum specifications for combined error, temperature effect, and creep. The units we use are rated for -10°C to +40°C (as of the January 2025 datasheet, at least). And the 3310 carries OIML R60 certification, which is the international reference for load cells used in scales where accuracy is not optional.
When I called our Vishay rep to place the order, she confirmed four-week lead time, then offered an expedite because it was a redesign. I didn't ask about the premium. Three weeks of delay were already on my head; a few hundred dollars of expedite was nothing.
Victor ran the same cold-room test on the Vishay units. Worst zero drift: 0.05 kg. Fifty grams on a scale rated for 150 kg. I wrote "acceptable" in the test report and underlined it. In our team, "acceptable" is how I say "flawless."
The product eventually shipped in January 2023, three weeks late. The clinics we kept in touch with have reported zero drift issues since.
Now the cost math, which I hate but will own: the cheaper load cells saved us $3,800. The switch back cost us $4,800 in rework, retesting, and expedited shipping. That's a $1,000 net loss before you count the delay, the lost engineering time, and the credibility hit with our beta customers. The unit price was never the real cost. The total cost of ownership — base price, setup, rework, shipping, and the risk of a failed product — is what actually matters, and the lowest quoted price rarely wins that calculation.
Why Phones Are Strong and Load Cells Matter
Somewhere in the middle of those three weeks, a friend asked me, "why are phones so strong now?" He'd dropped his phone on concrete, and it survived. I almost laughed — I had load cell drift on my mind, and he's asking about phone durability.
But the connection is real. Phones are strong now because engineers have measurement tools precise enough to see how materials deform, crack, and absorb impact. Drop test rigs use force sensors — the same strain gauge technology inside a load cell — to measure exactly what happens when a phone hits the ground. A strain gauge is just a tiny foil grid that changes electrical resistance when you stretch it. Fifty years ago, that was exotic aerospace technology. Today it's in a $500 phone, a $150 bathroom scale, and the blood pressure monitor your doctor uses.
That's the part of the industry that keeps evolving. What was best practice in 2020 may not apply in 2025. Ten years ago, precision load cells at consumer price points were almost unheard of. Now they're expected. But the fundamentals — temperature compensation, maximum vs. typical specs, traceable calibration — haven't changed, and they won't. Physics doesn't care about procurement pressure.
My checklist has one new item since this project: Did anyone run a temperature sweep? That question has caught 11 potential errors in the last 18 months. (Not a huge number, but each one would have been a disaster.)
If you're choosing a load cell for anything — a medical scale, a retail scale, an industrial weighing system — start by looking at the 3310. It's not flashy. It comes from a brand that doesn't need a pitch from me. But it's honest, and honest components make for boring, reliable success.
And if a vendor ever says their part is "same specs as Vishay," ask for three things: the maximum temperature drift, the OIML certificate number, and a test report with real data. If they hesitate — run.
I wish I had.