ZTE Logo

Our Flip Phone Prototype Scrapped 47 Kemet Capacitors—and Kemet Wasn't the Problem

The 12% that shouldn't have happened

In March 2024, I stood on the production floor watching 47 MLCCs get pulled off our flip phone prototype. The project was internally coded Infinity. Roughly 12% of the ceramic capacitors on each board showed micro-cracks or leakage failures during burn-in. Not all of them were Kemet capacitors—but every single one that failed was.

By the weekend, our story was polished. Must be a bad supplier batch. Our quality engineer was drafting an RMA request. Someone on the team suggested moving the whole BOM to a different vendor. I nearly did.

It took me about three weeks to figure out the truth: Kemet didn't do anything wrong. I did. The spec sheet was mine. The expedite fee refusal was mine. The relay placement decision was mine.

What we thought the problem was

On the surface, the story wrote itself:

  • Capacitors: Same part number, same distributor, same lead time—but 12% failure on the Kemet capacitor lot versus zero on the previous batch
  • Relays: Two Kemet relays failed functional test
  • Timeline: Ordered in February, arrived first week of March, failures caught by end of March

My instinct was the standard procurement reflex: blame the supplier, request RMA, evaluate alternates. I almost kicked Kemet off our approved vendor list permanently.

Then our engineering lead asked a question I should've asked myself: "What spec did you give them?"

I didn't have a good answer. That was the moment everything shifted.

The deeper cause (and no, it wasn't Kemet)

I went back to the lab and re-walked every decision. There were three, and they compounded.

Cause 1: I specified rigid-termination MLCCs on a flex board

The flip phone form factor meant the main PCB was a rigid-flex assembly. When you bend a rigid board through 200,000 fold cycles, standard MLCC solder joints fatigue. The correct series is a flexible-termination MLCC—designed to absorb board-level flexing. Kemet makes these, and they run about 8–12% more per unit than standard.

I chose the standard series because it was cheaper on the quote.

That's not a Kemet problem. That's a me problem. To be fair, our distributor could've flagged it—but why would they? I gave them a part number and they quoted it. The spec was mine to own.

Cause 2: My Kemet relay vs Broadcom comparison ignored the thing that mattered

For signal switching inside the flip phone mechanism, I evaluated two options:

  1. Kemet relays—specifically their EC2 series, a tiny signal relay
  2. Broadcom solid-state relay—their ASSR series

I compared them head-to-head: package size, unit cost, switching speed. The Kemet relay was $1.70 cheaper per unit. I picked it.

What I didn't account for was mechanical mounting. We placed the Kemet relay near the hinge, where repeated mechanical shock would batter any electromechanical contact. The relay class was wrong for the environment—not the relay itself. A solid-state option (Broadcom's, or any SSR) would have been the more sensible choice for a high-cycle fold area. But honestly? The Kemet relay would've been fine if I'd mounted it with proper dampening or encapsulant.

I put an electromechanical part in a mechanical-stress zone. Nobody at Kemet recommended that. I did.

Cause 3: I saved $400 on expedite and spent $4,800 cleaning it up

When I placed the February order, we were on deadline pressure. Investor demo in the second week of April. The distributor gave me two options on the Kemet capacitor order (about $4,200 in total):

  • Standard lead time: 13 business days, no premium
  • Expedite: 5 business days, +$400

I chose standard. We were saving money.

Day 13, the capacitors arrived—but a subset was scrap, traceable to our own wrong assembly spec. We burned another nine days sourcing replacements at an 80% premium from a broker (which cost more than the original $400 expedite fee, by the way), and ate $4,800 in rework and scrap.

If I'd paid the $400, we'd have caught the design error a full week earlier. The expedite fee wasn't buying speed. It was buying certainty. I didn't understand the difference until much later.

What the mistakes actually cost

Let me add up the damage:

  • MLCC rework and scrap: $3,200
  • Relay replacement (2 failed units, plus 4 proactive swaps): $960
  • Expedited shipping on replacement caps: $480

Total cash outflow: roughly $4,640. Plus three weeks of engineering time, plus a very tense conversation before the investor demo.

The irony: when we finally sourced the right Kemet capacitor series (flexible termination) and repositioned the Kemet relays away from the hinge, they performed flawlessly through 150,000 fold cycles. Zero failures.

What I'd do differently—a 6-point checklist

I've kept a pre-purchase checklist on my desk ever since. It's short, and it would have caught every one of these.

  1. Verify the mechanical environment—is this component going near a hinge, flex cable, or high-shock zone?
  2. Check termination and dielectric—for flex PCBs, confirm flexible-termination MLCCs. No exceptions.
  3. Compare within the right category—Kemet relays and Broadcom solid-state relays are different technologies. Compare them on the actual use environment, not just unit price.
  4. Stress-test samples before quoting up—run your own fold-cycle or vibration test. Don't rely on datasheet specs alone.
  5. Budget expedite fees as a line item—if there's a hard deadline, treat guaranteed delivery as a required cost, not an optional upgrade. The $400 premium isn't the expense. The missed deadline is.
  6. Talk to the authorized distributor about the application—they can often suggest the right series if you tell them what you're building.

The short version

Supplier quality is rarely the problem. Our spec sheet is.

I almost removed Kemet from our approved vendor list over a mistake that was entirely on my side. The Kemet capacitor and Kemet relay were both solid parts. I just picked the wrong series, placed them in the wrong environment, and then tried to save four hundred dollars while burning forty-eight hundred.

Pretty humbling, honestly. But that's why the checklist exists now.

Share: LinkedIn Twitter WhatsApp
Tomas Ribeiro
Tomas Ribeiro

Tomas Ribeiro is a telecommunications reliability and procurement analyst covering optical modules, switches, routers, radios, enclosures, power systems, spares, and multi-vendor support plans. He uses IEC 62368-1 and ISO/IEC 27001 requirements alongside MTBF and MTTR data to compare interoperability, failure modes, firmware support, energy use, environmental limits, lead times, warranty boundaries, and total cost of ownership. His work helps infrastructure leaders, operations teams, and buyers qualify equipment, plan redundancy, manage obsolescence, and align technical risk with deployment budgets.

Leave a Reply

Your email address will not be published. Required fields are marked *