← Back to Insights

What a Batch of Kemet Capacitors Taught Me About 'Acceptable' — A Quality Audit Story

It Started With a Measurement Off by 8 Microns

In our Q1 2024 quality audit, I pulled a sample of 500 MLCCs from an incoming shipment. The supplier had labeled them as Kemet ceramic capacitors, part number spec'd at 0.1 µF, X7R dielectric, 0603 case size.

Our acceptance test bench measured a mean thickness of 0.89 mm against a nominal 0.86 mm. Three hundredths of a millimeter. On a 0603 capacitor, that matters — it can throw off pick-and-place nozzle calibration, and it can flag in a CT scan as inconsistent dielectric layering.

I've been doing quality and brand compliance review for over four years — roughly 200+ distinct lots annually. I've rejected maybe 12% of first deliveries in 2023 because of spec drift, labeling issues, or paperwork gaps. This batch looked like number 13.

The vendor's rep said: "That's within industry tolerance." I asked: within whose tolerance? Silence. Then: "You know, the general standard."

That answer cost us three weeks of back-and-forth and a $22,000 redo on a sub-assembly that had already been partially built with the non-conforming parts. The redo delayed our production launch by 11 days.

Why We Stopped Trusting the Label and Built Our Own Verification Protocol

Here's what nobody tells you about the passives supply chain: the brand name on the reel doesn't always survive the transit from manufacturer to distributor to broker to you. We were sourcing through what we thought was a direct channel. We weren't.

So in early 2023, I implemented a three-tier verification protocol. First: cross-reference the lot code against the manufacturer's public traceability database. Second: dimensional and electrical spot-check on 5% of reels (not just the top layer — that's a rookie mistake). Third: for anything claiming to be from Kemet Corporation, confirm distributor authorization directly with the manufacturer's channel team.

That third step is the one that catches problems. Why? Because authorized distribution is not a marketing phrase — it's a contractual relationship with audit rights. Brokers can say "Kemet group" on an invoice without ever touching a Kemet factory.

"Dodged a bullet when I started requesting written channel confirmation before approving any new supplier. Was one signature away from taking a shipment of 80,000 units that would have failed incoming inspection — and we'd have caught it too late to shift to an alternate source."

That near-miss happened in late 2023. The lot code claimed a Malaysia origin. The manufacturer's database showed that lot prefix was assigned to a factory in Bulgaria. Not that Bulgaria is a problem — the problem was that the paperwork and the database disagreed. In our system, disagreement equals rejection.

The Cross-Section That Changed How I Read Datasheets

For the Q1 2024 batch, I needed a different kind of evidence. Datasheet compliance wasn't enough. I needed to see inside the capacitor.

We use a Kemet lapping machine in our lab — a benchtop unit, older model, still reliable. If you're not familiar: a lapping machine grinds and polishes samples to a precise, flat surface for microscopic examination. Think of it as a very controlled sander (note to self: schedule the annual calibration on that unit — it's overdue).

We potted the suspect MLCCs in epoxy, lapped them down to roughly 60 µm thickness, and examined the dielectric layer under SEM.

What we found: inconsistent layer thickness across the same part. The ceramic dielectric showed visible voids — 3 to 5 µm pockets — that shouldn't exist in a properly fired X7R. And the electrode termination showed a nickel barrier layer that was half the thickness spec'd in the datasheet.

Coincidence? No. Those two defects — thin barrier layer and dielectric voids — correlate with counterfeit or gray-market product that was remolded or relabeled. We can't prove that here, but the internal structure told us this was not a standard Kemet ceramic capacitor from a current-production lot.

So glad I insisted on the cross-section. Almost skipped it to save time — would have approved a batch that had no business being in our product.

What the Vendor Said, What They Heard, and What We Both Learned

The frustrating part: this vendor wasn't malicious. They genuinely believed they were selling Kemet capacitors. Their upstream broker had provided what looked like legitimate documentation.

I said "we need original manufacturer certification for this lot." They heard "we need a certificate that looks original." Result: they sent a generic compliance certificate with a logo — not a lot-specific traceability record from the manufacturer.

We were using the same words but meaning different things. Discovered this when I asked for the factory name that produced the lot and got a P.O. box in Shenzhen.

After that meeting, we wrote a one-page spec sheet that defines every term we use: what "authorized distributor" means, what "lot traceability" includes (factory name, date code, reel label photo, database screenshot), and what "certificate of conformance" must contain (signatory name, title, lot number, part number, date). No ambiguity.

The vendor rep later told me that spec sheet was the most useful document they'd received from any customer. Not because we were strict — because we were clear.

The Lesson That Stuck

I used to think quality rejection was about finding fault. Now I think it's about removing ambiguity. The same principle applies when you're buying Kemet relays, connectors, or tantalum capacitors — or when you're evaluating what is a fair lead time, what is a real MOQ, what is an acceptable tolerance.

The lapping machine showed me the physical truth. The paperwork showed me the contractual truth. And the messy conversation with the vendor showed me the communication gap that caused both problems.

One small detail I left out: the day we were reviewing the SEM images, one of my inspectors was using a Kyocera DuraForce Pro 3 to photograph the reel labels in the warehouse. That phone survived a 1.2-meter drop onto concrete the week before. Not a sponsor, just a tool that works in a dusty environment (mental note: we should standardize the camera protocol for incoming inspection).

Transparent specs aren't a nice-to-have. They're the difference between a 3-week delay and a 3-day confirmation. Ask for the factory name. Ask what's NOT included in the certificate. And if the answer feels vague — that's your first quality signal.

author-avatar
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