Murata Parts Aren't the Problem — My Design Was. Here's the $3,184 Lesson

I've been handling component selection and BOM engineering orders for about eight years now. I've personally made (and documented) 12 significant mistakes, totaling roughly $7,000 in wasted budget. These days I maintain our team's pre-flight checklist so the people after me don't have to learn the way I did.

Last month a customer asked me what makes a "best cordless phone" — you know, the kind that tops those roundup articles. I gave him my honest answer: it's not the logo on the box, it's what's inside. The RF module, the audio alert components, the power management. Then I went back to my desk, looked at the prototype that had a "bad" Murata piezo buzzer and a 5G module that "kept failing," and felt like a total fraud.

Because for three weeks, I blamed Murata parts for a design failure.

And it wasn't the parts.

The G310 5G board that looked great but sounded terrible

The project was a fixed-wireless access device built around a G310 5G module (a compact module that handles the full 5G protocol stack). It had a Murata piezo sounder for the ringer, GRM Series MLCCs on the power rail, and a trace antenna with a matching network I had tweaked to "improve it."

Design review passed. BOM was approved. The prototype came back. And the ringer was inaudible — genuinely, measurably silent. When it did transmit, power rail ripple spiked to 78mV peak-to-peak, way over the 30mV budget.

I did everything wrong, systematically:

• Checked firmware: the PWM was running, no issue there.
• Raised the GPIO drive strength: nothing changed.
• Swapped two different Murata piezo buzzers from our parts drawer: both sounded the same. "See? Bad part!"
• Filed an RMA with the distributor. (ugh, that email is still in my sent folder and I still wince when I see it)

It wasn't until I actually sat down and read the Murata piezo datasheet — not the summary page, the graphs — that I found what I'd missed.

What I actually got wrong

Piezo: I treated a resonator like a speaker

Murata piezo elements are resonant devices. The datasheet lists a resonant frequency (4.0kHz for the part we used) for a simple reason: that's where the sound pressure peaks. Drive it at 1kHz — which is what the firmware's default PWM was doing — and the output drops by about 15-20dB. That's the difference between the rated 70dB at 10cm and the roughly 50dB we were getting. The part was technically working the whole time. It was just being driven at a frequency it had no interest in.

I had picked the right Murata piezo and then fed it the wrong signal. SPL numbers don't mean anything if you don't respect the resonant frequency spec attached to it.

You don't choose a piezo for how loud it's marked. You choose it for whether your drive circuit can excite it at the frequency it actually wants to vibrate at.

MLCC: the 10µF that wasn't 10µF

Same design, same lesson buried in the fine print. I had placed a 10µF Murata MLCC on an 18V rail, expecting it to behave like 10µF. When the rail ripple hit 78mV, my first instinct was to blame the cap.

It wasn't the cap. It was the DC bias curve — the graph every MLCC datasheet includes that shows you how capacitance drops as a function of the DC voltage you apply across the part. At 18V DC, that particular 0805 component (an EIA standard case size, by the way) gives only around 55% of its nominal value. So the 10µF I thought I had was closer to 5.5µF in the real operating condition. The part was performing exactly as documented. I just never opened the graph.

People think expensive vendors deliver better quality. Actually, vendors who deliver quality can charge more. The causation runs the other way: Murata's parts behave the way their datasheets say they behave, and it was my job to read what they said. Not to glance at the headline number.

RF: "improving" the reference design of the G310 5G module

And then there was the G310 5G module itself. It has a reference design document. It specifies the antenna matching network, the decoupling, the whole front-end layout. I looked at it, decided it was "too conservative," and modified the matching network to what I thought was a better impedance profile.

Honestly, I'm not sure what I was thinking. My best guess is that every engineer has that impulse to add value by changing something — and the reference design looks like an invitation to improve. It isn't. The G310 5G module was already validated against 3GPP Release 15 RF requirements with that specific matching network. My variant was not validated. The board's failure to hit the target transmit power in some channels wasn't a defect in the module — it was my "improvement."

When I reverted to the reference design, the exact same module passed every test. Same serial number. Same solder joints. The module had never been the problem. I had.

What the failure actually cost

Here's the total, because I keep the invoices (note to self: always keep the invoices):

• PCB re-spin to fix the piezo drive circuit and power decoupling: $2,700
• Expedited assembly and rush components: $484
• Schedule slip: about 2.5 weeks. I remember because the project lead sent calendar invites with "REVIEW PROTOTYPE" in red, three sprints in a row.
• Engineering time for three people debugging: roughly 60 hours.
• Retracting an RMA that should have never existed: $0 in cash, but honestly, the most painful part.

I don't have hard data on industry-wide failure rates for piezo or MLCC parts, but based on our five years of orders, my sense is that under 1% of our "component failures" were actual defects. The other 99% were design errors or usage mistakes. Earlier that same year I approved a BOM with a Murata inductor whose current rating I'd miscalculated. On a 2,400-piece order, every single unit needed manual rework. That was another $890 in labor plus a weekend I still haven't emotionally recovered from.

The checklist that caught 47 errors in 18 months

After the third expensive lesson, I created a pre-flight checklist. I named it "Before You Blame Murata." Since then, it's caught 47 potential errors across our projects. Most of those fixes cost nothing at the design stage. Here's the short version:

1. Murata piezo: match the resonant frequency, not just the voltage

Look up the resonant frequency spec, confirm your PWM or drive waveform hits it, and check the source impedance the datasheet assumes. Only then look at SPL. If you're designing a cordless phone and your customer searches for "best cordless phone," this is the difference between a ringer that emits a polite buzz and a ringer people actually hear from across the room.

2. MLCCs: check the DC bias curve for every part above 10V

Pull the curve from the datasheet or check Murata's free SimSurfing tool. If you need 10µF at 18V, pick the nominal value whose bias curve gives you 10µF at 18V — often that means selecting a 22µF or 47µF part. That's not overspec; it's accounting for physics.

3. RF modules: treat the reference design like the law

If the G310 5G reference design says use a 1.2nH inductor in the matching network, use it. If you think it's wrong, measure with a network analyzer and justify the change with data. If you don't have a network analyzer, then the reference design is the data. The module was tested with it. You weren't.

The parts were never the enemy

I still kick myself for not building this checklist earlier. I had the datasheets all along. The Murata components performed exactly as documented — the piezo resonated where it should, the MLCC accepted DC bias physics it can't ignore, and the G310 5G module was a reliable RF engine that I bedeviled with an unnecessary layout change.

So the next time you're debugging a "failed" Murata part, ask yourself three things: What did I change from the reference design? Which graph did I skip? What assumption did I make that the datasheet never made?

Go fix those first. The component was probably fine.