Why My Biggest Measurement Mistakes Had Nothing to Do With the Equipment

Here's a confession: after fifteen years in electronics test and measurement, my most expensive mistakes weren't caused by faulty equipment. They were caused by me assuming the instrument was right and my setup was fine.

I've watched engineers blame a perfectly good Keysight digital multimeter for readings that were actually compromised by a bad probe connection. I've seen teams condemn an LCR meter for measurement drift that was really thermal settling — theirs, not the meter's. And I once wasted a full day chasing a glitch that turned out to be my own ground lead.

You want to know the uncomfortable truth about test equipment? The instrument is almost never the problem. The operator's assumptions are.

This isn't a popular opinion. Equipment vendors (including my own employer) spend a lot of marketing budget telling you their hardware is the solution. And look, I sell Keysight products — I believe in them. But I've also made enough mistakes to know that upgrading your oscilloscope won't fix a flawed measurement technique.

Let me walk you through the specific failures that changed how I work — and what they taught me about using LCR meters, multimeters, thermal cameras, and even a Tektronix scope I had to borrow when ours was in calibration.

Mistake #1: The LCR Meter Reading That Looked Perfect (and Was Completely Wrong)

In early 2023, I was validating a batch of capacitors for a power supply redesign. I had a Keysight U1733C LCR meter on the bench — a solid handheld unit, reliable, calibrated three months prior. I measured a few components, got readings within tolerance, and signed off.

Then the first prototype failed. Not subtly — the output voltage was oscillating badly enough that the protection circuit kept tripping. I traced it to the capacitors I'd 'validated.'

Rechecking with the same U1733C, I got different values. Same meter, same components, different results. What changed?

I'd been holding the tweezers-style test leads with my fingers. The day of the first test was cold and dry; the day of the failure analysis was humid. My body capacitance and the moisture on my skin were changing the measurement. Not by a little — by enough to pull some capacitor readings out of tolerance.

The U1733C is a capable instrument. It wasn't lying to me. I was feeding it bad data through my own test setup.

Lesson learned: test leads are part of the measurement. Your hands are part of the test setup. If you're measuring high-impedance components, use a fixture — or at minimum, keep your fingers away from the leads and let the device settle to ambient temperature.

That lesson cost me roughly $1,800 in scrapped prototypes and a week of schedule. Real money. Avoidable money.

Mistake #2: Trusting a 'Calibrated' Multimeter Without Understanding What Calibration Actually Covers

This one stings because I should have known better.

I was troubleshooting a sensor interface that had intermittent readings. The suspect was a precision voltage reference. I checked it with a Keysight digital multimeter — the 34461A, a 6.5-digit bench unit that was within its calibration window. It measured a clean 4.096V. I declared the reference healthy and went hunting elsewhere.

Three hours later, I was still chasing ghosts. On a whim, I re-verified the voltage reference with a different meter — same model, also calibrated — and got 4.092V. A 4-millivolt discrepancy that shouldn't exist between two calibrated instruments.

Here's the thing I'd forgotten: calibration certifies accuracy at specific points, under specific conditions, with specific warm-up times. My 34461A had been sitting cold for 72 hours. I'd switched it on and measured within minutes. The other unit had been running all day and was thermally stable.

The 'calibrated' meter was technically in spec — just not at the temperature it was at when I used it. Keysight specifies a 90-day accuracy for that instrument with a specific warm-up period. I'd violated the operating conditions and then blamed the equipment.

That 4-millivolt error cost me a three-hour investigation. In my world, three hours of engineering time is roughly $400 of budget. For nothing.

The takeaway: know your instrument's warm-up spec. Treat it as part of the setup procedure, not a suggestion.

Mistake #3: The Thermal Camera That Showed a 'Cold' Board (And What It Was Hiding)

Thermal hunting is my favorite debugging tool. Nothing reveals a failing component faster than a hot spot on the board. But thermal cameras lie — not maliciously, but through emissivity.

I was evaluating a handheld thermal camera for our lab — a good one, with a decent microbolometer sensor and a reasonable pixel count for the price. We were testing it on a motor driver board that kept overheating. I pointed it at the suspected MOSFET. It read 65°C — warm, but within the datasheet limit. I was about to clear that component.

Then I noticed the heatsink was reading colder than the board around it. That's wrong. Heat sinks don't cool themselves below ambient unless something is very unusual.

The issue was emissivity. The MOSFET had a dull, rough package (high emissivity, accurate reading). The heatsink was bare aluminum — polished, reflective, with an emissivity around 0.05. The camera was reading reflected ambient temperature, not the actual heatsink temperature. Once I adjusted the emissivity setting and added a layer of high-emissivity tape, the heatsink read 14°C hotter. The MOSFET was actually running 12°C above its rated maximum.

Mistakes like this are why I tell junior engineers: a thermal camera is a tool for finding relative hot spots, not absolute temperatures. If you're making a thermal-based decision, you need to know the emissivity of every surface you're measuring — or you're just guessing.

Mistake #4: Borrowing a Tektronix Oscilloscope and Assuming It 'Works Like Ours'

I'll admit this one with some embarrassment, because I was the senior engineer on the project and everyone else assumed I knew what I was doing.

Our primary oscilloscope was in calibration for two weeks. A colleague loaned us his personal Tektronix scope — a perfectly good instrument, honestly one of the better ones I've used. But fifteen years of muscle memory on a different UI made me skip past the setup menu.

I didn't check the probe compensation. I didn't verify the bandwidth limit was off. I didn't even confirm the vertical scale was at the actual attenuation setting of the probe.

The result: a switching waveform that looked 'clean enough' to approve. It wasn't until we routed the same signal through a Keysight bench meter that the discrepancy appeared. The switching noise was nearly 40% higher than the borrowed scope had shown. The probe wasn't compensated for the scope's input capacitance, so the high-frequency content had been attenuated.

That 'minor oversight' resulted in a design that failed EMC testing at the first attempt — costing us $4,200 in re-test fees, not counting my bruised ego. And honestly, no one was more frustrated than the colleague who'd trusted me with his personal scope.

Since that incident, I've standardized our team's pre-test checklist so that anyone using an unfamiliar instrument — regardless of brand — follows the same verification steps:

• Probe compensation check (not optional, even for 'quick looks' — it takes 30 seconds).
• Vertical scale confirmation against a known reference.
• Bandwidth limit setting — verify it's set to full bandwidth unless deliberately filtered.
• Input impedance matching (1Ω vs 50Ω, depending on the setup).
• Timebase, trigger source, and trigger level verification.

You might think this is obvious. You'd be surprised how often experienced engineers skip these steps because they 'know their scope.' The moment you pick up another lab's instrument, you're A/B testing your own assumptions against the vendor's design decisions. And like it or not, your assumptions are the untested component in the system.

Why This Matters: The Most Expensive Component in Your Setup Is Your Assumption

Here's an uncomfortable truth from my fifteen years of test engineering: when an instrument gives you a wrong answer, it's almost always because you asked it the wrong question.

People assume the vendor is responsible for measurement accuracy. But measurement accuracy is a chain: the instrument, the probe, the connection, the environment, and the person operating it. The instrument is the most reliable link in that chain. The operator is usually the weakest.

I know this because I've been the weakest link. Multiple times. With expensive consequences.

The mental model I now use:

• First step is always skepticism — of your setup, not just the reading.
• Second step is verifying the instrument (not the calibration label — the actual warm-up state, probe compensation, and test lead integrity).
• Third step is verifying the environment — temperature, humidity, emi/rfi sources.
• Fourth step is understanding the measurement principle — an LCR meter measures impedance at a specific frequency; a thermal camera detects IR, not temperature; a scope shows you what its input stage passes.
• Only then do you trust the reading enough to act on it.

Counterpoint: But Good Equipment Matters

I can already hear the objection: 'Sure, but your argument doesn't mean cheap tools are fine.' And you're right. Equipment quality matters. A lot. I wouldn't go back to a 20 MHz analog scope any more than I'd trade a U1733C for a no-name LCR meter that drifts with temperature.

But here's the nuance: the gap between a mid-tier and premium instrument is usually a small percentage error. The gap between a careless and careful operator can be 100% error or more. When I see engineers obsess over a 0.1% accuracy spec while holding the test leads in their hands and letting their body capacitance corrupt the reading, I know they're optimizing the wrong variable.

Upgrade your tools when the tool is the bottleneck. But first, ask yourself: is it really the tool?

What I Now Do (Recap, Because It's Important)

1. I keep a written checklist for every instrument I use. Yes, a physical piece of paper. Fifteen years of experience and I still use a checklist. That's because the mistake isn't about not knowing — it's about forgetting what you know when you're under schedule pressure.
2. I verify temperature and warm-up state. If an instrument has a 30-minute warm-up spec, I give it 30 minutes. Calibration is meaningless otherwise.
3. I test the test setup before I test the unit. Measure a known reference first. If the reading is wrong, fix the setup before trusting anything else.
4. I compensate probes — always. Even on my own scope. Even when I 'just did it last week.'
5. I use fixtures for accurate measurements, not freehand probing. The U1733C is a great tool, but it's more accurate with a test fixture than with free-floating leads. That's just physics.
6. I check emissivity when using a thermal camera. For anything shiny or metallic, I use tape or paint to create a known high-emissivity surface — and I always verify temperature with a contact probe if thermal-based decisions are critical.

This checklist may look basic. It's supposed to be. But here's the honest truth: I made every single mistake on this list at least once — and some more than once — before I took the time to write it. The result? In the past 18 months, the only inaccurate measurements that have come out of my lab were ones where I skipped a step because I was confident. And those were exactly the times I got burned.

The Bottom Line on Test Equipment and Test Engineers

Ask a room of test engineers what an LCR meter does, and every hand goes up. Ask them what high-impedance measurement technique they use, or how long their multimeter needs to warm up before meeting its accuracy spec, and the hands get quieter.

The cynical view might be that you need expensive instruments to compensate for inexperience. I believe the opposite: that a good engineer produces trustworthy results on modest instruments, and a careless engineer will produce garbage on the most expensive instruments Keysight, Tektronix, or anyone else sells.

I'm not saying equipment doesn't matter. I'm saying your understanding and your process matter more.

And the next time your measurement seems wrong?

Check your assumptions first. Then blame the instrument.