My name is Jack, and I manage test equipment purchases for a small RF lab. I've made eleven significant purchasing mistakes since 2017—totaling somewhere around $60,000 of wasted budget. This is the story of mistake number nine, and the one that finally forced me to build a checklist.
It started with a 5G mmWave board. The board's uplink frequency was 28 GHz. Our old generator stopped at 6 GHz. So I did what any reasonable person would do: I searched for a Keysight 40 GHz signal generator, found one with enough headroom, and started writing the justification.
If you've ever typed best multimeter for home use and ended up comparing lab-grade meters with more digits than your tax return, you already know where this is headed. The problem isn't the product. It's the question you're asking.
The Problem I Thought I Was Solving
On the surface, this was a frequency problem. I needed to generate a signal at 28 GHz. The lab needed a generator that could reach 28 GHz with margin. A 40 GHz source gave us 12 GHz of headroom. That sounded exactly right.
But I was solving a gap on the spec sheet, not a gap in the lab. That's a dangerous way to buy precision instruments.
The instrument I ordered was a Keysight 40 GHz signal generator with an analog output. It was a beautiful piece of engineering, and it generated exactly the kind of clean continuous-wave signal it was designed to produce. The board didn't need a clean CW signal. It needed a modulated signal with a defined bandwidth. An analog generator, even a very good one, was the wrong tool.
I should have known that before I clicked 'request a quote.' Instead, I learned it after the invoice. Even after I approved the PO, I kept second-guessing. What if the 40 GHz range wasn't enough? What if I should have asked for better phase noise? The ten weeks until delivery were stressful.
The Real Problem Wasn't the Frequency Range
I'm not an RF design engineer—I'm the person who approves the PO and then apologizes when the equipment doesn't work. But I've learned a few things since that purchase, and the first one is this: a microwave measurement is a system, not a box.
Here's something vendors won't tell you: the datasheet frequency range applies to the source's own output connector. It doesn't include the cable, the adapter, the attenuator, or the probably-battered jack on your device under test.
The 40 GHz output connector is typically a 2.4 mm precision connector. Our board had an SMA jack. You can't connect an SMA cable to a 2.4 mm port without an adapter. Every adapter is a discontinuity. At 28 GHz, even a good adapter can change your power reading by a few tenths of a dB, and a questionable one can turn a clean signal into a reflection nightmare.
So the first deep problem was not the generator's top frequency. It was the connector mating strategy between the generator and the DUT.
The second deep problem was the modulation capability. A 40 GHz signal generator has a frequency range, but it also has a modulation bandwidth and a waveform type. If the DUT needs 5G NR, a vector signal generator with internal baseband generation is usually the right starting point. I was looking at the frequency column of the datasheet and ignoring everything else.
The third deep problem was that a source is only one half of a measurement system. We didn't have a spectrum analyzer that could verify a 40 GHz signal. So we were about to own a generator whose output we couldn't check. That's like buying a precision torque wrench and then using it by feel.
If you look up Keysight Technologies, Inc., you'll find application notes about connector care and signal integrity. I should have read them before ordering. They might have saved me a very expensive lesson.
What the Purchase Actually Cost
After the base unit, the 40 GHz option, the high-stability timebase, the calibration report, and the necessary accessories, the number looked a lot different from the starting-at price in the brochure. Nobody was hiding it from me. I just didn't ask the right questions.
I'm a fan of transparent pricing. Not because it makes a quote look pretty, but because a line-item list forces you to confront the full system. The vendor who lists every option and fee up front—even when the total looks higher—is usually the one that costs less in the end. The uncomfortable question is not 'what is the price?' It's 'what is NOT included?'
Annual calibration of that 40 GHz instrument was quoted at roughly $3,000 when I last checked in Q4 2024. That's more than a very good handheld multimeter. You're paying for traceable uncertainty at frequencies that most labs will never use. If you don't need 40 GHz, you are paying to certify a range you're going to ignore.
And then we made the adapter mistake. The generator arrived. We connected it to the board through a 2.4 mm-to-SMA adapter. The measured output was 1.7 dB low. We spent two weeks troubleshooting the generator—phase noise, cables, connectors, everything. The problem was the adapter. It wasn't expensive. It just wasn't part of the original quote, and I hadn't thought to ask about it.
The two weeks of troubleshooting, the delay on the engineering schedule, and the credibility loss with my team were the real bill. The generator had become a symptom of a deeper problem: we bought equipment before we defined the measurement.
The Checklist That Would Have Stopped Me
Take it from someone who learned the hard way. If you're considering a Keysight 40 GHz signal generator—or any microwave source—answer these five questions first. Write them down. Force yourself to sit with them.
- What is the highest frequency at the DUT's RF jack, and what connector type is on that jack? Not the data rate, not the carrier frequency. The actual connector.
- Does your signal need modulation? CW, swept, or modulated? If modulated, what bandwidth and waveform?
- Do you already have a calibrated instrument to verify the output? If not, put the analyzer in the budget before the generator.
- What is the annual calibration cost, multiplied by the years you plan to own it?
- What is NOT included in the quote? Cables, adapters, software licenses, training, calibration, and warranty options all have a way of appearing after the PO is approved.
That list has caught 23 potential mistakes since I started using it. I'm not saying it fixes everything. It won't tell you which vector signal generator to buy or how to model a cable. But it will stop you from buying a box that doesn't connect to the thing you're trying to test. Trust me on this one.
What Does Best Multimeter for Home Use Have to Do With This?
More than you'd think.
The phrase best multimeter for home use is a trap for the same reason 40 GHz was a trap for me. It encourages you to shop by max spec instead of by actual use. People ask for the highest accuracy meter within their budget and then check the same three things: batteries, wall outlets, and Christmas lights. You do not need six digits of resolution to check whether a socket is alive.
For home use, the best multimeter is one with a fused current input, a CAT III 600V safety rating at minimum, a readable display, and enough accuracy to tell you about 120 volts. A basic meter from any reputable brand will last for years. I own a Keysight handheld for bench work, and I like it. But if I were only doing home repairs, I'd spend the extra money on good probes and quality batteries, not on another digit of resolution.
The same logic applies in reverse. If your lab genuinely tests at 40 GHz, a Keysight 40 GHz signal generator is a legitimate investment. Just don't buy it because you want headroom. Buy it because you've traced the signal path, accounted for connectors, hired a calibration budget, and verified that the rest of your measurement chain can keep up.
The 40 GHz generator still sits in our lab. Once we bought the right analyzer and the right 2.4 mm accessories, it became useful. But I could have made that purchase nine months later, after the measurement plan was ready, and saved myself a lot of grief.
Bottom line: if you can't trace the signal from the generator's connector to the DUT's jack, you haven't finished the measurement design.
My name is Jack. I document my mistakes so the next person doesn't have to.
