Two Ways to Face a Deadline
On a Thursday afternoon in March 2024, 36 hours before a product launch, the compliance lab called with bad news: our new 5G module had failed radiated emissions at 3.7 GHz. The first instinct on the engineering call was unanimous—rerun the test on the legacy spectrum analyzer and hope the lab's setup was the problem.
That analyzer showed a clean spectrum. The certified lab still showed a fail.
I'm a senior test engineer at a wireless module manufacturer. I've handled 200+ rush measurement requests in 12 years, including same-day turnarounds for certification-critical clients. And based on our internal data from those 200+ requests, I can tell you that this is the moment where most teams discover a painful truth: what your instrument is averaging away might be exactly the problem you're looking for.
This article compares two approaches to the same urgent measurement problem. Approach A is the legacy bench—the analyzer or LCR meter that's been on the rack for years, still calibrated, still "good enough." Approach B is the modern Keysight setup, specifically the Keysight MXA signal analyzer for network-level measurements and the Keysight impedance analyzer E4990A for component-level characterization. I'll compare them across three dimensions: network visibility, component characterization, and time-to-trusted-data.
When I'm triaging an urgent test request, the first question I ask isn't "how many hours do we have?" It's "how much do you trust the data you're already looking at?" If the answer is "probably enough," the risk just went up.
What Is Networks? The Blind-Spot Test
Let's start with a question that sounds too basic to ask: what is networks? In electronic measurement, a network is simply a path for a signal—a transmission line, a wireless channel, a filter, even a power distribution plane on a PCB. In 2005, most of those paths carried narrowband signals with straightforward modulation. A spectrum analyzer that displayed amplitude versus frequency was enough to tell you whether things were OK.
That changed, and the industry has quietly split in two. A modern 5G NR signal carries hundreds of subcarriers across a 100 MHz channel (400 MHz in millimeter-wave bands), using modulation like 256-QAM that is unforgiving to small distortions. Wi-Fi 6E and Wi-Fi 7 push similar complexity into higher bands. The question is no longer "is there a signal?" The question is "is the signal healthy enough that a receiver can decode it?"
Here's the surface illusion: people assume an old spectrum analyzer can still see everything because it sweeps wide. The reality is that sweeping and demodulating are different acts. A legacy analyzer can show you the envelope—the overall shape of the spectrum. It cannot show you the error vector magnitude of a 256-QAM subcarrier, or a transient intermodulation spike that appears only when the signal is fully loaded. The Keysight MXA signal analyzer sweeps, demodulates, and captures events in real time.
When I compared our legacy analyzer and the MXA side by side on that failing 5G module, I finally understood why "the signal looks clean" is not a measurement. It's a hope. The MXA gave us something defensible.
Keysight Impedance Analyzer E4990A vs. the LCR Meter That Worked Fine
The second dimension is component-level, and it has an equally persistent myth: that a reliable LCR meter is enough to characterize passives. It took me about three years and several unexplained EMC failures to understand that a single capacitance value at 1 kHz tells you almost nothing about how that part behaves at 100 MHz.
The Keysight E4990A impedance analyzer covers 20 Hz to 120 MHz—or rather, 120 MHz is the top of the range with the wide-frequency option; there are versions that stop at 10, 20, 30, or 50 MHz. That range matters because it lets you see the physics that a single-frequency LCR reading hides. Instead of one capacitance value, you get a full equivalent circuit: series and parallel capacitance, resistance, inductance, phase angle, and an impedance sweep across frequency.
A real example: last quarter, a customer called with an urgent request. Their IoT device was failing conducted emissions, and the decoupling capacitor on the switching regulator measured "fine"—100 nF, low dissipation factor on their LCR meter. When they ran the same component on the E4990A, the sweep showed self-resonance at 12 MHz, right where the regulator was switching. The capacitor wasn't decoupling at that frequency. It was acting as an inductor. Swapping to a component with a higher resonant frequency fixed the design in two days.
The old LCR meter was technically accurate. Put another way: it wasn't wrong, it was blind. An LCR meter gives you a number. The E4990A gives you the behavior.
The Emergency Metric: Time-to-Trusted-Data Isn't Infinity
Now for the dimension that matters most when a deadline is collapsing: how long until you have a measurement you can defend to a third party?
The legacy workflow looks like this: power on, wait for warm-up, run open/short calibration, configure the sweep by hand, measure, interpret, and then—if the result is questioned—do it all over again. When a number doesn't match your expectation, the default move is to remeasure, and remeasuring eats the exact hours you don't have.
The modern workflow is different. The MXA's internal calibration runs quickly. Its measurement applications set up 5G NR and Wi-Fi tests automatically. The E4990A produces the equivalent-circuit model in one run. Neither instrument removes engineering judgment, but both remove the grunt work that legacy equipment pushes onto the operator.
From the outside, passing and failing look binary. In measurement terms, they're not infinity apart—it's often a few dB of transient headroom that an old analyzer averages away. Based on our internal data from 200+ urgent requests, the old-vs-new difference is roughly 35 to 45 minutes per measurement. That doesn't sound critical until you multiply it by three failing samples and a certification deadline.
Here's the conclusion that surprised me: when an emergency happens, the modern instrument is actually the cheaper option. We paid $3,800 for an expedited MXA rental in March 2024 (which, honestly, felt like a lot at the time). The alternative—waiting for the next certification slot and delaying the launch by a quarter—was estimated at $60,000 in lost channel revenue. The math wasn't close.
I have mixed feelings about saying this, because part of me wants to encourage teams to squeeze more life out of existing equipment. Another part remembers the overtime and failed test runs from the years when we didn't trust our own measurements. The compromise I've landed on: keep the old analyzer for narrowband checks, but never put it between a critical deadline and a compliance decision.
Company Overview: Keysight's 85-Year Head Start
A quick company overview, for context. Keysight's roots go back to 1939, when Bill Hewlett and David Packard founded what later became HP in a Palo Alto garage. The electronic measurement side of that business spun off as Agilent Technologies in 1999, then became Keysight Technologies in 2014. That lineage matters for a practical reason: the company's instruments, calibration labs, and people have helped define how electronic measurements are performed for more than eight decades.
For an engineer in an emergency, the two most useful outputs of that history are traceability and standards alignment. Keysight's calibration services are ISO/IEC 17025 accredited, and calibration certificates are traceable to NIST. When a compliance engineer questions your data, "our analyzer is calibrated and traceable" carries substantially more weight than "we checked it ourselves and it seemed fine."
Current Keysight products also track the standards your certification lab actually uses—3GPP 5G NR test cases, IEEE 802.11be, and the signal-integrity requirements showing up in AI and data-center hardware. The name on the front panel shouldn't be your only criterion, but the credibility it brings is exactly what a disputed result needs.
So What Should You Actually Choose?
If you're deciding between these two Keysight instruments, the answer depends on where your risk lives:
- If you qualify or debug components—capacitors, inductors, filters, materials—the Keysight E4990A impedance analyzer is the tool that turns vague EMC issues into a specific, measurable component behavior.
- If you test radios, 5G modules, Wi-Fi devices, or anything that emits modulated signals, the Keysight MXA signal analyzer is the one that separates "signal present" from "signal compliant."
- If your legacy equipment handles simple narrowband tasks without issues, keep it. Old instruments aren't useless—they just have blind spots that show up under modern signals and tight deadlines.
And if you're facing a deadline and your plan relies on "I hope the old analyzer is still calibrated," treat that hope as a technical risk. Because it is.
In an emergency, the most expensive measurement on your bench is the one you can't defend.
What was best practice in 2020 isn't necessarily enough in 2025. The fundamentals haven't changed—you still need accurate measurement before the deadline. But what "accurate" means has evolved, and the instruments available now can see problems the old bench simply can't. It took me 12 years and 200+ urgent requests to learn that, and I hope it takes you less.
