5 Mistakes I Made with Keysight Test Equipment (And a Checklist to Avoid Them)

What This Checklist Will Save You From

This guide is for engineers and technicians who've just received a Keysight instrument—whether it's a 34461A multimeter for precision DC work, a PNA vector network analyzer for antenna tuning, or an 81150A pulse generator for ATE testing.

Here's what I've seen go wrong, in order of frequency:

1. Incorrect reference settings (especially when moving from lab to production)
2. Misconfigured power limits that damaged sensitive DUTs
3. Calibration intervals ignored—especially with the N93 series of power sensors
4. Improper averaging setup on signal analyzers for low-level signals
5. Assumptions about measurement bandwidth that led to erroneous pass/fail results
6. Ground loops introduced via USB/GPIB cabling, costing days of debug

Step 1: Verify Your Reference and Calibration State

Sounds basic, I know. But the number of times I've seen someone fire up a signal analyzer and assume the internal reference is accurate… I've lost count.

What I learned the hard way:

In Q3 2020, I was measuring phase noise on a 10 GHz oscillator using a Keysight E5052B. The results looked too good—about 4 dB better than the datasheet. I spent three days double-checking the setup before realizing the internal reference had drifted by 0.7 ppm because the instrument hadn't been recalibrated in 18 months.

Checklist item:

• Confirm the calibration sticker is within the valid window (usually 12 months for most Keysight equipment).
• If using an external reference (10 MHz), verify it's within the specified lock range. I've seen an external rubidium standard outputting 9.9999 MHz cause a 1 kHz frequency error at 10 GHz.
• For the N9030B and similar signal analyzers, run the self-calibration routine ("Cal" in the menu). It's fast and catches systematic errors.

I said "let's just use internal reference." They heard "skip calibration." Result: a $3,200 batch of modules failed final test because of a 2-dB offset. The customer caught it during acceptance testing.

Step 2: Set Input Protection and Power Limits Before Connection

This is the one that still makes me cringe. Two years ago, I connected a 30 dBm amplifier output directly to a spectrum analyzer without verifying the input attenuator setting. The analyzer's mixer was rated for +25 dBm max. I blew the front-end mixer—$4,700 repair bill and a two-week delay.

Checklist item:

• On any Keysight signal analyzer (CXA, EXA, MXA, PXA), set the reference level to at least 10 dB above expected peak power.
• For vector network analyzers, verify the source power is set to avoid compressing the DUT. Start at -10 dBm and increase slowly if needed.
• With the N93 series power sensors, use the 'zero and cal' feature before connecting high-power signals.
• If you're using a pulse generator like the 81150A or 81160A, check the output voltage against the DUT's absolute maximum ratings. I've seen a 3.3V CMOS input destroyed by a 5V pulse "just for a quick test."

Step 3: Configure Measurement Speed vs. Accuracy Correctly

Here's a common mistake: setting the fastest measurement speed because your production line needs throughput. Then wondering why your VSWR results fluctuate by 10% between runs.

What I discovered after a painful audit:

In January 2022, our team was certifying a new antenna design using a Keysight PNA-X. The return loss measurements showed inconsistent dips around 28 GHz. We spent two weeks chasing connector issues, cable flex, even temperature changes. Turns out, the measurement bandwidth (IFBW) was set to 10 kHz for speed—way too wide for that type of measurement.

Checklist item:

• For filter measurements and low-ripple passband characterization: IFBW ≤ 1 kHz.
• For antenna VSWR on a PNA: IFBW 100 Hz to 1 kHz depending on sweep points.
• For general spectrum measurements on a signal analyzer: set the RBW to match your signal's occupied bandwidth. Too wide = noise dominates. Too narrow = measurement time blows up.
• Use the "Auto" function as a starting point, but verify it's appropriate. I've seen "Auto" choose a 10 MHz RBW for a 100 kHz-wide LTE signal, which gives terrible sensitivity.

Never expected the default settings to cause the problem. Turns out, the surprise wasn't the hardware—it was my assumption that fast meant good enough.

Step 4: Match Impedance and Cable Type to the Measurement

We were using the same words but meaning different things. My colleague said "standard coaxial cable." I heard "RG-58." He meant semi-rigid with SMA connectors. Discovered this when the insertion loss at 40 GHz was 8 dB higher than expected. The entire dataset for a week had to be discarded.

Here's the thing about impedance:

Most Keysight RF equipment is 50 ohms. But I've seen people connect a 75-ohm cable because "it was in the drawer." The mismatch loss at 6 GHz is about 0.18 dB, but at millimeter-wave frequencies (28 GHz), it can exceed 1 dB. That's the difference between a passing and failing module.

Checklist item:

1. Verify cable impedance matches the instrument front panel labeling (50 or 75 Ω).
2. Use phase-stable cables for VNA measurements—especially above 10 GHz. Standard RG cables change phase with movement, causing measurement drift.
3. Torque connectors to specification. A loose SMA connection can add 0.5 dB loss at 26 GHz. I use a torque wrench for every connection now—after one too many "finger-tight" failures.
4. For the clear phone and similar consumer electronics RF testing, use appropriate micro-coaxial adapters. Don't force an SMA onto a U.FL connector—I've ripped pads off boards doing this.

Step 5: Use the Right Trigger and Sweep Mode

This is a subtle one that catches even experienced engineers. In 2023, we were testing burst signals from an 802.11ax device using a Keysight signal analyzer in swept mode. The analyzer was sweeping slowly, so it missed the burst most of the time. The result? We thought the device was dead. Spent an entire day swapping hardware before someone suggested using "free-run trigger" and adjusting the sweep time manually.

Checklist item:

• For bursted or pulsed signals (radar, Wi-Fi, 5G TDD): use "external trigger" or "video trigger" on the signal analyzer, not continuous sweep.
• For the 81150A pulse generator, ensure the trigger mode matches the test requirement of your N93-based power measurement.
• If using a PNA for time-domain measurements (TDR mode), set the frequency span correctly for the spatial resolution you need. A 20 GHz span gives about 5 mm resolution; a 1 GHz span gives about 100 mm.
• When uncertain, start with "automatic," then switch to manual only if you understand the trade-off. The automatic settings are surprisingly good—provided the input signal is continuous and stable.

What I Wish Someone Had Told Me Early On

Looking back, most of my expensive mistakes came from two things: rushing and assumptions. Rushing because deadlines are real, and assumptions because the equipment looked familiar.

Quick reference checklist:

• Is the calibration valid?
• Is the reference set correctly (internal vs external)?
• Are input power limits verified?
• Is IFBW/RBW appropriate for the signal?
• Are cables and connectors compatible and torqued?
• Is the trigger/sweep mode matched to the signal type?

Granted, working through this list adds 5-10 minutes to the initial setup. But compared to the alternative—a $4,700 repair, a $3,200 scrap batch, or a week of debugging—that's time well spent.

Industry standard for measurement confidence requires traceability to recognized standards (like NIST), but in daily practice, this checklist catches the errors that standards compliance alone won't. As of January 2025, Keysight's own application notes recommend calibrating at least every 12 months for bench instruments; for production-critical setups, every 6 months.