The Call That Changed Everything
In March 2024, I got a call at 9:47 PM on a Thursday. A client—a small RF design shop I'd worked with before—had just discovered their prototype validation was off by 12 dBm. Their big demo was in 72 hours. Normal turnaround for a full characterization? About two weeks.
I'm a test system integrator. My job is to panic so you don't have to. But this one had me sweating.
The client needed to verify power amplifier linearity across 20–40 GHz, measure harmonics, and check P1dB compression points. Their existing setup? A mismatched collection of benchtop gear from three different vendors—none of which spoke to each other. They had a spectrum analyzer from one brand, a signal generator from another, and a multimeter from yet another. The data was a mess.
They'd tried to save money by buying a "good enough" RF generator from a discount vendor. It was a mistake. The instrument was 1.5 dB off at 30 GHz. That's not measurement error; that's a fundamental flaw.
The Breakdown
By 10:00 PM, I was on the phone with our distributor, trying to figure out what we could get overnight. Here's what we needed:
- Reliable signal generation up to 40 GHz with known phase noise specs
- Automated switching between DUT, reference, and calibration paths
- High-precision power measurement that didn't drift—no guesswork
- Everything had to work together without custom drivers. We didn't have time for that.
I pulled up our inventory and started cross-referencing specs. The Keysight PSG signal generator was the obvious choice for the source. But the switching matrix? That was the bottleneck. The client had a manual switch setup—four DUTs, one cable. It took them 20 minutes per test point.
We needed a Keysight N1810UL RF switch matrix. Rated for DC to 67 GHz, low insertion loss (under 0.5 dB at 40 GHz), and repeatable to within 0.03 dB. That's the kind of spec that makes or breaks a production test.
Oh, and we needed a B1505A power device analyzer for the P1dB compression sweeps. At 30 V and 5 A pulsed. The B1505A can handle that like a champ.
I called the distributor back: "I need an N1810UL, a PSG with options 520 and 522, and a B1505A with the 2A ru module. Delivery by Friday noon."
The answer: "We can get the switch matrix and signal generator from the NJ warehouse. The B1505A? Maybe Monday."
Monday was 96 hours away. The demo was in 72.
The Decision Point
Here's where the transparent pricing thing really matters. The distributor gave me three options:
- Standard delivery: All gear arrives Monday. Cost: $12,400 (base prices). No rush fees. But we miss the demo deadline.
- Partial rush: Switch matrix and signal generator delivered Friday by 10 AM. B1505A arrives Monday. Cost: $13,600 (including $1,200 in rush fees). Client gets a partially tested prototype, but the P1dB data would be raw.
- Full rush: Everything delivered Friday by 2 PM. Cost: $14,800 ($2,400 in rush fees). Plus we'd need to pay for a courier to hand-carry the B1505A from the warehouse. Total: $15,200.
I told the client: "Option 3 is the only one that gets you a completed test by Saturday night. The cost is $15,200 all-in. No hidden fees for calibration certificates or software licenses. What's it worth?"
They said yes. I still have the email: "We can't afford to lose this contract. $15,200 is cheaper than losing the demo."
The Execution
Friday at 1:30 PM, three pallets arrived. We had the switch matrix rack-mounted and cabled to the DUTs by 3 PM. The N1810UL's isolation is better than 90 dB at 40 GHz—meaning no crosstalk between channels. That's not just a number; it's the difference between clean data and garbage.
We connected the PSG signal generator to the matrix's input port. Set up a 64-point sweep from 20 to 40 GHz. The PSG's phase noise at 40 GHz is −120 dBc/Hz at 100 kHz offset. That's Keysight spec. The budget signal generator they'd been using? −95 dBc/Hz. No wonder their measurements looked like static.
The B1505A arrived at 2 PM Saturday morning. I'd never calibrated a system this fast before. The analyzer's self-test took 30 minutes. We ran the P1dB compression sweeps from 9 AM to 2 PM. Sixty-four points per DUT, four DUTs, five power levels each. The B1505A's measurement repeatability is 0.01 dB. That's tight enough to see the difference between a good solder joint and a cold one.
By 7 PM Saturday, we had a complete report. The client's amplifier design was 1.8 dB low at 35 GHz. Turned out to be a bias network issue—a component value error in the simulation. We found it. We fixed it in the test fixture. The demo was saved.
The Reckoning
I still kick myself for not pushing them toward a standardized system earlier. If they'd bought a proper Keysight setup from the start—a PSG, a B1505A, and an N1810UL—they'd have spent $28,000 on the initial system. Instead, they spent $15,000 on piecemeal gear that couldn't talk to each other, then another $15,200 on the emergency resupply.
Total cost of the ad-hoc approach: $30,200. And they still had a design flaw that could have killed the project. The transparent, up-front approach would have cost $28,000 with no surprises.
This pricing was accurate as of Q4 2024. The market changes fast—especially for RF components—so verify current rates before budgeting. (Should mention: we also paid $800 in overnight shipping from the Keysight warehouse. That's on top of the $15,200.)
My experience is based on about 200 rush orders in the last five years. If you're working with a completely different scale—say, production-level systems with 100+ DUTs—your experience might differ significantly. But the principle holds: a properly spec'd system from a single vendor costs less in the end than a patchwork of "deals."
What I Learned
Three things, and I wish someone had told me this ten years ago:
- Lowest price is not lowest cost. The P1dB compression point—that's where the amplifier starts to distort. An instrument that costs 20% less but has 1.5 dB of error will make you fail every spec. The cost of failure—a missed demo, a lost client, a redesign cycle—far outweighs the purchase price.
- Time is the most expensive hidden cost. Those 20-minute manual switch times? They cost the client $3,200 in engineering time over the project's duration. The N1810UL automated the process, switching in under 100 microseconds per path. That's not just speed; it's data consistency.
- A single source matters. When the B1505A didn't arrive on time, we had to use an older analyzer from another vendor. The data didn't match. We spent three hours figuring out why. Turns out the impedance mismatch was different. With a unified system—Keysight signal source, Keysight switch matrix, Keysight power analyzer—the calibration is inherent. The trust is built in.
At the end of the day, the client landed that contract. The demo was a success. The amplifier design went into production. And I sleep a little better knowing I steered them toward a system that actually works—even if it took a 72-hour panic to get there.
If you're setting up a test bench and wondering which multimeter you need, or whether to buy a switch matrix—ask yourself: how much is your time worth? How much is certainty worth? Because the price of the instrument is not the price of the result.
