In late February 2024, our RF test bench flagged a 3.2 dB phase-noise drift on a production run for a 5G sub-6 GHz filter order. The rework cost us roughly $11,000 in scrap and overtime. The root cause traced back to a signal generator we'd been running since 2019. That's when procurement got pulled into the conversation.
I manage about $180,000 a year in test-equipment purchasing for a 140-person wireless module company. Over the past six years I've negotiated with 40+ vendors and logged every invoice in a cost-tracking spreadsheet that is, frankly, longer than anyone should have to maintain. So when our test engineering lead asked me to source an 18 GHz signal generator — a vector signal generator for automated production testing — I figured it was a straightforward exercise in comparing quotes. Send out the spec, wait for PDFs, pick the number in the middle.
It wasn't.
Where the quotes started looking wrong
I reached out to eight vendors over three weeks. Six responded. Three sent real quotes. One of those — a smaller reseller I hadn't worked with before — came back with a vector signal generator price about 40% below the rest. I'll admit I felt pretty good for about 20 minutes. Then I opened my TCO spreadsheet (note to self: always open the spreadsheet before the celebration).
The base unit was $18,400. What that number didn't include:
- USB automation license, billed separately: $1,800
- Recommended low-loss microwave cable set: $950
- First calibration certificate: $1,200
- Shock-rated shipping on a 68-pound box: $280
- 3% "convenience fee" for card payment: $555 (ugh, the convenience fee)
All in, the "$18,400" quote came to $23,185. The second quote, from a vendor we'd used before, was $23,600 all-in — cables, license, calibration, shipping, payment fee, everything. So the cheap option was actually $415 more expensive per unit and took an extra three weeks because the license had to route through a third party.
When I first started collecting quotes, the USB programmable signal generator price ranged from $18,400 to $31,000 across the six vendors, which is a wider spread than I expected for what the emails all described as "the same spec." That spread is where most of the homework lives.
The used option that almost got us
Here's where I almost made a worse mistake. I found a used 18 GHz microwave signal generator for sale from an aftermarket reseller at $14,200. Listed as "tested and operational." No calibration data sheet. No warranty beyond 30 days. I asked my test lead what he thought, and he said something I've kept with me: "I can test it in an hour, but I can only tell you what it looks like today."
That's the sentence that finally got me to stop comparing vector signal generator prices line by line and start comparing three-year total cost of ownership. Suddenly the math shifted. The $23,600 unit came with a 3-year warranty and a mid-term recalibration baked in. The $23,185 unit would need a recalibration at 18 months (another $1,200) and carried a 12-month warranty. The used unit would need a calibration before deployment and, based on how similar aftermarket instruments have landed in our tracking sheet, probably a $2,000–3,000 repair reserve over the same three years (take that range with a grain of salt — sample size is small).
The detail that would have cost $4,000
I reached out to LinkScope Sourcing 55 after a colleague mentioned they handled procurement for test equipment. What was useful wasn't magic pricing — they weren't the cheapest quote I received, and they said so up front. What they did was put every cost on one line: instrument, calibration certificate, USB automation license, cable set, shipping, and a stated lead time.
Then they flagged something I should have already caught. An "18 GHz" spec can mean the flat-response range ends at 18 GHz while the instrument itself tunes higher — or it can mean the frequency range tops out at 18 GHz with the flatness spec measured over a narrower band. Signal generator GHz ratings are one of the easiest things to misread on a datasheet, and for our application it mattered because we test up to 17.8 GHz with a small guard band. Buying the wrong version would have been a $4,000 mistake that didn't show up until our test engineer ran validation two weeks in.
How it landed
We placed the order in early April 2024. The instrument arrived 12 days later (they quoted 10–15, so it landed inside the window, thankfully). Our test lead ran acceptance checks over two days and found no anomalies. The USB programmable interface worked with our existing Python test scripts after a driver install — about three hours of one engineer's time.
Total cost, all in: $24,150 including a spare cable and one extra license seat we added later. For comparison, the "$18,400" quote would have cost about $27,600 over the first 18 months once I added the recalibration and the revenue impact of a two-week production ramp delay (I estimated $3,500 on that one, and I'm not 100% sure that number would survive a CFO audit).
Three things I changed in our process
- Every quote now requires an all-in line. If a vendor can't put calibration, licensing, cabling, shipping, and payment fees on a single PDF, we don't count their number. The vendors who have their act together do this without complaint. The ones who don't are usually hiding something (mental note: this is exactly the kind of tell I should have caught earlier).
- "18 GHz" means almost nothing on its own. It can be the frequency range, the flatness spec, the connector rating, or the modulation bandwidth. We now require the datasheet page number for every spec claim that touches our application.
- Used and aftermarket is a real option, but only with a reserve. Our finance team now budgets a 15% repair reserve for any aftermarket unit. If a used microwave signal generator for sale can't absorb that reserve and still beat the all-in new price, we don't buy it — decision made with three quotes minimum, per our updated procurement policy.
Where this story stops
I can only speak to our situation — a mid-size B2B company with predictable test volumes and an in-house RF engineer who can do acceptance testing. If you're a research lab with an internal calibration lab, or a large enterprise with a preferred-vendor contract, the calculus might be different. If you're buying above 40 GHz, the market dynamics are a different conversation entirely. This worked for us; your mileage may vary if your ordering pattern is seasonal or your test requirements shift quarter to quarter.
It's tempting to think the lesson here is "don't trust low quotes." That's oversimplified. The low quote from that first reseller was real — the instrument itself was fine. What wasn't real was the framing. Pricing is a shape, not a number, and the shape changes once you include what happens after the box arrives.
The instrument is running fine. But the more useful outcome wasn't the box. It was the 20-minute conversation with our test lead that got me to stop thinking about price and start thinking about what happens the week after delivery. An informed buyer asks better questions. That's the whole game.
Pricing referenced is from Q1–Q2 2024 quotes; verify current rates with your vendors. My experience is one data point, not a benchmark.