In January 2025, I was trying to settle a dumb argument: which cordless phone should a client deploy across four repair depots? The client wanted "the best cordless phone," and their definition was "the one that doesn't spew RF junk and actually has a working antenna." I had opinions. But I'd learned, three years earlier, not to trust opinions, so I dragged out the Anritsu Site Master, popped the case open, and set up a return-loss measurement at the antenna feed point.
Every phone read as an open circuit. Infinite VSWR. A perfect reflection—the kind you get when nothing is connected. I laughed, reconnected, got the same result. Then I made it worse: I told the client their phones were broken before checking my own test setup.
That cost me three days, a ton of credibility, and eventually roughly $2,300 in rework.
This is the story of a month where I blamed a good instrument for my own connector sins. If you do any RF or optical testing, the same mistakes are probably already hiding in your shop. I'll show you what they looked like on screen, what was actually happening, and the boring checklist that finally stopped the bleeding.
Let me explain "infinity," because it's the whole story. The Site Master measures return loss: how much of the transmitted signal reflects back from whatever is attached. Terminate it with a perfect 50-ohm load and almost nothing reflects. Leave the port open—or break a pin inside the connector—and everything reflects. Return loss drops to 0 dB, and VSWR goes to infinity. On a Smith chart, that's the far-right point: infinite impedance, total reflection.
So when every cordless phone showed infinite VSWR, the display wasn't saying "broken antenna." It was saying "the path is open somewhere between the instrument and the device." The instrument was right. I was wrong to blame the phones.
The device under test is innocent until proven guilty. The test path is usually the guilty party.
Actually, I should mention: I've watched a senior colleague chase a "fault on the line" for an entire day when his own test cable was the fault. It's not just beginners who make this mistake. It's anyone who forgets that the instrument can only measure what it sees through the connector chain.
I traced the path: Site Master → adapter → test cable → the phone's antenna feed. The first suspect was the adapter. It was a cheap generic barrel that I'd bought because my usual supplier had a minimum order I was too stubborn to hit. Which is a good segue: small orders are how small shops learn who their real partners are. More on that later.
The adapter looked fine. The cable looked fine. I've been running this lab since 2021; I know what a worn RF connector looks like—bent center pins, burned plating, a dented dielectric. None of that was visible. So I did what you do when visible inspection fails: I got a connector gauge.
The center pin on the cable's Type N male was recessed. Not dramatically. Visually unnoticeable. But the military interface spec for Type N (MIL-STD-348) is precise about pin position, and this pin was too deep by just enough that the female contact in the adapter never touched it. The connector mated perfectly by feel. Electrically, it was an open. Hence: infinity.
The cable had been mated and unmated maybe four hundred times. I swap field crews, and nobody logged connector cycles. When did I last verify it? I didn't. It said "GOOD" on the label I'd taped to it. Labels lie.
Everything I'd read about connector maintenance focused on visible damage. My experience: the most dangerous faults are invisible—pin depth, contamination, torque. You can't eyeball any of them.
Once the pin depth was fixed, I switched to a borrowed spectrum analyzer to check the phones' transmit power and harmonics. (The Site Master's main job is return loss and cable analysis, not spectrum sniffing—different tool, same connector mistakes.) The readings jumped around between -40 dBm and -20 dBm in ways that made no sense. No, wait—first I blamed the bench power supply. Maybe it was humming. That wasn't it either.
Here's what it actually was: the female barrel adapter between the test cable and the phone had been tightened by hand so many times that the slotted female contact was deformed. Somebody (me) had cranked it like a lug nut, not like a precision RF component. An overtightened slotted contact spreads and no longer pinches the pin. It makes intermittent contact. DECT cordless phones live at 1.88–1.90 GHz in most regions (1.92–1.93 GHz in the US, per ETSI EN 301 406 / 47 CFR Part 15 Subpart D). At 1.9 GHz, a 0.1-mm air gap is the same thing as a break.
It looked fine. I could have inspected the contact under 10x magnification in fifteen seconds and seen the spread gap. I didn't. I only believed in contact-level inspection after ignoring it once. The once was this month.
Two weeks later, a two-person fiber contractor asked me to look at a batch of SFP optical transmitters they'd bought online. Twelve units. Small order, small ticket. On my bench sat the Anritsu MS9710C, a benchtop optical spectrum analyzer I'd picked up used—it covers the telecom wavelengths (600–1750 nm), so it's the right tool for verifying center wavelength and spectral shape.
The first SFP measured "wrong." Center wavelength off by 4 nm, distorted spectrum. I was about to tell the contractor every SFP was bad. Then I caught myself: I'd skipped the setup. The MS9710C needs a clean reference connection and a correct baseline, like any optical instrument. I'd plugged in my test patch cord without inspecting the end face. Contamination on the fiber connector was so bad that the analyzer was measuring the patch cord's reflection, not the SFP's light.
The fiber inspection standard IEC 61300-3-35 has defect limits that would have caught this in thirty seconds with a fiber scope. The actual fix was a minute of cleaning. The contractor's SFPs were fine. I had almost cost him a week of false troubleshooting and a needless replacement order.
Should mention: I rushed because the order felt small. Twelve units, small money, "not worth the full prep." Then it almost ate an entire week. Small project, same physics, same stakes. The person with a twelve-unit order deserves the same rigor as a Tier 1 operator. I wish I'd internalized that without the adrenaline.
Let me be concrete, because "lessons learned" without numbers is just blogging.
Total: north of $2,300, plus one relationship that almost didn't survive. All of it caused by connectors.
I'm not going to write a sermon about calibration culture. Here's the checklist I now run before every measurement session. It takes ten minutes, and in the 18 months since I've kept track, it has caught 47 potential errors. I count them because I used to be the source of those errors.
Bottom line: the instrument is the last thing to blame. The Site Master and the MS9710C both survived a month of my abuse and were accurate the entire time—my connectors were the liars. That doesn't mean the gear never needs care; it's a test instrument with a calibration interval. Anritsu's guidance (verify the current interval at anritsu.com) is a normal annual cycle, and my test cables are now on a tighter one. So is my ego.
When I was starting out, the suppliers who treated my small orders seriously—who asked about my calibration, threw in a loaner cable—are the ones I still buy from today. My customers get the same respect now. The size of the order doesn't change the size of the obligation.
The best cordless phone, by the way? Once the test path was fixed, three of the four models were fine, and the ranking changed. But that's the boring ending. The exciting ending is the one where I check the connectors first.