Published on 10/13/2026 • Updated on 10/19/2026
The device worked on the bench. Two kilometers of range in field trials with the antenna on a short pigtail. Then the production enclosure shipped with the antenna mounted remotely on a mast, connected by eight meters of thin coax cable, and range collapsed to a few hundred meters. Support tickets called it a firmware bug. It was not.
⚠️ Watch out: This is a composite, illustrative scenario built from common IoT deployment failures, not a report about any specific customer or product. Every number is an example chosen to show the diagnostic method. We publish it this way because the failure pattern is real and extremely common, and the method generalizes.
On This Page
The setup in our example: a LoRa-class sensor at 868 MHz, 14 dBm transmit power, a small whip antenna that delivered strong range with a 30 cm pigtail during trials. The production mounting moved the antenna to a mast and connected it with 8 m of thin RG174-class coaxial cable assembly. Reported symptoms: range down by more than 80 percent, RSSI readings 10 to 14 dB lower than trials at the same locations, and one unit in ten far worse than the rest.
Before touching hardware, redo the arithmetic with the as-built feedline. Our RF link budget guide has the full method; the short version for this example:
The math alone explained most of the collapse. That is always step one: if the as-built numbers predict the failure, you are hunting a design problem, not a defect.
Key idea: Long, thin feedlines eat sub-GHz links alive. Loss roughly doubles every time you double the run, and thin micro-coax classes can cost about 1 dB per meter at 868/915 MHz. See our RG174 coaxial cable guide and the LoRa/LoRaWAN cable article for cable-class numbers.
With the budget pointing at the feedline, the next move is to isolate each element instead of swapping parts at random. The sequence that works: measure the radio output directly (power at the module port), then sweep the feedline assembly on a VNA for insertion loss and VSWR, then check the antenna match at the far end of the installed feedline. In the example, the feedline loss matched the prediction, but the one-in-ten bad units showed a VSWR over 3:1. The cause turned out to be a connector mix-up on a procurement substitution.
| Rank | Culprit | How it presents | Fix |
|---|---|---|---|
| 1 | Feedline loss from long thin coaxial cable | Uniform range loss across all units | Shorter run, fatter low-loss cable, or move the radio to the antenna |
| 2 | SMA vs RP-SMA mix-up | Units connect but show terrible VSWR; some nearly dead | Inspect contacts; our SMA vs RP-SMA guide shows the difference |
| 3 | Antenna mismatch at operating band | High VSWR at band edges; detuned by mast or enclosure | Measure installed match; retune or relocate the antenna |
| 4 | Water ingress at outdoor joint | Loss that grows over weeks, worse after rain | Seal per our waterproofing guide; replace wet cable |
| 5 | Antenna placement and ground plane | Range varies with orientation and mounting surface | Mount per antenna maker rules; clear of metal |
Building an IoT product and need the feedline done right? We build antenna cables to your length and sweep every one.
In the example, the production change was twofold: swap the 8 m thin coax for a short 1.5 m LMR-195-class run by relocating the gateway electronics closer to the mast, and lock the connector specification (with a photo in the work instruction) to stop the SMA/RP-SMA substitution. The verification step matters as much as the fix: every new feedline assembly was VNA-swept for VSWR and insertion loss before installation, and field RSSI was re-measured at the original trial points. Range recovered to within 10 percent of the bench trials, and the residual gap traced to mast detuning, handled with a corrected antenna mount.
Notice the order: diagnose with numbers, fix the dominant term first, verify with measurement, then chase the residual. Swapping antennas first would have changed nothing.
How range hunts go wrong
Why did my IoT device lose range when I moved the antenna to a mast?
The longer coaxial cable link almost always explains it. Thin coax classes cost around 1 dB per meter at 868/915 MHz, so an 8 m run can cut range by more than half before anything else goes wrong.
How much does coaxial cable length affect LoRa range?
Every 6 dB of added loss halves range. At roughly 1 dB per meter for thin coax at 868 MHz, a few extra meters is a serious range tax. Use low-loss cable or move the radio near the antenna.
Can the wrong SMA RF connector kill my signal even if it screws on?
Yes. SMA and RP-SMA bodies thread together but the center contacts do not mate, leaving an open or intermittent connection. Our SMA vs RP-SMA identification guide shows what to check.
Should I measure the antenna or the coaxial cable first?
Sweep the feedline assembly alone first; it is the most common culprit and the easiest to isolate. Then measure the antenna match at the end of the installed feedline.
Is it better to extend the antenna coaxial cable or the data cable?
Almost always extend the data or power side and keep the RF feedline short. Low-voltage data runs tolerate length far better than RF at sub-GHz and above.
What RSSI drop means I have a cable problem?
Compare against the as-built link budget. A uniform drop across all units matching predicted feedline loss points to design; a scattered pattern across units points to defects like connector substitutions.
Range loss after a mechanical redesign is almost never mysterious. Rebuild the link budget with the as-built feedline, isolate radio, feedline and antenna with real measurements, and fix the dominant loss term first. In most IoT deployments the feedline is the budget, and a verified short run of the right cable gets the range back.
Need assemblies built and verified this way?
SigmaRF builds standard and custom coaxial cable assemblies at our Vadodara, India volume facility and our Cleveland-area precision facility in the USA. Every assembly is VNA-swept for VSWR and insertion loss across 100 MHz to 40 GHz before it ships, and lead times run as short as 2 days.