An outdoor PoE run rarely fails because of the switch. It fails because the jacket is being asked to do three jobs at once — block ultraviolet, keep water out of the core, and survive the temperature those two jobs create — and most cable is good at one of them. The copper is never the problem. The copper is what tells you there is one.

Last month I stood on a warehouse roof in Bao'an with a thermal camera and a bad attitude. Eight perimeter cameras, working since March, dropping video between noon and four. The integrator had already replaced two cameras and a switch port. The cable was 100 metres of black Cat5e tied to a conduit every two metres, unshaded, on a white membrane roof. Jacket surface at 13:20: 61 °C. Air: 34 °C. That reading explained four months of service calls, and it took two visits to take it, because my first instinct was the power budget.

Three ohms, and a camera that only fails after lunch

Here is the arithmetic I should have done on day one. Copper resistance rises 0.393 percent per degree Celsius, referenced to 20 °C. A 100 m run of 24 AWG has a loop resistance of 18.8 Ω at 20 °C — 9.38 Ω per 100 m per conductor, two conductors in series. At 61 °C the same run measures 21.8 Ω. Nothing changed. It just got three ohms more expensive to push current through.

Put a load on it. A 22 W PTZ with the IR array on draws about 0.47 A at 47 V. At 20 °C that current loses 8.8 V in the cable; at 61 °C it loses 10.3 V. With the PSE at 52 V under load, the camera sees 43.2 V in the morning and 41.8 V at half past one. The 802.3at floor is 42.5 V. The camera does not care about my March commissioning sheet.

Cable temperatureLoop resistance, 100 mDrop at 0.47 AVolts at camera (52 V PSE)Verdict
20 °C18.8 Ω8.8 V43.2 VPass, 0.7 V spare
40 °C20.3 Ω9.5 V42.5 VOn the line
61 °C21.8 Ω10.3 V41.7 VFail
70 °C22.5 Ω10.6 V41.4 VFail, and the jacket is next

That is the whole failure mode. Not a bad camera, not a weak port, not firmware. A cable that passed acceptance because acceptance happened in spring. I have written about how category and heat interact, but that was bundles heating each other indoors. This is a single cable cooked from outside by a star.

Why a black jacket ends up at 61 °C

People assume the cable is hot because it carries power. It does not. Do the sums: 0.47 A through 18.8 Ω is 4.2 W over 100 metres, 0.042 W per metre. A 6 mm cable intercepts about 0.006 m² per metre, so at noon it takes roughly 6 W of sunlight per metre, most of it absorbed by a black polyethylene jacket. The sun delivers about a hundred times more heat per metre than the PoE load. Switch the port off and the jacket stays at 61 °C. That is why "just lower the current" never fixed anybody's afternoon outage.

The roof matters as much as the sun. A white membrane throws light back up at the underside of the cable, so black cable on a white roof runs hotter than the same cable on gravel. Add a conduit it is tied against, no air between ties, and a jacket that is black precisely because carbon black is the cheapest way to make polyethylene UV-stable, and 61 °C on a 34 °C day stops being remarkable.

Measure it properly. An infrared gun at default emissivity reads shiny black fairly well and light grey badly, and it will read the sky if your angle is wrong. I tape a thermocouple under the jacket at the midpoint for twenty minutes, with a second one in shade. Under 5 °C of difference and the sun is not your problem — go and look at the terminations. The conductor lags the surface in the morning and catches up by mid-afternoon, so I design against the jacket number.

Water does not kill the link. It kills the conductor.

Water enters outdoor cable by two routes, and only one involves rain hitting a hole. The obvious one is a split jacket, a bad gland, or a connector crimped without a boot and left facing up. The other is breathing: warm damp air diffuses slowly through polyethylene, condenses at night, and never leaves. Two or three seasons and there is a film of water inside a cable that was never cut.

What it does is undramatic. It sits at the low point, and bare copper in water with oxygen becomes oxide that takes up more volume than the metal it came from. The first symptom is not a dead camera. It is a link that comes up at 100 Mbps instead of 1 Gbps, or a 4K stream that stutters at dusk, because water changes the dielectric and the impedance along that section. By the time the camera drops, the conductor at the bottom of the loop has been green for a year.

Corrosion is rarely symmetrical. One conductor of a pair goes first, and unequal DC resistance pushes the PD's isolation transformer into a flux offset it was not designed for — 802.3 limits resistance unbalance for exactly this reason. In the field you see a port reporting less current than the device draws, or a device running noticeably warm.

Then winter arrives. Water expands about 9 percent when it freezes, and a jacket that has been slightly porous for two years splits at the low point the first hard freeze. Now there is a hole, and next spring there is more water. One winter is enough.

The cheap countermeasures still win. A drip loop before the cable enters any enclosure, glands facing down, no jacket slit above a box. And if the run ends in a sealed wall cabinet, remember the cabinet is an oven too, and the cable entry is where heat and water meet.

Gel, dry tape, or a second jacket

Three constructions get sold as "outdoor cable" and they are not interchangeable.

ConstructionWater blockingTerminationHandlingWhere it belongs
Gel-filled coreBest. Blocks migration along the coreSlow. Gel off every conductor before punchdownStiff, wide bend radiusDirect burial, wet trenches, long horizontal runs
Dry water-blocking tapeGood, until compressed or repeatedly flexedFast, clean handsLike standard cableAerial spans, trays, draining conduit
Dual jacket, LLDPE over PVCMechanical and UV; water between jackets can still travelFastThicker, tough against crushRooftops, gravel, foot traffic
Plain PE jacket, no blockingNoneFastNormalA three-metre drop to an outdoor AP
Indoor CM cable in conduitNoneFastNormalNowhere outdoors

The mistake I see most is the last row. Somebody runs indoor Cat6 through 40 mm PVC because it looks protected, and what they have built is a bathtub with a cable in it. Underground conduit always holds water. A non-blocked cable inside has made the failure slower to find, not less likely.

The gel trade-off lands on whoever holds the crimper. Residual compound at the contact interface raises contact resistance, attracts grit, and gives you exactly the kind of two-ohm termination that eats a volt at 80 metres. If installers will not wipe every conductor, dry-blocked cable beats gel cable in practice even though it is worse on paper.

What the ratings on the box do and do not mean

CMX means outdoor. It does not mean aerial, does not mean tray-rated, and does not mean ten summers. "UV resistant" without a test standard is an adjective, not a specification. The standards that mean something are ASTM G154 for fluorescent UV, ASTM G155 for xenon arc, and IEC 60068-2-5 for solar radiation. When a supplier quotes 720 hours, ask which lamp at which irradiance — those hours do not convert cleanly into years on a roof. Treat it as pass or fail. And a jacket rated to 75 °C is not a promise of full power at 75 °C, because ampacity tables assume an ambient of maybe 30 °C and a black jacket in July is past that before noon.

Rodents do not read datasheets

Every outdoor catalogue has a paragraph on rodent resistance and most of it is about an additive. Bittering agents and capsaicin do work for a while. They also weather, and a squirrel in a roof void is not tasting your cable, it is wearing its teeth down on it. I have never seen an additive claim survive three years where there is a real rodent population.

Mechanical works. Steel tape or interlocked armour, at roughly 1.6 to 2 times the cost of the same cable unarmoured — I have quoted both often enough to know the multiplier is in that band, not the "slightly more" band sales suggests. Or conduit, where the run is short. Or route where nothing walks: high on a wall, inside a covered tray. And pack the penetrations, because a cable through a wall with a bead of sealant around it is a hole with a cable in it. Copper mesh then sealant holds up far better than steel wool, which rusts into nothing in one wet season.

And if the run leaves the building it needs surge protection at the entry, which is only as good as its bond to earth. I covered what 4 kV and 6 kV figures actually mean separately. A jacket discussion that skips the grounding is half the outdoor story.

What I write in the spec now

The list I hand to a supplier or installer. Ten minutes, and it has saved me three return visits.

  1. Name the jacket compound, not the marketing term. UV-stabilised LLDPE or MDPE with carbon black. If the answer is "outdoor grade", ask again.
  2. Ask for the UV standard and hours, then judge whether the number means anything at your latitude. A roof in Guangdong is not a roof in Rotterdam.
  3. Match the listing to the route: CMX for exposed outdoor, armour for anything with teeth or mowers, direct-burial rated only if it is going in the ground.
  4. Get the water-blocking construction in writing — gel, dry tape, or dual jacket — and match it to horizontal or vertical.
  5. Check the jacket rating against the hottest surface the cable touches, not against air temperature. A roof membrane at 2 p.m. is a surface.
  6. Specify 23 AWG near 100 m. Loop resistance falls from 18.8 Ω to 15.0 Ω, which returns 1.8 V at 0.47 A — often the whole margin between a camera that survives July and one that does not.
  7. Shade what you can. White split conduit over a black jacket has measured 10 to 15 °C cooler for me on the same roof, and costs less than moving the run.
  8. Leave a drip loop and a service loop at every entry, gland facing down. Free, and it prevents the most common water failure.
  9. Measure loop resistance on all four pairs at handover, then again at 13:00 on a hot day, and write both down. One number says the cable is connected. Two say whether it still will be in August.
  10. Check PD voltage at the far end with a tester, not the switch's own figure. The switch tells you what it sends. You care what arrives.

Where I got it wrong

I spent two visits on that roof replacing things before I put a thermocouple on the cable. Cameras dropping in the afternoon and recovering overnight looks exactly like a thermal fault inside the camera, and cameras are cheaper to swap than cable. Swapping is also more satisfying. You leave having done something.

The earlier mistake was dumber. On another job I specified gel-filled Cat6 for a twelve-metre vertical riser in a leaking shaft because it was "the outdoor one". By August there was compound in the bottom junction box and the punchdowns were sitting in it. Gel is designed for horizontal burial, where gravity is not working against the compound for years; cheap compound on a hot vertical run is not always as thixotropic as the datasheet implies. I use dry-blocked cable for anything vertical now.

Questions I get asked on site

Can I just paint the cable white, or wrap it in tape?

Paint on polyethylene flakes within a season and takes the UV stabiliser with it. Tape traps moisture against the jacket and is worse than nothing. A white split conduit is the version that works — 10 to 15 °C cooler by my measurements.

Is direct-burial cable fine inside conduit?

Yes, but not for the reason people give. The conduit keeps shovels and rocks out, not water. The blocking in the cable keeps water out, and a non-blocked cable in a flooded conduit is worse off than in free air, because the water cannot drain.

How do I tell whether heat is my problem?

Log PD voltage at the far end at 09:00 and 14:00 on a clear day, same load. An afternoon reading a volt or more lower, with a healthy morning, means copper temperature. If both are bad it is termination or length, and you should be at the jacks. Stable voltage with a flapping link means water, and power was never the issue.

Would a higher PoE class help?

No, it makes it worse. A higher class lets the device draw more current, and more current through hot copper means more drop. More power at the far end comes from a shorter run, a larger conductor, four-pair delivery so the current splits, or a mid-span extender that re-powers the device locally. A bigger port does none of those.

Do I need shielded cable outdoors?

Only with a real interference source, like a VFD or a long parallel run beside mains. A shield does nothing for heat or water, and adds a bonding requirement that gets skipped half the time. An unbonded shield is an antenna.

If you have a run that behaves differently in August than it did in March, I would like the numbers — length, gauge, load, two voltage readings. That comparison is worth more than any datasheet, and it is how I learned most of what is above.