How far poe over ethernet can feed a device is not set by the wattage printed on the switch — it is set by how much resistance the copper cable can carry. The hard ceiling for twisted-pair Ethernet is 100 meters, and the af/at/bt standards all run inside that line; PoE did not invent its own rule. Past the limit, the gigabit data signal fails first, not the power — but the voltage margin thins out too. Choosing cable is about balancing resistance, voltage drop and bandwidth along that 100-meter line.
The 100 m limit comes from the Ethernet channel spec (90 m permanent link + 10 m patch), set by data integrity — not by PoE.
Voltage drop is ΔV = I × R_loop. Cat6 (23 AWG) has about 20% lower loop resistance than Cat5e (24 AWG), which matters most at 90W bt.
af/at is fine on Cat5e; bt should be on Cat6, and do not run it to the full 100 m.
Never use CCA (copper-clad aluminum); in high-temperature industrial sites leave about 20% power margin.
Many people think a short PoE range means "the electricity can't get there." In fact the 100 m boundary comes from the Ethernet channel spec: 90 m horizontal permanent link plus 5 m patch at each end equals a 100 m channel. It is the physical limit of the data signal, because gigabit (1000BASE-T) is sensitive to return loss, crosstalk and delay, and past 100 m negotiation drops or the link breaks.
Power is less fragile. PoE's nominal 44–57V DC rides the line loss on voltage margin alone, so purely in terms of "electrons" it could go farther. The problem is the data fails first and the whole link is dead, so af, at and bt all lock their nominal distance at 100 m — nobody gave PoE a separate green light.
When cable carries power, current flows through the copper core and resistance turns part of the voltage into heat — that is the voltage drop. Drop depends on only two things: loop resistance R and current I, by ΔV = I × R_loop.
Wire gauge decides R. Cat5e is mostly 24 AWG, about 9.4Ω/100m per conductor, so a go-and-return loop is about 19Ω. Cat6 is mostly 23 AWG, about 7.3Ω/100m, loop about 15Ω — roughly 20% lower than Cat5e. Do not dismiss that 20%; the bigger the power, the more it shows.
Take a small 802.3af load: the PD gets 12.95W at about 0.32A, drop on Cat5e is about 6V, leaving around 42V at the end — stable. Move to a 802.3bt 90W high-current class and the current is several times higher; the Cat5e drop jumps to a dozen volts, the end voltage approaches or drops below the floor, and the camera reboots repeatedly. Cat6's shorter loop leaves visibly more margin. When you buy an industrial PoE switch, look at the cable-gauge recommendation alongside the wattage, not just the port count.
| Cable | Gauge | Loop R (100m) | Small-load af/at | High-load bt (90W) | Advice within 100m |
|---|---|---|---|---|---|
| Cat5e | 24 AWG | ≈19Ω | Easily fed | Larger drop, thin end margin | af/at enough, leave margin for bt |
| Cat6 | 23 AWG | ≈15Ω | Easily fed | More margin, lower dropout risk | bt first choice, especially long runs |
| Cat6A | 23 AWG | ≈15Ω | Easily fed | Plenty of margin, strong immunity | Noisy / long bt runs |
Every project has a run that will not make 100 m — old plants, campus lights, cross-building cameras. Three approaches, cheapest first:
Drop speed for distance. Gigabit is the first to fail negotiation past 100 m; dropping to 10/100M usually buys a few more tens of meters. The cost is collapsed bandwidth — camera bitrates stack up and stall — so it only suits low-bitrate points or pure-power spots.
Add a PoE extender / repeater. Keep each link segment under 100 m; string one extender to reach 200 m, two to reach 300 m. The extender reshapes both data and power, but it is a power hog — it eats PoE itself, so budget for its share or the far end ends up hungrier.
Fiber to the far end + local injection. Fiber ignores the 100 m limit and easily reaches kilometers. Put a small switch or poe injector at the far end to power APs and cameras locally, bypassing the copper distance wall entirely. Campus, cross-building and wild monitoring basically all go this route. Where only one legacy device lacks power, a poe splitter can tap a non-PoE port to feed it without a full switch.
Two more remedies: 802.3bt uses all four pairs so current spreads and per-pair drop shrinks; or use Cat6A and shorten the horizontal link. In industrial switch selection, write the extender or fiber plan into the design for long-run points up front — it is easier than fixing it later.
Warning: An extender is a "power hog." It draws PoE for itself, so if you do not reserve its share in the budget, the far end gets less power than without it. Count the extender as a load.
Avoid CCA (copper-clad aluminum). Aluminum-core resistance is several times copper; 100 m drop is huge, bt cannot feed it, and at high current it heats — a safety hazard. Most "cheap engineering-grade cable" is exactly this.
Patch cords count toward the 100 m. 5 m patch + 95 m horizontal + 5 m patch already exceeds it; negotiation drops and you blame the switch.
Heat raises resistance. Resistance climbs with temperature; a 60°C summer cabinet adds drop over the room-temperature figure, and end devices drop out. In industrial sites do not budget to the max — leave about 20% margin.
Don't treat power distance as transmission distance. PoE can "coast on voltage" a bit farther, but the data signal will not play along — do not gamble the link on power margin.
Measure the horizontal link, add both patch cords, keep the total under 100 m.
af/at is fine on Cat5e; bt or long runs go straight to Cat6 / Cat6A.
Reject CCA; insist on pure copper 23/24 AWG.
Past the limit, prefer an extender (reserve power for it) or fiber-to-end + local injection.
In hot industrial sites leave 20% margin; do not calculate to the nameplate.
Even a 24 port poe switch budgets whole-unit watts, not a simple sum of port labels — read the nameplate budget, not the port count.
Rayin's industrial PoE switches are built with this logic in mind, leaving real margin on cable matching so long-run points land more reliably.
1. Why is PoE capped at 100 m? 100 m is the twisted-pair Ethernet channel ceiling (90 m horizontal + 10 m patch), set by data signal integrity, not by PoE alone. Past it gigabit drops negotiation and the link dies, so af/at/bt all lock at 100 m.
2. Does Cat5e vs Cat6 make a big PoE difference? The difference is resistance. Cat5e loop is about 19Ω/100m, Cat6 about 15Ω (20% lower). Small af loads differ little, but at 90W bt the Cat5e drop gets so large the end device drops out; Cat6 has far more margin.
3. Can you still PoE past 100 m? Yes, with help: drop to 10/100M for a few tens of meters, string a PoE extender keeping each segment under 100 m, or run fiber to the far end and inject locally. Hard-pulling copper past the limit basically drops the link.
4. Why is CCA cable banned for PoE? Aluminum-core resistance is several times copper, drop is huge, bt cannot feed it, and it heats at high current — a safety risk. PoE, especially bt, must use pure copper.
Customer Manager at Shenzhen Rayin Technology Co., Ltd.
