Cat5e is a perfectly good PoE cable for 802.3af and 802.3at at the full 100 m channel length, and it becomes a compromise — occasionally an unsafe one — once you move to 802.3bt. The deciding factor is not bandwidth. It is DC resistance and the heat that resistance produces: a thinner conductor turns part of your power budget into a heater inside the conduit, and the warmer the copper gets, the higher its resistance climbs. Cat6 buys a heavier conductor and a few watts of margin. Cat6a buys thermal headroom and shielding at the cost of diameter and stiffness. Choose by current, not by category number.
The confusion comes from mixing two unrelated specifications. Category rating is a data claim measured in megahertz and bit error rate; PoE is a low-frequency DC power circuit on the same four pairs. A cable can be excellent at 500 MHz and mediocre as a power conductor.
PoE is a DC circuit: resistance is the only thing that matters
Power leaves the switch, travels the pair, and arrives reduced by the voltage dropped across the round trip. That drop is Ohm's law: current multiplied by the loop resistance of the pair, out and back. The power lost in the cable is the current squared times that same resistance, dissipated as heat along the whole jacket rather than at one spot you can inspect.
The published figures are not marketing round numbers. They are budgets that already contain the cable loss the standard expects you to have.
| Standard | PSE output | Power at the device | Current | Pairs carrying power |
|---|---|---|---|---|
| 802.3af (Type 1) | 15.4 W | 12.95 W | 350 mA | 2 |
| 802.3at (Type 2) | 30 W | 25.5 W | 600 mA | 2 |
| 802.3bt Type 3 | 60 W | 51 W | 600 mA per pair | 4 |
| 802.3bt Type 4 | 90 W | 71.3 W | 960 mA per pair | 4 |
Read the 802.3at row as a worked example. 30 W leaves the port, 25.5 W is guaranteed at the device, so 4.5 W is reserved for the cable. At 600 mA that reserve corresponds exactly to a loop resistance of 12.5 Ω. Above that resistance the arithmetic no longer closes, and something gives — usually the device, which browns out or drops off the network under load. The full derivation is in our guide to calculating a PoE power budget.
The 12.5 Ω planning number versus what the drum measures
Twelve and a half ohms per pair is the figure the standards dimension the channel against. It is a planning model, not a datasheet property, and the gap between the model and the drum is where most 802.3bt projects get into trouble.
Take the conductor resistance limits for a 100 m run at 20 °C. A 24 AWG conductor — the usual Cat5e construction — is permitted up to 9.38 Ω per 100 m. Out and back through a pair gives a loop near 18.8 Ω. A 23 AWG conductor, typical of Cat6 and Cat6a, is permitted about 7.5 Ω per 100 m, giving a loop near 15 Ω. Both exceed 12.5 Ω at full length, which is precisely why the standards cap the channel at 100 m and why the power budget keeps a reserve.
Then add temperature. Copper resistance rises roughly 0.393 % per degree Celsius, so a cable sitting at 60 °C in a sunlit conduit instead of 20 °C carries about 16 % more resistance than the datasheet figure. An 18.8 Ω loop becomes 21.8 Ω on the same cable that was already over budget at room temperature. That compounding, not the category, decides whether a long run survives its third summer.
Heat is the failure mode nobody puts in the quote
Cable loss is small per metre and large in aggregate. One Cat5e run at 600 mA dissipates a few watts over 100 m, which the jacket sheds easily into open air. Put thirty-seven of those cables in one bundle inside a sealed riser and the same few watts each have nowhere to go. A rise of 10 to 20 °C above ambient is normal in dense trays, and it arrives silently.
Three properties decide how much heat a bundle tolerates. Conductor gauge sets how much is generated. Insulation sets how hot the plastic may get — PVC is commonly rated 75 °C, polyethylene and LSZH compounds often 90 °C or higher, and a cable run near its rating has no margin left. Installation practice sets how the heat escapes: conduit fill, tray spacing and whether the bundle sits in a ceiling void or in free air change the outcome far more than the category printed on the jacket.
802.3bt makes this worse in a way that surprises people. Two-pair PoE energises half the cable; four-pair PoE energises all of it at 600 mA or 960 mA per pair. Heat per metre roughly doubles at the same current when the second pair set is energised, and quadruples from 600 mA to 960 mA, because loss scales with the square of current. A bundle that was comfortable on 802.3at cameras can be over its thermal limit when the same tray is reused for 90 W access points or heated PTZ housings.
Cat5e, Cat6 and Cat6a side by side
The table uses conductor limits at 20 °C for a full 100 m run. The loss column is pure I²R arithmetic at maximum current; real links usually draw less.
| Cable | Typical conductor | Loop resistance, 100 m | Loss at 600 mA, 2-pair | Realistic PoE role |
|---|---|---|---|---|
| Cat5e | 24 AWG | ≈ 18.8 Ω | ≈ 6.8 W | 802.3af and 802.3at to 100 m; 802.3bt only on short runs or light loads |
| Cat6 | 23 AWG | ≈ 15 Ω | ≈ 5.4 W | Solid 802.3at at 100 m; 802.3bt Type 3 to roughly 80 m |
| Cat6a | 23 AWG, larger overall diameter | ≈ 15 Ω | ≈ 5.4 W | 802.3bt Type 3 and Type 4 at full length, with better heat spreading in bundles |
Note what the table does not say. Cat6a is not electrically superior to Cat6 for DC power — both use 23 AWG conductors and the loop resistance is effectively identical. What Cat6a adds is a larger jacket and, in shielded constructions, a foil or braid layer that spreads heat laterally. For 90 W work in dense trays that is a real benefit.
Where each cable earns its place, and how to stretch existing Cat5e
Existing Cat5e. If the building has Cat5e and the devices are 802.3af or 802.3at, keep the cable and spend the budget on a better switch. Cameras at 12.95 W and access points at 25.5 W need nothing heavier at 100 m. The one condition is solid copper with all four pairs terminated; the failure patterns are collected in our list of common PoE deployment mistakes.
New construction and mixed loads. Cat6 gives roughly 20 % lower loop resistance than Cat5e at the same length. For a surveillance project with 802.3at cameras plus a few Type 3 devices, Cat6 is the sensible default.
90 W and dense bundles. Type 4 at 71.3 W draws 960 mA per pair across all four pairs. Two parallel paths at that current and 15 Ω loop resistance dissipate more than 20 W in the cable itself, so a 90 W link at full length is a thermal design problem before it is a cabling problem. Cat6a, generous conduit and shorter runs are the combination that works. Distance is covered in extending PoE beyond 100 m.
Five measures recover margin from existing Cat5e without touching the cable:
- Shorten the run. Loss is linear in length, so a 60 m Cat5e run has about 40 % less loop resistance than a 100 m one.
- Reduce the load, not the distance. A device drawing 40 W instead of 60 W cuts cable loss by more than half, because loss follows the square of current.
- Split the bundle. Two thinner bundles run cooler than one thick one, which lowers resistance, which lowers loss.
- Move the power source closer. A midspan injector or a small PoE switch near the load replaces a 90 m power path with a 5 m one, and an active PoE extender resets the 100 m channel without new cable.
- Refuse CCA. Copper-clad aluminium has roughly 1.6 times the resistance of copper and degrades further as it warms. It is sold as Cat5e and it will fail a 600 mA link. Check the conductor with a knife before trusting the printing.
Frequently asked questions
Can I run 802.3bt over existing Cat5e?
For Type 3 loads at moderate distance, usually yes — the link works, but expect more loss and more heat than the standard's budget assumes. For Type 4 at 71.3 W over a full 100 m run inside a bundle, Cat5e is the wrong cable and no switch setting fixes it. Measure loop resistance first; if it is far above 12.5 Ω, shorten the run or change the cable.
Does Cat6 let me exceed 100 m?
No. The 100 m channel limit is a standard, not a cable property, and it applies to data as much as to power. Better cable reduces loss inside the channel; it does not extend the channel. At 150 m you need an extender, a media converter or fibre, not a higher category.
Does shielded cable run cooler?
Slightly, mainly because the foil or braid conducts heat away from the bundle rather than for any electrical reason. Conductor gauge, current and packing density still dominate. Buy shielded cable for electromagnetic reasons and treat the thermal benefit as a bonus.
How hot is too hot?
Compare the measured jacket temperature against the insulation rating, not against your comfort. A cable in a 45 °C riser with a 15 °C rise sits at 60 °C — fine on a 90 °C LSZH jacket, marginal on 75 °C PVC. Within 10 °C of the rating, re-bundle or derate.
Does PoE degrade the data link?
Not directly, but heat does. Higher conductor temperature raises insertion loss slightly, which matters most on long runs already near the limit. On a 95 m Cat5e run carrying 600 mA, the combination of length and warmth is where intermittent link drops come from.
Next step
The cable question resolves quickly once the current is known: add up the per-port wattage, find the highest class on the run, and check loop resistance against distance. Send us the cable category already installed, the longest run in metres and the device list with class ratings, and our engineers will tell you whether the existing plant will carry the load or where to split it. Browse our PoE switches, extenders and media converters for hardware that matches the topology, or contact our team with your site plan and we will return a bill of materials you can install against.