The 100 m limit is not a PoE rule. It comes from the Ethernet channel specification: 90 m of permanent link plus 10 m of patch cords, the distance at which 1000BASE-T equalization and the loss budget were defined to work. PoE inherits it because power and data share the same cable, and both fall away at the same distance.
Extending past it is routine — gates, poles, outbuildings, car parks — but the design question is not only how to get signal there. Every metre past the limit also costs watts, and the two problems have to be solved with one answer.
Why does PoE stop at 100 metres?
Two independent limits arrive at roughly the same distance, and it is worth knowing which one is binding in a given design.
The signal limit is the familiar one: attenuation and crosstalk in Category cable push 1000BASE-T beyond the equalizer's range somewhere past 100 m. That limit applies to the data, not to the power, which is why every extender has to re-drive the Ethernet signal — a device that only boosts voltage will not bring the link up.
The power limit is arithmetic. Cat5e 24AWG has about 9.4 Ω of loop resistance per 100 m on a two-pair PoE path. At 350 mA that is 3.3 V of drop and 1.2 W of loss; at 600 mA it is 5.6 V and 3.4 W. The standard's 15.4 → 12.95 W and 30 → 25.5 W gaps exist precisely to absorb that at 100 m. Past 100 m the gaps are gone and the PD starts absorbing the deficit.
How does each extra segment change the power budget?
An extender is not a transparent pipe. It consumes power itself — typically 1.5–3 W for a two-port unit — and the cable on both sides of it drops voltage. Each hop therefore costs the port the cable loss for the full length plus the device's own consumption.
| Total length | Hops | Wire loss at 350 mA | Wire loss at 600 mA | Net from a 15.4 W port | Net from a 30 W port |
|---|---|---|---|---|---|
| 100 m | 0 | 1.2 W | 3.4 W | 14.2 W | 26.6 W |
| 200 m | 1 | 2.3 W | 6.8 W | 11.1 W | 21.2 W |
| 300 m | 2 | 3.5 W | 10.1 W | 7.9 W | 15.9 W |
| 400 m | 3 | 4.6 W | 13.5 W | 4.8 W | 10.5 W |
The figures assume 2 W per extender and a cable at 20 °C. Read the fifth column as the answer to "can this stay on an af port": at 200 m the net is 11.1 W, below the 12.95 W a Class 3 PD expects, so a single extender on an af port has already broken the link by the standard's own arithmetic. On a 30 W port the same 200 m is comfortable, and 300 m still feeds a 12.95 W camera with margin.
Voltage is usually the harder constraint. At 300 m and 600 mA the drop is about 17 V; from a 50 V PSE the far end sits near 33 V, below the 42.5 V an at PD requires and below the 37 V af floor. The fix is a PSE regulating nearer 56–57 V, which buys roughly 18 m per volt at 600 mA and 30 m per volt at 350 mA. A 57 V port at 300 m and 350 mA delivers about 47 V — comfortably inside the af window, and that contrast is why an af-class camera at 300 m is a normal design while an at-class camera at the same distance is not.
When is fibre the better answer?
Fibre removes the distance problem outright: multimode reaches 550 m and single-mode tens of kilometres, and the link is immune to the electrical environment around it. Where a camera pole sits 400 m from the cabinet, one fibre pair with a media converter at each end is a cleaner design than three cascaded copper devices.
The trade is cost per drop and the need for power at the far end. A media converter needs its own supply unless it is fed by PoE, which means the far end needs a local PoE switch or a power feed — and the saving you expected from fibre can disappear in the trench. Fibre wins clearly when several drops share one trench, when the run crosses between buildings on separate electrical grounding, or when the distance is past about 500 m and no practical number of copper hops will hold up.
Grounding deserves its own sentence. Copper between two buildings at different ground potentials carries current on the pairs and shield, and every cascaded extender adds another point where a surge can enter the run. Fibre is dielectric, has no such path, and is the standard answer wherever lightning exposure is a real design input.
Repeater vs extender vs passthrough switch
A midspan injector adds PoE to a switch port that has none. It solves a power problem on a short run and gains no distance at all.
A repeater or extender has two ports: it receives the Ethernet signal and PoE on one side, re-drives the data, and presents the link on the other. Passive units simply pass the DC through and still lose voltage on both segments; active units regenerate it, which is what a 200 m or 300 m design actually needs. Passive units are widely sold as 100 m extenders and deliver appreciably less than the table above suggests.
A PoE-powered field switch is a different animal: one upstream port feeds a device with several downstream ports and its own budget. It is electrically a single hop however many cameras hang off it, which is why a small switch at the base of a pole is usually a better answer than three extenders along the run.
| Option | Distance gain | Power cost | Best use |
|---|---|---|---|
| Cascade extender | About +100 m per hop | 1.5–3 W per hop plus cable loss | One or two drops past 100 m |
| PoE-powered field switch | One hop, many ports | Upstream port only | Clusters of cameras far from the cabinet |
| Fibre with media converter | Effectively unlimited | Local power at the far end | Long runs, building to building |
| Midspan injector | None | Adds one device | Non-PoE switch on a short run |
How many hops before the design becomes a liability?
One hop is routine and two are defensible. Past three, the design is usually telling you something.
The reasons accumulate rather than multiply. Each unmanaged extender removes link diagnostics, so a port that drops at 02:00 leaves no log entry and no counter. Three units in series with 100,000-hour MTBF each give a chain MTBF near 33,000 hours — roughly four years of continuous running before the first failure, and the first failure takes down every camera downstream.
There is also a standards point worth stating plainly. A PD is designed and tested for a 100 m channel. Hand it a 300 m path and nothing in the standard protects the installation: the PD's input voltage range is the only safety net, and its failure mode is intermittent rather than obvious. Cascaded extenders can also break LLDP and two-event classification, so a managed switch reports a powered camera as Class 0 without knowing what it is or what it needs.
A practical rule: one hop for any 802.3at load, two hops only when the load is af-class and the PSE regulates at 56–57 V, and three hops only for devices with a wide input range and a maintenance route that does not involve climbing a tower in the rain.
FAQ
Can I go past 100 m with a high-quality Cat6 cable?
Not as a design assumption. Cat6 with 23AWG conductors lowers loss, which helps the voltage budget, but the data limit is a channel specification and does not extend with cable quality alone. Some links come up at 120–150 m; a design that depends on it will not survive a cable change or a temperature swing.
Does a PoE extender need its own power supply?
A cascaded extender is normally fed by PoE from the upstream segment and passes power downstream, consuming a small amount for itself. That is why every hop has to be subtracted from the port budget — the extender is a load, not a wire.
What happens to the data rate through an extender?
A good extender re-drives the link at full rate on both sides, so a 1000BASE-T camera stays at 1000BASE-T. Passive units that merely bridge the pairs can negotiate a lower rate or drop the link entirely under load.
Is fibre or a cascade of extenders better for 400 m?
Fibre, if there is any way to power the far end, because it removes the distance limit and the grounding path in one step. Extenders are the pragmatic answer when the far end has no power and the load is a single af-class camera.
Do extenders affect PoE class negotiation?
They can. Simple units pass the detection and classification handshake through, but some break two-event classification or LLDP, so the switch treats a Class 4 camera as Class 0 and allocates the af minimum. Check the reported class at the switch after commissioning rather than assuming the port behaves as designed.
