Two numbers, and only one of them is on the box
Power over Ethernet is sold on port count and quietly limited by budget. A switch with eight PoE ports each capable of 30 watts is not a 240 watt switch; it is usually a 120 or 130 watt switch that can put 30 watts down any one port as long as the others are not all doing the same. The budget is the shared pool, and it is the number that decides how many cameras you can actually run.
The second number is what the cable takes on the way. PoE sends power down the same twisted pairs that carry data, at somewhere between 44 and 57 volts, and copper has resistance. The current squared times the loop resistance is turned into heat inside the wall, and both terms work against you: longer runs have more resistance, and thinner conductors have more resistance per foot.
The classes, and what each one actually guarantees
The standards are written from both ends. Each type specifies what the switch must be able to put out and what the device is guaranteed to receive after the worst allowable cable, and the gap between those two figures is the cable loss the standard budgets for.
| Type | At the port | Guaranteed at the device | Pairs used | Typical loads |
|---|---|---|---|---|
| 802.3af, Type 1 | 15.4 W | 12.95 W | 2 | Fixed cameras, doorbells, phones |
| 802.3at, Type 2 | 30 W | 25.5 W | 2 | Cameras with heaters or strong infrared, access points |
| 802.3bt, Type 3 | 60 W | 51 W | 4 | Pan-tilt-zoom cameras, multi-sensor units |
| 802.3bt, Type 4 | 90 W | 71.3 W | 4 | Large PTZ units with heaters, displays |
Those guaranteed figures are not arbitrary. Work a Type 2 run at the standard worst-case channel resistance of 12.5 ohms: 30 watts at 50 volts is 0.6 amps, and 0.6 squared times 12.5 is 4.5 watts of loss, leaving exactly 25.5. Type 4 at its 6.25 ohm four-pair limit gives 71.3. The standards are this same arithmetic run against the worst cable they permit.
Which means real installations usually do better. A hundred metres of 24 AWG solid copper is about 8.4 ohms on a two-pair run, not 12.5, so a Type 2 camera at the end of a full-length run receives close to 27 watts rather than the 25.5 it is promised. Good cable buys margin the standard does not assume.
Where the loss actually bites
The loss is proportional to length and to resistance per foot, so the two things that make it hurt are long runs and thin conductors. Thirty feet of 24 AWG on a Type 2 camera loses about a quarter of a watt, which is nothing. Three hundred feet of the same cable loses close to three watts, and three hundred feet of 26 AWG loses more than four. Copper-clad aluminium, which is sold cheaply and looks identical once terminated, has roughly 55 percent more resistance than solid copper of the same gauge and turns a comfortable run into a marginal one.
Four-pair PoE halves the effective resistance because the current is shared over twice as many conductors, which is part of why the higher-power types were defined that way rather than just pushing more current down two pairs.
These runs are carrying data as well, and the 100 metre channel limit governs both, which the AV cable run calculator translates into route length once slack and service loops are counted. The devices at the far end are sized by the pixel density calculator and their streams by the camera bandwidth calculator, and for loss on ordinary mains circuits rather than on PoE, the voltage drop calculator is the equivalent tool.
What happens when the budget runs out
This is the failure worth understanding before it happens, because it does not look like a power problem. A switch that has committed its full budget stops granting power to further devices, and when the load rises past the budget on a switch already running, it sheds ports. How it chooses is firmware behaviour: some go by port priority if you configured it, most go by port number, and the result is a camera going dark for reasons that look nothing like a power fault from the app. Worse, it can oscillate — a port comes up, the budget is exceeded, something drops, the budget frees, the port comes up again.
The margin you leave is what prevents this, and there is a specific reason to leave more than feels necessary. Cameras do not draw their class maximum most of the time. They draw it on the coldest night of the year with the heater running and the infrared illuminators at full output, which is precisely the night the system needs to work. Size for that night. This calculator uses class maxima throughout for that reason.
Injectors, midspans and the second switch
When one or two devices break the budget, a single-port injector on those runs solves it without replacing anything: the injector supplies power from its own wall adapter and the switch handles data only. When the whole system is over, a second switch or a larger one is cleaner than a rack of injectors and their individual power bricks, each of which is a thing to lose and a thing to fail.
One warning applies to both. Passive PoE injectors, which put a fixed voltage on the cable without the negotiation the standards define, will damage a device that was not expecting it. Standards-compliant equipment negotiates before power flows, which is what makes it safe to plug anything into a PoE port. Passive equipment does not negotiate, and matching it to the wrong device is how hardware dies. If a supply does not state which standard it implements, read the silence as the answer.
Questions people ask
How many cameras can one PoE switch run?
Divide the switch budget by the class draw of the cameras, then leave margin. A 130 watt switch running Type 2 cameras at 30 watts each covers four comfortably, five if nothing else changes, and eight only if the cameras are Type 1 at 15.4 watts. The port count is almost never the limit — the shared budget is. Read the budget figure off the spec sheet rather than multiplying the port count by the per-port maximum, because the two are rarely the same number, and leave twenty percent or so spare for the cold night when heaters and infrared push every camera toward its class maximum at once.
How far can PoE go?
The specified channel is 100 metres, about 328 feet, and that includes the patch leads at both ends, not just the cable in the wall. Within it, power is not usually what fails first — data errors are, because Ethernet timing is what the distance limit was written around. Power loss over a full-length run of decent copper is a few watts, which the standards already budget for. Beyond 100 metres you are outside the specification, and the honest answers are a small switch or media converter part way along, or fibre for the long leg with a powered switch at the far end. Cable sold as extending PoE beyond the limit exists and is a compromise rather than a fix.
Does cable length reduce the power a camera gets?
Yes, and by a knowable amount. The loss is the current squared times the loop resistance of the run, so it grows with length and with thinner conductors. A Type 2 camera drawing 0.6 amps over 300 feet of 24 AWG loses around 2.8 watts in the cable, so about 27 watts of the 30 sent actually arrives. That is still above the 25.5 watts the standard guarantees, which is why full-length runs work. Drop to 26 AWG or use copper-clad aluminium and the same run gets much closer to the edge, which is where an intermittently rebooting camera at the end of a long run usually comes from.
What is the difference between 802.3af, at and bt?
They are successive versions of the same standard supplying more power. Type 1 under 802.3af puts 15.4 watts on the port and guarantees 12.95 at the device, over two pairs, which runs most fixed cameras. Type 2 under 802.3at raises that to 30 and 25.5, which covers cameras with heaters and infrared illuminators and most access points. 802.3bt adds Type 3 at 60 and 51 watts and Type 4 at 90 and 71.3, both using all four pairs, which halves the resistance and is how they carry the extra current. Devices and switches negotiate down to whatever they both support, so a Type 1 camera on a Type 4 port takes 15.4 watts and nothing more.
Can I use copper-clad aluminium cable for PoE?
It carries data and it will pass power, and it is a poor choice for PoE specifically. CCA has roughly 55 percent more resistance than solid copper at the same gauge, so every run loses proportionally more and runs hotter under load, which matters most in a bundle of powered cables in a conduit where the heat has nowhere to go. It is also more brittle at terminations and tends to fail at punch-downs over time. Beyond the engineering, it falls outside what many wiring specifications and installation codes permit for this use, and whether it is allowed in your installation is a question for local rules rather than for a calculator.