What Is PoE: One Cable, Four Wattage Tiers
PoE sends DC power down the same copper pair that carries data.
That is the whole trick, and it has been standardised since 2003. A camera, a phone, an access point or a small screen can run with no wall adapter, because the network cable already reaches it.
The useful part for a buyer is not the idea. It is the four power tiers, and the gap between what a switch promises and what the device actually receives.
The Short Answer
PoE is short for Power over Ethernet. A power sourcing device, usually a switch, injects a common DC voltage onto the cable. The powered device takes it off with a centre tap on its transformer.
Four tiers exist today, and each has two numbers. One is what the source must supply. The other is what the device may rely on receiving.
| Tier | Standard | At the source | At the device |
|---|---|---|---|
| Type 1 | 802.3af, 2003 | 15.4 W | 12.95 W |
| Type 2 | 802.3at, 2009 | 30 W | 25.5 W |
| Type 3 | 802.3bt, 2018 | 60 W | 51 W |
| Type 4 | 802.3bt, 2018 | 90 W | 71.3 W |
The first tier runs at a minimum of 44 V DC and 350 mA. That is the floor of the whole system, and it is why the numbers look odd rather than round.
The gap between the two columns is not slack. It is the power the standard expects to lose in the cable, and it is calculated for a run of up to 100 metres.
Where the Power Rides
A standard Ethernet cable holds four twisted pairs. PoE uses two of them, or all four, depending on the tier and the design.
Two arrangements are common in the older tiers. One puts power on the same pairs that carry data. That works because the voltage is common to both wires, while the data signal is differential. The other puts power on the spare pairs.
The fast tiers were held back by design, not by copper. Type 2 explicitly does not allow all four pairs to carry power at the same time.
So Type 3 and Type 4 arrived with a different rule. They may use all four pairs, in parallel, which halves the current in each conductor for the same total power.
That change is the reason the later tiers exist at all. More pairs, less current per wire, and less heat in a bundle.
Nothing Gets Power Until It Says Hello
A PoE port does not switch on blind. Before any voltage appears, the source looks for a signature.
The signature is a resistor placed across the power pairs inside the device. The classic value is 25 kΩ, and the source supplies power only when it finds something close to it. That signature check is what keeps a laptop safe.
That single resistor is why a laptop does not fry when someone plugs it into a PoE switch. The port measures, finds no signature, and stays off.
After detection comes a short conversation about class, which tells the source how much power to reserve. Modern equipment can negotiate further up the stack, so a device may ask for more or less once it has booted.
The order matters for a buyer. Detection, then class, then power. Any product that skips the first step is not running standard PoE.
The Word Passive Should Stop You
Plenty of hardware is sold as PoE and is not. Passive PoE, sometimes called always-on, puts 24 V or 48 V on the pairs with no detection and no negotiation.
It works, and it is cheap. It also means the voltage is present before anyone agrees on it, and a device that expected a different tier can be damaged.
Two questions separate the two worlds. Does the port test for a signature before powering up. And does it negotiate a class.
If the answer to either is no, the product is a DC injector in an Ethernet jacket. That can be fine for one device from one maker, and it is not a standard.
| Standard PoE | Passive, always on | |
|---|---|---|
| Checks before power | Signature, then class | Nothing |
| Safe with an unknown device | Yes | No |
| Voltage on the pairs | Negotiated, within the tier | Fixed, from the moment it is plugged in |
| Cost | Higher | Lowest |
What Actually Arrives at the Device
The gap between source and device power is the part buyers miss, and it is deliberate. The standard budgets the loss rather than pretending it does not happen.
Type 1 halves the story in one line. A port supplies 15.4 W, and the device is promised 12.95 W. The difference is 2.45 W, spent as heat in 100 metres of cable.
At the top tier the allowance is larger in absolute terms, because the current is far higher. A Type 4 port supplies 90 W and the device is promised 71.3 W.
Current per pair is capped as well. Type 3 allows up to 600 mA on a pair, and Type 4 up to 960 mA.
Those caps explain the cable rule. Type 3 and Type 4 need at least Cat 5. Thinner or poorer copper turns part of that current into heat instead of delivering it.
Why a short cable is not free margin
A buyer sometimes assumes a two-metre run gives plenty of spare power. It does give margin, but the margin is already inside the standard’s numbers.
The device figure is defined for the worst allowed cable. Shortening the run recovers a little of the budget. It does not create a bigger power supply.
So a 12.95 W camera on a 15.4 W port, with a short cable, still cannot draw 15 W. The port and the device class decide that.
An installer can check the real figure in a minute. A small inline meter between the port and the device shows the watts actually drawn, and the number is often far below the class.
That reading is useful twice. It confirms the device is inside its class, and it shows how much of the switch budget is really in use.
Heat in a bundle
One cable in open air sheds heat easily. Forty cables in a closed tray do not.
The current per pair is low by power-cable standards, which is why PoE is safe to run in a data bundle. It is still current, and the standard sets the limits with normal installation in mind.
If a project pushes the top tier through a dense bundle in a hot ceiling, the installation rules matter as much as the switch rating.
The Port Rating Is Not the Budget
A datasheet that says 90 W per port describes one port. It does not describe the box.
Every PoE switch has a total power budget, and the supply inside it is smaller than the port count multiplied by the port rating.
Take a 24-port unit rated 90 W a port. Full load would be over two kilowatts. Almost no switch is built that way, because most installations never reach it.
So the number that matters is the budget, not the port. A 24-port switch with a 400 W budget will happily run cameras and phones, and it will refuse the fourth laptop.
The refusal is not always obvious. Some switches shed the lowest priority port when the budget runs out, which looks like a random device dropping off the network.
One more property belongs in the same sum. Power comes from one box, so one uninterruptible supply can hold every device through a short outage.
That is a real advantage over a shelf of wall adapters. It only works if the switch itself sits on the backup supply.
What to add up before ordering
Add the class figure, not the idle draw, for each device. The class is what the port reserves.
Then compare that total with the switch budget, and leave room for the next device. A plan that fits exactly on paper has no room for a camera added in year two.
Mid-span injectors are the usual fix. A single-port injector sits between a non-PoE switch and one device, and it carries the full tier on that one cable.
What the Marketing Names Promise
Catalogues use four names for three standards, and one of them is a brand rather than a standard.
PoE usually means Type 1. PoE+ means Type 2. Both are old enough that the name has become casual.
PoE++ and 4PPoE both point at the 802.3bt tiers. The second name describes the mechanism: four pairs.
Then there is UPoE, which is a vendor name from before the standard caught up. It delivered 60 W over four pairs and became the model for Type 3.
None of these names states a wattage you can design to. Two products can both say PoE++ and differ by 30 W at the device.
A line that says 802.3bt without a type leaves 30 W of ambiguity on the table. Type 3 or Type 4, and the port figure, remove it.
So an order should never use the marketing name alone. The standard number and the type number are what a factory can build to.
What the Device Side Must Declare
A powered device is not a passive load. It has to present a signature, declare a class, and survive the moment the port switches on.
That last point is where cheap designs fail. A large capacitor on the input draws a rush of current when the voltage appears. The port may read that rush as a fault and shut down.
The power arrives on the same wires that carry the data, so the device has to separate the two again. A transformer does that job, and the centre tap is where the DC comes out.
That transformer is also the isolation barrier. If a cheap design skips it, the network port and the mains side end up sharing metal, and no amount of software fixes that.
So the device needs a controlled start, a signature resistor, and a class that matches its real draw. Get those wrong and the product works on the bench with a bench supply and fails on a real switch.
One more thing belongs in the datasheet. The device should say what it draws in normal use, not only the class ceiling. A camera that idles at 4 W and peaks at 12 W fits a Type 1 port. The class tells the switch which budget to reserve.
Five Lines for the Order
Everything above fits into five lines, and they belong in the same sheet you hand a supplier for any other part.
One. Name the standard and the type, not the marketing word.
Two. State the device wattage you need, and the port wattage you expect to sit behind.
Three. Say whether the product must work on two pairs, four pairs, or both.
Four. Say that passive always-on power is not acceptable, unless it truly is expected.
Five. Set the cable and installation limits: length, category, and the bundle conditions at the site.
Ask for the datasheet line that states the standard and the type, and keep it with the drawing. A marketing name cannot be tested, and a type can.
The electrical side of the same product still needs its own file. A PoE circuit board is judged on workmanship by the assembly rules. Its insulation is checked by a high-voltage insulation check before anything is shipped.
For a product going into Europe, the safety mark question is separate again, and the EU paperwork is where that trail starts.
Send me the PoE line from your spec, or the sheet you were given. I will say whether the tier is named by standard or by marketing word, and whether the device figure matches a real port. No cost, no obligation.
Check My PoE SpecQuestions Buyers Ask
Does PoE work on any Ethernet cable?
Not any. The early tiers tolerate older cable, but Type 3 and Type 4 need at least Cat 5. On thin or aluminium-heavy cable, part of the power turns into heat before it reaches the device.
Can a PoE port damage a device that does not use PoE?
Not on a standard port. The port looks for a signature resistor first and stays off when it finds none. Damage happens with passive always-on injectors, which skip that check.
What is the difference between PoE+ and PoE++?
PoE+ is Type 2, at 30 W from the port and 25.5 W at the device. PoE++ covers the 802.3bt tiers, at 60 W and 90 W from the port. The names alone do not tell you which of the two you are buying.
Can I run the full 90 W over 100 metres?
The standard allows the 100 metre run and budgets the loss inside it. What the cable and the installation can take is a separate question, especially in a hot bundle.
Does a short cable give me extra power?
It recovers a little of the budgeted loss. It does not raise the port rating or the device class, so the ceiling stays where the hardware puts it.
What happens if the device draws more than its class?
The port protects itself. Current above the allowed level trips the port and the device loses power, which usually shows up as a camera that reboots on a cold night.
Questions & Comments
We read every one and reply within 24 hoursIf the spec says PoE++ and stops there, send me the line. I will say which tier it really is, and what reaches the device.