Power over Ethernet or PoE technology describes a system to pass electrical power safely, along with data, on Ethernet cabling. The IEEE standard for PoE requires Category 5 cable or higher for high power levels, but can operate with category 3 cable for low power levels.[1] Power is supplied in common mode over two or more of the differential pairs of wires found in the Ethernet cables and comes from a power supply within a PoE-enabled networking device such as an Ethernet switch or can be injected into a cable run with a midspan power supply.
The original IEEE 802.3af-2003[2] PoE standard provides up to 15.4 W of DC power (minimum 44 V DC and 350 mA[3][4]) to each device.[5] Only 12.95 W is assured to be available at the powered device as some power is dissipated in the cable.[6]
The updated IEEE 802.3at-2009[7] PoE standard also known as PoE+ or PoE plus, provides up to 25.5 W of power.[8] The 2009 standard prohibits a powered device from using all four pairs for power.[9] Some vendors have announced products that claim to be compatible with the 802.3at standard and offer up to 51 W of power over a single cable by utilizing all four pairs in the Category 5 cable.[10]
Numerous non-standard schemes had been used prior to PoE standardization to provide power over Ethernet cabling. Some are still in active use.
Contents |
This technology is especially useful for powering IP telephones, wireless LAN access points, cameras with pan tilt and zoom (PTZ), remote Ethernet switches, embedded computers, thin clients and LCDs.
All these require more power than USB offers and very often must be powered over longer runs of cable than USB permits. Field-spliced outdoor category 5 Ethernet cable can power radios and other low-power devices, for instance, through over 100 m of cable, an order of magnitude further than USB's theoretical maximums.
In addition, PoE uses only one type of connector, an 8P8C modular connector (often called RJ45), whereas there are numerous types of USB connectors and each new USB standard has added more.
PoE is presently deployed in applications where USB is unsuitable and where AC power would be inconvenient, expensive[note 1] or infeasible to supply.
However, even where USB or AC power could be used, PoE has several advantages over either, including:
Some types of devices with PoE include:[11]
Power sourcing equipment (PSE) is a device such as a switch that provides ("sources") power on the Ethernet cable. The maximum allowed continuous output power per cable in IEEE 802.3af is 15.40 W. A later specification, IEEE 802.3at, offers 25.50 W.
When the device is a switch, it's called an endspan. Otherwise, if it's an intermediary device between a non PoE capable switch and a PoE device, it's called a midspan. An external PoE injector is a midspan device[12]
A powered device (PD) is a device powered by a PSE and thus consumes energy. Examples include wireless access points, IP Phones, and IP cameras.
Many powered devices have an auxiliary power connector for an optional, external, power supply. Depending on the PD design, some, none, or all power can be supplied from the auxiliary port,[13][14] with the auxiliary port sometimes acting as backup power in case of PoE supplied power failure.
Most advocates expect PoE to become a global longterm DC power cabling standard and replace "wall wart" converters, which cannot be easily centrally managed, waste energy, are often poorly designed, and are easily vulnerable to damage from surges and brownouts.
Critics of this approach argue that DC power is inherently less efficient than AC power to push through long cables, and this is made worse by the thin conductors of Ethernet. A typical 48-port Ethernet switch has a 50 W to 80 W power supply allocated for the traditional Ethernet switch and transceiver IC. Over and above this it requires typically a 370 W (for 802.3af) to 740 W (for 802.3at) power supply allocated solely for PoE ports, permitting a maximum draw on each. This can be quite inefficient to supply through long cables. However, where this central supply replaces several dedicated AC circuits, transformers and inverters, and prevents expensive human interventions (AC installations) the power loss of long thin DC cable is easily justifiable. Power can always be introduced on the device end of the Ethernet cable (radically improving efficiency) where AC power is available.
Beyond the inherent advantages of an optimized inject-anywhere AC-and-DC cabling infrastructure, the switches themselves often contain "active", "smart", or "managed" power management features to further reduce AC draw of all devices combined.
Multi-protocol teaming standards (G.9960, G.hn, and IEEE P1905) and handoff standards (IEEE 802.21) generally rely on simulating Ethernet features in other media. These standards enable more optimal energy and bandwidth management solutions than would otherwise be possible. For instance, networking on existing AC power lines to an outlet where a PoE router is plugged, making it capable of moving a gigabit per second to every device, with minimal wiring, participating fully in both AC and DC device power demand management. Or, letting a session migrate from a high-power Ethernet switch to a low-power power-over-ethernet wireless routing when the need for bandwidth is low and there is no need for power on the Ethernet cable to be supplied to the device.
By late 2011, managed switches with both powered and unpowered Ethernet ports, from Netgear and others, featured many significant energy management features, some of which are proprietary but are likely to migrate into the eventual standard. For instance, "auto power-down and cable-length detection" allowing lower signal strength to be used.[15] Given that such features were available in switches selling for under US$250, power savings alone could justify some users switching to a security, VoIP or wireless AP infrastructure based on power over Ethernet, as they would pay for it very quickly.
Power advantages are a major sales appeal of powered over unpowered Ethernet or unpowered alternatives (such as strictly wireless sensor networks, which must in practice rely on batteries if they draw more than their own solar capacity). Advertising for power-over-Ethernet devices usually cites its "green" features including less packaging and improvements over previous models.
After integration with the IEEE 802.3az Energy-Efficient Ethernet (EEE) standard, the energy management capabilities of the combined standard are expected to be formidable. Pre-standard integrations of EEE and PoE (such as Marvell's EEPoE outlined in a May 2011 white paper [1]) claim to achieve a savings upwards of 3 watts per link, extremely significant across the tens of millions of new links shipped each year. These losses are especially significant as higher power devices come online. Marvell claims that:
Standards-based power over Ethernet is implemented following the specifications in IEEE 802.3af-2003 (which was later incorporated as clause 33 into IEEE 802.3-2005) or the 2009 update, IEEE 802.3at. A phantom power technique is used to allow the powered pairs to also carry data. This permits its use not only with 10BASE-T and 100BASE-TX, which use only two of the four pairs in the cable, but also with 1000BASE-T (gigabit Ethernet), which uses all four pairs for data transmission. This is possible because all versions of Ethernet over twisted pair cable specify differential data transmission over each pair with transformer coupling; the DC supply and load connections can be made to the transformer center-taps at each end. Each pair thus operates in common mode as one side of the DC supply, so two pairs are required to complete the circuit. The polarity of the DC supply may be inverted by crossover cables; the powered device must operate with either pair: spare pairs 4–5 and 7–8 or data pairs 1–2 and 3–6. Polarity is required on data pairs, and ambiguously implemented for spare pairs, with the use of a diode bridge.
| Property | 802.3af (802.3at Type 1) | 802.3at Type 2 |
|---|---|---|
| Power available at PD[note 2] | 12.95 W | 25.50 W |
| Maximum power delivered by PSE | 15.40 W | 34.20 W |
| Voltage range (at PSE) | 44.0–57.0 V[16] | 50.0–57.0 V[16] |
| Voltage range (at PD) | 37.0–57.0 V[17] | 42.5–57.0 V[17] |
| Maximum current | 350 mA[18] | 600 mA[18] per mode |
| Maximum cable resistance | 20 Ω (Category 3) | 12.5 Ω (Category 5) |
| Power management | Three power class levels negotiated at initial connection | Four power class levels negotiated at initial connection or 0.1 W steps negotiated continuously |
| Derating of maximum cable ambient operating temperature | None | 5°C with one mode (two pairs) active |
| Supported cabling | Category 3 and Category 5[1] | Category 5[1][note 3] |
| Supported modes | Mode A (endspan), Mode B (midspan) | Mode A, Mode B |
Two modes, A and B, are available. Mode A delivers phantom power on the data pairs of 100BASE-TX or 10BASE-T. Mode B delivers power on the spare pairs. PoE can also be used on 1000BASE-T Ethernet in which case, there are no spare pairs and all power is delivered using the phantom technique.
Mode A has two alternate configurations (MDI and MDI-X), using the same pairs but with different polarities. In mode A, pins 1 and 2 (pair #2 in T568B wiring) form one side of the 48 V DC, and pins 3 and 6 (pair #3 in T568B) form the other side. These are the same two pairs used for data transmission in 10BASE-T and 100BASE-TX, allowing the provision of both power and data over only two pairs in such networks. The free polarity allows PoE to accommodate for crossover cables, patch cables and auto-MDIX.
In mode B, pins 4–5 (pair #1 in both T568A and T568B) form one side of the DC supply and pins 7–8 (pair #4 in both T568A and T568B) provide the return; these are the "spare" pairs in 10BASE-T and 100BASE-TX. Mode B, therefore, requires a 4-pair cable.
The PSE, not the powered device (PD), decides whether power mode A or B shall be used. PDs that implement only Mode A or Mode B are disallowed by the standard.
The PSE can implement mode A or B or both. A PD indicates that it is standards-compliant by placing a 25 kΩ resistor between the powered pairs. If the PSE detects a resistance that is too high or too low (including a short circuit), no power is applied. This protects devices that do not support PoE. An optional "power class" feature allows the PD to indicate its power requirements by changing the sense resistance at higher voltages. To stay powered, the PD must continuously use 5–10 mA for at least 60 ms with no more than 400 ms since last use or else it will be unpowered by the PSE.[19]
There are two types of PSEs: endspans and midspans. Endspans are Ethernet switches that include the power over Ethernet transmission circuitry. Endspans are commonly called PoE switches. Midspans are power injectors that stand between a regular Ethernet switch and the powered device, injecting power without affecting the data.
Endspans are normally used on new installations or when the switch has to be replaced for other reasons (such as moving from 10/100 Mbit/s to 1 Gbit/s or adding security protocols), which makes it convenient to add the PoE capability. Midspans are used when there is no desire to replace and configure a new Ethernet switch, and only PoE needs to be added to the network.
| Stage | Action | Volts specified [V] |
|
|---|---|---|---|
| 802.3af | 802.3at | ||
| Detection | PSE detects if the PD has the correct signature resistance of 19–26.5 kΩ | 2.7–10.0 | |
| Classification | PSE detects resistor indicating power range (see below) | 14.5–20.5 | |
| Mark 1 | Signals PSE is 802.3at capable. PD presents a 0.25–4 mA load. | — | 7–10 |
| Class 2 | PSE outputs classification voltage again to indicate 802.3at capability | — | 14.5–20.5 |
| Mark 2 | Signals PSE is 802.3at capable. PD presents a 0.25–4 mA load. | — | 7–10 |
| Startup | Startup voltage [20][21] | > 42 | > 42 |
| Normal operation | Supply power to device [20][21] | 37–57 | 42.5–57 |
IEEE 802.3at capable devices are also referred to as "type 2". An 802.3at PSE may also use layer2 communication to signal 802.3at capability.[22]
| Class | Usage | Classification current [mA] |
Power range [Watt] |
Class description |
|---|---|---|---|---|
| 0 | Default | 0–4 | 0.44–12.94 | Classification unimplemented |
| 1 | Optional | 9–12 | 0.44–3.84 | Very Low power |
| 2 | Optional | 17–20 | 3.84–6.49 | Low power |
| 3 | Optional | 26–30 | 6.49–12.95 | Mid power |
| 4 | Valid for 802.3at (Type 2) devices, not allowed for 802.3af devices |
36–44 | 12.95–25.50 | High power |
Class 4 can only be used by IEEE 802.3at (type 2) devices, requiring valid Class 2 and Mark 2 currents for the power up stages. An 802.3af device presenting a class 4 current is considered non-compliant and, instead, will be treated as a Class 0 device.[24]
| TLV Header | MED Header | Extended power via MDI | |||||
|---|---|---|---|---|---|---|---|
| Type (7 bits) |
Length (9 bits) |
TIA OUI (3 octets) |
Extended power via MDI subtype (1 octet) |
Power type (2 bits) |
Power source (2 bits) |
Power priority (4 bits) |
Power value (2 octets) |
| 127 | 7 | 00-12-BB | 4 | PSE or PD | Normal or Backup conservation | Critical, High, Low |
0–102.3 W in 0.1 W steps |
The setup phases are as follows:
The rules for this power negotiation are:
Cisco manufactured WLAN access points and IP phones many years before there was an IEEE standard for delivering PoE. Cisco's original PoE implementation is not software upgradeable to the IEEE 802.3af standard. Cisco's original PoE equipment was capable of delivering up to 10 W per port. The amount of power to be delivered is negotiated between the endpoint and the Cisco switch based on a power value that was added to the Cisco proprietary Cisco Discovery Protocol (CDP). CDP is also responsible for dynamically communicating the Voice VLAN value from the Cisco switch to the Cisco IP Phone.
Under Cisco's pre-standard scheme, the PSE (switch) will send a Fast Link Pulse (FLP) on the transmit pair. The PD (device) connects the transmit line to the receive line via a low pass filter. And thus the PSE gets the FLP in return. And a common mode current between pair 1 and 2 will be provided resulting in 48 V DC[26] and 6.3 W[27] default of allocated power. The PD has then to provide Ethernet link within 5 seconds to the auto-negotiation mode switch port. A later CDP message with a type-length-value tells the PSE its final power requirement. A discontinued link pulses shuts down power.[28]
PowerDsine, now a Microsemi brand, have been selling midspan power injectors since 1999 with its proprietary Power over LAN solution. Several companies such as Polycom, 3Com, Lucent and Nortel utilize PowerDsine's Power over LAN.[29]
Numerous devices exist which make use of positive power received on pins 4 and 5 of the Ethernet cable, with negative return on pins 7 and 8. Power over Ethernet injectors exist which convert 110-240 V 50/60 Hz AC power to 9 V, 15 V, 24 V, 48 V or 55 V DC with currents ranging from 0.5 A to 2.0 A. In the common "passive" PoE system the injector does not communicate with the powered device to negotiate its wattage requirements, but merely supplies power. Passive DC to DC injectors also exist which convert a 9 V to 36 V DC input power source to a stabilized 24 V 1 A or 48 V 0.5 A PoE feed with pins +4,5 and -7,8. These DC to DC PoE injectors are used in a variety of different telecom applications.[30]
Category 5 cable uses 24 AWG conductors, which can safely carry 360 mA at 50 V according to the latest TIA ruling. The cable has eight conductors (only half of which are used for power) and therefore the absolute maximum power transmitted using direct current is 50 V × 0.360 A × 2 = 36 W. Considering the voltage drop after 100 m, a PD would be able to receive 31.6 W. The additional heat generated in the wires by PoE at this current level (4.4 watts per 100 meter cable) limits the total number of cables in a bundle to be 100 cables at 45 °C, according to the TIA. This can be somewhat alleviated by the use of Category 6 cable which uses 23 AWG conductors.
| PINS on Switch | 10/100 DC on Spares (mode B) | 10/100 Mixed DC & Data (mode A) | 1000 (1 Gigabit) DC & Bi-Data (mode B) | 1000 (1 Gigabit) DC & Bi-Data (mode A) |
|---|---|---|---|---|
| Pin 1 | Rx + | Rx + DC + | TxRx A + | TxRx A + DC + |
| Pin 2 | Rx - | Rx - DC + | TxRx A - | TxRx A - DC + |
| Pin 3 | Tx + | Tx + DC - | TxRx B + | TxRx B + DC - |
| Pin 4 | DC + | unused | TxRx C + DC + | TxRx C + |
| Pin 5 | DC + | unused | TxRx C - DC + | TxRx C - |
| Pin 6 | Tx - | Tx - DC - | TxRx B - | TxRx B - DC - |
| Pin 7 | DC - | unused | TxRx D + DC - | TxRx D + |
| Pin 8 | DC - | unused | TxRx D - DC - | TxRx D - |
|
|||||||||||||||||||||||