Power over Ethernet (PoE)
Running power over data cabling has been commonplace right from the outset of data transmission over copper. Why?
Because all transmission over wires requires a certain amount of power be it 5v signalling or 50v ring circuits.
The difference with a technology such as Power over Ethernet (PoE)
is that the aim is to power the end device
using the same transmission path, or at least the cable, as the data. This enables the elimination
of a separate cable solely for power use. In a building system environment it is normal practice to use shielded
twisted pair data cable to provide signalling and 24v power. This STP cable is often of 22AWG grade and can
carry current levels such as 1 ampere significant distances. Work has also been done on using 24AWG UTP
within building system environments, the idea being to unify structured cabling systems for data, voice
and building systems applications (See The Use of SCS in Intelligent Buildings
for more information). Using 24AWG cable requires careful thought in the design as the
smaller cross-sectional area imposes some limitations on current level and distances allowed.
The advantages of using the same cable to transmit power and data
include the elimination of AC power and DC transformers at the device location, consequently reducing power
infrastructure requirements, plus the
simplification and central control of power distribution by the Ethernet switching device.
One of the main application drivers for PoE was IP Telephony. The principle aim here was to replace
the traditional analogue/digital handset with an IP-based equivalent. Of course, in order to do this
the IP phone had to operate off one cable mimicking the original telephone. Manufacturers began
working on their proprietary versions of PoE in the late 1990s, Ethernet had, by then, become the dominant LAN
technology. Cisco was one of those manufacturers that began including PoE capability within its switches in 2000.
Other applications that gained benefit from having PoE included wireless access points, IP security cameras, card readers,
handheld devices and print servers.
802.3af refers to the point at which the 802.3 committee approved clause 33 (June 12th 2003). This clause
is entitled Data Terminal Equipment (DTE) Power via Media Dependent Interface
and defines the characteristics of the Powered Device (PD)
and the Power Sourcing
. 802.3af applies to the MDI interfaces 10BaseT, 100BaseT and 1000BaseT interfaces only
running on 100Ω balanced cabling. The MDI interface also acts as a Powered Interface (PI)
In plain English this means that network devices (PD) may be powered, as well as communicate via the LAN interface (PI),
receiving this power from a LAN switch (endpoint PSE) or inline alternative (Midspan PSE).
You can download sections of the 802.3 standard at
The defined characteristics are as follows:
- Classification of devices based on their power requirements (PSE)
- A power source to add power to the 100 Ω balanced cabling system (PSE)
- The characteristics of a powered device's load on the power source and the structured cabling
- A protocol allowing the detection of a powered device (PSE)
- A method for scaling power back to the detect level when power is no longer required (PSE)
- Supply of 48VDC and up to 350mA so that the power initiated does not exceed 15.4W and the power that is delivered
does not exceed 12.95W
- Supported on Category 3, 5, 5e and 6 cable
The following diagrams illustrate the alternative power supply arrangements on the 4-pair cabling used
to transmit 10BaseT, 100BaseT and 1000BaseT. Bear in mind that 10BaseT and 100BaseT use
pairs 2 and 3 for data whereas 1000BaseT uses all four pairs. Alternative A refers to providing power on the
same pair as the data signal whereas Alternative B refers to the supply of power on the unused
pair (pairs 1 and 4).
With Endpoint PSE Alternative A the power is supplied on the same pair as the data signalling from the point of view
of 10BaseT and 100BaseT. These pairs terminate on pins 1,2 and 3,6 of the RJ45. The term
Phantom Power is used to describe the power running over the same wires as the data signal.
This is always going to be the case with 1000BaseT because it uses all four pairs
and this is why PoE may be used alongside Gigabit Ethernet in this scenario. The power
source is the LAN switch (or hub) and the device is able to receive its power on the data pair or the unused pair.
With Endpoint PSE Alternative B the power is supplied on the unused pairs
to form the Power Interface (PI), the data signalling pair is untouched.
Again this is from the point of view of 10BaseT and 100BaseT where the unused pairs terminate on pins 4,5 and 7,8
of the RJ45, 1000BaseT cannot be run in this scenario!
With Midspan PSE Alternative B the power is supplied on the unused pair and this is carried out by a PSE
that sits within the cable path between the switch and the device. This is referred to as Inline Power.
Note that the LAN switch is not capable of providing power. Again be aware that because a midspan device
delivers power on the unused data pair i.e. the PI, such a device may not be used with Gigabit Ethernet because it uses
all four pairs to send data! The Midspan device may be a Power Injector or a Powered Patch Panel.
Taking the standard RJ45 pin assignment (T568B wiring):
A table may be drawn up showing the pinouts for Power over Ethernet:
||MDI Alternative A
||MDI-X Alternative A
||AII Alternative B
It is perfectly possible to have Alternative A and Alternative B PoE operational for the same device, with one
acting as backup for the other to provide seamless failover. It is also possible for this arrangement to cause
problems with incompatible signalling. To aid this power resilience the PD must be capable of sustaining a voltage of up to 57v
plus be able to cope with polarity reversing on the PI e.g. the MDI/MDI-X scenario.
Clause 33 of the 802.3 standard provides great detail on the following requirements:
- State diagrams for the PSE and PD covering the order of events and decision making including
the PD detection cycle and backoff algorithm
- Permissable current and voltage levels
- Electrical validation circuit diagrams
- The supply of power by the PSE only occurs once detection and a request for power has been sent by the PD
- Power supply is independent of any data link status
- Power output specifications for each class
- Power off voltage (idle)
- Power turn on time
- Input Capacitance
- Return Loss
- Insertion Loss
- Near End Crosstalk (NEXT)
- Compatibility with Category 3 cable as well as higher grade Category 5/5e/6
The above list is not exhaustive. The complexity of parameters that have to be considered is due to the importance
of not only maintaining equipment safety and reliability, but also the integrity of the data path that has to be
PSE Classification of devices is an optional component of 802.3af. Classification is useful when managing power
load. The following table lists the classes:
||Minimum PSE Power output level (W)
||Range of maximum power used by the PD (W)
||0.44 to 12.95
||0.44 to 3.84
||3.84 to 6.49
||6.49 to 12.95
A PoE connection takes the following steps when initiating:
- Detection - A pulse to measure the powered device for the correct signature resistance of 15–33 kΩ,
for this pulse the voltage range needs to be 2.7-10.0
- Classification - Next, detect which class of device the resistor indicates, for this the voltage range needs
to be 14.5-20.5
- Startup - The powered device starts up when the voltage is greater than 42
- Normal operation - Supply power to device when the voltage range is 36-57
IEEE 802.3af uses DC powered device detection, however originally Cisco prestandard Power over
Ethernet implementation used AC powered device detection.
DC detection differs from Cisco's AC detection in that AC detection transmits a low frequency AC signal on the transmit pair
and waits for the
same signal to be returned on the receive pair. DC detection uses a DC Current and detects the presence of a
powered device by measuring the load applied by the powered device.
Being limited to a small discrete power supply levels of 4W, 7W and a maximum of 15.4W is a nuisance if you need to
power motorised cameras such as Pan-Tilt-Zoom (PTZ) cameras,
or if you want to power small thin clients, or multiple radio Access Points such as those required in 802.11n.
This is especially considering that the maximum power that can be drawn by the PD (at the application) is actually 12.95W.
In September 2005 the IEEE set up the 802.3at committee to produce a standard for higher Power over Ethernet. Initially
this was designated PoE Plus.
You can keep up to date with the latest developments in the
Public area for IEEE 802.3at
Backwards compatiblity with 802.3af is required so there is a low power mode for the legacy PDs. There is a limit to how
much power may be drawn across the 24AWG wires before electrical damage occurs due to overheating within connectors and
cable bundles, plus signal interference issues. This means that multiple pairs may need to be used to deliver power.
At the moment, 802.3at limits the number of pairs that may carry power to two.
Currently a current limit of 720mA is being considered allowing 29.5W per pair,
however Draft 3.0 of 802.3at is looking to reduce this to 600mA giving 25W per pair, or 50W per device.
802.3at are also looking at Category 5 cables and above in order to fix the specification and not have to
worry about supporting Category 3 cabling.
With 802.3at the maximum power that can be delivered to the application is 50W. The first detection pulse or
Classification Pulse from the PSE will be the same as 802.3af to which the 802.3af PDs will respond normally. There
will then be a second PoE Plus pulse that will cause a 802.3at PD to respond as a Class 4 device and draw the Class 4 current.
Once this has happened there will be a data exchange where information such as duty-cycle, peak and average power needs
will be shared.
Other features to be catered for in 802.3at include
Dynamic power assignment leading to a more efficient power supply designs and consequent power saving.