Ethernet, once thought to be the sole province of computer networks, has found itself linked to applications and devices far beyond its original scope. Aside from the normal computer to computer communications, Ethernet today links many diverse devices, including cameras, motor controls, VoIP phones, intercom paging and public-address systems, industrial devices for sensing, monitoring, and control, and wall clocks synchronized to network time protocols, to name just a few.
In many cases, Ethernet-connected devices must also connect to power sources of some type, typically standard ac power. But unless wall outlets are handy, this usually means running power, along with the Ethernet cabling, to the device, which adds cost and complexity. So one can see the advantage of a device that only needs one cable connection to provide both signal and power. Such was the case when IEEE created the original Power-over-Ethernet (PoE) 802.3af standard in 2003.
A new standard
The 802.3af standard details a method of transmitting up to 15.4 W of electrical power at a minimum of 44 V and 350 mA, along with Ethernet signals, in a single Cat-5 (or earlier Cat-3) cable. However, only 12.95 W is available for the device due to power losses in the cable. As more devices demanded greater power, the IEEE created a new standard, 802.3at, in 2009. The updated standard, called PoE Plus or PoE+, almost doubled the available power to 25.5 W and applied new terminology to the equipment used for PoE applications.
Power-sourcing equipment (PSE) identifies those pieces of equipment that supplies power to the Ethernet system. The source may be built into equipment, such as routers or switches, which are called endspans. Or it can be an add-on pass-through connection known as a midspan that adds PoE capability to non-PoE devices.
Any device that draws power from the Ethernet is known as a powered device or PD. In many cases, PDs also have auxiliary power connectors that supply the devices external power in case of PoE failures.
When 802.3at was adopted, IEEE split PDs into two types. Type 1 devices draw less than 13 W of power, making them compatible with the older standard. PDs that need more than 13 W, up to the 25.5-W maximum, are classified as Type 2 devices. A PSE that handles Type 2 devices can also run Type 1 devices, for backward compatibility.
A PoE power path is defined as having three components: the power of the PSE, the power delivered to the PD, and the power sent on to the application, which is obviously less than that delivered to the PD. This has led to some confusion as different manufacturers market different power levels. For example, a switch maker might hype its product’s PSE capacity, while a PD vendor would tout the power needed by the PD (the power delivered to the PD). And a sensor maker using PoE would be more concerned with the power available to the application to operate the sensor. When comparing power ratings, always be sure to verify which power point the manufacturer is speaking about.
Benefits of PoE
One major advantage of PoE is that it eliminates “wall warts,” the plug-in power transformers and supplies typically needed in remote situations. These transformers are notoriously inefficient, often poorly designed, and can be easily damaged by surges and brownouts.
Many PoE installations have been sold on its “green” merits alone. However, it’s possible to lose that efficiency without carefully planning the installation.
For example, a standard 48-port Ethernet switch with PoE capability typically uses a power supply between 60 to 80 W that provides power for the switch electronics. But it may need an additional 370 W (for 802.3af) or 740 W (for 802.3at) to supply maximum power to all PoE devices connected to the switch. As Cat-5 Ethernet cabling is usually 24-awg wire (23 awg for Cat-6), long runs tend to lose a greater portion of power. However, the advantage of not needing to run ac power sources to the PD justifies these losses. In addition, when ac power is readily available, the PD can be externally powered, improving efficiency at the PD and reducing the load on the PSE. The trade-off may be a slightly higher loss at the PD power supply.
PSEs may contain active, smart, or managed power features that reduce the draw of all powered devices. For instance, automatic power-down and cable-length detection lets the switch accept lower signal strengths, reducing power needs. And power can be shutdown to devices that do not need the PoE feature.
How it works
Delivering power from the PSE to the PD depends on the type of Ethernet connection. For example, 10Base-T and 100Base-TX connections use only two of the four pairs in a standard Cat-5 Ethernet cable. In such cases, PoE can be sent using either Mode A or Mode B transmissions. Mode A sends power in what’s known as “phantom” mode. PoE power is injected into the center-tap of the two active pair transformers in the PSE and removed via similar connections in the PD. Pins 1 and 2 of a standard 8C8P (aka RJ-45) connector share one polarity, while pins 3 and 6 provide access to the opposite polarity.
Mode B, on the other hand, relies on the two remaining inactive pairs as direct power wires, keeping each pair a single polarity. Pins 4 and 5 become one polarity, while pins 7 and 8 provide the return. Diode bridge rectifiers steer incoming voltages for proper polarity within the PD.
When working over 1000Base-TX, all four wire pairs transmit signals. Therefore, both Mode A and B are implemented using the phantom technique.
The PSE determines if the system runs Mode A, B, or both. It does this by detecting a 25-kΩ resistor between the powered pairs. If the PSE detects a resistance that is too high or low, no power is applied to the circuit. This protects the PSE from trying to power shorted wires, an open circuit, or a non-PoE-compliant Ethernet connection.
To stay powered, a PD must continuously use 5 to 10 mA for at least 60 msec but no more than 400 msec since its last use. Some PDs incorporate an optional “power class” feature that lets the PD indicate its power needs.
Negotiating power demands between the PSE and PD follows a specific sequence of operations. First the PSE tests the PD to make sure it’s properly connected and a good device. If the PSE is satisfied, the PD is powered. The PD then sends the PSE two pieces of information: its maximum power needs and the amount of power it’s requesting to use. The PSE responds with the maximum power it lets the PD use. The PD now uses the power specified by the PSE.
The negotiations are carried out according to a set of rules that starts with the PD never requesting more power than permitted by 802.3af (the 13-W level). It may also never draw more than the maximum power allocated by the PSE. The PSE may deny (turn off) any PD that draws more power than the PSE’s allowed maximum. But the PSE cannot reduce the power given to a PD that’s in use. Finally, a PSE may request reduced power via conservation mode, which usually happens when the system switches to a battery-powered supply in a power outage.
A number of nonstandard PoE implementations have been created by different manufacturers. Some were developed before the standards were created, but still find use today. Others were developed as a method to supply more power to the PD than permitted by the standard.
Cisco’s original PoE scheme for their WLAN access points and IP phones was developed many years before the IEEE standard. While the original Cisco PoE could only deliver 10mW, it is not upgradable to the 802.3af standard.
Another incompatible PoE make is PowerDsine, now a Microsemi brand. It’s original “Power over LAN” setup was created in 1999 and is used by a number of different companies, including Polycom, 3Com, Lucent, and Nortel.
A new entry in the nonstandard category is Linear Technology’s LTPoE++, which promises up to 90 W of power delivered to the PD. The advantage of the LTPoE++, however, is that it is backward compatible to both 802.3af and 802.3at standards. This lets their system work within a standards-compliant architecture, while still supplying more power to devices that can use the LTPoE++ architecture.
The LTPoE++ system comes in selectable power levels of 35, 45, 70, and 90 W with all power levels capable of running the two power standards. The higher power levels expand the field of Ethernet-powered devices, enabling PoE designs that were only possible before with external power available.
The use of PoE continues to grow, along with the power demands of the PDs in the system. One can only assume that the next IEEE standard will double the power levels again. Until then, proprietary techniques such as the LTPoE++ format can fill the power gap where needed.