A primer on Power over Ethernet

The release of the IEEE standard for Power over Ethernet (PoE) in June of 2003 has been a boon to Internet Protocol (IP) devices. By integrating almost 13 Watts of power and the data needed for application specific devices, the barrier to entry was lowered so that the economically infeasible became feasible. Infrastructure installation for devices such as IP cameras, wireless access points or IP telephones has become as simple as terminating wires in a punch block.

PoE hardware is segregated into two categories: equipment that provides power to an Ethernet link, and devices that utilize that power. Power sourcing equipment (PSE) is the name given to the equipment that powers a link. Powered device (PD) is an equally unimaginative name provided to the devices that use the power put on the Ethernet link. PSEs and PDs, while symbiotically linked, are best evaluated separately.

When first considered, PoE seems a trivial technology. Standard telephones have provided power and data on the same wiring for over a hundred years. However, with Ethernet, there was a huge installed base in place before anyone thought to add power. If the existing Ethernet wires were to become energized, every deployed piece of equipment would have been damaged or destroyed. To prevent this pitfall, the task force that authored the standard developed a procedure to detect devices that were capable of receiving power.

The detection process is performed by a PSE prior to applying power to a link. Using at least two low-voltage and current-limited levels, the PSE inspects the link looking for the PD’s unique signature which is a 25K Ohm resistive slope. These signals are required to be between 2.8V to10V, and must not be capable of sourcing more than 5mA of current. While only two test probes are required by the standard, invalid signatures can appear as valid, under certain conditions. The best practices of today use three and sometimes four separate probe signals, which are allowed by the standard.

If the PSE successfully detects a PD, then it may power the link. If the PSE does power the link, it must do so within 400ms, or it must start the detection process all over. Once a link is powered, the PSE must monitor the link to ensure that there is no over current event and that the PD still requires power. Over current almost certainly indicates a fault in either the cabling or the PD, However, a PD may no longer require power because a user attached a local power supply to the PD or disconnected the PD from the link. PoE can not tell the difference between a disconnected or a “non-requesting” PD. The end result, however, is the same. Power is removed from the port by the PSE.

In addition to detecting a powering a PD, a PSE may also attempt to interrogate a PD about the amount of power that it will require through a process called classification. While not required by the specification, most PDs provide useful classification information approximating their power into either quarter-, half- or full-power devices. PSEs use this information to allocate their power supplies efficiently.

On the PD side of the cable, things are much the same. When the PD detects and input voltage in the range 2.7V to 10.1V, it must connect a load which is equivalent to 25K Ohms. If the PD input is in the range of 14.5V to 20.5V, the PD can connect a constant current load – which is indicative of the amount of power the PD will consume. Note that reporting this classification via the current load is optional. However, every integrated PD IC on the market today supports this feature, and it is useful information to report. Once the input port voltage is in the range of 36V to 42V, the PD power supply turns on. The PD turns off when the input drops into the 36V to 30V range. Unlike the PSE, there is no requirement in the PD to follow succession of states. The PD is a purely voltage driven function.

Real world applications and the engineers that design them must be concerned about such matters as surge protection, noise immunity and radiated spectrum. While the IEEE specification does include some requirements for isolation, these are not to be confused with the safety requirements of the intended market of the end equipment. These requirements are largely beyond the scope of the IEEE PoE specification and are addressed by other regulations and requirements. One notable exception is that the IEEE specification does require that the PSE meet the requirements of IEC 60950 for a Limited Power Source.

With currently available integrated circuits, hardware designers are freed from concerning themselves with all the details of the IEEE specifications explained here. This article should provide enough information to allow designers to feel comfortable with the PoE technology they need to implement their new applications.

About the Author

Michael McCormack is the business development manager for PoE products at Texas Instruments. In addition to his product duties, Michael was editor for the original IEEE Std 802.3af drafts, and currently chairs the P802.3at Task Force which is revising the IEEE PoE standard. To read more about PoE, visit www.ti.com/poe.