[For a copy of this article in PDF format, which displays figures and equations,
Requires Adobe Acrobat Reader,
The recently ratified IEEE 802.3af Power over Ethernet (PoE) standard provides a uniform method of powering Ethernet devices over CAT-5 Ethernet cables formerly used only for data.
PoE eliminates the need for local 120-Vac power when installing a WLAN access point, IP phone or other application. This is helpful when installing devices that aren't near normal sources of power, such as above a false ceiling, on the outside of a building or rooftop, or high on a wall.
CAT-5 Ethernet wiring can be engineered and installed at lower cost than traditional ac power infrastructure because it's low voltage and energy limited. PoE allows an IT department to more efficiently manage multiple appliances from a central point. An IT manager can remotely turn off misbehaving devices, reset “locked-up” devices, or detect when a device's power supply fails. Multiple devices with critical functions can be protected by a common UPS to ensure operation during a power outage.
What are PSE and PD?
In the IEEE 802.3af specification, power sourcing equipment (PSE) inserts power onto a cable (link). The PSE may be located at a natural cable head, such as a router or switch, or it may be a separate device located midspan between the switch and the powered device (PD). The PD is the natural termination of this link and could be an Internet protocol phone, a wireless access point or a point-of-sale terminal. This opens up the possibility of creating applications that don't even use the Ethernet cable for communication, such as a wireless security camera. A PD can draw up to 12.95 W from the Ethernet cable.
IEEE 802.3af puts all of the intelligence into the PSE, which possesses considerable processing power and an administrative interface. The PD functionality can be implemented as a simple, inexpensive hardware solution.
Fig. 1 shows a simplified block diagram of a PD electrical interface. Power can enter the PD with either polarity on pairs 1-2 and 3-6 (numbers refer to the RJ-45 pins) or on pairs 4-5 and 7-8. In Fig. 1, the PD is a 10/100baseT solution where the “spare” pairs don't carry data. The data interface is the usual and customary isolation transformer, often packaged with common-mode chokes, but with an added center tap on the transformer side that faces the RJ-45 jack. The data flow through the transformers to Ethernet PHY. The 48-Vdc power is extracted from the center taps and passes through a PoE interface block that can be viewed as a specialized hot-swap. Following the interface is a dc-dc converter block that may have isolation. The dc-dc block converts the 36-V to 57-V input into the voltages required by the application circuits.
Phantom powering is the method for mixing data and power on the two-pair interface. The method applies voltage at the PSE between RX and TX pairs, with common-mode current flowing in each pair. Fig. 2 demonstrates how the transformer operates when used in a phantom power application. DC flux cancellation is achieved in each transformer when the (equal) common mode current in each wire of a pair flows through an oppositely phased end of the winding. The current into or out of the center tap is then twice the current in each wire of the pair. The differential data pulses are unaffected by the superposition of the dc current, and flow is unhindered through the transformer. One method of inserting the power at the PSE is to drive a set of transformer center taps with a voltage source. The transformers must:
- Have resistive and turns balance
- Be able to carry the required current
- Be able to accept the inevitable DC imbalances
- Still have all the characteristics necessary for proper operation of the data interface.
Fortunately, some commercially available devices meet all these requirements.
There are two unused pairs in 10/100baseT systems. Power can be directly inserted in these pairs (particularly by a midspan PSE) without the need for isolation transformers. Power from an external ac-powered supply can be brought in and ORed into either the PoE interface input or output, or into the dc-dc converter.
The dc-dc converter may require isolation between the incoming power and the application electronics. The standard can be interpreted to mean that 1500 Vac of isolation is required between the power front end and all other metallic interfaces. An external (dc) power source that connects to the PoE front end must be isolated from frame ground and meet a 1500-Vac test.
The powering process consists of a sequence of events: detection, classification (optional) and powering. This set of events is shown in Fig. 3.
Tables 1 and 2 summarize some key parameters of IEEE 802.3af specifications at the PD's Ethernet interface. For complete details, refer to the IEEE specification. The timing relationships of all the parameters have been ignored for brevity. Fig. 4 is a pictorial representation of Table 1.
Detection: Normally, the PSE will periodically start the detection sequence to determine when a new device is plugged in. In detection, the PSE probes the PD's port with two voltages that fall between 2.7 V to 10.1 V (ΔV > 1V). The PSE checks that the incremental resistance (ΔV/ΔI) is within a guard-banded range of 23.75 kΩ to 26.25 kΩ. Detection prevents powering devices that can't accept PoE in order to avoid damaging them or driving one PSE output from another.
Classification: Classification is used by the PSE to keep track of the PD power requirements; this information allows power allocation and rationing. The PSE applies a voltage between 15.5 V and 20.5 V (14.5 V to 20.5 V at the PD) and measures the PD current draw. Table 2 shows the details on the required PD current for each class and how the PSE will interpret it. Classification is optional for both the PSE and the PD. A PD should read as Class 0 if not specifically enabled for this function.
Classification will be critical to the future of PoE; if every output of a 96-port PSE were designed to deliver the full 15.4 W (at the cable head end) for each of its outputs, 1478 W would be delivered. The cost of supplying all of this power infrastructure is significant. It multiplies into cost of the supply, cost of the power and ac wiring, cooling, rack space, packaging, and possibly UPS size. Classification allows design of a statistically sized power source or expandable source, and it allows the means to prevent it from being overloaded.
|Detection V-I slope||2.7 V to 10.1 V||23.75||26.25||kV|
|Input capacitance during detection||2.7 V to 10.1 V||0.05||0.12||µF|
|Classification voltage range||See class currents||14.5||20.5||V|
|Must start voltage||42||V|
|Must turn off voltage||30||V|
|Maximum average input current||37 V |
|Inrush current||Bulk cap. charge, 50 mS||400||mA|
|Input capacitance during operation||5||µF|
|Minimum dc operating current||10||mA|
Startup and Normal Operation: Once past the detection and classification stages, the PSE can apply operational power to the PD. The PD should start up at voltages between 30 Vdc and 42 Vdc. A large input current, referred to as inrush current, will be drawn as the usual bulk input capacitor is charged. The PD must limit the current to control the voltage drop, while preventing the dc-dc converter from starting during this period. The controller needs to guarantee that it has: 1) enough hysteresis to cover the source voltage droop due to the inrush and operating current (assuming the 20 Ω maximum feed impedance); or 2) can charge the bulk capacitor from the 400-mA current-limited PSE within 50 ms. The normal operational range of 36 V to 57 V allows for a wide-range PSE (44 V to 57 V) source plus the ohmic cable-voltage loss of 8 V during an inrush of 400 mA. Some latitude exists in the choice of startup voltage based on operating and inrush current. Generally, this latitude lets legacy devices ensure operation with compliant PSEs. The standard provides that if a PSE chooses to remove power from the PD, it must restart the powering process with detection.
Maintain Power Signature: The PD must indicate to the PSE that there is a valid load present during normal operation. This is referred to as the maintain power signature (MPS). The MPS is accomplished by maintaining the minimum dc current (dropouts are allowed) and the ac impedance below the value implied by the 5-µF minimum bulk capacitor specification. The PD must provide the proper signature for at least 75 ms with dropouts of less than 250 ms to keep the PSE from removing power. This behavior allows detection of PD removal to protect against voltage on exposed cables (safety) and plugging a non-compliant device into a powered cable. The PSE output is current-limited and has a timed turnoff, providing fault protection against a shorted cable or PD.
The PD designer can choose between discrete and IC-based solutions when implementing the PoE interface. New integrated circuits, such as the TPS2370, provide “turnkey” solutions that implement all of the PoE functions. The integrated solutions require only a few simple choices, and the design is done. Discrete approaches were used in various pre-specification implementations, but they seem outdated and component-intensive (especially with classification). The new ICs also provide a means to coordinate the inrush function with the downstream dc-dc converter. This is important because it provides a smart way to keep the converter from starting during the inrush, possibly preventing a startup. The technique of using a long time delay on the dc-dc startup always carries certain risks.
|Class||Class Current (mA)||Power Range (W)||Notes|
|0||0 to 4||0.44 to 12.95||Default|
|1||9 to 12||0.44 to 3.84||Optional|
|2||17 to 20||3.84 to 6.49||Optional|
|3||26 to 30||6.49 to 12.95||Optional|
|4||36 to 44||Not Allowed||Treat as Class 0|
There is always the possibility that a PD design won't operate with all compliant PSEs despite the existence of a well-defined specification. The Power over Ethernet Consortium, formed in conjunction with the University of New Hampshire Interoperability Lab, provides an industrywide resource (www.iol.unh.edu/consortiums/poe/index.html). Some manufacturers of PoE interface circuits may provide documentation of interoperability of their reference solution, thus saving the cost of testing to their customers who choose not to participate directly.
For more information on this article, CIRCLE 341 on Reader Service Card