Power Over Ethernet – LAN Wiring Does Double Duty

Adopting PoE will end your AC wall-plug dependency.

Why not distribute power for Ethernet devices via LAN wiring? What a great idea. In fact, Cisco and Lucent are among the companies that have been offering Ethernet equipment with a PoE capability for several years. However, these proprietary PoE systems are not necessarily compatible with each other. The IEEE 802.3af PoE standard that ensures interoperability was only ratified in 2003.¹

Not having to connect Ethernet devices to a separate AC supply is very convenient. But, there are many more compelling reasons for adopting PoE than mere convenience.

Perhaps the strongest argument is that PoE inherently groups together network-device power requirements, allowing backup by a central UPS. As more enterprise phone systems convert to Ethernet-based VoIP, UPS-supported PoE promises the reliability of traditional POTS.

Cost is yet another big part of the equation. In the longer term, replacing POTS with IP phones eliminates the need to maintain a company's entire analog phone system. Only one technology must be supported, and IP phones don't require major wiring-closet changes when departments are reorganized.

Industrial systems also will benefit from PoE. Here, Ethernet is displacing traditional RS-485 and RS-232 networks with the concept of smart device servers. An embedded PC located at the remote sensor provides a bridge between the proprietary field-device serial data and the IP network. Standard IT tools then can directly communicate with the field device and facilitate access to the Internet if required.²

PoE Is Not Wireless
Two physical-layer implementations of PoE are supported by 802.3af, both of which apply to PSE within the network endpoint. In Figure 1a, DC power is provided as a common-mode voltage on the center taps of isolating transformers at each end of a twisted-pair circuit. Figure 1b connects DC to unused pairs in the LAN cable. The standard specifies that the two wires in a pair should be connected together so that one provides the drive current and the other functions as the return path.

Figure 1a. Endpoint PSE, Alternative A Source: IEEE 802.3af-2003

Figure 1b. Endpoint PSE, Alternative B Source: IEEE 802.3af-2003

Figure 1c. Midspan PSE, Alternative B Source: IEEE 802.3af-2003

PoE also may be accomplished by a midspan PSE, which only can provide power via unused pairs (Figure 1c). This means that use of midspan PSEs only is possible in 10BASE-T and 100BASE-TX networks.

Cat 5 copper cable contains four color-coded unshielded twisted pairs. 10BASE-T and 100BASE-TX Ethernet use one pair for each direction of a full duplex system. That leaves two unused pairs, usually reserved for network expansion. If no growth is planned for your network, those two pairs can be used to power network devices. The wiring will have to be changed later should the pairs be needed for another communications channel or if you upgrade to 1000BASE-T GE, which uses all four pairs.

Endpoint PSEs must source power via either the A or B alternative method but not both. PDs must accept power from either alternative, regardless of the voltage polarity. However, a PD cannot accept power from both the A and B methods simultaneously.

Unwanted Power
PoE is working now in many networks and will become the norm over time. In the interim, some legacy devices simply cannot use power provided via the LAN cable, and others may be damaged by a relatively high common-mode voltage. To avoid these problems, 802.3af specifies an analog detection and monitoring scheme to ensure that power is applied only when appropriate.

To start the process, the PSE applies a voltage from 2.8 V to 10.1 V for a period of not more than 500 ms to verify that the resistance presented by the PD lies in the acceptable 23.75-kΩ to 26.25-kΩ region (Figure 2). If so, the PD is requesting power but has not yet received it from the PSE.

Figure 2. Detection, Classification, Turn On, and Total Cycle Timing Relationships Source: IEEE 802.3af-2003A value outside of the guard bands on either side of the acceptable range, <12 kΩ on the low side and >45 kΩ on the high side, means that the PD will not accept power. Applying two different voltages during the detection period facilitates measurement of the PD resistance as the ratio of ΔV/ΔI and permits the PD capacitance to be determined.

Following the 500-ms detection period is a short classification time. PDs optionally may be classified according to the amount of power they are intended to draw. The PSE output is raised to between 15.5 V and 20.5 V for 75 ms and the result-ing current measured. PD classes 1, 2, and 3 correspond to power no lower than 4 W, 7 W, and 15.4 W, respectively. Class 0 indicates the default maximum level of 15.4 W, and class 4 is reserved for future use.

The purpose of classification is to support power monitoring. For example, should a PD draw more power than allowed by its classification, the associated PSE may remove power from it. If the PSE does not provide the classification function, all PDs it is powering are assumed to be class 0. After detection, the PSE must apply power to the PD within 400 ms.

After power has been applied to the PD, the PSE continues to monitor the AC, DC, or both the AC and DC characteristics of the PD impedance. The DC MPS is defined as a minimum current and a duty cycle. The PSE will continue to provide power if the DC current is greater than 10 mA for at least 60 ms out of every 460-ms period. This type of specification allows the PD to conserve power while at the same time maintaining its continuous supply.

To determine the presence of the AC MPS, a 5-Hz probing signal, at an amplitude greater than 2.5 V, is temporarily added to the PD input voltage and the complex impedance measured. According to the specification, [The] impedance shall have [a] non-negative resistive component and a net capacitive reactive component.• The absolute value of the impedance must be less than 27 kΩ for power to be maintained and greater than 1,980 kΩ for power to be removed.

The length of the disconnect detection period, T_MPDO, during which the lack of an MPS condition may be determined, is 400 ms maximum, 300 ms minimum. This is the period of time for which the AC probing signal is applied. After T_MPDO, the PSE will continue to supply power to the PD or remove it within 500 ms.

Separate sections of the specification present operating requirements for PDs and PSEs. In some cases, the PD performance may seem more tightly controlled than that of the PSE. For example, compared to the 60 ms out of 460-ms duty cycle that the PSE must allow as a valid DC MPS, a PD must draw current equal [to] or above the minimum input current for a duration of 75 ms followed by an optional MPS dropout for no longer than 250 ms.

This case is an example of biasing the specifications so values that could be inconclusive are eliminated. If the PD conforms to its specification, there's no way that the PSE can misinterpret the condition as a lack of valid MPS. The signature-generation spec is more stringent than the detection spec.

In other cases, there are grey areas where IEEE 802.3af states that the PSE may perform in either of two ways. For example, a PD that presents a detection resistance of 20 kΩ does not satisfy the 23.75-kΩ to 26.25-kΩ requirement, but neither does it fall outside the lower guard-band edge of 12 kΩ. The PSE may or may not recognize this value as acceptable although the PD clearly does not meet its specification.

There are many more sections to the standard than have been considered here. Because a PSE performs the functions of a power supply, its output specifications must relate to transient behavior, noise, ripple, stability, and fault tolerance. In addition, the PSE needs to provide isolation between PDs that it may be powering as well as between the powered circuits and ground. The PD specifications reflect similar needs for stability and isolation at the driven end of the circuit.

Test Implications of 802.3af
Clause 33 of IEEE 802.3af deals with PoE. The document states: Clause 33 utilizes the existing MDIs of 10BASE-T, 100BASE-TX, and 1000BASE-T without modification and adds no significant requirements to the cabling.•

Annex 33C is titled Recommended Test Configurations and Procedures. The operation of the PoE capability defined in Clause 33 can be tested in its entirety by following the suggested methods in Annex 33C. Test circuits and representative waveforms are presented as well as all the relevant signal levels, timings, and tolerances.

Building a test set that implements only the procedures of Annex 33C is not sufficient today. For several years, proprietary versions of PoE systems have been sold, and a test capability needs to address the protocols associated with installed equipment as well as the new 802.3af protocols. Because there are many other wiring-plant parameters that must be measured while installing, commissioning, and troubleshooting, PoE tests most likely will be added to the feature set of existing types of test instruments.

As an example of an alternative PD detection scheme, Cisco Catalyst Switches and the company's WS-PWR-PANEL PSE use separate detection schemes, both of which differ from the 802.3af specification. The switch starts the detection process by sending a special FLP on the PD's receive pair. PDs that are compatible with Cisco's FLP system have an internal relay that connects the PD's receive pair to its transmit pair in the absence of power. If the switch receives the looped-back FLP signal, it assumes the PD can accept power and applies it. Applying power causes the relay to open, and normal operation follows.

The WS-PWR-PANEL provides power independent of any network endpoint functions. It corresponds to the 802.3af provision for midspan PSEs that have only PD detection and power capabilities and are not part of a network endpoint such as a switch. Cisco refers to the company's proprietary PoE equipment as having in-line power capability. Based on this terminology, the WS-PWR-PANEL is referred to as an IPPP.

To begin the PD detection process, the IPPP sends a 347-kHz loopback tone out of a port. If a PD capable of accepting in-line power is attached, it will loopback the tone. The IPPP must detect 16 transitions within a 50-ms period to confirm the PD's presence. Power then can be applied.³

Thad Ward, a product manager at Acterna, commented, A PoE test tool must provide both physical and operational testing. On the physical level, it should:

Support all defined power detection, classification, and MPS modes. Emulate CC or CP operation.

Be compatible with common 10/100/1000-Mb/s Ethernet speeds.

Provide selectable A or B wiring alternatives.

Functionally, the test instrument must verify that the provision of DC power does not adversely affect transmission/reception of real-world traffic.

Ixia's PoE Product Manager Dean Lee agreed. Typically, a PoE-capable Ethernet switch is designed to deliver power to a large number of ports. Verifying the 802.3af compliance of individual ports is essential for the PSE manufacturer. More importantly, testing to determine the effectiveness of power management over a large number of ports is critical for the PSE designer.

In addition,• he continued, it is crucial to test the data-plane performance when the PoE is enabled. The combination of PoE and data-plane testing will ensure an error-free data plane when the power is delivered via the same path.

Many instrumentation manufacturers plan to introduce PoE test capabilities, but to date only Ixia has done so. The company's solution is based on the new PLM1000T4-PD Load Module that emulates up to four PDs when investigating PD detection, class negotiation, AC MPS, and power dissipation. Up to 16 modules may be operated in a single chassis and multiple chassis daisy-chained for high port-density applications.

This is a benchtop or rack-mounted solution. Other Ixia traffic-generation and analysis modules may be installed in the same chassis with the load modules to form a complete PoE test set. In addition to validating PSE operation by emulating PDs, the load modules feature multiple programmable PD detection, classification, and AC MPSs. These provisions support testing proprietary PoE implementations.

There is a need for portable field test equipment, but to date none is available. However, while preparing this article, we received many positive indications from test-instrument manufacturers that suggest this situation will change soon. Companies presently providing Cat 5/6, Ethernet, and IP network testers are eager to enter the PoE test market. It's just a matter of time.

References
1. IEEE 802.3af, Data Terminal Equipment (DTE) Power via Media Dependent Interface (MDI) free from http://standards.ieee.org/getieee802/index.html
2. Jackson, M., Impact of Power-over-Ethernet on Industrial-Based Networking,• July 2004, www.poweroverethernet.com/poe/content/view/full/1523/
3. Understanding the Cisco IP Phone 10/100 Ethernet In-Line Power Detection Algorithm,• Cisco Document ID: 15263, www.cisco.com/en/US/products/hw/phones/ps379/products_tech_note
09186a00801189b5.shtml