Avaya Power Over Ethernet Practical Manual

A Practical Guide to Power Over Ethernet (PoE) by Avaya

Release 1.1 Avaya Labs

ABSTRACT

This paper will discuss the highlights in the Power over Ethernet standard (IEEE 802.3af) and the earlier pre-standard PoE implementations. Common implementation issues are also addressed to help you avoid unpleasant surprises.

External posting: www.avaya.com.

Application Note

October 2006 COMPAS ID 122875

TRADEMARK NOTICE

Avaya and the Avaya Logo are trademarks of Avaya Inc. and may be registered in certain jurisdictions. All trademarks identified by ® or ™ are registered trademarks and trademarks respectively of Avaya Inc. All other trademarks are the property of their respective owners.

NOTICE While reasonable efforts were made to ensure the

information in this document was complete and accurate at the time of printing, Avaya can assume no responsibility for any errors. Changes and corrections to the information contained in this document may be incorporated into future releases.

Comments or questions may be emailed to: afunguy@Avaya.com

A Practical Guide to Power over Ethernet (PoE)

Document Summary

This document contains basic, but practical information concerning Power over Ethernet. Powering devices like Access Points (APs), IP Telephones, Web Cameras, etc. through the Ethernet cable is a natural evolution in the networking industry. The Central Office in the PSTN historically powered analog phones for many decades. The PBX has also powered digital phones through the wires. Appliances, like Access Points and Web-based Cameras are positioned in hard to reach locations, like ceilings, and are therefore prime candidates for power through the Ethernet cable that connects them. There are benefits in powering IP telephones through the Ethernet cable. One is convenience and another is the ability to monitor devices.

The first part of this paper describes the basics of the IEEE 802.3af standard that was ratified in June 2003. The second part of this paper describes practical differences in Power over Ethernet (PoE) implementations of Avaya and Cisco equipment compared to the standard and to each other. This paper is not intended to point out shortcomings of any other vendor, but it is beneficial to reveal subtle issues in vendor specific implementations of PoE that can cause issues during and after installation. Since many customers use a Cisco data infrastructure, Cisco specific examples are cited and compared to Avaya’s products. A PoE checklist is included in Appendix A as a quick check to aid your implementations.

Table of Contents

A Practical Guide to Power over Ethernet (PoE) Document Summary .........................................3

A Practical Guide to Power over Ethernet (PoE) Paper .................................................................4

1 Introduction.............................................................................................................................4

2 Definitions...............................................................................................................................4

3 IEEE 802.3af – how it works ...............................................................................................5

4 Static vs. Dynamic Power Allocation..................................................................................7

5 PoE Implementation Details ................................................................................................9

6 Avaya Configuration ...........................................................................................................12

7 Cisco Configuration.............................................................................................................14

8 Future Directions .................................................................................................................22

A Practical Guide to Power over Ethernet (PoE)

1 Introduction

Voice over Internet Protocol (VoIP) is the convergence of traditional voice onto an IP data network to provide better application integration by using common protocols. This lowers costs by using one network infrastructure and even melding separate support staffs into one. Other real-time traffic streams, such as uncompressed video and streaming audio are also converging onto data networks.

Convergence began with basic VoIP and, over time, has expanded traditional features and functions including powering IP endpoints. The idea to send power through cables to an end device is not new. Home analog telephones have been powered from the Central Office (CO) for scores of years. PBX type switches have also provided power to analog and digital phones in business offices for decades. The evolution to send power over a Cat-5 type cable is then a natural extension of an existing idea applied to a new platform – the data network.

It is interesting that the IEEE eventually based their standard on the Avaya (Lucent Technologies at that time) PoE scheme. This means that all Avaya products have been standards compliant even before the standard was ratified. This also means that proprietary implementations of PoE, like Cisco, had to convert from their vendor specific methods to the standards-based protocol. Older Cisco chassis-based switch blades required daughter cards or total blade replacement to provide compliant power to the newer

standard. Adhering to the standard, titled IEEE 802.3af, is important because it allows power

interoperability from any vendor’s products. You can have confidence in Avaya’s products knowing that Avaya led the IEEE standards committee with the ultimate solution for Powering devices over the Ethernet cable. No one implements Power over Ethernet (PoE) better than Avaya. dynamically apply the right amount of power to ANY vendor’s products that are standards compliant. Please see section 5 of this paper for details.

Avaya can intelligently and

2 Definitions

Before technologies are discussed, some important definitions must be learned.

End-Span. An End-Span, or Endpoint PSE is an Ethernet switch that is capable of sending Power over

Ethernet up to 100 meters over copper twisted pair to an endpoint device like an IP telephone.

Mid-Span. A device that lies between the Ethernet switch (or hub) and injects Power through the Ethernet

(copper) cable to the endpoint device. A Mid Span is not an access device like a switch or a hub. Its only purpose is to inject power from the middle of the link between an access device (switch) and the endpoint – hence the name “Mid” span.

Ohm. A measure of resistance using the Omega symbol - Ω PD – Powered Device. A powered device is an endpoint that requires power. Many devices can

accommodate power in more than one way such as a local transformer or Power over Ethernet. Examples of devices include IP telephones, Access Points, Web Cameras, magnetic card readers, etc.

PoE – Power over Ethernet is any scheme, either proprietary or standards-based, that defines how to send

power through an Ethernet cable.

PSE – Power Sourcing Equipment. A PSE is a device that sends Power over Ethernet to the PD. End-

Span and Mid-Span provide PoE, but there can also be other devices that send power to PDs.

Watt. A measure of power derived by multiplying current (Amperes) with resistance (Ohms). One Watt

equals one Ampere across one Ohm’s worth of resistance.

3 IEEE 802.3af – how it works

The phases between connecting a PD, power flow to the PD and power termination are:

PD (Powered Device) Detection

PD Classification or actual power determination

Power management, termination and PD re-discovery

PD Detection: The goal of this phase is to detect a valid PD. When a PD is connected to a PSE (Power

Sourcing Equipment), a small DC current is sent from the PSE. A PSE is typically a switch (End-Span) or a power injector (Mid Span). If the resistive load of a PD is sensed at 25K Ω (+/- 1.25K Ω), the device may meet the 802.3af requirements for power. Not only does the resistive load need to meet the 25K Ω level, but the PD must also meet the resistance range during a short voltage ramp-up period. The idea is to only power devices that are valid PDs and to protect non-PDs that could be damaged if they received power.

In a short amount of time, voltage is increased from 2.7 Volts to 10.1 Volts in 1.0 Volt increments. The resistance must remain between the range of 23.75K Ω and 26.25K Ω or power is removed almost immediately. resistive range and that same range during a voltage ramp-up period. Any device that does not present this resistive signature, will not receive power from the PSE. Additionally, any “signature” in the zones

23.75 – 26.25 K Ω presents an invalid or non-compliant signature and power will not be applied. See appendix B for specifics on invalid and non-compliant signatures. Capacitance and inductance values are

also measured when validating a PD to the IEEE 802.3af standard.

This is generally known as the 802.3af resistive signature – adhering to both an initial

beyond

Table 1 – Parameters for a valid PD power signature

Valid PD Detection Ramp-up Conditions Resistive Signature Range

Initial Resistive load for a potential PD 23.75K Ω – 26.25K Ω

Required resistance during voltage ramp-up 2.7V – 10.1V 23.75K Ω – 26.25K Ω

Input capacitance 2.7V – 10.1V 0.05µF – 0.12µF

Input Inductance 2.7V – 10.1V ≤ 100µH

Note: Power in the 802.3af standard is expressed in watts assuming 48 volts to the PD. By design, a PD must accept a voltage variance. The standard mandates that “any PD will withstand 0 to 57 volts indefinitely without damage”.

The IEEE 802.3af standard allows power on either the spare pairs or the signaling pairs. Ethernet uses signaling wire pairs 1,2 & 3,6. Power over these used pairs is sometimes called phantom power or inline power. Most End-Span PSEs (switches) use the inline power scheme, but they can also use the spare pairs 4,5 & 7,8 as an option.

Mid-Span PSEs use only the spare pairs 4,5 & 7,8 to transmit power. All PDs are required to accept power on either spare or signaling pairs. All Avaya products comply with this requirement and can be powered three ways; from the signal pair, the spare pair or one pair (7,8) as a further option to support external power supplies like a “brick” type transformer.

PD Classification: After detection, the goal of this next phase is to supply appropriate power. This can be

done based on the IEEE 802.3af power class of the PD, or more intelligent algorithms can sense how much power is actually needed.

The IEEE 802.3af standard defines, but does not mandate, the use of power classes. This means a PD may advertise one, and only one power class. This is important for non-Avaya PSEs that logically allocate power from the total power pool and is discussed later in this paper. The following chart describes power (in watts) required by PDs and provided by PSEs.

Table 2 – IEEE 802.3af power definitions for each class of PSE and PD devices

Class Usage Minimum Power from PSE Power Range at PD

0 Default 15.4w 0.44w – 12.95w

1 Optional 4.0w 0.44w – 3.84w

2 Optional 7.0w 3.84w – 6.49w

3 Optional 15.4w 6.49w – 12.95w

4 Reserved Treat as Class 0 Future Use

The preceding chart not only describes the power ranges for endpoints, but also the power minimums for PSEs. The difference between the maximum PD power and the maximum PSE delivered power is needed to “push” the power across a potential 100 meters (328 feet) of copper twisted pair between the PSE and the PD. There is a natural attenuation of power from the PSE to the PD as the power encounters the impedance of the copper wire. Therefore, more power is needed from the PSE so the full PoE class maximum wattage is delivered to the PD.

IEEE 802.3af PD power class is derived from a classification signature measured in milli-Amps. The chart below describes the values that a PD presents to a current under a constant voltage. The values returned to the PSE then classify that PD into an IEEE power class.

Table 3 – Values used to assign PDs into IEEE 802.3af classes

Classification Signature Conditions Values

Current for Class 0 14.5V – 20.5V 0 – 4 mA

Current for Class 1 14.5V – 20.5V 9 – 12 mA

Current for Class 2 14.5V – 20.5V 17 – 20 mA

Current for Class 3 14.5V – 20.5V 26 – 30 mA

Current for Class 4 14.5V – 20.5V 36 – 44 mA

PD Management, Termination and Rediscovery: This final stage is steady state power management

including continual power parameter sensing and detecting PD absence or violation of required parameters. Steady state powering involves a continual sensing of valid parameters. An example is a sudden short circuit of the wire pairs carrying power. This condition is detected and power is almost instantly halted by the PSE. If a PD is drawing power and is removed from the PSE port, power is again halted because another device could use that same PSE port and be damaged by erroneously receiving power. Lastly, if resistance, capacitance or inductance lies in the values listed in Appendix B, power is removed because the signature is now invalid or non-compliant. This is the end of the 802.3af material. The remaining sections are vendor specifics on implementing IEEE 802.3af PoE.

4 Static vs. Dynamic Power Allocation

Static Power Allocation

This method is used by many vendors because it is cheaper to implement. If a device advertises itself as a class-1 endpoint but needs only 1-Watt of power, the PSE will allocate 4-Watts of power to the port serving that device. This is done by logically allocating 4-Watts from the available power pool. Similarly, a class-3 device needing 8-Watts will be allocated 15.4-Watts of power. The problems with allocating the top of the class range power are:

More power is reserved than is needed. A 1-Watt device will be allocated 4-Watts because the top of the range for class-1 is 3.84-Watts plus a little more to travel over a possible 100 meters of Cat-5 cable.

Reserved power can deplete the total power pool even though it is not used. The resulting penalty from logically reserving more power than the PD requires is that available power is all physical switch ports are used. For example, if a customer is using IP phones and those phones are PoE class-3, each physical port will be ready to send out 15.4-watts. If each phone only requires 7-watts, but 15.4-watts are reserved, 8.4-watts per port will logically consume part of the total power pool. Ten of these physical ports used will strand 84 watts of power from the power pool. 24 ports will strand 201.6 Watts. 96 ports will strand 806.4 Watts. This is a worst case example.

The assumption that each port connected to a Class-3 device should be ready to provide 15.4-Watts at any time doesn’t make sense with IP telephones or any known standards compliant device. Very few if any devices operationally vary in power needs more than one Watt. An on-hook phone taking 8 watts will almost never require more than 9-Watts in the off-hook state. Very few if any PDs require more than 11 watts, so the assumption of having to apply 15.4 watts rarely happens.

“logically” exhausted before

Static power allocation results in a brute-force application that is wasteful and unintelligent. It is easy to calculate and deliver power based on the top of the power class, but it results in wasting or “stranding” power logically from the total power pool. The practical results are:

9 Incurring low port density. A 48-port switch or switch blade may only provide power to 32 ports

because power was logically exhausted.

9 Buy larger power supplies. Systems with more than one power supply or chassis based switches

usually have options to buy higher wattage supplies. A larger power supply can mitigate or even overcome the logical reservation limit, but at a cost of needlessly buying a large power supply you don’t really need. Larger power supplies also cost more to run.

9 Buy more switches. If larger power supplies don’t solve the problem, buying more switches may be

an alternative but can be very expensive. Many fixed switches have an internal power supply that cannot be upgraded, so buying more may be a solution to low port density.

9 Manually configure a power ceiling for each port. Today, many vendors have implemented a

feature that allows the administrator to manually set a power limit for each port. While this stops the wasting of stranded logical power, it is a manual process requiring the switch administrator to know the power needs of every device. Furthermore, the administrator must manually apply the power limit to each port and be ready to change that limit as devices move or are changed with other PDs.

Note: As you will see, Cisco and others use Static power allocation, but Cisco has created commands that can put a power limit on any number of ports to solve issues of low port density, buying larger power supplies or buying more switches.

Manual configuration and maintenance is still a cost using this method.

Dynamic Power Allocation

This method is more expensive to implement in a switch or Mid-Span, but is less expensive to operate and allows full port density. Dynamic power allocation senses how much power is required and delivers that exact amount of power to the port leading to the PD. Dynamic, means that as the power requirements change, the power delivered to that port also changes accordingly. Since dynamic power allocation doesn’t depend on the advertised PD class, the algorithm isn’t confined to delivering 15.4-Watts for every port. Therefore, the size of the power supply can be smaller than a static power allocation design.

Unlike the preceding section on static power allocation, if a device advertises itself as a class-1 endpoint but needs only 1-Watt of power, the PSE will allocate 1-Watt of power to the port serving that device. There is no logical reservation from the available power pool. Similarly, a class-3 device needing 8-Watts will be allocated 8-Watts of power. There are no known problems with dynamically allocating the exact amount of power needed. In fact the advantages over the static allocation scheme are:

No more power is reserved than is needed. A 1-Watt device will be allocated 1-Watt.

The total power pool is not affected by reservations. There is no wasted power because there is no concept of reserving power from a pool. There is no logical pool to consider. Each device is given the amount of power it requires.

The assumption that each port connected to a Class-3 device should be ready to provide 15.4-Watts at any time still doesn’t make sense with IP telephones or any other standards compliant device. Again, very few if any devices operationally vary in power needs more than one Watt. A phone taking 8 watts will almost never require more than 9-Watts. If a PD required 12.95 watts, 15.4-Watts would be supplied by the

802.3af specification. If, theoretically, all ports required 15.4-Watts, you could run out of available power due to a smaller power supply. This is not a known event in industry to date because very few if any Class3 PDs require a full 12.95-Watts.

Dynamic power allocation results in an intelligent application that is applied to any vendor’s PD. The practical results of this method are:

9 Full port density. A 48-port switch or Mid-Span will provide power to all 48 ports.

9 No need to buy larger power supplies. Operational costs are lower.

9 No need to buy more switches to compensate for a logical reservation of power. The equipment

footprint and operational/maintenance requirements are kept to a minimum.

9 There is no manual configuration needed by a network administrator. Errors are avoided and time

is saved because there is no manual effort required to administer power for PDs.

Changes can be

administered automatically or with a minimum of administrator effort.

+ 17 hidden pages

Avaya Power Over Ethernet Practical Manual

Specifications and Main Features

Frequently Asked Questions

User Manual