2002-06-04 IEEE C802.16-02/05

Project

IEEE 802.16 Broadband Wireless Access Working Group<http://ieee802.org/16>

Title

IEEE Standard 802.16: A Technical Overview of the WirelessMAN™Air Interface for Broadband Wireless Access

Date Submitted

2002-06-04

Source(s) Roger Marks Voice: +1-303-497-3037NIST Fax: +1-303-497-3037325 Broadway mailto:r.b.marks@ieee.orgBoulder, CO 80305

Re: IEEE Std 802.16

Abstract The broadband wireless access industry, which provides high-rate networkconnections to stationary sites, has matured to the point at which it now hasa standard for second-generation wireless metropolitan area networks. IEEEStandard 802.16, with its WirelessMAN™ air interface, sets the stage forwidespread and effective deployments worldwide. This article overviews thetechnical medium access control and physical layer features of this newstandard.

This article was written by Carl Eklund, Roger B. Marks, Kenneth L. Stanwood, and Stanley Wang. It was published in

IEEE Communications Magazine

, June 2002, pp. 98-107.” For more details, see:

<

http://www.comsoc.org/ci1/Public/2002/Jun/index.html

>

Purpose

This document will improve the effectiveness of participants in IEEE Working Group 802.16 by providing an overview of the published base standard upon which the Working Group is currently developing several amendments.

Notice

This document has been prepared to assist IEEE 802.16. It is offered as a basis for discussion and is not binding on the contributing individual(s) or organization(s). The material in this document is subject to change in form and content after further study. The contributor(s) reserve(s) the right to add, amend or withdraw material contained herein.

Patent Policy and Procedures

The contributor is familiar with the IEEE 802.16 Patent Policy and Procedures (Version 1.0)<http://ieee802.org/16/ipr/patents/policy.html>, including the statement “IEEE standards may include the known use of patent(s), including patent applications, if there is technical justification in the opinion of the standards-developing committee and provided the IEEE receives assurance from the patent holder that it will license applicants under reasonable terms and conditions for the purpose of implementing the standard.”

Early disclosure to the Working Group of patent information that might be relevant to the standard is essential to reduce the possibility for delays in the development process and increase the likelihood that the draft publication will be approved for publication. Please notify the Chair <mailto:r.b.marks@ieee.org> as early as possible, in written or electronic form, of any patents (granted or under application) that may cover technology that is under consideration by or has been approved by IEEE 802.16. The Chair will disclose this notification via the IEEE 802.16 web site <http://ieee802.org/16/ipr/patents/notices>.

2002-06-04 IEEE C802.16-02/05

Copyright Permission

Copyright ©2002 Institute of Electrical and Electronics Engineers, Inc. Reprinted, with permission, from

IEEE Communications Magazine

, June 2002, pp. 98-107. This material is posted here with permission of the IEEE. Internal or personal use of this material is permitted. However, permission to reprint/republish this material for advertising or promotional purposes or for creating new collective works for resale or redistribution must be obtained from the IEEE (contact pubs-permissions@ieee.org).

By choosing to view this document, you agree to all provisions of the copyright laws protecting it.

IEEE Communications Magazine • June 200298

IEEE Standard 802.16:A Technical Overview of theWirelessMAN™ Air Interface forBroadband Wireless Access

0163-6804/02/$17.00 © 2002 IEEE

ABSTRACT

The broadband wireless access industry, whichprovides high-rate network connections to sta-tionary sites, has matured to the point at whichit now has a standard for second-generationwireless metropolitan area networks. IEEE Stan-dard 802.16, with its WirelessMAN™ air inter-face, sets the stage for widespread and effectivedeployments worldwide. This article overviewsthe technical medium access control and physicallayer features of this new standard.

INTRODUCTION ANDMARKET OPPORTUNITIES

IEEE Standard 802.16-2001 [1], completed inOctober 2001 and published on 8 April 2002,defines the WirelessMAN™ air interface specifi-cation for wireless metropolitan area networks(MANs). The completion of this standard her-alds the entry of broadband wireless access as amajor new tool in the effort to link homes andbusinesses to core telecommunications networksworldwide.

As currently defined through IEEE Stan-dard 802.16, a wireless MAN provides networkaccess to buildings through exterior antennascommunicating with central radio base stations(BSs). The wireless MAN offers an alternativeto cabled access networks, such as fiber opticlinks, coaxial systems using cable modems, anddigital subscriber line (DSL) links. Becausewireless systems have the capacity to addressbroad geographic areas without the costly infra-structure development required in deployingcable links to individual sites, the technologymay prove less expensive to deploy and may

lead to more ubiquitous broadband access.Such systems have been in use for several years,but the development of the new standard marksthe maturation of the industry and forms thebasis of new industry success using second-gen-eration equipment.

In this scenario, with WirelessMAN technolo-gy bringing the network to a building, users insidethe building will connect to it with conventionalin-building networks such as, for data, Ethernet(IEEE Standard 802.3) or wireless LANs (IEEEStandard 802.11). However, the fundamentaldesign of the standard may eventually allow forthe efficient extension of the WirelessMAN net-working protocols directly to the individual user.For instance, a central BS may someday exchangemedium access control (MAC) protocol data withan individual laptop computer in a home. Thelinks from the BS to the home receiver and fromthe home receiver to the laptop would likely usequite different physical layers, but design of theWirelessMAN MAC could accommodate such aconnection with full quality of service (QoS).With the technology expanding in this direction, itis likely that the standard will evolve to supportnomadic and increasingly mobile users. For exam-ple, it could be suitable for a stationary or slow-moving vehicle.

IEEE Standard 802.16 was designed toevolve as a set of air interfaces based on a com-mon MAC protocol but with physical layer spec-ifications dependent on the spectrum of use andthe associated regulations. The standard, asapproved in 2001, addresses frequencies from10 to 66 GHz, where extensive spectrum is cur-rently available worldwide but at which theshort wavelengths introduce significant deploy-ment challenges. A new project, currently in theballoting stage, expects to complete an amend-

Carl Eklund, Nokia Research Center

Roger B. Marks, National Institute of Standards and Technology

Kenneth L. Stanwood and Stanley Wang, Ensemble Communications Inc.

TOPICS IN BROADBAND ACCESS

Portions are U.S. Govern-ment work, not subject toU.S. Copyright.

IEEE Communications Magazine • June 2002 99

ment denoted IEEE 802.16a [2] before the endof 2002. This document will extend the air inter-face support to lower frequencies in the 2–11GHz band, including both licensed and license-exempt spectra. Compared to the higher fre-quencies, such spectra offer the opportunity toreach many more customers less expensively,although at generally lower data rates. This sug-gests that such services will be oriented towardindividual homes or small to medium-sizedenterprises.

THE 802.16 WORKING GROUPDevelopment of IEEE Standard 802.16 and theincluded WirelessMAN™ air interface, alongwith associated standards and amendments, isthe responsibility of IEEE Working Group802.16 on Broadband Wireless Access (BWA)Standards (http://WirelessMAN.org). The Work-ing Group’s initial interest was the 10–66 GHzrange. The 2–11 GHz amendment project thatled to IEEE 802.16a was approved in March2000. The 802.16a project primarily involves thedevelopment of new physical layer specifica-tions, with supporting enhancements to thebasic MAC. In addition, the Working Grouphas completed IEEE Standard 802.16.2 [3](“Recommended Practice for Coexistence ofFixed Broadband Wireless Access Systems”) toaddress 10–66 GHz coexistence and, throughthe amendment project 802.16.2a, is expandingits recommendations to include licensed bandsfrom 2 to 11 GHz.

Historically, the 802.16 activities were initiatedat an August 1998 meeting called by the NationalWireless Electronics Systems Testbed (N-WEST)of the U.S. National Institute of Standards andTechnology. The effort was welcomed in IEEE802, which opened a Study Group. The 802.16Working Group has held weeklong meetings atleast bimonthly since July 1999. Over 700 individ-uals have attended a session. Membership, whichis granted to individuals based on their atten-dance and participation, currently stands at 130.The work has been closely followed; for example,the IEEE 802.16 Web site received over 2.8 mil-lion file requests in 2000.

TECHNOLOGY DESIGN ISSUES

MEDIUM ACCESS CONTROLThe IEEE 802.16 MAC protocol was designedfor point-to-multipoint broadband wirelessaccess applications. It addresses the need forvery high bit rates, both uplink (to the BS)and downlink (from the BS). Access and band-width allocation algorithms must accommo-date hundreds of terminals per channel, withterminals that may be shared by multiple endusers. The services required by these end usersare varied in their nature and include legacytime-division multiplex (TDM) voice and data,Internet Protocol (IP) connectivity, and packe-tized voice over IP (VoIP). To support thisvariety of services, the 802.16 MAC mustaccommodate both continuous and bursty traf-fic. Additionally, these services expect to beassigned QoS in keeping with the traffic types.The 802.16 MAC provides a wide range of ser-vice types analogous to the classic asyn-

chronous transfer mode (ATM) service cate-gories as well as newer categories such asguaranteed frame rate (GFR).

The 802.16 MAC protocol must also supporta variety of backhaul requirements, includingboth asynchronous transfer mode (ATM) andpacket-based protocols. Convergence sublayersare used to map the transport-layer-specific traf-fic to a MAC that is flexible enough to efficient-ly carry any traffic type. Through such featuresas payload header suppression, packing, andfragmentation, the convergence sublayers andMAC work together to carry traffic in a formthat is often more efficient than the originaltransport mechanism.

Issues of transport efficiency are alsoaddressed at the interface between the MACand the physical layer (PHY). For example, themodulation and coding schemes are specified ina burst profile that may be adjusted adaptivelyfor each burst to each subscriber station. TheMAC can make use of bandwidth-efficient burstprofiles under favorable link conditions but shiftto more reliable, although less efficient, alterna-tives as required to support the planned 99.999percent link availability.

The request-grant mechanism is designed tobe scalable, efficient, and self-correcting. The802.16 access system does not lose efficiencywhen presented with multiple connections perterminal, multiple QoS levels per terminal, and alarge number of statistically multiplexed users. Ittakes advantage of a wide variety of requestmechanisms, balancing the stability of con-tentionless access with the efficiency of con-tention-oriented access.

While extensive bandwidth allocation andQoS mechanisms are provided, the details ofscheduling and reservation management are leftunstandardized and provide an importantmechanism for vendors to differentiate theirequipment.

Along with the fundamental task of allocatingbandwidth and transporting data, the MACincludes a privacy sublayer that provides authen-tication of network access and connection estab-lishment to avoid theft of service, and it provideskey exchange and encryption for data privacy.

To accommodate the more demanding physi-cal environment and different service require-ments of the frequencies between 2 and 11 GHz,the 802.16a project is upgrading the MAC toprovide automatic repeat request (ARQ) andsupport for mesh, rather than only point-to-mul-tipoint, network architectures.

THE PHYSICAL LAYER10–66 GHz — In the design of the PHY speci-fication for 10–66 GHz, line-of-sight propaga-tion was deemed a practical necessity. With thiscondition assumed, single-carrier modulationwas easily selected; the air interface is designat-ed “WirelessMAN-SC.” Many fundamentaldesign challenges remained, however. Becauseof the point-to-multipoint architecture, the BSbasically transmits a TDM signal, with individu-al subscriber stations allocated time slots serial-ly. Access in the uplink direction is bytime-division multiple access (TDMA). Follow-ing extensive discussions regarding duplexing, a

While extensive

bandwidth

allocation and

QoS mechanisms

are provided, the

details of

scheduling and

reservation

management

are left

unstandardized

and provide an

important

mechanism for

vendors to

differentiate their

equipment.

IEEE Communications Magazine • June 2002100

burst design was selected that allows both time-division duplexing (TDD), in which the uplinkand downlink share a channel but do not trans-mit simultaneously, and frequency-divisionduplexing (FDD), in which the uplink and down-link operate on separate channels, sometimessimultaneously. This burst design allows bothTDD and FDD to be handled in a similar fash-ion. Support for half-duplex FDD subscriberstations, which may be less expensive since theydo not simultaneously transmit and receive, wasadded at the expense of some slight complexity.Both TDD and FDD alternatives support adap-tive burst profiles in which modulation and cod-ing options may be dynamically assigned on aburst-by-burst basis.

2–11 GHz — The 2–11 GHz bands, bothlicensed and license-exempt, are addressed inIEEE Project 802.16a. The standard is in bal-lot but is not yet complete. The draft current-ly specifies that compliant systems implementone of three air interface specifications, eachof which provides for interoperability. Designof the 2–11 GHz physical layer is driven bythe need for non-line-of-sight (NLOS) opera-tion. Because residential applications areexpected, rooftops may be too low for a clearsight line to a BS antenna, possibly due toobstruction by trees. Therefore, significantmultipath propagation must be expected. Fur-thermore, outdoor-mounted antennas areexpensive due to both hardware and installa-tion costs.

The three 2–11 GHz air interface specifica-tions in 802.16a Draft 3 are:• WirelessMAN-SC2: This uses a single-carri-

er modulation format.• WirelessMAN-OFDM: This uses orthogonal

frequency-division multiplexing with a 256-point transform. Access is by TDMA. Thisair interface is mandatory for license-exempt bands.

• WirelessMAN-OFDMA: This uses orthogo-nal frequency-division multiple access witha 2048-point transform. In this system, mul-tiple access is provided by addressing a sub-set of the multiple carriers to individualreceivers.Because of the propagation requirements, the

use of advanced antenna systems is supported.It is premature to speculate on further

specifics of the 802.16a amendment prior to itscompletion. While the draft seems to havereached a level of maturity, the contents couldchange significantly in balloting. Modes couldeven be deleted or added.

PHYSICAL LAYER DETAILSThe PHY specification defined for 10–66 GHzuses burst single-carrier modulation with adap-tive burst profiling in which transmission param-eters, including the modulation and codingschemes, may be adjusted individually to eachsubscriber station (SS) on a frame-by-framebasis. Both TDD and burst FDD variants aredefined. Channel bandwidths of 20 or 25 MHz(typical U.S. allocation) or 28 MHz (typicalEuropean allocation) are specified, along withNyquist square-root raised-cosine pulse shapingwith a rolloff factor of 0.25. Randomization isperformed for spectral shaping and to ensure bittransitions for clock recovery.

The forward error correction (FEC) used isReed-Solomon GF(256), with variable block sizeand error correction capabilities. This is pairedwith an inner block convolutional code to robust-ly transmit critical data, such as frame controland initial accesses. The FEC options are pairedwith quadrature phase shift keying (QPSK), 16-state quadrature amplitude modulation (16-QAM), and 64-state QAM (64-QAM) to formburst profiles of varying robustness and efficien-cy. If the last FEC block is not filled, that blockmay be shortened. Shortening in both the uplink

� Figure 1. The downlink subframe structure.

Prea

mbl

e

Broadcastcontrol

DIUC = 0TDM

DIUC a

Prea

mbl

e

DL-MAP UL-MAP

TDMDIUC b

TDM portion

Prea

mbl

e

Prea

mbl

e

TDMADIUC d

TDMADIUC e

Prea

mbl

e

TDMADIUC f

Burst start points

Prea

mbl

e

TDMADIUC g

TDMDIUC c

TDMA portion

The PHY

specification

defined for

10–66 GHz uses

burst

single-carrier

modulation with

adaptive burst

profiling in which

transmission

parameters,

including the

modulation and

codling schemes,

my be adjusted

individually to

each subscriber

station on a

frame-by-frame

basis. Both TDD

and burst FDD

variants are

defined.

IEEE Communications Magazine • June 2002 101

and downlink is controlled by the BS and isimplicitly communicated in the uplink map (UL-MAP) and downlink map (DL-MAP).

The system uses a frame of 0.5, 1, or 2 ms.This frame is divided into physical slots for thepurpose of bandwidth allocation and identifica-tion of PHY transitions. A physical slot isdefined to be 4 QAM symbols. In the TDD vari-ant of the PHY, the uplink subframe follows thedownlink subframe on the same carrier frequen-cy. In the FDD variant, the uplink and downlinksubframes are coincident in time but are carriedon separate frequencies. The downlink subframeis shown in Fig. 1.

The downlink subframe starts with a framecontrol section that contains the DL-MAP forthe current downlink frame as well as the UL-MAP for a specified time in the future. Thedownlink map specifies when physical layer tran-sitions (modulation and FEC changes) occurwithin the downlink subframe. The downlinksubframe typically contains a TDM portionimmediately following the frame control section.Downlink data are transmitted to each SS usinga negotiated burst profile. The data are transmit-ted in order of decreasing robustness to allowSSs to receive their data before being presentedwith a burst profile that could cause them to losesynchronization with the downlink.

In FDD systems, the TDM portion may be fol-lowed by a TDMA segment that includes an extrapreamble at the start of each new burst profile.This feature allows better support of half-duplexSSs. In an efficiently scheduled FDD system withmany half-duplex SSs, some may need to transmitearlier in the frame than they receive. Due totheir half-duplex nature, these SSs lose synchro-nization with the downlink. The TDMA preambleallows them to regain synchronization.

Due to the dynamics of bandwidth demandfor the variety of services that may be active, themixture and duration of burst profiles and thepresence or absence of a TDMA portion varydynamically from frame to frame. Since therecipient SS is implicitly indicated in the MAC

headers rather than in the DL-MAP, SSs listento all portions of the downlink subframe they arecapable of receiving. For full-duplex SSs, thismeans receiving all burst profiles of equal orgreater robustness than they have negotiatedwith the BS.

A typical uplink subframe for the 10–66 GHzPHY is shown in Fig. 2. Unlike the downlink,the UL-MAP grants bandwidth to specific SSs.The SSs transmit in their assigned allocationusing the burst profile specified by the UplinkInterval Usage Code (UIUC) in the UL-MAPentry granting them bandwidth. The uplink sub-frame may also contain contention-based alloca-tions for initial system access and broadcast ormulticast bandwidth requests. The access oppor-tunities for initial system access are sized toallow extra guard time for SSs that have notresolved the transmit time advance necessary tooffset the round-trip delay to the BS.

Between the PHY and MAC is a transmis-sion convergence (TC) sublayer. This layer per-forms the transformation of variable lengthMAC protocol data units (PDUs) into the fixedlength FEC blocks (plus possibly a shortenedblock at the end) of each burst. The TC layerhas a PDU sized to fit in the FEC block current-ly being filled. It starts with a pointer indicatingwhere the next MAC PDU header starts withinthe FEC block. This is shown in Fig. 3.

The TC PDU format allows resynchroniza-tion to the next MAC PDU in the event that theprevious FEC block had irrecoverable errors.

� Figure 2. The uplink subframe structure.

Accessburst

Accessburst

Collision CollisionBandwidthrequest

Bandwidthrequest

SS transitiongap

Tx/Rx transitiongap (TDD)

Initialmaintenanceopportunities

(UIUC = 2)

Requestcontention

opps(UIUC = 1)

SS 1scheduled

data(UIUC = i)

SS Nscheduled

data(UIUC = j)

� Figure 3. TC PDU format.

Transmission convergence sublayer PDU

P MAC PDU which has startedin previous TC PDU

First MAC PDU,this TC PDU

Second MAC PDU,this TC PDU

IEEE Communications Magazine • June 2002102

Without the TC layer, a receiving SS or BSwould potentially lose the entire remainder of aburst when an irrecoverable bit error occurred.

MEDIUM ACCESS CONTROL DETAILSThe MAC includes service-specific convergencesublayers that interface to higher layers, abovethe core MAC common part sublayer that car-ries out the key MAC functions. Below the com-mon part sublayer is the privacy sublayer.

SERVICE-SPECIFIC CONVERGENCE SUBLAYERSIEEE Standard 802.16 defines two general ser-vice-specific convergence sublayers for map-ping services to and from 802.16 MACconnections. The ATM convergence sublayeris defined for ATM services, and the packetconvergence sublayer is defined for mappingpacket services such as IPv4, IPv6, Ethernet,and virtual local area network (VLAN). Theprimary task of the sublayer is to classify ser-vice data units (SDUs) to the proper MACconnection, preserve or enable QoS, andenable bandwidth allocation. The mappingtakes various forms depending on the type ofservice. In addition to these basic functions,the convergence sublayers can also performmore sophisticated functions such as payloadheader suppression and reconstruction toenhance airlink efficiency.

COMMON PART SUBLAYERIntroduction and General Architecture — Ingeneral, the 802.16 MAC is designed to supporta point-to-multipoint architecture with a centralBS handling multiple independent sectors simul-taneously. On the downlink, data to SSs are mul-tiplexed in TDM fashion. The uplink is sharedbetween SSs in TDMA fashion.

The 802.16 MAC is connection-oriented. Allservices, including inherently connectionless ser-vices, are mapped to a connection. This providesa mechanism for requesting bandwidth, associat-ing QoS and traffic parameters, transporting androuting data to the appropriate convergence sub-layer, and all other actions associated with thecontractual terms of the service. Connections are

referenced with 16-bit connection identifiers(CIDs) and may require continuously grantedbandwidth or bandwidth on demand. As will bedescribed, both are accommodated.

Each SS has a standard 48-bit MAC address,but this serves mainly as an equipment identifi-er, since the primary addresses used duringoperation are the CIDs. Upon entering thenetwork, the SS is assigned three managementconnections in each direction. These three con-nections reflect the three different QoSrequirements used by different managementlevels. The first of these is the basic connec-tion, which is used for the transfer of short,t ime-critical MAC and radio l ink control(RLC) messages. The primary managementconnection is used to transfer longer, moredelay-tolerant messages such as those used forauthentication and connection setup. The sec-ondary management connection is used for thetransfer of standards-based management mes-sages such as Dynamic Host ConfigurationProtocol (DHCP), Trivial File Transfer Proto-col (TFTP), and Simple Network ManagementProtocol (SNMP). In addition to these man-agement connections, SSs are allocated trans-port connections for the contracted services.Transport connections are unidirectional tofacilitate different uplink and downlink QoSand traffic parameters; they are typicallyassigned to services in pairs.

The MAC reserves additional connections forother purposes. One connection is reserved forcontention-based initial access. Another isreserved for broadcast transmissions in thedownlink as well as for signaling broadcast con-tention-based polling of SS bandwidth needs.Additional connections are reserved for multi-cast, rather than broadcast, contention-basedpolling. SSs may be instructed to join multicastpolling groups associated with these multicastpolling connections.

MAC PDU Formats — The MAC PDU is thedata unit exchanged between the MAC layers ofthe BS and its SSs. A MAC PDU consists of afixed-length MAC header, a variable-length pay-load, and an optional cyclic redundancy check(CRC). Two header formats, distinguished bythe HT field, are defined: the generic header(Fig. 4) and the bandwidth request header.

Except for bandwidth request MAC PDUs,which contain no payload, MAC PDUs containeither MAC management messages or conver-gence sublayer data.

Three types of MAC subheader may be pre-sent. The grant management subheader is usedby an SS to convey bandwidth managementneeds to its BS. The fragmentation subheadercontains information that indicates the presenceand orientation in the payload of any fragmentsof SDUs. The packing subheader is used to indi-cate the packing of multiple SDUs into a singlePDU. The grant management and fragmentationsubheaders may be inserted in MAC PDUsimmediately following the generic header if soindicated by the Type field. The packing sub-header may be inserted before each MAC SDUif so indicated by the Type field. More detailsare provided below.

� Figure 4. Format of generic header for MAC PDU.

HT

= 0

(1)

EC (

1)

Type (6)

Rsv

(1)

CI (

1) LENmsb (3)

EKS(2)

Rsv

(1)

LEN lsb (8) CID msb (8)

CID Isb (8) HCS (8)

IEEE Communications Magazine • June 2002 103

Transmission of MAC PDUs — The IEEE802.16 MAC supports various higher-layer pro-tocols such as ATM or IP. Incoming MAC SDUsfrom corresponding convergence sublayers areformatted according to the MAC PDU format,possibly with fragmentation and/or packing,before being conveyed over one or more connec-tions in accordance with the MAC protocol.After traversing the airlink, MAC PDUs arereconstructed back into the original MAC SDUsso that the format modifications performed bythe MAC layer protocol are transparent to thereceiving entity.

IEEE 802.16 takes advantage of incorporat-ing the packing and fragmentation processeswith the bandwidth allocation process to maxi-mize the flexibility, efficiency, and effectivenessof both. Fragmentation is the process in which aMAC SDU is divided into one or more MACSDU fragments. Packing is the process in whichmultiple MAC SDUs are packed into a singleMAC PDU payload. Both processes may be ini-tiated by either a BS for a downlink connectionor an SS for an uplink connection.

IEEE 802.16 allows simultaneous fragmenta-tion and packing for efficient use of the band-width.

PHY Support and Frame Structure — TheIEEE 802.16 MAC supports both TDD andFDD. In FDD, both continuous and burst down-links are supported. Continuous downlinks allowfor certain robustness enhancement techniques,such as interleaving. Burst downlinks (eitherFDD or TDD) allow the use of more advancedrobustness and capacity enhancement tech-niques, such as subscriber-level adaptive burstprofiling and advanced antenna systems.

The MAC builds the downlink subframe start-ing with a frame control section containing theDL-MAP and UL-MAP messages. These indicatePHY transitions on the downlink as well as band-width allocations and burst profiles on the uplink.

The DL-MAP is always applicable to the cur-rent frame and is always at least two FEC blockslong. The first PHY transition is expressed in thefirst FEC block, to allow adequate processingtime. In both TDD and FDD systems, the UL-MAP provides allocations starting no later thanthe next downlink frame. The UL-MAP can,however, allocate starting in the current frame aslong as processing times and round-trip delaysare observed. The minimum time betweenreceipt and applicability of the UL-MAP for anFDD system is shown in Fig. 5.

Radio Link Control — The advanced technolo-gy of the 802.16 PHY requires equally advancedradio link control (RLC), particularly the capa-bility of the PHY to transition from one burstprofile to another. The RLC must control thiscapability as well as the traditional RLC func-tions of power control and ranging.

RLC begins with periodic BS broadcast ofthe burst profiles that have been chosen for theuplink and downlink. The particular burst pro-files used on a channel are chosen based on anumber of factors, such as rain region and equip-ment capabilities. Burst profiles for the downlinkare each tagged with a Downlink Interval Usage

Code (DIUC). Those for the uplink are eachtagged with an Uplink Interval Usage Code(UIUC).

During initial access, the SS performs initialpower leveling and ranging using ranging request(RNG-REQ) messages transmitted in initialmaintenance windows. The adjustments to theSS’s transmit time advance, as well as poweradjustments, are returned to the SS in rangingresponse (RNG-RSP) messages. For ongoingranging and power adjustments, the BS maytransmit unsolicited RNG-RSP messages com-manding the SS to adjust its power or timing.

During initial ranging, the SS also requests tobe served in the downlink via a particular burstprofile by transmitting its choice of DIUC to theBS. The choice is based on received downlinksignal quality measurements performed by theSS before and during initial ranging. The BSmay confirm or reject the choice in the rangingresponse. Similarly, the BS monitors the qualityof the uplink signal it receives from the SS. TheBS commands the SS to use a particular uplinkburst profile simply by including the appropriateburst profile UIUC with the SS’s grants in UL-MAP messages.

After initial determination of uplink anddownlink burst profiles between the BS and aparticular SS, RLC continues to monitor andcontrol the burst profiles. Harsher environmen-tal conditions, such as rain fades, can force theSS to request a more robust burst profile. Alter-natively, exceptionally good weather may allowan SS to temporarily operate with a more effi-cient burst profile. The RLC continues to adaptthe SS’s current UL and DL burst profiles, everstriving to achieve a balance between robustnessand efficiency. Because the BS is in control anddirectly monitors the uplink signal quality, theprotocol for changing the uplink burst profile foran SS is simple: the BS merely specifies the pro-file’s associated UIUC whenever granting the SSbandwidth in a frame. This eliminates the needfor an acknowledgment, since the SS will alwaysreceive either both the UIUC and the grant orneither. Hence, no chance of uplink burst profilemismatch between the BS and SS exists.

In the downlink, the SS is the entity thatmonitors the quality of the receive signal andtherefore knows when its downlink burst profile

� Figure 5. Minimum FDD map relevance.

Round-trip delay + Tproc

Framecontrol

Downlinksubframe

Uplinksubframe

Frame n – 1

DL-MAP n – 1UL-MAP n

DL-MAP nUL-MAP n + 1

DL-MAP n + 1UL-MAP n + 2

Frame n Frame n + 1

IEEE Communications Magazine • June 2002104

should change. The BS, however, is the entity incontrol of the change. There are two methodsavailable to the SS to request a change in down-link burst profile, depending on whether the SSoperates in the grant per connection (GPC) orgrant per SS (GPSS) mode (see “BandwidthRequests and Grants”). The first method wouldtypically apply (based on the discretion of theBS scheduling algorithm) only to GPC SSs. Inthis case, the BS may periodically allocate a sta-tion maintenance interval to the SS. The SS canuse the RNG-REQ message to request a changein downlink burst profile. The preferred methodis for the SS to transmit a downlink burst profilechange request (DBPC-REQ). In this case,which is always an option for GPSS SSs and canbe an option for GPC SSs, the BS responds witha downlink burst profile change response(DBPC-RSP) message confirming or denyingthe change.

Because messages may be lost due to irrecov-erable bit errors, the protocols for changing anSS’s downlink burst profile must be carefullystructured. The order of the burst profile changeactions is different when transitioning to a morerobust burst profile than when transitioning to aless robust one. The standard takes advantageof the fact that an SS is always required to listento more robust portions of the downlink as wellas the profile that was negotiated. Figure 6shows a transition to a more robust burst pro-file. Figure 7 shows a transition to a less robustburst profile.

Uplink Scheduling Services — Each connec-tion in the uplink direction is mapped to ascheduling service. Each scheduling service isassociated with a set of rules imposed on the BSscheduler responsible for allocating the uplinkcapacity and the request-grant protocol betweenthe SS and the BS. The detailed specification ofthe rules and the scheduling service used for aparticular uplink connection is negotiated atconnection setup time.

The scheduling services in IEEE 802.16 arebased on those defined for cable modems in theDOCSIS standard [4].

Unsolicited grant service (UGS) is tailoredfor carrying services that generate fixed units ofdata periodically. Here the BS schedules regular-ly, in a preemptive manner, grants of the sizenegotiated at connection setup, without anexplicit request from the SS. This eliminates theoverhead and latency of bandwidth requests inorder to meet the delay and delay jitter require-ments of the underlying service. A practical limiton the delay jitter is set by the frame duration. Ifmore stringent jitter requirements are to be met,output buffering is needed. Services that typical-ly would be carried on a connection with UGSservice include ATM constant bit rate (CBR)and E1/T1 over ATM.

When used with UGS, the grant managementsubheader includes the poll-me bit (see “Band-width Requests and Grants”) as well as the slipindicator flag, which allows the SS to report thatthe transmission queue is backlogged due to fac-tors such as lost grants or clock skew betweenthe IEEE 802.16 system and the outside net-work. The BS, upon detecting the slip indicatorflag, can allocate some additional capacity to theSS, allowing it to recover the normal queuestate. Connections configured with UGS are notallowed to utilize random access opportunitiesfor requests.

The real-time polling service is designed tomeet the needs of services that are dynamic innature, but offers periodic dedicated requestopportunities to meet real-time requirements.Because the SS issues explicit requests, theprotocol overhead and latency is increased, butthis capacity is granted only according to thereal need of the connection. The real-timepolling service is well suited for connectionscarrying services such as VoIP or streamingvideo or audio.

The non-real-time polling service is almostidentical to the real-time polling service exceptthat connections may utilize random accesstransmit opportunities for sending bandwidthrequests. Typically, services carried on theseconnections tolerate longer delays and are ratherinsensitive to delay jitter. The non-real-timepolling service is suitable for Internet access witha minimum guaranteed rate and for ATM GFRconnections.

A best effort service has also been defined.Neither throughput nor delay guarantees areprovided. The SS sends requests for band-width in either random access slots or dedi-cated transmission opportunit ies . Theoccurrence of dedicated opportunities is sub-ject to network load, and the SS cannot relyon their presence.

� Figure 6. Transition to a more robust burst profile.

BS SS

Yes

No

RNG-REQ or DBPC-REQchange to DIUC k

DL data at DIUC n

DL data at DIUC k

DL data at DIUC k

RNG-RSP or DBPC-RSP

C/(N + I)too low

for DIUC n

Send DL dataat DIUC k

Continuemonitoring

DL datathroughDIUC n

Monitor DLdata

only throughDIUC k

IEEE Communications Magazine • June 2002 105

Bandwidth Requests and Grants — The IEEE802.16 MAC accommodates two classes of SS,differentiated by their ability to accept bandwidthgrants simply for a connection or for the SS as awhole. Both classes of SS request bandwidth perconnection to allow the BS uplink schedulingalgorithm to properly consider QoS when allocat-ing bandwidth. With the grant per connection(GPC) class of SS, bandwidth is granted explicitlyto a connection, and the SS uses the grant onlyfor that connection. RLC and other managementprotocols use bandwidth explicitly allocated to themanagement connections.

With the grant per SS (GPSS) class, SSs aregranted bandwidth aggregated into a single grantto the SS itself. The GPSS SS needs to be moreintelligent in its handling of QoS. It will typicallyuse the bandwidth for the connection thatrequested it, but need not. For instance, if theQoS situation at the SS has changed since the lastrequest, the SS has the option of sending thehigher QoS data along with a request to replacethis bandwidth stolen from a lower QoS connec-tion. The SS could also use some of the band-width to react more quickly to changingenvironmental conditions by sending, for instance,a DBPC-REQ message.

The two classes of SS allow a trade-offbetween simplicity and efficiency. The need toexplicitly grant extra bandwidth for RLC andrequests, coupled with the likelihood of morethan one entry per SS, makes GPC less efficientand scalable than GPSS. Additionally, the abilityof the GPSS SS to react more quickly to theneeds of the PHY and those of connectionsenhances system performance. GPSS is the onlyclass of SS allowed with the 10–66 GHz PHY.

With both classes of grants, the IEEE 802.16MAC uses a self-correcting protocol rather thanan acknowledged protocol. This method usesless bandwidth. Furthermore, acknowledged pro-tocols can take additional time, potentiallyadding delay. There are a number of reasons thebandwidth requested by an SS for a connectionmay not be available:• The BS did not see the request due to

irrecoverable PHY errors or collision of acontention-based reservation.

• The SS did not see the grant due to irrecov-erable PHY errors.

• The BS did not have sufficient bandwidthavailable.

• The GPSS SS used the bandwidth for anoth-er purpose.In the self-correcting protocol, all of these

anomalies are treated the same. After a timeoutappropriate for the QoS of the connection (orimmediately, if the bandwidth was stolen by theSS for another purpose), the SS simply requestsagain. For efficiency, most bandwidth requestsare incremental; that is, the SS asks for morebandwidth for a connection. However, for theself-correcting bandwidth request/grant mecha-nism to work correctly, the bandwidth requestsmust occasionally be aggregate; that is, the SSinforms the BS of its total current bandwidthneeds for a connection. This allows the BS toreset its perception of the SS’s needs without acomplicated protocol acknowledging the use ofgranted bandwidth.

The SS has a plethora of ways to requestbandwidth, combining the determinism of uni-cast polling with the responsiveness of con-tention-based requests and the efficiency ofunsolicited bandwidth. For continuous band-width demand, such as with CBR T1/E1 data,the SS need not request bandwidth; the BSgrants it unsolicited.

To short-circuit the normal polling cycle, anySS with a connection running UGS can use thepoll-me bit in the grant management subheaderto let the BS know it needs to be polled forbandwidth needs on another connection. The BSmay choose to save bandwidth by polling SSsthat have unsolicited grant services only whenthey have set the poll-me bit.

A more conventional way to request band-width is to send a bandwidth request MACPDU that consists of simply the bandwidthrequest header and no payload. GPSS SSs cansend this in any bandwidth allocation theyreceive. GPC terminals can send it in either arequest interval or a data grant interval allocat-ed to their basic connection. A closely relatedmethod of requesting data is to use a grantmanagement subheader to piggyback a requestfor additional bandwidth for the same connec-tion within a MAC PDU.

In addition to polling individual SSs, the BSmay issue a broadcast poll by allocating a requestinterval to the broadcast CID. Similarly, thestandard provides a protocol for forming multi-cast groups to give finer control to contention-based polling. Due to the nondeterministic delaythat can be caused by collisions and retries, con-

� Figure 7. Transition to a less robust burst profile.

BS SS

Yes

No

RNG-REQ or DBPC-REQchange to DIUC m

DL data at DIUC n

DL data at DIUC m

RNG-RSP or DBPC-RSP

C/(N+I) highenough for

DIUC m

Send DL dataat DIUC m

Start monitoringDL data through

DIUC m

IEEE Communications Magazine • June 2002106

tention-based requests are allowed only for cer-tain lower QoS classes of service.

Channel Acquisition — The MAC protocolincludes an initialization procedure designed toeliminate the need for manual configuration.Upon installation, an SS begins scanning its fre-quency list to find an operating channel. It maybe programmed to register with a specified BS,referring to a programmable BS ID broadcast byeach. This feature is useful in dense deploymentswhere the SS might hear a secondary BS due toselective fading or when the SS picks up a side-lobe of a nearby BS antenna.

After deciding on which channel or chan-nel pair to attempt communication, the SStries to synchronize to the downlink transmis-sion by detecting the periodic frame pream-bles. Once the physical layer is synchronized,the SS will look for the periodically broadcastDCD and UCD messages that enable the SSto learn the modulation and FEC schemesused on the carrier.

Initial Ranging and Negotiation of SS Capa-bilities — Upon learning what parameters touse for its initial ranging transmissions, the SSwill look for initial ranging opportunities byscanning the UL-MAP messages present in everyframe. The SS uses a truncated exponentialbackoff algorithm to determine which initialranging slot it will use to send a ranging requestmessage. The SS will send the burst using theminimum power setting and will try again withincreasingly higher transmission power if it doesnot receive a ranging response.

Based on the arrival time of the initial rang-ing request and the measured power of the sig-nal, the BS commands a timing advance and apower adjustment to the SS in the rangingresponse. The response also provides the SS withthe basic and primary management CIDs. Oncethe timing advance of the SS transmissions hasbeen correctly determined, the ranging proce-dure for fine-tuning the power can be performedusing invited transmissions.

All transmissions up to this point are madeusing the most robust, and thus least efficient,burst profile. To avoid wasting capacity, the SSnext reports its PHY capabilities, including themodulation and coding schemes it supports andwhether, in an FDD system, it is half-duplex orfull-duplex. The BS, in its response, can deny theuse of any capability reported by the SS.

SS Authentication and Registration —Each SS contains both a manufacturer-issuedfactory-installed X.509 digital certificate andthe certificate of the manufacturer. These cer-tificates, which establish a link between the 48-bit MAC address of the SS and its public RSAkey, are sent to the BS by the SS in the Autho-rization Request and Authentication Informa-tion messages. The network is able to verify theidentity of the SS by checking the certificatesand can subsequently check the level of autho-rization of the SS. If the SS is authorized tojoin the network, the BS will respond to itsrequest with an Authorization Reply containingan Authorization Key (AK) encrypted with the

SS’s public key and used to secure further trans-actions.

Upon successful authorization, the SS willregister with the network. This will establish thesecondary management connection of the SSand determine capabilities related to connectionsetup and MAC operation. The version of IPused on the secondary management connectionis also determined during registration.

IP Connectivity — After registration, the SSattains an IP address via DHCP and establishesthe time of day via the Internet Time Protocol.The DHCP server also provides the address ofthe TFTP server from which the SS can requesta configuration file. This file provides a standardinterface for providing vendor-specific configura-tion information.

Connection Setup — IEEE 802.16 uses theconcept of service flows to define unidirectionaltransport of packets on either downlink or uplink.Service flows are characterized by a set of QoSparameters such as latency and jitter. To mostefficiently utilize network resources such as band-width and memory, 802.16 adopts a two-phaseactivation model in which resources assigned to aparticular admitted service flow may not be actu-ally committed until the service flow is activated.Each admitted or active service flow is mapped toa MAC connection with a unique CID.

In general, service flows in IEEE 802.16 arepreprovisioned, and setup of the service flows isinitiated by the BS during SS initialization. How-ever, service flows can also be dynamically estab-lished by either the BS or the SS. The SS typicallyinitiates service flows only if there is a dynamical-ly signaled connection, such as a switched virtualconnection (SVC) from an ATM network. Theestablishment of service flows is performed via athree-way handshaking protocol in which therequest for service flow establishment is respond-ed to and the response acknowledged.

In addition to dynamic service establishment,IEEE 802.16 also supports dynamic servicechanges in which service flow parameters are re-negotiated. Like dynamic service flow establish-ment, service flow changes also follow a similarthree-way handshaking protocol.

Privacy Sublayer — IEEE 802.16’s privacy pro-tocol is based on the Privacy Key Management(PKM) protocol of the DOCSIS BPI+ specifica-tion [5] but has been enhanced to fit seamlesslyinto the IEEE 802.16 MAC protocol and to bet-ter accommodate stronger cryptographic meth-ods, such as the recently approved AdvancedEncryption Standard.

Security Associations — PKM is built aroundthe concept of security associations (SAs). TheSA is a set of cryptographic methods and theassociated keying material; that is, it contains theinformation about which algorithms to apply,which key to use, and so on. Every SS establishesat least one SA during initialization. Each con-nection, with the exception of the basic and pri-mary management connections, is mapped to anSA either at connection setup time or dynami-cally during operation.

In general, service

flows in IEEE

802.16 are

preprovisioned,

and setup of the

service flows is

initiated by the

BS during SS

initialization.

However, service

flows can also be

dynamically

established by

either the BS or

the SS.

IEEE Communications Magazine • June 2002 107

Cryptographic Methods — Currently, thePKM protocol uses X.509 digital certificates withRSA public key encryption for SS authenticationand authorization key exchange. For trafficencryption,the Data Encryption Standard (DES)running in the cipher block chaining (CBC) modewith 56-bit keys is currently mandated. The CBCinitialization vector is dependent on the framecounter and differs from frame to frame. Toreduce the number of computationally intensivepublic key operations during normal operation,the transmission encryption keys are exchangedusing 3DES with a key exchange key derived fromthe authorization key.

The PKM protocol messages themselves areauthenticated using the Hashed MessageAuthentication Code (HMAC) protocol [6] withSHA-1 [7]. In addition, message authenticationin vital MAC functions, such as the connectionsetup, is provided by the PKM protocol.

SUMMARY AND CONCLUSIONThe WirelessMAN™ air interface specified inIEEE Standard 802.16 provides a platform forthe development and deployment of standards-based metropolitan area networks providingbroadband wireless access in many regulatoryenvironments. The standard is intended toallow for multiple vendors to produce interop-erable equipment. However, it also allows forextensive vendor differentiation. For instance,the standard provides the base station with aset of tools to implement efficient scheduling.However, the scheduling algorithms that deter-mine the overall efficiency will differ from ven-dor to vendor and may be optimized forspecific traffic patterns. Likewise, the adaptiveburst profile feature allows great control tooptimize the efficiency of the PHY transport.Innovative vendors will introduce cleverschemes to maximize this opportunity whilemaintaining interoperability with compliantsubscriber stations.

The publication of IEEE Standard 802.16 is adefining moment in which broadband wirelessaccess moves to its second generation and beginsits establishment as a mainstream alternative forbroadband access. Through the dedicated serviceof many volunteers, the IEEE 802.16 WorkingGroup succeeded in quickly designing and forg-ing a standard based on forward-looking tech-nology. IEEE Standard 802.16 is the foundationof the wireless metropolitan area networks ofthe next few decades.

ACKNOWLEDGMENTSAs lead PHY editors of IEEE Standard 802.16-2001, Jay Klein and Lars Lindh played keyroles in the completion of the 10–66 GHzphysical layer discussed here. Mr. Klein alsochaired the PHY Task Group that led itsdevelopment. Mr. Lindh and Ken Peirce eachprovided a helpful technical review of thismanuscript.

REFERENCES[1] IEEE 802.16-2001, “IEEE Standard for Local and Metropoli-

tan Area Networks — Part 16: Air Interface for FixedBroadband Wireless Access Systems,” Apr. 8, 2002.

[2] IEEE P802.16a/D3-2001: “Draft Amendment to IEEE Stan-dard for Local and Metropolitan Area Networks — Part 16:Air Interface for Fixed Wireless Access Systems — MediumAccess Control Modifications and Additional Physical LayersSpecifications for 2–11 GHz,” Mar. 25, 2002.

[3] IEEE 802.16.2-2001, “IEEE Recommended Practice for Localand Metropolitan Area Networks — Coexistence of FixedBroadband Wireless Access Systems,” Sept. 10, 2001.

[4] SCTE DSS 00-05, Data-Over-Cable Service InterfaceSpecification (DOCSIS) SP-RFIv1.1-I05-000714, “RadioFrequency Interface 1.1 Specification,” July 2000.

[5] SCTE DSS 00-09, DOCSIS SP-BPI+-I06-001215, “BaselinePrivacy Plus Interface Specification,” Dec. 2000.

[6] H. Krawczyk, M. Bellare, and R. Canetti, “HMAC: Keyed-Hashing for Message Authentication,” IETF RFC 2104,Feb. 1997.

[7] Federal Information Processing Standards Publication180–1, “Secure Hash Standard,” Apr. 1995.

BIOGRAPHIESCARL EKLUND (carl.eklund@nokia.com) is a senior researchengineer with Nokia Research Center, Helsinki, Finland. Hechaired the MAC Task Group that developed the IEEE802.16 medium access control protocol and served as alead MAC editor of IEEE Standard 802.16-2001. He receivedhis M.Sc. in engineering physics from Helsinki University ofTechnology in 1996. He is currently a guest researcher atthe National Institute of Standards and Technology (NIST),Boulder, Colorado.

ROGER B. MARKS [F] (marks@nist.gov) is with NIST, Boulder,Colorado. In 1998 he initiated the effort that led to theIEEE 802.16 Working Group on Broadband Wireless Access,chairing it since inception. He served as technical editor ofIEEE Standards 802.16-2001 and 802.16.2-2001. Hereceived his A.B. in physics in 1980 from Princeton Univer-sity and his Ph.D. in applied physics in 1988 from Yale Uni-versity. Author of over 80 publications, his awards includethe 1995 IEEE Morris E. Leeds Award (an IEEE TechnicalField Award) and the Broadband Wireless Hall of Fame. Hedeveloped the IEEE Radio and Wireless Conference andchaired it from 1996 through 1999.

KENNETH L. STANWOOD (ken@ensemble.com) is currentlyprincipal member of technical staff and manager of sys-tems engineering at Ensemble Communications, San Diego,California, where he was the primary designer of the MACand transmission convergence layers of Ensemble’s propri-etary Adaptix™ broadband wireless access system. He grad-uated with a B.S. degree in mathematical sciences fromOregon State University in 1983 and an M.S. in computerscience from Stanford University in 1986. He has beenheavily involved in the IEEE 802.16 10–66 GHz project(serving as a lead MAC editor) as well as its Europeancounterpart, ETSI BRAN HIPERACCESS. He is technical work-ing group chair for the Worldwide Interoperability forMicrowave Access (WiMAX) Forum, which is dedicated toproducing test specifications and system option profiles toensure interoperability of systems built to IEEE Standard802.16. He chairs 802.16’s Task Group c, which is creatingprofiles for 10–66 GHz 802.16 systems.

STANLEY WANG (stanley@reddotwireless.com) is currently adirector of RedDot Wireless Inc., San Jose, California,where he is leading the development and implementationof MAC protocols. Before joining RedDot Wireless, he waswith Ensemble Communications Inc., where he worked onIEEE 802.16 standards, serving as a lead MAC editor ofIEEE Standard 802.16-2001. He received his Ph.D. in com-puter engineering in 1994 from the University of SouthernCalifornia. He was with the faculty of the Computer Sci-ence Department at California State University-San Marcosfrom 1994 through 2000, where he received the Harry E.Brakebill Outstanding Professor Awards in 1996 and histenure in 1999.

Through the

dedicated service

of many

volunteers, the

IEEE 802.16

Working Group

succeeded in

quickly designing

and forging a

standard based

on forward-look-

ing technology.

IEEE Standard

802.16 is the

foundation of

the wireless

metropolitan

area networks of

the next few

decades.