WP 1, D2 – Functional System Parameters description

BROADWAY, IST-2001-32686 WP1-D2 version 1, date 10-10-02 Programme: IST2.5 RF constraints ..............................................................................................................................402.5.1. Phase noise...........................................................................................................................402.5.2. Switching time between 5 and 60GHz ....................................................................................413. First assumptions on the whole system................................................................................................413.1 Dual mode operations ...................................................................................................................413.2 HIPERSPOT and BROADWAY classes........................................................................................414. HIPERSPOT parameters and requirements ..........................................................................................414.1 HIPERSPOT functional requirements ............................................................................................414.2 HIPERSPOT parameters...............................................................................................................425. Open questions in system design .........................................................................................................435.1 Multiple access issues...................................................................................................................435.2 Ad hoc network: to which extent?..................................................................................................435.2.1. Basic Scenario Considerations ...............................................................................................435.2.1.1. Neighbor Discovery Frequency .......................................................................................435.2.2. Traffic Management..............................................................................................................445.2.2.1. Special Cases .................................................................................................................445.2.3. One-hop Scenarios ................................................................................................................445.2.3.1. One MT inside the 60 GHz range of the AP......................................................................445.2.3.2. One MT outside the 60 GHz range of the AP....................................................................445.2.3.3. Two MTs that belong to the same 60 GHz range...............................................................445.2.4. Multi-hop Scenarios ..............................................................................................................455.2.5. Internetworking Issues...........................................................................................................455.3 Baseband enhancements to be investigated.....................................................................................456. References.........................................................................................................................................463

BROADWAY, IST-2001-32686 WP1-D2 version 1, date 10-10-02 Programme: ISTRevision HistoryEvery time something is changed or added to this document, please note the modification in the followingtable.Date Revision Description ContributorsFrom01/04/02 to30/09/0230/09/2002 Rev.1Rev. 0 Many iterations between partners All partnersVersion ready to be delivered to the EuropeanCommissionAll partners4

BROADWAY, IST-2001-32686 WP1-D2 version 1, date 10-10-02 Programme: IST1.1 Scenario 1: vendors hot spot coverage1.1.1. DescriptionAn important scenario for BROADWAY application is the coverage of physically well-defined areas in anindoor or outdoor very dense urban environment. This is a requirement from vendors that will attract peopleto come close to their advertising areas by offering certain radio-based services for free. The applicationswill mainly be free content, typically advertising. Sessions will be quite short.1.1.2. Functional requirementsRequirement /ConstraintValue / ReferenceCommentsEnvironment Indoor, limited outdoor Vendor areas, shopping mallsCoverage radius 3-10m Limited by the wallsMax density ofusers100 users for 100m 2Aggregate datarate500Mbit/sUser data rate 10 to 100 Mbit/sMobilityReduced user mobility buthigh environment mobilityTypicalapplicationsPassive user, advertisement,free contentExpected levelof serviceLowSecurityLowRoutingMostly centralized butmultihop allows larger rangeCostAlmost free for the user,medium for the vendorMisc.Broadcast and downlinkfavored1.2 Scenario 2: public Internet access1.2.1. DescriptionIn this scenario, the main objective is to provide Internet anywhere anytime , that is for example in publiclocations, such as airports, railway stations, concert halls, coffee shops. This scenario differs from theprevious one mainly in the expected QoS, which is expected to be higher here, since typical applications willbe web browsing, file transfers or remote office applications.1.2.2. Functional requirementsRequirement /ConstraintValue / ReferenceCommentsEnvironment Indoor / outdoor Ex.: Railway stations, airports, cyber cafe, concert hall,on board trains6

ABCDEF RSTUVWXYZ GHIJ KLMNOPQab cd efgh ijk lmno pqBROADWAY, IST-2001-32686 WP1-D2 version 1, date 10-10-02 Programme: ISTon board trainsCoverage radiusMax density ofusersAggregate datarateUser data rateMobilityTypicalapplicationsExpected servicelevelSecurityRoutingMisc.20m100 users for 100m 2500Mbit/s10 to 100 Mbit/sReduced user mobility buthigh environment mobilityWeb browsing, remoteoffice applications, filetransferModerate (private use)High (professional use)Medium (private use)High (professional use)Mostly centralized butmultihop allows larger rangeDownlink favoredConcert hall: 20m, train: 15m (half the length of thecarriage)Railway station: huge and directive!1.3 Scenario 3: high density residential dwellings and flatsdeployment1.3.1. DescriptionExtending HIPERLAN/2 in the 60GHz range where radio signals hardly penetrate concrete/brick walls willinherently grant no interference with neighboring cells (e.g. other room/office, other apartment). In areas,where there is good (e.g. line-of-sight) condition between communicating radio terminals , the devices willmigrate to the 60GHz radio frequency band, which will offload the HIPERLAN/2 radio resources from datatraffic and interference significantly. Neighboring homes will become really private on a physical level.Wireless Home Network(IP/Wireless 1394)60GHz60GHzDVDVTRCable TVPSTNReceiver /Network-Interface-Unit5GHz60GHzPrinter5GHz60GHzSatelliteAudio ( CD )IP/IEEE1394Figure 1.3-1: Illustration of scenario 3 in a home environment includingmigration path from existing WLAN system (HIPERLAN/2) to BROADWAY1.3.2. Functional requirementsRequirement / Value / Reference Comments7

BROADWAY, IST-2001-32686 WP1-D2 version 1, date 10-10-02 Programme: ISTConstraintEnvironmentIndoorDensity ofdevices10 devices for 100m 2 Not really users but transmitting or receiving devicesAggregate datarateLow cost: 50Mbit/sHigh cost: 200Mbit/sUser data rateHighly dependant on theapplicationGiven by DV stream, wireless IEEE1394MobilityReducedTypicalapplicationsWeb browsing, multimediacontents, gamesExpected servicelevelModerateSecurity High security needed Partially provided by the use of the 60GHz frequencyRoutingNo multihopMostly peer to peerCost Low Compared to other scenariosMisc. Low directivity access point Directivity easy to operate1.4 Scenario 4: Corporate environment1.4.1. DescriptionAs already existing WLANs have a lot of professional applications, the HL/2 offload mechanism at 60GHzhas a great interest in the corporate environment. This scenario will require a high quality of service. In thisspecific environment, high cost infrastructure equipment will be accepted. Therefore, interesting scenariossuch as directive antennas with reflectors, or Single Frequency Networks, could be investigated.1.4.2. Functional requirementsRequirement /ConstraintEnvironmentDensity ofdevicesAggregate datarateUser data rateMobilityTypicalapplicationsExpected levelof serviceSecurityRoutingCostMisc.Value / ReferenceIndoor50 devices for 100m 2500Mbit/sHighly dependent on theapplication3m/sOffice applications, databaseaccess, file transferHighHigh security neededMainly centralized,multihop usefulHighDirective Access Point,Eventually reflectorsComments8

BROADWAY, IST-2001-32686 WP1-D2 version 1, date 10-10-02 Programme: IST1.5 Scenario 5: Campus environment1.5.1. DescriptionThe 5/60GHz dual mode operation, particularly in combination with ad hoc networking functionality, is wellsuited in a campus environment, both indoor and outdoor. The availability of WLAN connectivity in suchenvironments is taking off rapidly (University of Twente, University of Bremen, Telenor office campus),both for business/office and college purposes.An important characteristic of the campus scenario is collaborative working and related forms of peer-topeerinteractivity, which justifies ad hoc networking functionality. In the case of a university campusstudents will appreciate wireless access to information sources (knowledge databases, lecture series, etc) andthe ability to have peer-to-peer content exchange in its own right or to form ad hoc project groups. Especiallystudents are expected to use this type of connectivity for informal/entertainment type of applications (funcontent). Like in the public outdoor case, environmental parameters will be challenging.1.5.2. Functional requirementsRequirement /ConstraintEnvironmentCoverage radiusDensity ofdevicesAggregate datarateUser data rateMobilityTypicalapplicationsExpected servicelevelSecurityRouteValue / ReferenceOutdoor/Indoor50 devices for 100m 2500Mbit/sHighly dependant on theapplication3m/sWeb browsing, educationalapplications, file transfer,database access, gamingModerateNot an importantrequirementCentralized: AP-MTPeer-to-peer9CommentsIndoor: lecture rooms; cantina facility; libraryOutdoor: in between office type buildings and open areas(green fields or paved squares)Depends on campus dimensions. 5 GHz umbrellacoverage seems feasible.Typical applications: downloading lecture material;remotely monitoring college sessions (live or delayed);fun applicationsVery low mobility of users. Surrounding population maybe moving.Security is generally not an issue in case of educationalcontent. Privacy requirements might be applicable thoughPeer-to-peer communication in (small) groups(collaborative work, gaming)

BROADWAY, IST-2001-32686 WP1-D2 version 1, date 10-10-02 Programme: IST1.6 Comparison between scenariosThe following table is a synthetic description of the five scenarios presented above, which allows to point outtheir common characteristics and their respective particularities.FeaturesScenariosPublic(I)Public(II )Office Home Campus(extra)Indoor/outdoorOutdoor/IndoorIndoor/OutdoorIndoor Indoor Indoor&outdoorStructure Streets, open Airports, Railway office rooms, Residential sized Indoor: class/collegeareas,stations, Coffee open-plan offices roomsOutdoor: open areasshopping malls shopsPopulation Density High Moderate/highModerate Low Moderate/HighUser behaviorWalking/ At rest At rest At rest At rest/on the moveOn the movePopulationModerate/ Low/Low Low ModerateDynamicsHighModerateSocialLimited Limited Moderate High HighInteractionApplications(broadband)Passive userAdvertisements,Web browsingStreaming contentOfficeapplicationsWeb browsingMultimedia contentWeb browsingEdu applicationsFree content Remote office Database access (incl streaming) File transferapplicationsFile transferFile transfer GamesDatabase accessGamingSecurity/Privacy level Low Low-High(depends on use)ExpectedLowModerate (privateService leveluse)High (professionaluse)High High Low-ModerateHigh (professionaluse)Moderate(private use)ModerateSession time Short

BROADWAY, IST-2001-32686 WP1-D2 version 1, date 10-10-02 Programme: ISTFrequency GHz54 55 56 57 58 59 60 61 62 63 64 65 6654.25Japan Licensed BandUnlicensed BandFor low power FWA For Wireless LAN (BRAN)EuropeTX power-Ant. Gain100mW -20dBi(licensed)10mW -47dBi(unlicensed)(57dBm-eirp)15dBW-eirp(45dBm-eirp)USUnlicensed Band500mW -BW(MHz) / 100MHzDensity (ave.):9µW/cm 2 @3mDensity (peak):18µW/cm 2 @3m(40dBm-eirp)Figure 2.1-1: Worldwide frequency panorama around 60GHzA further advantage is that, assuming the RF front end will use an Intermediate Frequency (IF) around55GHz, ghost frequencies will be located in forbidden or restricted access band, and therefore will not bepolluted.2.2 Propagation conditions: theoretical aspectsThe aim of this section is to investigate characteristics describing basic wave propagation phenomena such as• surface roughness,• Fresnel zone extensions,• Fresnel’s reflection and reflection coefficients,• penetration loss through building materials,• attenuation due to reflection,• atmospheric effects,• free-space and two-path propagation,• knife-edge diffraction,• and shadowingWhere appropriate, comparisons between 60 GHz and 5 GHz are given.2.2.1. Surface RoughnessThe Rayleigh and Fraunhofer roughness criteria provide useful tools as to whether a surface should beconsidered smooth or rough as far as reflection and transmission is concerned. Fig. 2.2-1 depicts a twodimensional vertical cut through an irregular surface onto which a plane electromagnetic wave is incident.With reference to Fig. 2.2-1, the path difference ∆ s between the points 1-1’ of ray 1 and the points 2-2’ ofray 2 is∆ s = 2 ∆zcos θwith ∆ z the height difference between the point A and B of reflection and θithe zenith angle of incidenceof the rays. For a monochromatic plane wave the phase difference ∆ ϕ between the reflected or scattered ray1 and 2 is∆ ϕ = k ∆s= k z cos θ ,02 0∆( )i( )i11

BROADWAY, IST-2001-32686 WP1-D2 version 1, date 10-10-02 Programme: ISTwhere k0= 2πλ0denotes the wave number and λ0the wavelength of the wave.1∆z2θiθiθiθiAB2'1'Due to the fact that the height irregularity ( x y)Fig. 2.2-1: Geometry for roughness criterion.z , of the surface is a stochastic function of the horizontaldimensions x and y only the square root of the mean square of the phase differences ∆ ϕ can bedetermined.The operator ( )( )∆ ϕ2 = 2 k 0σ cosθ i.2⋅ denotes the average of the enclosed quantity and σ = ∆z is the surface roughness. Forsmall surface roughness σ the phase deviation between reflected or scattered ray becomes small. Hence thecomplex field strengths in the far field of the two rays are nearly in phase. The resultant far field nearlycoincides with the one from a perfectly smooth plane surface.2.2.1.1. Rayleigh CriterionAccording to Rayleigh, the average phase errorHence2∆ ϕ must remain smaller the 222 0( θ )πR i∆ϕ = k σ cos

BROADWAY, IST-2001-32686 WP1-D2 version 1, date 10-10-02 Programme: ISTFig. 2.2-2: Surface roughness versus angle of incidence according to the Rayleigh criterion.2.2.1.2. Fraunhofer CriterionFraunhofer defines a surface to be smooth if the average phase erroror equivalentlyπ .2∆ ϕ remains smaller the 82= 2k0σ cos( θ )πF i∆ϕ

BROADWAY, IST-2001-32686 WP1-D2 version 1, date 10-10-02 Programme: ISTFig. 2.2-3: Surface roughness versus angle of incidence according to the Fraunhofer criterion.The Fraunhofer criterion is more stringent than Rayleigh’s in qualifying surfaces as smooth. The angle ofincidence must be quite gazing to allow surfaces still to appear smooth for large values of the surfaceroughness σ .F2.2.2. Fresnel ZonesFresnel zones are areas enclosed by ellipsoids whose focal points coincide with the location of thetransmitting and receiving antenna. These ellipsoids are the geometrical location of all points whosedistances to the two focal points sum up to a value that exceeds the transmitter-receiver antenna separationthby an integer multiple of half a wavelength. Hence, the n Fresnel zone encloses all radio paths from thetransmitter (Tx) to the receiver (Rx) for which the additional path length does not exceed n times half awavelength.The concept of Fresnel zones applies to both, line-of-sight (LOS) path and trajectories with refle ctions.• For a propagation path to be LOS, it is necessary that the link is designed to maintain adequate clearancebetween the optical LOS path and the terrain being crossed. It is not sufficient to plan for an opticalground clearance because up to 6 dB diffraction loss will be experienced if the optical clearance is small.• For waves suffering a reflection from a plane and smooth surface of an obstacle, that surface must belarge enough to prevent fringing effects to become dominant.2.2.2.1. Fresnel Clearance for LOS PathsA radio path requires a certain amount of clearance around the central ray if the signal expected from freespacepropagation is to be received. The required clearance is quoted in terms of Fresnel zones.14

BROADWAY, IST-2001-32686 WP1-D2 version 1, date 10-10-02 Programme: IST2 nd Fresnel zone 3 rd Fresnel zone1 st Fresnel zoneTxdd 11d 12RxFig. 2.2-4: Fresnel zones around the direct ray between transmitter and receiver.The shapes of the first four Fresnel zones are depicted in Fig. 2.2-4. For a given Fresnel clearance, noobstruction should exist in the three-dimensional volume produced by the ellipsoid around the central ray.Fig. 2.2-5 and Fig. 2.2-4 show the shapes of the 1 st - and 2 nd -Fresnel ellipses for a horizontal transmitter toreceiver separation of 5 m and 10 m.a) b)Fig. 2.2-5: 1 st Fresnel ellipses for transmitter-receiver separations of a) 5 m, and b) 10 m.a) b)Fig. 2.2-6: 2 nd Fresnel ellipses for transmitter-receiver separations of a) 5 m, and b) 10 m.The length of the minor axis M of the Fresnel ellipses are given by15

BROADWAY, IST-2001-32686 WP1-D2 version 1, date 10-10-02 Programme: IST⎛ n ⋅ ⎞M = ⎜ ⎟ + n ⋅ λ ⋅ d ,⎝ 2 ⎠thwhere n denotes the n -Fresnel ellipse, λ the wavelength, and d (major axis) the distance between thetransmitting and receiving antenna located at the focal points of the Fresnel ellipses.2.2.2.2. Fresnel Clearance for Reflectionλ 2thThe principle of Fresnel clearance applies also to reflection. The n Fresnel zone on a reflecting surfaceencloses all points on the interface for which the additional path length to the receiver is less than n timeshalf a wavelength. The Fresnel zones on the surface of the obstacles are ellipses (see Fig. 2.2-7). As a rule ofthumb, at least the first of these Fresnel zones must be fully contained in the surface of the obstacle.Otherwise Fresnel’s reflection and transmission coefficient starts to predict the reflected and transmitted fieldwrongly. Hence, the effect of wave reflection may be overestimated if the size of the reflecting surface is notlarge enough.The needed size of a surface to fully reflect the incident wave can be estimated from Fig. 2.2-5 by drawing aline across the ellipses that corresponds to the location of the obstacle’s surface with respect to thetransmitting and receiving antenna. Dependent on the orientation of the obstacle’s surface with respect to thedirect ray between the transmitter and receiver the size of the first Fresnel zone might vary vastly.With reference to Fig. 2.2-7 the major axis dMAof the Fresnel ellipse within the plane and smooth surface ofthe wall is determined bydMA( d − d )= 2 , whereMLdM⎛⎜⎝h⎞⎟⎠Tx= ⎜ ⋅ dhhTx+ h ⎟ , andRxd L2 2nλd = ( hTx+ hRx) + dh, C1= − d ,2C2 27h − TxC3= .2− C6± C6− 4C5C7= with2CC52C3= ,2C1C4dh= ,C1C5 1 C 4C6= 2CC , and2= − ,3 416

BROADWAY, IST-2001-32686 WP1-D2 version 1, date 10-10-02 Programme: ISTHorizontel CutTxRxd hh Rxh TxdWalld Ld Md MA1 st Fresnel zone2 nd Fresnel zone3 rd Fresnel zoneFig. 2.2-7: Fresnel zones for reflection.2.2.3. Fresnel’s Reflection and Transmission CoefficientA radio wave can be reflected, scattered, and absorbed by obstacles it meets. If the illuminated surface of theobstacle is plane, smooth (see section 2.2.1) and large enough (see section 2.2.2) the incident wave will bereflected. The general situation is illustrated in Fig. 2.2-8. The line A-A’ divides the space into two parts ofdifferent electrical properties.IncidenceReflectionθ iθ rAA'θ tRefraction17

BROADWAY, IST-2001-32686 WP1-D2 version 1, date 10-10-02 Programme: ISTFig. 2.2-8: Reflection and refraction at a plane-medium boundary.The wave meets this interface at an incident angle θi. Owing to the different wave impedance of the twomedia, part of the energy will be reflected at an angle θrequal to θi. The remainder will penetrate into theadjacent matter at an angle θt, which, in general, will not equal θi. The change in direction of thepenetrating ray is referred to as refraction.The relative portion of energy reflected from or penetrated into the adjacent matter depends on frequency,the angle of incidence, the polarization of the wave, and the electric and magnetic properties of the twomedia.For the process of wave reflection and transmission, the plane of incidence must be introduced. It is definedas the plane formed by the normal unit vector to the plane boundary formed by the two media and the vectorpointing into the direction of incident wave propagation.To examine reflections and transmissions at oblique angles of incidence for a general wave polarization, it ismost convenient to decompose the electric field into its parallel and perpendicular components (relative tothe plane of incidence) and analyze each one individually. The resultant reflected or refracted field will thenbe the vector sum of the two components.2.2.3.1. Parallel PolarizationFor parallel polarization of the field, Fresnel’s reflection ( ρ ) and transmission ( τ////) coefficients for planewave incident on plane, smooth, and infinitely extended surfaces are dete rmined by [11]ρτ////EEir1r22 2 2( θ ) − ε µ ε µ −εµ sin ( θ )r r i r r r r r rir r 1 21 1 2 2 1 1= = =, andEEiHHµµεεr1coscosr22 2 2( θ ) + ε µ ε µ −εµ sin ( θ )ir1r1r2r2r1r1( θ )r1rr2rit F2t1 2= = =,iZZF1HHiµεcos22 2 2( θ ) + ε µ ε µ − ε µ sin ( θ )iεµr1εr1µr2cosr2r1r1iiwhereEi,Er,Et,Hi,Hr,electric and magnetic fields, respectively, withF1respectively. Further, εr1andr2Htdenote the incident, reflected, and refracted magnitudes of the complexZ and ZF2the intrinsic impedance of the matter 1 and 2,ε are the complex relative permittivities,µ , andr1µ complex relativer2permeabilities of medium 1 (incident and reflected fields) and 2 (refracted field), respectively. Theserelations are a consequence of the boundary conditions for the electric and magnetic fields at the intersectionof the two media.2.2.3.2. Perpendicular PolarizationFor perpendicular polarization of the field, Fresnel’s reflection ( ρ ) and transmission ( τplane wave incident on plane, smooth, and infinitely extended surfaces are determined by [11]⊥⊥) coefficients forρ⊥E=Eri=HHriµ=µr2r2εεr1r1coscos2 2 2( θ ) − ε µ ε µ − ε µ sin ( θ )i2 2 2( θ ) + ε µ ε µ − ε µ sin ( θ )ir1r1r1r1r2r2r2r2r1r1r1r1ii, andτE=Z=H=2 εr1µr2cos( θi)2 2 2( θ ) + ε µ ε µ −εµ sin ( θ )t F2t⊥.EiZF1Hi µ εr r1cos2i r1r1r2r2r1r1i18

BROADWAY, IST-2001-32686 WP1-D2 version 1, date 10-10-02 Programme: IST2.2.3.3. Fresnel’s Reflection and Transmission Coefficient - Building MaterialsThe subsequent sub-sections are aimed at the dependence of parallel (//) and perpendic ular (-/-) componentof Fresnel’s reflection coefficient on the zenith angle θiof waves incident on a plane and smooth interfacebetween two media. The matter containing the inc ident and the reflected wave is filled with air while theother half space is homogeneously filled with the respective building material.Fresnel’s reflection and transmission coefficients for the parallel and perpendicular components of theelectric field do not show an explicit frequency dependency. The frequency dependent behavior is caused bydispersive matters exhibiting frequency dependent complex relative permittivities ε and permeabilities µ(c.f. Table 2.2-1). In this sense, materials showing frequency independent real relative permittivities εrandpermeabilities µrwhose conductivity σ are not vanishing are also dispersive becauseεr= εrσ− jωε 0with j , ω , and ε0denoting − 1 , the angular frequency, and the permittivity of vacuum, respectively.rrBrickFig. 2.2-9: Fresnel coefficients for parallel (//) and perpendicular (-/-) polarisation of a plane, smooth surface of brick.(e r ’=4.12 / e r ’’=0.16) [5 GHz] and (e r ’=3.95 / e r ’’=0.073) [60 GHz].ConcreteFig. 2.2-10: Fresnel coefficients for parallel (//) and perpendicular (-/-) polarization of a plane, smooth surface of concrete.(e r ’=5.5 / e r ’’=0.18) [5 GHz] and (e r ’=6.4954 / e r ’’=0.4284) [60 GHz].19

BROADWAY, IST-2001-32686 WP1-D2 version 1, date 10-10-02 Programme: ISTGlassFig. 2.2-11: Fresnel coefficients for parallel (//) and perpendicular (-/-) polarization of a plane, smooth surface of glass.(e r ’=3.98 / e r ’’=1.0764) [5 GHz] and (e r ’=4.7 / e r ’’=0.1551) [60 GHz].PlasterboardFig. 2.2-12: Fresnel coefficients for parallel (//) and perpendicular (-/-) polarization of a plane, smooth surface ofplasterboard. (e r ’=2.02 / e r ’’=0.05328) [5 GHz] and (e r ’=2.6 / e r ’’=0.0364) [60 GHz].ChipboardFig. 2.2-13: Fresnel coefficients for parallel (//) and perpendicular (-/-) polarization of a plane, smoothsurface of chipboard. (ε r ’=2.88 / ε r ’’=0.4879) [5 GHz] and (ε r ’=2.78 / ε r ’’=0.1362) [60 GHz].Table 2.2-1: Electrical properties of several building materials.MATERIALS 5GHz 60GHz20

BROADWAY, IST-2001-32686 WP1-D2 version 1, date 10-10-02 Programme: IST[thickness, specified] : er' er'' tan(d) n er' er'' tan(d) nBrick (made of chalk, with holes) 4,1200 0,16000 0,038830 2,0301-j0,0394 3,9500 0,07300 0,018481 1,9875-j0,018364Brick (without holes) 3,3000 0,01000 0,003031 1,8165-j0,00275 3,2600Brick wall 3,5600 0,34000 0,095505 1,888-j0,08999Concrete (one year) 5,5000 0,18000 0,032727 2,3455-j0,03837 6,4954 0,42840 0,065954 2,55-j0,084Concrete (40 years) 4,6000 0,24000 0,052173 2,145-j0,05593Concrete (50 mm) 11,4700 0,29593 0,025800 3,387-j0,04369Aerated concrete 2,2600 0,10170 0,045000 1,5037-j0,03381Aerated concrete (50 mm) 3,6600 0,12481 0,034100 1,9134-j0,03261Wood (a) 2,0500 0,29600 0,144390 1,4354-j0,1031Wood (b) 1,6500 0,23500 0,142424 1,2877-j0,09124wood 1,5-1,64 0,09-1,115 0,06-0,68 1,2253-j0,03672Plain wood (19 mm, 1 layer) 2,0687 0,41388 0,200770 1,445-j0,143Plain wood (19 mm, 2 layer) 1,9515 0,36830 0,188770 1,40-j0,131PlywoodPlasterboard (5 mm, 1layer) 2,4845 0,06211 0,025000 1,57-j0,0199Plasterboard (5 mm, 2 layer) 1,8427 0,00000 0,000000 1,357-j2e-7Plasterboard (5 mm air gap, 1 layer) 2,5410 0,61179 0,240770 1,6-j0,19Plasterboard (5 mm air gap, 2 layer) 2,1333 0,00299 0,001400 1,46-j0,00010,0364-Plasterboard 2,0200 0,05328 0,026376 1,42139-j0,018742 2,6-3,08 0,05544 0,014-0,018 1,61249-j0,01286Plasterboard (9 mm) 2,3700 0,12012 0,050682 1,54-j0,039Plasterboard (12 mm) 2,7000 0,05346 0,019800 1,643-j0,0162660,1362-Chipboard 2,8800 0,48790 0,170000 1,703-j0,14323 2,78-3,15 0,1795 0,049-0,057 1,66783-j0,04083Chipboard (22 mm, wood) 2,8500 0,15875 0,055700 1,6889-j0,04699Thermolite block (100 mm, 1 layer) 2,9396 0,27250 0,092700 1,716-j0,0794Thermolite block (100 mm, 2 layer) 4,9913 0,32094 0,064300 2,23-j0,0718Cealing board (8.8 mm, rock wool) 1,5876 0,01260 0,007937 1,26-j0,005Floor board (24.8 mm, syntheticresin) 3,9135 0,32868 0,083986 1,98-j0,083Stone 6,72-8,550,0336-0,5643 0,005-0,066 2,5923-j0,00648Marble 11,5600 0,08092 0,007000 3,4-j0,0189990,09223-Tiles 4,01-8,58 0,7807 0,023-0,091 2,0026-j0,023027Clear glass (4 mm, 1layer) 2,9591 0,00041 0,000138 1,72-j0,00019Clear glass (4 mm, 2 layer) 6,3919 0,00032 0,000050 2,528-j6e-5Meshed glass (5 mm, 1 layer) 8,0213 0,00024 0,000030 2,83-j4,6e-5Meshed glass (5 mm, 2 layer) 3,4808 0,57364 0,164800 1,87-j0,15330,1551-Glass 5,9800 1,07640 0,180000 2,455-j0,2192 4,7-6,13 0,50266 0,033-0,082 2,1682-j0,03576Acrylic glass 2,5200 0,03024 0,012000 1,5874-j0,0095245Plexiglas 2,7600 0,01800 0,006522 1,6613-j0,00541732.2.4. Frequency Dependent Loss FactorPlane waves are the simplest solution of the wave equation. The electric field ( x)propagating in the direction of the positive x -axis is of the formEj k xk x k x( ) j '''−− −x E e = E e e= ,where E0denotes the complex envelope of its amplitude and'k and''k its real and imaginary part.0022ω ε0εrµ0µr− jωµ0µr= ω ε0εrµ0kE of such a wave' ''= k − j k the complex wave number with' ''k = k − j k =µ .r21

BROADWAY, IST-2001-32686 WP1-D2 version 1, date 10-10-02 Programme: ISTFor the investigation of the frequency dependence of the propagation loss, the loss factorpower)L F= e' '−2 kdLF(related tohas been evaluated as a function of the distance d . Fig. 2.2-14 through Fig. 2.2-18 show the behavior ofObviously, the water content of these materials has a dominant influence onto the propagation loss of a wavepropagating within them.2.2.4.1. BrickLF.Fig. 2.2-14: Loss factorLF within brick. (e r ’=4.12 / e r ’’=0.16) [5 GHz] and (e r ’=3.95 / e r ’’=0.073) [60 GHz].2.2.4.2. ConcreteFig. 2.2-15: Loss factorLF within concrete. (e r ’=5.5 / e r ’’=0.18) [5 GHz] and (e r ’=6.4954 / e r ’’=0.4284) [60 GHz].22

BROADWAY, IST-2001-32686 WP1-D2 version 1, date 10-10-02 Programme: IST2.2.4.3. GlassFig. 2.2-16: Loss factorLF within glass. (e r ’=3.98 / e r ’’=1.0764) [5 GHz] and (e r ’=4.7 / e r ’’=0.1551) [60 GHz].2.2.4.4. PlasterboardFig. 2.2-17: Loss factorFL within plasterboard. (e r ’=2.02 / e r ’’=0.05328) [5 GHz] and (e r ’=2.6 / e r ’’=0.0364)[60 GHz].23

BROADWAY, IST-2001-32686 WP1-D2 version 1, date 10-10-02 Programme: IST2.2.4.5. ChipboardFig. 2.2-18: Loss factorLF within chipboard. (e r ’=2.88 / e r ’’=0.4879) [5 GHz] and (e r ’=2.78 / e r ’’=0.1362) [60 GHz].Table 2.2-2: Additional path loss for common building materials [1].2.2.5. Attenuation due to ReflectionRadio energy incident on a plane and smooth interface between two media (1 and 2) of different electricalproperties become partly reflect at an angle equal to the one of incidence. The rest of the energy istransmitted into the adjacent media 2 causing loss of energy for the reflected waves.S i, verticalMaterialLoss factor / dB12cm thickness brick wall >2030cm thickness concrete pillar >20Plasterboard stud partition wall 13.2Wooden door 9.6Glass door with wire mesh 4.2Moveable hardboard office partition 4.7Metal filing cabinet >20Person 14.6S iS tθ i θ rS i, horisontalS t, horisontalS t, verticalMedium 2ε r2>1, µ r2=124Medium 1 (Air)ε r1 = 1, µ r1 = 1θ t

BROADWAY, IST-2001-32686 WP1-D2 version 1, date 10-10-02 Programme: ISTFig. 2.2-19: Pointing ve ctor at interface between two plane, smooth media.To quantify this loss of energy, the frequency dependence of the magnitudeS ,of the power fluxdensity St ,being transmitted perpendicularly into the adjacent media 2 is evaluated for bothverticalpolarizations of the incident plane wave having the electric field strength Ei. For the parallel polarized fieldtverticalSt, vertical,//=2τ//EZF22icos( θ )twhere for the perpendicular polarized oneS2 2= τ⊥Eit,vertical,cos( θt)Z.⊥F2Here, τ//and τ⊥are Fresnel’s transmission coefficients (see section 2.2.3.1 and 2.2.3.2) for the parallel andperpendicular component of the incident electric field, respectively, θtthe zenith angle of the refracted wavewith respect to the interface of the two media, and ZF2. the intrinsic impedance of the media containing therefracted wave which is given byZ F 2=µε εwith εr2the relative permittivity of the medium 2. Snell’s law of refraction statesβ00r2( θ ) β sin ( )1sini=2θ twith β1and β2the phase constants of media 1 and 2. The cosine of the zenith angle of the refracted wavetakes on the formcos⎛ β ⎞=⎜ ⎟.⎝ ⎠1 2( θ ) 1 − ⎜ ⎟ sin ( )tθ iβ2The phase constants are related to the permittivity and permeability of the respective media by⎛2 ⎞1 ⎜ ⎛ σ ⎞ ⎟β = 2πf µε 1 + ⎜ ⎟ + 1 ,2 ⎜⎟⎝ 2πfε⎠⎝⎠where µ = µ0µrand ε = ε0εrwith µr, εr, and σ the relative permeability, relative permittivity, andthe conductivity of the matter, respectively.The magnitude Siof the power flux density of the incident wave is given by2S =iEZ2iF1with ZF1the intrinsic impedance of the media 1.The magnitude of the power flux density St, vertical, //for parallel polarized fields normalized by Sitakes onthe formS25( )t,vertical,// 2 r2= τ//cos θt,Siεr1ε

BROADWAY, IST-2001-32686 WP1-D2 version 1, date 10-10-02 Programme: ISTwhile the one for the perpendicular polarized field isS( )t,vertical,⊥ 2 r2= τ⊥cos θt.Siεr1These two quantities are measures of the loss that plane waves suffer in the process of being reflected from aplane and smooth interface between the two media.Obviously, for grazing angle of incidence ( θ = 90°) all materials tend to reflect perfectly while foriperpendicular incidence ( θ = 0°) the largest amount of energy is transferred into the adjacent media.iεTable 2.2-3: Reflection attenuation factor a values of some common materials [2].2.2.6. Atmospheric EffectsA LOS link at mm-wave frequencies suffers from two major atmospheric effects:1. attenuation due to rain, and2. attenuation due the oxygen.MaterialaPlasterboard 0.98Chipboard 0.8Aerated concrete blocks 0.842.2.6.1. Scattering by RainThe extinction cross section cextis a handy quantity to investigate the power removed from the incidentwave as scattered and absorbed power Pais given by the product of the magnitude of the incident powerflux density Siwith the extinction cross section cext.P = c S .aextiThe extinction cross section is therefore the sum of the scattering cross sectionsection ca.c = c + c .extsacsand the absorption cross26

BROADWAY, IST-2001-32686 WP1-D2 version 1, date 10-10-02 Programme: ISTFig. 2.2-20: Extinction cross-section cext versus a) drop diameter and b) frequency [3].When an electromagnetic wave propagates through a rain cell it encounters a great many water droplets withdifferent radii. Since cextstrongly depends on the radius r = D 2 of an equivalent rain drop ( D denotes thediameter of an equivalent rain drop being spherical in shape), it is necessary to take into account the dropsize distribution. Let N ( r) dr be the number of drops per unit volume with radii in the interval r to r + dr .The total power P removed from a wave with power density2ESi= ,2 Zby the drops in a volume element of unit cross-sectional area and thickness dz along z is (propagation inz -direction assumed)dPdz= −SF0∞∫ ( ) ( )icextr N r dr .0Here, E is the electric field at the location of the rain drop, andfree space. As a result of the power loss, the power flow decay at a rate∞∫0( ) ( )A=2α= c r N r drextZ ≅ 377 Ω the intrinsic impedance ofF 02 α iswhere A is the specific attenuation per unit length along the propagation path. Combining the aboveequations one yieldswhose solution isdPdz= −A( z) Si27

BROADWAY, IST-2001-32686 WP1-D2 version 1, date 10-10-02 Programme: IST∫A( z') dz'0P( z) = P( 0 ) e .The cumulative effect of attenuation suffered by an electromagnetic wave penetrating a rain cell isconsidered in the subsequent section.2.2.6.2. Rain AttenuationFig. 2.2-21 shows the specific attenuation due to the sum of the contribution of each individual rain dropwithin a rain cell for different values of the rain rate [3]. The results have been obtained assuming an−1−3exponential distribution for the drop-size with N = 8000 mm .z−0mA light drizzle corresponds to a rain of 0.25 mm/h, light rain corresponds to about 1 mm/h, moderate rain to4 mm/h, and heavy rain to 16 mm/hClouds of ice crystals and snow do not cause appreciable attenuation, even if the rate of fall exceeds125 mm/h. This is due to the much-reduced loss of ice compared to water. Attenuation by concentric spheresof different dielectric constant, for example, melting ice spheres, may approach that of rain.Fig. 2.2-21: Specific attenuation due to rain [3].Table 2.2-4: Rain Attenuation for different rain rates [4].Rain attenuation / dB Rain Rate / mm/h25 5060 GHz 10.1 17.963 GHz 10.4 18.266 GHz 10.6 18.5Rain attenuation is considered negligible for small distances (< 200 m).2.2.6.3. Oxygen absorptionThe oxygen absorption has a peak around 60 GHz and the additional path loss can be modeled like [4]:28

BROADWAY, IST-2001-32686 WP1-D2 version 1, date 10-10-02 Programme: ISTγ⎧ − − ≤ ≤3.2615.10 0.104( f 60) , 60 f 630[ dB / km ]( f[ GHz ])= ⎨2.25 1.27⎩11.35 + ( f −63) −5.33( f − 63) , 63< f ≤66Although γ 0 ranges from 15.1 dB/km (at 60 GHz) to 1.7 dB/km (at 66 GHz), the additional path loss isnegligible for indoor propagation.2.2.6.4. LOS Wave PropagationWhile transmitting information over a LOS path two situations must be distinguished:1. free-space propagation, and2. two-path propagation.The frequency dependence of the path loss for both situations are addressed in this section.Free-Space PropagationFor a path to be LOS, it is necessary that the link is designed to maintain adequate clearance between theradio path and the terrain being crossed. This is due to the fact that even if there is clearance between thedirect ray and the terrain such that one can optically see the transmitting from the receiving antenna, up to6 dB diffraction loss can be experienced if the clearance is too small.The path clearance criteria is specified in terms of radius of the first Fresnel ellipsoid F1at the mostsignificant path obstruction, i.e. the path obstruction that would cause loss of LOS if the antenna heightwould be lowered. Commonly, the first Fresnel zone should be free of any obstacle for Friis formula to hold.According to Friis [5], the received signal power PRxis given in terms of the transmitted power PTXby2PTxλPRx= G2 TxGRx4 π d 4 πassuming perfect match between the feeding cable and the antenna at the transmitter and receiver site. Thegain of the transmitting and receiving antenna are GTx, and GRx, respectively, with d the distance betweenthem and λ the wavelength. The path loss L is therefore given by2P ⎛ ⎞Tx4 π d ⎛ 4π⎞L = =⎜⎟ ⎜2⎟ .PRx⎝ GTxGRx⎠ ⎝ λ ⎠Fig. 2.2-22 depicts the path loss L versus transmitter-receiver separation d using directive antennas with again of 20 dBi at both the transmitting and receiving site.29

BROADWAY, IST-2001-32686 WP1-D2 version 1, date 10-10-02 Programme: ISTFig. 2.2-22: Path loss L versus distance d .GTx = GRx = 20 dBi.For other types of antennas, the corresponding gains can be subtracted from the path loss shown in Fig.2.2-22. As becomes obvious from Fig. 2.2-22 path loss increases tremendously with increasing frequency.This is due to the reduction of the effective area Aeof the receiving antenna with increasing frequency. Theeffective area Aeiof an isotropic radiator is2λAei= .4 πTwo-path PropagationThe two-path wave propagation model is an often-applied assumption within a flat environment. As can beseen from Fig. 2.2-23 two propagation paths exist between the transmitting and receiving antenna, a directand an indirect one.ReceivingAntennaTransmittingAntennadirect pathd 1d 2h Rxindirect pathh TxdL yL xFig. 2.2-23: Plane Earth wave propagation.The available receiving power in terms of the radiated transmitting power PTxfor parallel (vertical) PRx, //and perpendicular (horizontal) PRx, ⊥polarization of the antennas are given byPRx,//2− j k0d1⎛ λ ⎞e= ⎜ ⎟ P G G + ρ⎝ ⎠− j k 0 d 2Tx Tx Rx//4 πd1d2e2, andP⎛ ⎞e= ρ⊥⎝ 4 ⎠d2− j k 0 d1λeRx,⊥⎜ ⎟ PTxGTxGRx+πd1− j k0d2where k = 2π 0λ is the wave number with ρ and ρ denoting Fresnel’s reflection coefficients for the//⊥parallel and perpendicular polarized field, respectively. The two expressions neglect the difference in gainbetween the direct and the indirect path from the transmitting to the receiving antenna. This is a veryconvenient simplification for transmitter-to-receiver separations d1is much larger than either of the heightshTxand hRxof the transmitting and receiving antenna, respectively. Further more, it is assumed that the areaLx ⋅ L yfully contains at least the first Fresnel ellipse as required for the surface to reflect (see section2.2.2.2).22,30

BROADWAY, IST-2001-32686 WP1-D2 version 1, date 10-10-02 Programme: ISTThe usual form of the two-path model found at many places within the literature is based on the assumptionthat d1>> hTxand d1>> hRx. This allows to simplify the amplitude terms in the above equations by settingd 1≅ d 2≅ d which neglects the slight path difference between direct and indirect trajectory. Further more,for d1>> hTxand d1>> hRxthe angle of incidence of the indirect ray with the ground is very small.Therefore, Fresnel’s reflection coefficients for parallel ( ρ ) and perpendicular ( ρ ) polarized fieldsbecome approximately –1 independent of the values of the complex relative permittivity of the ground andthe air. Using these approximations within the above expressions for the available received power yieldsPRx⎛= ⎜⎝λ4 πd2⎞⎟⎠PTxGTxGRx//2 ⎛ k4 sin ⎜⎝which is independent of the polarization of the field. Hence, the path loss L is given byPL =PTxRx=λ2GTx2 24πd2⎛kGRxsin ⎜⎝0hTxd0hhTxdRxh⎞⎟⎠.Rx⎞⎟⎠⊥Fig. 2.2-24: Path loss ve rsus distance d forAgain, ifd >> hTx1andRxa) b)h = 0.5 m and h = 2 m, floor of brick . a) parallel, b) perpendicularRxTxpolarization. GTx = GRx = 20 dBi.d1>> h , the argument of the sinus function becomes very small. Hence,2⎛ksin ⎜⎝which inserted into the expression for the path loss yields0hRxTxdhTxL =Rx⎞ ⎛ k⎟ ≅ ⎜⎠ ⎝TxRx0h2TxTxdh2RxRx4P d= .P G G h hObviously, the path loss L as predicted by the two-path model with ideal reflection of the indirect wave onthe ground is independent of the choice of frequency and increases with the 4 th power of the transmitter-toreceiverseparation d .Two-path Propagation, Rough SurfaceThe interaction of electromagnetic waves with plane surfaces of large extension ( Lx>> λ , Ly>> λ ) andsmall roughness factor ( σR

BROADWAY, IST-2001-32686 WP1-D2 version 1, date 10-10-02 Programme: ISTextent compared to the situation with specular reflection. This is due to partial destructive interference of thereflected signal with the scattered ones.ReceivingAntennaTransmittingAntennadirect pathd 1d 2h Rxindirect pathh TxdRough surfacescatteringL yL xFig. 2.2-25: Rough surface scattering.A well known and proven approximation for the reduction of the field strength caused by scatteringcompared to reflection is by modifying Fresnel’s reflection coefficients ρ and ρ for the parallel andperpendicular polarized fields by a roughness factor [1]://⊥ρS2 2( σ λ ) cos ( θ )2−8πFi= e,whereσF, and θidenote the surface roughness according to Fraunhofer’s criterion, and the zenith angle ofincidence, respectively. The modified reflection coefficientsincident field areρmod//= ρ ⋅//ρ Sandρmodρ and⊥mod//= ρ ⋅⊥Inserting the surface roughness criterion according to Fraunhofer (see also 2.2.1.2)σFλ

BROADWAY, IST-2001-32686 WP1-D2 version 1, date 10-10-02 Programme: ISTHuygen'ssecondarysourceszr T(y,z)r R(y,z)AbsorbingScreenH > 0Transmitterd Tydx 0d RReceiverxFig. 2.2-26: Knife -edge diffraction.Assuming that the transmitting and receiving antenna radiate and receive, respectively, cylindrical waves(2D case) and that the magnitude of the height H of the infinitely thin absorbing screen is much smallerthan the distancesdTandd of the transmitter and receiver, respectively, to this screen ( H

BROADWAY, IST-2001-32686 WP1-D2 version 1, date 10-10-02 Programme: ISTB1+0.926 v= , and2 + 1.792 v + 3.104 v( )2A v1= .2 + 4.142 v + 3.492 v + 6.67 v( v) 23The diffraction loss LDrelative to the one of the corresponding free-space path is defined with respect toreceived signal power. Hence⎛ EFS⎞L =⎜⎟D.⎝ ERx⎠In Fig. 2.2-27 height H of the absorbing screen is relative to the direct ray between the transmitter and thereceiver. For H > 0 the receiver is in the shadow region of the screen, for H < 0 the receiver is “opticallyvisible” from the transmitter site, and for H = 0 the receiver is just at the shadow boundary from thetransmitters point of “view”.At the shadow boundary ( H = 0 ), the relative diffraction loss LDincreased already to 6 dB independent of25the selected frequency. Within the visibility region ( H < 0 ) LDfluctuates around a value of 0 dB caused byconstructive and destructive interferences of secondary waves emerging from areas close to the knife-edge.Deep302into the shadow region ( v >> 0 ) the relative diffraction loss LDis proportional to v and, hence,keeping the parameters of the geometry ( dT, dR, and H ) constant results in a f dependency of the35relative diffraction loss. It should be noted that the relative diffraction loss LDdoes not include anyf = 5 GHzfrequency dependent effects caused f = by 60 the GHz transmitting or receiving antenna. The absolute propagation path402loss due to a trajectory suffering one knife-edge diffraction along its path would involve an other fdependency caused by the free-space wave propagation process resulting in an overall frequency dependency3 2 4 6 8 10of f .Path Loss L rel. to Free Space / dBPath Loss L Drel. to Free Space / dB05101520Knife-Edge DiffractionDistance d R/ ma)Knife-Edge Diffractionf = 5 GHzf = 60 GHz-3 -2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5 3Height H / m2b)34

BROADWAY, IST-2001-32686 WP1-D2 version 1, date 10-10-02 Programme: ISTFig. 2.2-27: Diffraction lossLD relative to free space versus a) screen height H and b) receiver distance to screen d R for H =0.5 m.2.2.8. ShadowingAt 60 GHz the radio signal barely passes through wall (cf. Table 2.2-2) and coverage is basically confined toa single room. Thus shadowing by furniture and people need to be considered.Table 2.2-5: Additional loss due to shadowing [1].Additional loss / dBFurniture 8.2Human 4.2The attenuation by furniture or persons blocking the direct path is not always as severe as might be predictedfrom the material losses measured at close range (see section 2.2.4). The author of [1] concludes that in thesecases only a fraction of possible multipaths is obstructed by the item of furniture or the person and thusreflected paths contribute markedly to the received signal.2.3 Propagation models2.3.1. Narrowband, Path LossThe basis for such a model is the free-space equationor in logarithmic form asP ⎡Rxλ ⎤L = = GTxGRx⎢ ⎥PTx⎣4πd ⎦2⎛ P ⎞Rx⎛ λ ⎞L/ dB=− log10⎜ ⎟=−20log10⎜ ⎟−10log10GTx−10log10GRx⎝ PTx⎠ ⎝4πd ⎠where L is the path loss at distance d. P Tx is the power supplied to the transmitting antenna with a gain of G Txin the direction of the receiving antenna. P Rx is the power at the receiving antenna of gain G Rx . λ is thewavelength of propagation.Considering a frequency of 60 GHz:( )L/ dB= 68.1+ 20log d / m −10log G − 10log G10 10 T10As the above equation describes propagation in free-space and indoor propagation is affected by reflectionsand attenuations caused by walls, floors, furniture, people, etc., an inverse exponential law with the distance−nbetween antennas is introduced; i.e. P Rx∝ d , where n depends on the environment. Values for n havebeen presented in the literature, ranging from 1.4 to 2.5. As it is not totally clear how this parameter dependson the scenario, we propose to use n = 2, i.e. free-space propagation, as the “middle course”.Values for n < 2, i.e. better than free space, are commonly explained by wave guiding effects in corridors andreverberation effects in rooms, causing raised power levels by multipath contributions.The small-scale fading caused by multipath is typically modeled with a Rician distribution for LOS andRaleigh for non-LOS. The Rician K-factor is found to be highly variable over a large range, with the highestvalues in LOS conditions.Large-scale fading results from the shadowing by obstacles. The shadowing effects are usually found followa log-normal distribution about the local median of the path loss.R35

BROADWAY, IST-2001-32686 WP1-D2 version 1, date 10-10-02 Programme: IST2.3.2. Wideband, Impulse ResponsesDuring the ACTS AC006 MEDIAN project a measurement based, wideband model has been proposed forsystem simulation purposes.2.3.2.1. Measurement ResultsThe measurements were performed in a library at IMST premises.The results reported here focus on an indoor environment with large metal reflectors (bookshelves) supposedto cause large multipath contributions, resulting in a kind of worst case LOS scenario. Emphasis is put onLOS or almost LOS conditions (i.e., no strict LOS, but with still a dominant path e.g. by diffraction) , whichis felt to be the most relevant practical application case.Fig. 2.3-1 shows two sets of measured impulse responses at two LOS positions. Note that due to the constantpower flux design of the AP antenna, the gain is comparable for both pos itions, despite the different distancebetween MT and AP. Multipath decay is slower for positions farther from the AP.a) b)Fig. 2.3-1: Measured impulse responses at two positions a) far from and b) near to AP.The effect of mispointing at short distances is clearly visible in (b). Moving the MT antenna 25 cm hereresults in a 10 dB drop of the direct path signal, and, as the multipath is approximately constant, acorresponding degradation of the C/M ratio. In Fig. 2.3-2 two averaged power delay profiles (PDP) for twoLOS positions far (a) and near (b) the AP are given. Multipath contributions are visible approx. 30 dB belowthe LOS level. The envelope of the multipath shows a linear (on dB scale) decay behavior. Some additionalpeaks are visible indicating the existence of additional discrete echoes.a) b)Fig. 2.3-2. Averaged power delay profiles at two LOS positions. The vertical lines denote the 50 % - 99 % energy window ofthe noise normalized PDP.36

BROADWAY, IST-2001-32686 WP1-D2 version 1, date 10-10-02 Programme: ISTFor these two positions, a comparison of the empirical probability density function (pdf) of the PDPamplitude and the maximum likelihood estimate (MLE) of a Rician pdf for a 5 ns delay window of 501PDP’s is shown in Fig. 2.3-3. In both cases, the power ratio of the constant and stochastic part, the K-factor,is sufficiently low to effectively reduce the Rician to a Rayleigh pdf, i.e., real and imaginary parts of thecomplex impulse response essentially follow a Gaussian distribution, confirming the proposition in [7].The match between the empirical and MLE Rician pdf is significantly better for the position near the AP.This may be due to the fact that the spread of multipath directions-of-arrival may be much higher for thenearer position than for the remote position. At that position, multipath contributions are supposedly morelikely to arrive from the general direction to the AP.(a) b)Fig. 2.3-3. Empirical pdf and maximum likelihood estimate of Rician pdf for a 5 ns delay window in the multi -path region of501 PDP’s for two LOS positions.Several other authors have performed related radio channel measurements for several antenna combinations,e.g. omni–omni (two biconical horns) and high-gain–high-gain (two horns) [8], omni–omni (biconical hornand monopole) in a nearly empty office room [9]. In [7], measurements are reported with an omni constantpower flux lens antenna at AP and a standard horn antenna in an office equipped with tables, chairs andheating radiators, resulting in a well-behaved channel with few multipath contributions. They found that, inLOS condition, dynamic reflectors (people moving outside LOS) did not influence static measurementresults for the lens–horn combination.2.3.2.2. Model DescriptionIn [10], Saleh and Valenzuela propose an indoor multipath model based on the assumption of rays arriving inclusters, with clusters and rays within clusters forming Poisson arrival processes. The above results areconsistent with a simple one-cluster version of this model. The model may be applied in the following way.For LOS conditions, the delay, amplitude, and phase of the direct, i.e. free-space, signal component may bedetermined using the geometric distance between AP and MT and the corresponding antenna gains. 1The multipath part is modeled by a single cluster of echoes with, according to [10], exponentially distributedinter arrival times τ with a mean arrival rate λ. The mean echo amplitude β follows an exponential decayrelationship β 2 (τ) = β 2 (0) exp(-τ / γ) with the multipath power decay coefficient γ. Echo amplitudes areRayleigh and phases are uniformly distributed. Because there is only one cluster, the parameters Λ and Γ thatdescribe cluster arrival and decay behavior in [10] are not required here. Therefore, the first cluster is to startimmediately after the direct path.The mean arrival rate λ is determined according to the following considerations. Gaussian-like behavior ofthe impulse response in the multipath region suggests a superposition of a large number of rays with quasicontinuoustimes-of-arrival, which may not be resolved by the measurement. Therefore, each echo producedby the model represents a set of physical echoes, with the Rayleigh distribution of echo gains accounting forthe superposition, so the mean arrival rate does not reflect physical arrival times of single echoes, but sets ofechoes. We suggest to select λ in accordance with the measurement or simulation bandwidth, i.e. λ ≅ 1 / ∆τ(or slightly larger) where ∆τ is the delay resolution of the measurement or simulation under consideration.For our measurements, theoretically ∆τ = 1 / 960 MHz ≅ 1 ns, and we propose a value of λ = 1 ns -1 .1 Depending on how strictly the LOS condition is given, the direct path amplitude may be subject to aparticular, e.g., Rician, distribution.37

BROADWAY, IST-2001-32686 WP1-D2 version 1, date 10-10-02 Programme: ISTThe number M of multipath echoes is to be selected such that M > λτ, where τ is the delay required for themultipath echo amplitude β(0) to fall below the noise level under consideration.2.3.2.3. Model ParametersTable 2.3-1 summarises the parameters β(0) and γ. The values of β(0) are given in decibels relative to thedirect path amplitude. The extraction of β(0) and γ is based on linear regression within a particular energywindow of the averaged PDP. In our case, the used window extends from 50 % to 99 % of the averagedPDP's that have been normalised to the noise level. The approach is illustrated in Fig. 2.3-2.Table 2.3-1: Model parameters.β(0) / dB(max)γ / nsr / mMT Antenna Position = x Mean = x MeanHorn cx -29.8 -24.9 -27.4 12.0 14.3 13.2 4.82cy -29.0 -19.1 -24.1 11.2 10.7 11.0 3.90cz -28.5 -19.9 -24.2 8.5 9.8 9.1 2.27dy -19.9 -19.9 9.3 9.3 3.02ew -29.0 -20.0 -24.5 12.5 14.6 13.6 5.81ex -27.5 -18.5 -23.0 10.2 10.0 10.1 4.05ey -27.7 -24.6 -26.1 8.9 10.6 9.7 2.42ez -29.2 -24.8 -27.0 8.5 10.8 9.6 2.28fz -17.6 -5.6 -11.6 10.6 13.4 12.0 3.95Mean -26.5 -19.7 -23.3 10.2 11.8 10.9Dipole Array cx -34.5 -34.5 11.6 11.6 4.82cy -33.5 -33.5 10.0 10.0 3.90cz -29.9 -29.9 7.5 7.5 2.27dy -23.3 -23.3 9.1 9.1 3.02ew -35.0 -35.0 12.2 12.2 5.81ex -30.6 -30.6 9.4 9.4 4.05ey -30.2 -30.2 8.6 8.6 2.42ez -31.5 -31.5 8.2 8.2 2.28Mean -31.1 -31.1 9.6 9.6Total Mean -28.6 -19.7 -25.8 9.9 11.8 10.5Notes.1. The values of β(0) are given in dB relative to the maximum value in the averaged PDP (cf. Fig. 2.3-2).2. The polarization combination is indicated by = and x for co- and cross-polarization, resp.AP polarization is always vertical.3. r denotes the distance between MT and AP antenna.38

BROADWAY, IST-2001-32686 WP1-D2 version 1, date 10-10-02 Programme: IST2.3.2.4. Comparison of Measured and Modeled Impulse Responsesa) b)Fig. 2.3-4. Comparison of a measured a) and a modeled b) impulse response. White Gaussian noise has been added to themodeled impulse response.A comparison of a typical measured and modeled (direct path gain = -66 dB, β(0) = -28.5 dB(max),γ = 8.5 ns, λ = 1 ns -1 , M = 150) impulse response is shown in Fig. 2.3-4. White Gaussian noise (70 dB belowdirect path) has been added to the modeled impulse response. A Hann window has been applied to bothresponses.2.3.2.5. Conclusions from the MeasurementsIn LOS and almost LOS conditions, the channel is dominated by the direct path. Multipath contributionsshow an essentially Gaussian behavior with an exponential amplitude decay that fits well into the frameworkof the statistical indoor model by Saleh and Valenzuela. A simplified one-cluster version of this mode laugmented by a direct path of free-space propagation is found to match the measurement results well, withmodifications to the direct path required for almost LOS conditions.The model may be used to generate typical impulse responses for simulation purposes. We assume thatblocking of the direct path e.g. by a person passing through the beam may be covered by our model as wellby dropping the direct path. However, this assumption has not been backed up by measurements.Note that so far only the indoor case has been investigated. A perspective is to find an outdoor propagationmodel too.2.4 Link budget: what seems to be achievable?In order to check whether the application scenarios requirements were achievable, a preliminary link budgetstudy has been carried out, allowing to have a first estimation of how much bit rate could be achieved with areasonable range.For this purpose, the necessary C/I to achieve a Bit Error Ratio of 10 -4 has been computed using the Unionbound [13] for different sets of constellation and coding rate. The necessary received power has then beendeducted from these C/I values for different bandwidths, and the range has been obtained using the abovepath loss formula for indoor and outdoor environment.Of course, assumptions on some baseband and RF parameters had to be made for carrying out this study.Realistic values have been chosen, however they are only examples of possible choices since all basebandand RF parameters are not fixed yet. The chosen parameters for this study were:• OFDM parameters:- Carrier spacing: 625kHz- guard interval size: 800ns- oversampling rate: 0.75• RF parameters:- transmit power: 10dBm- noise figure: 8dB39

BROADWAY, IST-2001-32686 WP1-D2 version 1, date 10-10-02 Programme: IST- hardware margin: 2dB- antenna gain: 3dB Tx, 3dB Rx- thermal noise: -174dBm/HzThe path loss exponent has been chosen as 2.25 for outdoor environment, and 2.2 for indoor environment.Each OFDM subcarrier is supposed to be affected by an independent Rayleigh coefficient.The results for different bandwidths, constellations and code rates have been summarized on the table below.Constellation QPSK 16-QAM 64-QAM QPSK 16-QAM 64-QAM QPSK 16-QAM 64-QAMCoding rate 1/2 9/16 3/4 1/2 9/16 3/4 1/2 9/16 3/4Bandwidth (MHz) 80 80 80 160 160 160 320 320 320FFT size 128 128 128 256 256 256 512 512 512Bit rate (Mbit/s) 40 90 180 80 180 360 160 360 720Spectral efficiency (bit/s/Hz) 0.500 1.125 2.250 0.500 1.125 2.250 0.500 1.125 2.250C/I shift compared to QPSK-SNR (dB), AWGN 7 16 7 16 7 16C/I shift compared to QPSK-SNR (dB), BRAN-C 11 17 11 17 11 17C/I for Rayleigh Channel (Union Bound) 6.307 17.307 23.307 6.307 13.307 22.307 6.307 13.307 22.307C/I for AWGN Channel (Union Bound) 3.515 10.515 19.515 3.515 10.515 19.515 3.515 10.515 19.515Outdoor propagation at 60GHzDistance (m) for Rayleigh Channel (Union Bound) 571.4 288.2 183.2 483.5 307.7 156.3 400.5 249.1 119.7Distance (m) for AWGN Channel (Union Bound) 664.2 449.3 244.2 566.5 371.2 195.4 478.6 302.8 151.5Indoor propagation at 60GHzDistance (m) for Rayleigh Channel (Union Bound) 19.4 6.1 3.3 14.2 6.8 2.7 10.3 5.0 1.9Distance (m) for AWGN Channel (Union Bound) 26.0 12.5 4.9 19.0 9.1 3.6 13.9 6.7 2.6Indoor propagation at 5GHzDistance (m) for Rayleigh Channel (Union Bound) 36.3 16.5 10.7 29.3 17.7 9.3 23.6 14.2 7.5Distance (m) for AWGN Channel (Union Bound) 44.4 26.8 14.0 35.8 21.6 11.3 28.8 17.4 9.1This preliminary study shows that bit rates around 350 Mbit/s can be obtained at a still reasonable range (6mindoor), but it seems difficult to consider much higher rates unless using enhancements (for exampleMIMO).Another valuable information provided by this study is that the best way to get a given bit rate is to use ahigh enough bandwidth with a small size constellation (e.g. QPSK), which allows a definitely higher rangethan using a high size constellation. For example, the three following ways to get 160-180Mbit/s arecompared:- 180Mbit/s with 64-QAM, R=3/4, 80MHz with 128 carriers: range 3.3m- 180Mbit/s with 16-QAM, R=9/16, 160MHz with 256 carriers: range 6.8m- 160Mbit/s with QPSK, R=1/2, 320MHz with 512 carriers: range 10.3m2.5 RF constraints2.5.1. Phase noiseOscillators phase noise is an important issue in OFDM systems because it may lead to a loss of orthogonalitybetween subcarriers. Therefore, the subcarrier spacing has to be chosen carefully in order to keep phase noiseeffects negligible. A study on phase noise in function of different carrier spacing has been carried out and is40

BROADWAY, IST-2001-32686 WP1-D2 version 1, date 10-10-02 Programme: ISTreported in [16]. It results from this study that the same subcarrier spacing as HIPERLAN/2 (i.e. 312.5kHz)would be achievable.2.5.2. Switching time between 5 and 60GHzSince the BROADWAY system will have to switch very often from 5GHz to 60GHz, or from 60GHz to5GHz, it is important to have a first guess of the switching time. Indeed, Work-Package 2 will need thatvalue when considering different system scenarios.The switching time limitation comes from the switch selecting the two different parts of the front-end (5 and60GHz). Assuming that the HMC224MS8 Hittite module is used in the RF front-end (see [16]), a switchingtime of 10ns is then achievable.3. First assumptions on the whole system3.1 Dual mode operationsBroadway will be a hybrid dual frequency system based on a tight integration of HIPERLAN/2 OFDMhighly spectrum efficient technology at 5GHz and an innovative ad hoc extension of it at 60GHz, namedHIPERSPOT. The global system will be cell-based, with dual-mode Access Points able to worksimultaneously at 5 and 60GHz, so that they can handle peer to peer connections at 60GHz with MobileTerminals located within their 60GHz range to offload the 5GHz band, while still being a 5GHz AccessPoint. Mobile Terminals will also be dual mode, but will work in only one band at a time.With these assumptions, it is obvious that the Access Points will have two RF Front Ends and two separatePHY parts. On the other hand, Mobile Terminals are expected to have dual-mode RF and PHY parts in orderto reduce their cost. Moreover, a seamless vertical hand-over mechanism, controlled at 5GHz, will have to beimplemented in order to allow the offloading of the 5GHz band into the 60GHz band.3.2 HIPERSPOT and BROADWAY classesThe HIPERSPOT part of Broadway will consist of two subsystems at 60GHz:· HIPERSPOT/C (HS/C): a HL/2 Compatible mode, which shall offer mainly the same transmissioncapabilities as HL/2 but at 60GHz. It is intended to keep the HL/2 compatible mode at 60GHz as close to theHL/2 PHY layer specification as possible, but some changes might be required in order to make low-costdevices possible.· HIPERSPOT/E (HS/E): an innovative Extension of the HL/2 at 60GHz in order to cope with specialenvironments and offering significantly higher data rates by exploiting the higher bandwidths available at60GHz. This mode is expected to use OFDM enhancements that will be studied in WP3.Based on these two HIPERSPOT modes, BROADWAY will define two different but compatible classes ofmobile terminals:· Class A: will be derived from limited extensions of the HL/2 hardware transposed in the 60GHz space.Those devices will support HL/2 and HS/C.· Class B: standing for high-end devices. This class will support HL/2, HS/C and HS/E.4. HIPERSPOT parameters and requirements4.1 HIPERSPOT functional requirementsBased on both the initial desired scenarios and the previous sections on constraints we will have to deal with,we derive realistic functional requirements for HIPERSPOT. This section is expected to be a guidelinefor the other work-packages.Requirement Value / Reference CommentsMax grossSee link budget section: more than this seems difficult to360Mbit/s(PHY) data rateachieve with a reasonable range41

BROADWAY, IST-2001-32686 WP1-D2 version 1, date 10-10-02 Programme: ISTMobility ofDevicesMobility ofEnvironmentExcess delay tocope withRangeUp to 3 m/sClass A: up to 15 m/sClass B: Up to 15 m/s100-150nsIndoor: 0.5 to 15 metersHL/2, IEEE802.11a assumptionWe should separate the “mobility of devices” and the“mobility of environment”.See section 2.3These parameters will be used for the design of theOFDM parameters.Naturally limited by the wallsDepends on the data rate (see section 2.4)4.2 HIPERSPOT parametersThe following tables contain first values, or range of values, that will be taken as working assumption for theproject. Note that these values are first assumptions, and some of them may change during the project if itturns out that different values would be better.Table 4.2-1: HIPERSPOT system parametersParameter Value CommentsBand59..62 GHzThis band is available worldwideChannelBandwidthsGuardBandwidthbetweenchannels20 MHz … 320 MHzPossibly 40 MHz20 MHz is the HL/2 channel size. For compatibilityreasons, a multiple of 20MHz will be used.The guard band between channels will be large in order toreduce the constraints on the filters. This value is aninitial guess.Table 4.2-2: HIPERSPOT baseband parametersParameter Value CommentsModulationOFDMOFDM is chosen because of its considerable advantagesfor high data rate systems (spectral efficiency, easyequalization, etc.), and for compatibility with existing5GHz WLANs (HL/2, IEEE802.11a) .The 60 GHz channels are expected to have an excessdelay of approx. 100…150 ns. Assuming very shortfilters in the TX/RX, a guard interval size of 200 nsseems to be sufficient (e.g. in the case where only veryGuard Intervalbroad channels are used, the filters become short).50 – 800 nssizeHowever, too short TX/RX filters are expected to lead toperformance drawbacks. In this case, either the guardinterval must be enlarged or means must be provided indigital in order to do some compensation. The 800 nsguard interval might be kept for compatibility reasonswith HL/2.This carrier spacing of 312.5 kHz is used in HL/2. At aCarrier spacing 312.5-2500 kHzfirst guess, this value should be reasonable for 60 GHz,however a phase noise study has to be carried out todetermine if a larger carrier spacing is required.FFT size 64-1024PossibleconstellationsBPSK, QPSK, 16-QAM, 64-QAMThese are the HL/2 constellations. Other constellations(e.g. 8-PSK) may be investigated too.42

BROADWAY, IST-2001-32686 WP1-D2 version 1, date 10-10-02 Programme: ISTNote that Table 4.2-2 contains only the main OFDM parameters. A whole deliverable [14] produced byWork-Package 3 is devoted to the description of baseband requirements and parameters. We refer the readerto it for further details.5. Open questions in system design5.1 Multiple access issuesThe choice of the MAC protocol for Broadway has a great importance because it determines the realthroughput available for the users. Since the aim of Broadway is to supply the user with very highthroughput, the selected MAC will have to be able to deal with raw physical layer bit rates of severalhundred of Mbit/s in hot spot scenarios, i.e. when a large number of users access the medium. [Is physicallayer bit rate really dependent on number of users??]On the other hand, Broadway is expected to be as compatible as possible with existing 5GHz WLANs,namely HIPERLAN/2 and IEEE802.11a. However, since the MAC protocols of these two standards aretotally different, it will not be possible to ensure backward compatibility at the MAC layer with both ofthem. Nevertheless, it is desirable to have a certain level of compatibility with one of them. The choice is noteasy since IEEE802.11 MAC protocol relies on CSMA/CA, which is rather suitable for ad hoc networking,while HIPERLAN/2 MAC protocol is based on TDD-TDMA, which generally ensures a better QoS.Since the choice had to be made as soon as possible , because it has an impact on all the WP2 studies, WP2carried out an analytical comparison between IEEE802.11 and HIPERLAN/2 MAC protocols in terms ofachievable throughput. The complete study will be inserted in [15]. The main conclusion of this study is thatthe CSMA/CA approach implemented as in IEEE802.11 does not support high bit rates. The enhancedimplementation of IEEE802.11e offers much higher throughputs, but still not high enough for theBROADWAY project. Therefore, a direct extension of an existing implementation of a CSMA/CA schemecould not be chosen for Broadway. On the other hand, the TDD/TDMA scheme of HL/2 allows highthroughputs, and even if TDD/TDMA has certain drawbacks for ad hoc networking, it seems at the time tobe the best candidate for the BROADWAY system.5.2 Ad hoc network: to which extent?Under the framework of the BROADWAY project, several ad hoc networking issues will be considered inWork-Package 2. In particular, all scenarios that are possible to consider in one mode of operation (5 or 60GHz) can be considered also for the other. The first and most important consideration is that the defaultmode of operation is the 5 GHz mode. The 60 GHz mode of operation will be used to offload the trafficwhenever necessary. For this reason the AP (Access Point) of BROADWAY will be the main “coordinator”of all the control operations that will be required for the proper system operation. However, part of the APfunctionality can be transferred to certain MTs (Mobile Terminals) known as Cluster Heads (CH).In the following sections a number of connectivity scenarios considered in the BROADWAY project arepresented.5.2.1. Basic Scenario ConsiderationsAn important assumption is that the AP of BROADWAY can operate simultaneously at 5 GHz and 60 GHz.The AP is responsible for handling all kinds of communication (control and data) at 5 GHz. It is evident thatthere will be no such handling for MTs outside the range at 5 GHz. Each MT will start operating at 5 GHz asthe default initial mode. A period of “AP discovery” will follow, in which the MT will try to discover,whether it is in the range of an AP or not. If an AP is found, then an association phase starts that correspondsto particular operations required for the establishment of communication (e.g. exchanging Connection IDs,MAC ID).5.2.1.1. Neighbor Discovery FrequencyThe AP will periodically order each MT to switch to a specific frequency at 60 GHz and run a certainneighbor discovery algorithm. The results will be transferred to the AP. The latter will decide whether aparticular MT belongs to a “cluster” and whether the particular MT could play the role of the CH.43

BROADWAY, IST-2001-32686 WP1-D2 version 1, date 10-10-02 Programme: ISTIf the mobility of the environment is high, then the frequency of this phase should be increased. If themobility of the environment is comparably low then the frequency of the neighbor discovery phase should bedecreased.In addition, an event-driven mechanism should be considered, capable of triggering the neighbor discoveryphase. Using such a mechanism, rapid changes even for low mobility environments would be detected assoon as they take place.5.2.2. Traffic ManagementEach CH is responsible for its cluster to enable the transmission of data from one MT to another according tocertain traffic criteria (destination, source, shortest path, quantity of data etc.). Each CH will be responsiblefor the construction of the frame of its cluster and therefore the exact scheduling of the data packets will bedone according to specific criteria.5.2.2.1. Special CasesIf the MT does not belong to the range of an AP then it switches to a predetermined frequency at 60 GHz andtries to ‘discover’ other MTs located nearby. As a result communication among MTs without the presence ofAPs at 60 GHz can be initiated.5.2.3. One-hop ScenariosFor all following scenarios, the MTs will be considered inside the 5 GHz range area of the AP.5.2.3.1. One MT inside the 60 GHz range of the APIn this scenario the CH for the MT is the AP. The AP is responsible to decide whether the 60 GHz mode or 5GHz mode of operation will be used for data traffic. In case the 60 GHz band is to be used, the AP informsthe MT about the particular 60 GHz frequency that will be used for data communication.In case the 5GHz band is chosen, the operations are identical to the standard HIPERLAN/2 operations. If the60 GHz mode of operation is chosen, then data are forwarded directly to the AP .Figure 5.2-1: First scenario: 1-hop communication. The MT belongs to the 60 GHz cell of the AP.5.2.3.2. One MT outside the 60 GHz range of the APIn case there is one MT that needs to communicate with the AP, this is either because it wants to accessinformation from a host in the Internet or because access to another MT can be done only through the 5 GHzband. The 5 GHz band has to be used and this is the standard HIPERLAN/2 mode of operation at 5 GHz.5.2.3.3. Two MTs that belong to the same 60 GHz rangeFigure 5.2-2 depicts two MTs that need to communicate. The 5 GHz band can always be used but it possibleto use the 60 GHz band to offload the traffic.44

BROADWAY, IST-2001-32686 WP1-D2 version 1, date 10-10-02 Programme: ISTFigure 5.2-2: First scenario: 1-hop communication. Two MTs are in the 60 GHz cell.5.2.4. Multi-hop ScenariosFigure 5.2-3 presents the 2-hop communication case. More than 2-hop communication scenarios will beavoided since the 5 GHz will be by far more suitable for data transfer.In this scenario the AP elects a CH. This CH is responsible for communicating with the AP for allinformation needed (neighbors, switching period, etc.) and for handling all communication between the MTs.Figure 5.2-3: Second scenario: 2-hop communication.5.2.5. Internetworking IssuesThe BROADWAY network will be considered as a single IP subnet. As a result, routing will take place atlayer 2. Both at the AP and the MT's, access to the wireless communication capability offered by Broadwaywill be represented by a single network interface, irrespective to the frequency used and the instantaneousconnectivity to other nodes. (In addition, the AP may have one or more interfaces to fixed networks).. Theaccess network from the Internet to the AP is considered as a high capacity network with no limitationscompared to the bandwidth of the BROADWAY network.Special reserved IP addresses will be assigned for the operation with no AP present. The main requirementwill be to belong to a certain subnet5.3 Baseband enhancements to be investigatedThe studies on base-band issues will concentrate on several key modules (see [14] for details). Thus, themodulation will be evaluated and several innovative modifications of standard OFDM, like Zero-PaddedOFDM, Pseudo Random Prefix OFDM, etc. will be closely studied. Directly linked to the new modulations,simple channel tracking mechanisms will be examined. Another goal is to find optimum pilot structures forthe 60 GHz requirements. Considerable effort will be spent on ideas for reducing the power consumption of45

BROADWAY, IST-2001-32686 WP1-D2 version 1, date 10-10-02 Programme: ISTthe system. Here, it is intended to decrease the backoff of the A/D converter in the receiver and eventually togain one bit. Another way is to use smaller encoder structures for channel coding leading to far less costlydecoders.6. References[1] P. Nobles, “A Study Into Indoor Propagation Factors at 17 GHz and 60 GHz - Final Report,” study carriedout on behalf of the UK Radio Communications Agency, University of Swansea, Dec. 1997, available atwww.radio.gov.uk.[2] M. Fiacco, M. Parks, H. Radi, and S.R. Saunders, “Final Report: Indoor Propagation Factors at 17 and 60GHz,” study carried out on behalf of the UK Radio Communic ations Agency, University of Surrey, Aug.1998, available at www.radio.gov.uk.[3] M.P.M. Hall, L.W. Barclay, and M.T. Hewitt (Editors), “Propagation of Radiowaves”, The Institution ofElectrical Engineers, 1996.[4] L.M. Correia, P.O. Frances, “A Propagation Model for the Average Received Power in an OutdoorEnvironment in the Millimeter Wave Band,” in Proc. of VTC’94 – 44 th IEEE Vehicular TechnologyConference, Stockholm, Sweden, Jun. 1994.[5] H.T. Friis, “A note on a simple transmission formula”, Proc. IRE, Vol. 34, 1946.[6] A.M.D. Turkmani and A.F. de Toledo, “Radio transmission into, and within, mult istory buildin gs”, IEEProceedings-I, Vol. 138, No. 6, December 1993, pp. 577-584.[7] J. Borowski, S. Zeisberg, J. Hübner, K. Koora, E. Bogenfeld, and B. Kull, “Performance of OFDM andComparable Single Carrier System in MEDIAN Demonstrator 60 GHz Channel,” Conf. Rec. on the 3 ndACTS Mobile Summit, pp 653-658, Aalborg, Oct. 1997.[8] F. Poegel, S. Zeisberg, and A. Finger, “Comparison of Different Coding Schemes for High Bit RateOFDM in a 60 GHz Environment,” Proc. ISSSTA 96, pp. 122-125, Sep. 1996.[9] S. Zeisberg, B. Kull, F. Poegel, P. Pikkarainen, and A. Finger, “Channel Coding for Wireless ATM usingOFDM,” Conf. Rec. on the 2 nd ACTS Mobile Summit, pp 652-658, Granada, Nov. 1996.[10] A. A. M. Saleh and R. A. Valenzuela, “A Statistical Model for Indoor Multipath Propagation,” IEEE J.Select. Areas Communications, vol. SAC-5, No. 3, pp. 128-137, Feb. 1987.[11] N. Geng, W. Wiesbeck, “Planungsmethoden für die Mobilkommunikation, Funknetzplanung unter realenphysikalischen Ausbreitungsbedingungen”, Springer-Verlag Berlin, Heidelberg, New York, ISBN 3-540-64778-3, 1998.[12] M. Abramowith, I.A. Stegun, “Handbook of Mathematical Functions”, Dover Publ., New York, 1972.[13] J.G. Proakis, ”Digital Communications”, Third Edition, McGraw-Hill International Editions, 1995.[14] IST-BROADWAY WP3-D3 deliverable, “BROADWAY Baseband Requirements”, September 2002.[15] IST-BROADWAY WP2-D6 deliverable, ”Performances evaluation of adhoc algorithms forWLAN/WPAN networks ”, November 2002.[16] IST-BROADWAY WP4-D4 deliverable, “Architecture proposal for RF Front End with specialrequirements on dual 60GHz and 5GHz solution”, November 2002.46