RADIO NETWORK SOLUTIONS FOR HIPERLAN/2
`
`Johan Torsner and Goran Malmgren
`
`Ericsson Radio Systems AB, S—l6480 Stockholm, Sweden
`
`Abstract — HIPERLAN/Z is a wireless LAN system cur-
`rently being specified by ETSI BRAN for use in the 5
`GHz band. The system uses a channel spacing of 20
`MHz and will provide data rates up to 54 Mb/s (2.7
`bit/s/Hz). The typical application is communication
`between portable computers and a fixed data network.
`The high data rates together with the limited number of
`available
`frequency carriers
`(approximately nine)
`implies that deploying a full coverage, multi cell system
`is challenging. A potential radio network solution based
`on automatic frequency assignment, link adaptation and
`multi beam antennas is presented in this paper. Simula-
`tion results show that the presented solution provide a
`high peak data rate and system throughput.
`
`I.
`
`INTRODUCTION
`
`ETSI BRAN (European Telecommunications Standards
`Institute Broadband Radio Access Networks) develops
`standards and specifications for broadband radio access
`networks that cover a wide range of applications
`intended for different frequency bands. A system cur—
`rently being specified by BRAN is HIPERLAN type 2
`(ii/2) which will provide high speed (up to 54 Mb/s
`data rate) communications between portable computing
`devices attached to an IP, ATM or UMTS backbone net-
`work. H/2 will be capable of supporting multimedia
`applications and the typical operation environment is
`indoor with restricted user mobility. In Europe, the fre—
`quency band 5.15—5.25 GHz currently allocated to fill,
`will most likely be available also for PM Systems.
`Additional spectrum above 5.25 GHz may also be avaiL
`able for H/Z systems, which than will share the spec-
`trum with earth exploring satellites and/0r
`radar
`equipment. “/2 can also be deployed in the ll-Nl] hand
`in the US. We here assume that 200 MHz uninterfered
`
`bandwidth can be used for a H/2 system in both Europe
`and USA. Ongoing discussions may also result in ”/2
`being deployed in a similar frequency band in Japan.
`
`In order to support the required peak data rate of 25
`Mb/s, a channel spacingibandwidth of 20 MHz has been
`adopted in ETSI BRAN. With 200 MHz bandwidth,
`typically nine frequency carriers are available if a guard
`band of one channel is applied. Since H/2 will be used
`
`0-7803—5565—2/99/$10.00 © 1999 IEEE
`
`1217
`
`in an unlicensed band, it may occur that an operator can
`not utilize the entire frequency band due to interference
`from other operators in the close vicinity. Furthermore,
`a high radio quality is required to fulfil the peak data
`rate requirement of 25 Mb/s.
`
`The H/2 system is likely to be deployed in a wide range
`of environments, such as office buildings, exhibition
`hails, airports, industrial buildings and outdoor deploy—
`ment. Altogether, this implies that deploying a full cov—
`erage, multi—cell system is challenging. It is clear that
`the system must be able to adapt to different propaga-
`tion and interference environments.
`
`In this paper it is shown that a radio network solution
`based on automatic frequency planning together with
`link adaptation achieves a high peak data rate and sys-
`tem throughput. Further extensions using multi beam
`antennas seem appropriate in line of sight (LOS) envi-
`ronments. With the proposed solution it is possible to
`deploy full coverage, multi cell systems in both non line
`of sight (NLOS) and LOS environments.
`
`H. REQUIREMENTS ON HIPERLAN/Z
`
`The most important requirements on the H/2 system [i]
`are summarized here:
`
`- Radio Range: H/Z shall provide a range of 30 min a
`typical indoor environment and up to 150 m in a typ-
`ical outdoor or large open indoor (e.g. large factory
`hall, airport) environment.
`' Peak Data Rate: H/2 shall provide a peak input data
`rate of at least 25 Mb/s to the PHY layer.
`
`- Capacity and Coverage: In a single operated multi-
`cell environment the average system throughput (per
`AP) should at least be 20 Mb/s (input to the PHY
`layer). In 95% of the area the mobile terminal (MT)
`should be able to provide at least 4 Mbls as input to
`the PHY layer.
`Further, it is desired that no manual planning of the
`radio network shall be needed, is. frequency planning
`and adjusunent of radio parameters shoald be per-
`formed automatically by the system.
`
`Microsoft Corporation
`
`Exhibit 1014-00001
`
`Microsoft Corporation
`
`

`

`(BCH) is transmitted sequentially on each antenna ele-
`ment. Immediately after the BCH an announcement is
`made on all active antenna elements (i.e., those who
`will carry data in the upcoming MAC frame) conveying
`the structure of the MAC frame. This is done through
`the Frame Control cHannel
`(FCH) which grants
`resources for different connections. Note that one MT
`
`could have multiple connections, where each connec-
`tion carries a certain type of traffic. This enables the
`DLC to act accordingly to the tequired QOS.
`Fix Mason [1——2ns)
`. I
`
`.
`
`ii
`
`VaryingInnis!
`Varyinghomer
`Gall trains per connection
`Celltrains per connection
`Contention
`{duwnlnkl
`(uphill)
`Phase
`
`Broadcast
`Phase
`
`
`
`III. SYSTEM DESCRIPTION
`
`A typical [1/2 system consists of a number of access
`points (APs) connected to a backbone network, e.g. an
`ethernet LAN. An AP can use an omni antenna, a multi
`beam antenna, or a number of distributed antenna ele-
`ments. The system supports mobility between access
`points on the same backbone network, i.e. handover is
`made between APs.
`Mull: noun
`antenna:
`
`Dittubutai
`antenna:
`
`I
`
`R
`
`,
`
`We,
`
`
`Mcbtltl Termnal
`a.u. Laplcp comma:
`
` l
`
`
`!
`
`'anom crass
`
`it somea: i.
`
`Figure 2. MAC frame for HIPERLAN type 2
`
`The resource grants (RG5) in the FCH allocates a part
`of the frame for a certain connection. An absolute
`
`pointer is used together with a specification on the
`number of Short transport CHannels (SCH, for resource
`request and ARQ ACKINACK messages) and Long
`transport Cl-Iannels (LCH, for user data). Furthermore.
`the PHY mode is also specified per connection. The
`MAC frame length is fix, but the length is not yet
`decided. Most likely, the frame length will be 1—2 ms.
`
`IV. RADIO NETWORK
`
`Dynamic Frequency Selection (DFS)
`
`As no manual planning of the system shall be needed,
`the system must automatically allocate frequencies to
`each AP for communication. It is assumed that the
`
`dynamic frequency selection (DFS) reacts on slow
`changes in the radio environment, e.g. when a new AP
`is installed. The DFS does normally not react to move-
`ment of users or short term changes in the traffic distri-
`bution. The frequency selection is based on filtered
`interference measurements performed by the AP and its
`associated MTs. The communication will be halted
`
`when the AP performs interference measurements and
`no traffic will be scheduled to MTs that are ordered to
`
`measure. The DFS allows several operators to share the
`available frequency spectrum.
`
`"wt,
`mm”? It'd-J
`. m
`
`
`
`Figure 1. System Overview HIPERLAN type 2 (H12)
`
`Physical layer
`
`strong alignment has been achieved
`Currently a
`between three standardisation bodies,
`[EEE 802.11a
`(U.S), ETSI BRAN (Europe) and MMAC (Japan) on
`the PHY layer. All have adopted OFDM with 64 sub-
`carriers, where 48 subcarriers are modulated for data
`transmission and 4 subcarriers are used for pilot signals.
`The remaining 12 subcarriers are set to zero. in order to
`support link adaptation a number of PHY modes have
`been defined, where a PHY mode corresponds to a sig-
`nal constellation and code rate combination. These are
`shown in Table l. The demodulation is coherent.
`
`Details about the proposed physical layer can be found
`in [2].
`
`Data Link Control (DLC)
`
`A centralised controlled Data Link Control (DLC) layer
`is adopted in H/Z, which means that the access point
`controls how the resources are allocated in a MAC
`
`frame. Each MT requests capacity in future MAC
`frames when it has some data to send. This request is
`sent either in a Short transport CHannel (SCH), which
`is piggy hacked to the uplink data transmission, or in
`the random access slots. The general frame shown in
`Figure 2 can be utilised for different radio network
`solutions. In the switched muiti beam or distributed
`
`antenna configuration the Broadcast Control cHannel
`
`1218
`
`Microsoft Corporation
`
`Exhibit 1014-00002
`
`Microsoft Corporation
`
`

`

`Link Adaptation (LA)
`
`Multi beam antennas
`
`The radio quality, in terms of C/[ is highly dependent on
`the radio environment. The CI] level also changes over
`time, depending on the traffic in surrounding cells. in
`order to maximize the link throughput, a link adaptation
`scheme is used. where the PHY mode can be adapted to
`the time varying link quality. Link adaptation has been
`studied in e.g. [3]. Due to the rapid time variations of
`the link quality, it is deemed difficult to assign PHY
`mode based on the momentary link quality. Instead, the
`PHY mode is assumed to be adapted at a time interval
`significantly larger than the MAC frame duration (e.g.
`5—10 MAC frames).
`
`The link quality is estimated from measurements. e.g.
`C/I estimates from the training sequence in each MAC
`frame. Link adaptation is used in both uplink and down—
`link. where the AP measures the link quality on the
`uplink and signals to the MT which PHY mode to use
`for uplink communication. In a similar way, the MT
`measures the link quality on the downlink and signals a
`PHY mode suggestion to the AP for downlink commu-
`nication. However, the AP is responsible for the final
`PHY mode selection for both uplirtk and downlink.
`
`Multi beam antennas have been discussed in ETSI as a
`
`means to improve the link budget and increase the CII
`ratio in the radio network. The MAC protocol and the
`frame structure in HI2 allow multi beam antennas with
`up to 7 beams to be used. The beam selection is MT ini-
`tiated, i.e. each MT requests which beam that should be
`used for downlink communication, based on measure-
`ments on the broadcast fields that are transmitted
`
`if desired,
`sequentially in each beam. The AP can,
`select another beam for the uplink communication.
`40
`
`35
`
`30A .
`
`.,
`
`.
`
`25
`21:-
`
`'5
`
`I“
`
`,
`
`i
`
`nasal Lmll Abaptwon
`
`metasMbpsl
`
`..
`
`Mila'IMhml
`
`M4 -'18Mhi'l
`
`A
`
`'
`
`'1
`
`'
`
`Morthu‘sl"
`M1 16 Maps]
`
`
`
`
`
`
`
`mam
`
`M1
`
`
`
`
`
`
`
`7
`
`
`
`Table 1 Data rate (input to PHY layer), modulation and code rate for
`the modes Ml—M’t‘. Mode M7 is optional.
`
`Link simulations have provided throughput as a func-
`tion of CII for the defined PHY modes on a channel
`
`with 50 ns delay spread, see Figure 3. The results are
`only valid for ideal selective repeat ARQ. If no link
`adaptation is used in the system, i.e. a single PHY mode
`is used, the parameters for this mode should be selected
`to fulfil the requirements on range and data rate men,
`tioned in Section 11. However, no single PHY mode will
`fulfil both requirements. In the performance analysis.
`PHY mode M4 (18 Mbls) is used as reference, as this is
`the mode with highest data rate among the modes that
`fulfils the requirement of 30 111 range in a typical office
`environment. It is assumed that the range requirement is
`fulfilled if the channel packet error rate is below 10% in
`a noise limited environment.
`
`Figure 3, Throughput as a function of CH for the six mandatory
`PHY modes. The thick line indicates the throughput for ideal link
`adaptation.
`
`V. PERFORMANCE ANALYSIS
`
`Modelling of the radio network
`
`In this paper, an algorithm for distributed frequency
`assignment proposed in [4] has been used to develop a
`frequency plan prior to the simulations. The frequency
`plan is fix during the simulations.
`
`The link adaptation is modelled by updating the PHY
`mode every 10th MAC frame. The position of the
`receiver is fixed during the update interval, and the
`interferers are placed randomly each MAC frame. In
`each update interval the throughput for all PHY modes
`is estimated, and the mode achieving the highest
`throughput is used during the next update interval.
`
`The beam selection for multi beam antennas is mods
`
`elled by using the beam with the highest received
`downlink signal strength, i.e.
`lowest pathloss. As the
`interference, arising from other cells. is independent of
`the beam selection,
`this approach will also maximise
`the received downlink C/I ratio. The same beam is used
`
`for both uplink and downlink communication. In the
`simulations the system performance for a system with
`
`
`
`
`
`
`
`1219
`
`Microsoft Corporation
`
`Exhibit 1014-00003
`
`Microsoft Corporation
`
`

`

`omni antennas is compared to the performance for a
`system using multi beam antennas with 6 beams. The
`multi beam antenna has 90 degrees 3 dB lobewidth and
`6 dBi antenna gain.
`
`Simulation Environment
`
`For radio network evaluations. two "typical" environ—
`ments have been considered, an office building and an
`open exhibition hall. The office building is dominated
`by NLOS propagation, and the exhibition hall consists
`
`
`
`
`
`
`only of LOS propagation. l u u . «-
`
`
`
`
`
`
`
`
`
`
`
`Figurc 4. One floor in the simulated office building and the post
`tions of the Al’s (*).
`
`The office scenario included a building with five floors
`and a large number of MTs. Each floor was covered by
`eight Al’s which were located at the same position on
`every floor, see Figure 4. The average path loss (Le.
`without fast fading) between :1 MT and a AP was calcu—
`lated using the extended Keenan-Motley model {5]
`which includes the attenuation by the distance, walls
`and floors in the direct propagation path (the straight
`distance between the MT and AP). The model is based
`on 900 MHz propagation measurements. The frequency
`dependent term in the model as well as the wall and
`floor attenuation is however modified for 5 GHz.
`
`The exhibition hall scenario consisted of a large build-
`ing with one floor and no inner walls. The hall was cov-
`ered with l6 access points placed in a rectangular grid
`with a site to site distance of 60 m. It was assumed that
`
`very high capacity is needed in this environment.
`Therefore, the large number of installed access points.
`In this scenario, a LOS propagation model was used.
`Furthermore, lognormal fading with a standard devia—
`tion of 2 dB was added in order to model shadowing
`caused by e.g. people moving around in the environ-
`mentThe mobiles were randomly placed in the build-
`ings according to a uniform distribution. The simulation
`technique was static: each iteration corresponded to a
`
`1220
`
`traffic situation unrelated to the previous one, i.e. "snap-
`shots" were taken of the situation in the building. The
`system was analysed at a full
`load situation,
`i.e. all
`access points and terminals were active. Only results
`from the downlink is presented in this paper, as the
`downlink is assumed to carry the majority of the data
`traffic. The downlink interference arises from other APs
`
`as well as from other MTs, due to the unsynchronised
`TDD MAC frame. External interference was modelled
`
`by assuming that a second operator is present, using
`four of the nine carriers. The important simulation
`parameters are summarized in table 2.
`
`Number of frequencies
`
`System load
`
`Adjacent channel suppressron
`
`
`
`
`
`
`
`i Antennas (omni)
`
`
`
`Floor attenuation (office building)
`
`
`
`9 I S
`
`
`
`
`
`[00 %
`
`30 dB
`
`mm
`
`M...
`
`
`
`
`
`mamm-
`m
`
`20 dB
`
`Table 2. Important simulation parameters
`
`VI. SIMULATION RESULTS
`
`Figure 6 shows the downlink CH for a system with
`omni antennas. The CII is shown for 9 and 5 frequency
`reuse, corresponding to a single operator and a two
`operator scenario respectively. These distributions are
`the basis for the system throughput performance evalu-
`ation below.
`10:
`
`:Exhibitifin'i
`
`.
`
`. Office
`
`a
`
`i\ ’ ’enviroitmcnt
`
`Downlink Cl] with 9 and 5 frequency reuse for a system with court
`ammnas.
`
`Microsoft Corporation
`
`Exhibit 1014-00004
`
`Microsoft Corporation
`
`

`

`Note that the Cl] varies greatly between the exhibition
`hall and the office environmentThe impact on C/I with
`multi beam antennas is shown in Figure 7. The 10 per,
`Gentile of C/I is improved 2—3 dB in the exhibition hall.
`However, the mean C/I is improved more than 5 dB,
`which gives an significant improvement of throughput.
`
`
`
`on [on]
`
`spectrum in the exhibition hall.
`
`27.7 Mb/s 34.6 Mb/s
`
`Exhibition hall
`
`Office
`
`5 frequencies
`16.7 Mb/s
`
`33.7 Mb/s
`
`Table 4; System throughput with iink adaptation in a system with
`omni antennas
`
`Finally, the system throughput when multi beam anten-
`nas are used in the exhibition hall is shown in Table 5.
`
`The system throughput is increased, but the requirement
`is still not fulfiled in the two operator case
`
`_——
`Exhibition hall
`29.4 Mb/s
`18.9 Nib/S
`
`Table 5: System throughput with link adaptation in a system with
`multi beam antennas.
`
`Figure 5. Downlink Cll with 9 and 5 frequency reuse in the exhibi-
`tion hall for a system with omni antennas and multi beam antennas (6
`beams) respectively.
`
`VII. CONCLUSION
`
`By combining the previous shown link simulation
`results (Figure 3) with the C/I distributions (Figure 6
`and 7), the system throughput can be derived. The link
`quality (C/I) of each user is mapped on a corresponding
`throughput. The system throughput is then calculated as
`the mean throughput for all users, which corresponds to
`a scheduling strategy where each user is allocated the
`same amount of radio resources, in terms of transmitted
`OFDM symbols.
`
`
`The C/I level differs significantly between different
`environments. In the office environment, which is dom-
`inated by NLOS propagation, the system throughput is
`relatively high even without link adaptation or multi
`beam antennas. in the exhibition hall, which consists of
`LOS propagation, the system throughput is lower. How-
`ever, if link adaptation and multi beam antennas are
`used, the system throughput requirement of 20 Mb/s
`can almost be fulfilled even when two operators share
`the available spectrum. Notc that the DLC overhead is
`larger when multi beam antennas are used. The over—
`head also depends on the MAC frame length. There-
`fore,
`the results for multi beam antennas may differ
`slightly from the values presented here, as no DLC
`overhead is included in the simulations.
`
`
`
`Exhibition hall
`
`Office
`
`9 frequencies
`16.8 Mb/s
`
`5 frequencies
`12.9 Mbls
`
`17.8 Mb/s
`
`175 Mb/s
`
`Table 3: System throughput for PHY mode M4 (18 Mo's) m a system
`with omni antennas
`
`VIII. REFERENCES
`
`Note that the system throughput calculation does not
`consider the bursty nature of packet data traffic. It can
`be seen as the achievable data rate during e.g., a large
`file transfer. The system throughput for the fixed PHY
`mode M4 in a system with omni antennas is shown in
`Table 3. The throughput in the office environment is
`very close to the maximum value 18 Mbls which is nat-
`ural since the C11 level in the office environment is very
`high. both for 9 and 5 frequency reuse. In the exhibition
`hall, the interference from other APs degrade the per-
`formance slightly. When link adaptation is used (Table
`4) the system throughput is increased significantly. The
`system throughput is well above the requirement of 20
`Mb/s in all cases except when two Operators share the
`
`1221
`
`for
`"Requirements and Architectures
`[1] ETSI BRAN,
`Mreless Broadband Access", DTRJ’BRAN—GIODO2
`V0.13
`
`'Modulalion for HIPERLAN Type
`[2] Uwe Dettmar et 21.1.,
`2", to appear in Proc. IEEE 49th VTC‘99
`131 A. Furuskar et al, “System Performance of EDGE; a
`Proposal for enhanced Data Rate in Existing Digital Cel—
`lular Systems", in Proc. IEEE VTC’98
`{4} M. Almgren et al., “Slow Adaptive Channel Allocation
`for Automatic Frequency Planning”, in ICUPC 96
`[5] C.Tiimevik et al., "Propagation Models, Cell Planning
`and Channel Allocation for Indoor Applications of Cel~
`lular Systems", in ProclEEE 43rd VTC’93.
`
`Microsoft Corporation
`
`Exhibit 1014-00005
`
`Microsoft Corporation
`
`