air interface dimension ing tec

Upload: kenebook

Post on 07-Apr-2018

218 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/6/2019 Air Interface Dimension Ing Tec

    1/8

    Air Interface Dimensioning Techniques for Coverage

    and Capacity-Limited HSPA Based Networks

    Anis Masmoudi 1,21High Institute of Electronic and Communication (ISECS),

    University of Sfax,

    Sfax, Tunisia

    [email protected]

    Sami Tabbane 22 Mediatron Lab.,

    SupCom School, University of Carthage,

    Tunis, Tunisia

    [email protected]

    AbstractDimensioning and planning air interface of the

    evolution of UMTS networks based on HSPA technique is

    interesting for mobile operators. This paper deals with coverage-

    limited and capacity-limited dimensioning of HSPA based

    Beyond 3G networks with the necessary analytical support. The

    models established for coverage-limited case are exploited to

    generate abacuses useful mainly for initial dimensioning. Thecapacity-limited dimensioning case is also studied through

    different methods based on scheduling techniques. We suggest

    enhancing Fair Throughput scheduling technique to improve

    coverage and capacity dimensioning performance. Finally, we

    assess performance of the different introduced methods for

    HSPA dimensioning by comparing them to basic UMTS (Release

    99) dimensioning.

    Keywords-HSPA; air interface dimensioning; coverage-limited;

    capacity-limited; scheduling techniques; network performance;

    planning abacuses

    I. INTRODUCTIONThe fast challenge of mobile radio networks pushes

    operators to adapt their planning and engineering procedures

    to the new technologies and multimedia services. Radio

    interface planning and dimensioning is especially crucial since it

    remains a bottleneck for the whole process of network

    deployment. In this paper we investigate planning and

    dimensioning methods for HSPA (High Speed Packet Access)

    [1][7] Based Beyond 3G (B3G) mobile networks known as3.5G. This includes both coverage and capacity-limited cases.

    Dimensioning air interface consists in determining the number

    of radio sites to deploy through the calculation of cell size and

    capacity. Dimensioning performance is measured and

    evaluated in terms of capacity enhancement and sitesminimization or cell size maximization. Our approach is

    original since most of the literature about HSPA deals only

    with its performance through different scheduling techniques

    [7] but with neither consistent analytical study nor practical

    engineering rules for this evolution of 3G networks. In this

    paper, first we give a new definition of air interface coverage

    in HSPA based networks with the adequate proof and

    analytical support. The last established mathematical

    formulations are used to plot some figures as abacuses that

    help to dimension coverage-limited HSPA networks. To

    complete our dimensioning study, we present the capacity-

    limited case. In fact Fair Resource and Fair Throughput

    scheduling techniques are distinguished as case studies for

    capacity-limited dimensioning. We show also an introduced

    dimensioning principle based on an Enhanced Fair

    Throughput scheduling technique that we suggest in order toimprove capacity and coverage performance so as to obtain a

    more efficient dimensioning. Those scheduling based methods

    are mathematically modelled. Finally, the described

    dimensioning techniques introduced for HSPA evolution are

    evaluated by making a comparison between them and other

    methods used for the basic downlink Release 99 UMTS (with

    activated and deactivated power control, and Fair Power

    Partitioning: FPP). This comparison is accomplished through

    simulation and the assessment is performed in terms of the

    spectral efficiency and the allowed coverage.

    II. RADIO COVERAGE REFORMULATION IN HSPABASEDSYSTEMS

    The coverage concept in HSPA based UMTS (Universal

    Mobile Telecommunication System) networks is defined as the

    fact that the signal received by mobile guarantees a minimum

    required received SINR (Signal to Interference and Noise

    Ratio) or power level threshold at a given coverage

    probability as on the expressions (1)(4).

    The area coverage probability Fu is written as follows

    [8][9]:

    +=

    b

    baerf

    b

    baaerfFu

    .11.

    ..21exp)(1

    2

    12

    where

    2.

    0 rPxa

    = and2.

    log..10 10 enb = ,

    : Standard-deviation of the shadowing effect (in dB)x0: Mean threshold of the power sensitivity

    Pr: Average level of the power on the cell border

    n: Propagation coefficient

    e: Exponential constant

    The differencex0 Prrefers to the shadowing margin

    (1)

    (2)

    2010 17th International Conference on Telecommunications

    978-1-4244-5247-7/09/$26.00 2009 IEEE 196

  • 8/6/2019 Air Interface Dimension Ing Tec

    2/8

    erf: error function defined by :

    =x

    t dtexerf0

    2

    2)(

    The coverage probability Cu on the cell edge (border) is

    given by:

    [ ])(12

    1aerfCu =

    The previous definition is equivalent to the fact that a minimal

    given bit rate is guaranteed (at the same probability). That has

    the same significance as guaranteeing a given quality indicator

    parameter (called in HSPA as Channel Quality Indicator or

    simply CQI [1][5], [7]).

    In order to validate this last new definition of radio

    coverage (in HSPA), we calculate, for a given shadowing

    standard-deviation and for different values of the distance to

    node B (25 m to 2 Km with a step of 25 m), the probability

    that CQI is above the CQI threshold value CQI0 referring to a

    given service bit rate (assuming the correspondence table of

    the standard [10] between CQI and the Transport Block SizeTBS, and that instantaneous bit rate depends on the TBS

    through the Transmit Time Interval TTI denoted as TTIdelay

    whose value is specified by the standard: Eg. TTI is equal to 2

    ms in HSDPA: High Speed Downlink Packet Access). This

    calculation is repeated for four shadowing standard-deviation

    values (6 dB, 8 dB, 10 dB and 12 dB) and four services at

    different required nominal bit rates (64 Kb/s, 128 Kb/s, 384

    Kb/s and 2 Mb/s). The calculation of the probability is

    accomplished according to the following theoretical discrete

    distribution model valid in the DL of HSPA systems:

    ===

    +

    2

    )(

    2

    )(

    2

    1)(Pr

    1k

    k

    k

    dLnerf

    dLnerfkCQIobp

    where

    =

    1010

    )(

    10 101010intraratioTXinter IOffsetkCQIPI

    kd if

    +

    Offset

    CQI

    IPk

    ratio

    intraTX0 (Iinter

  • 8/6/2019 Air Interface Dimension Ing Tec

    3/8

    independent of the shadowing standard-deviation value

    (corresponds to the average power received by the node B

    which is the same as the one obtained with a deterministic

    propagation model).

    Figure 1. Coverage probability versus distance to node B for different services

    (having different nominal bit rates) and shadowing values

    III. ABACUSES FIGURES FORCOVERAGE-LIMITEDDIMENSIONING

    In this paragraph, we show examples of abacuses figures

    generated on the basis of the mathematical study achieved in

    the previous paragraph. So, those abacuses are valid for the

    coverage-limited case. The capacity-limited case is carried out

    in the next two paragraphs.

    Fig. 2 & 3 provide examples of abacuses respectively for

    categories 10 (having the highest offered bit rate) and 1 (withthe least provided bit rate) of mobile terminals [5] allowing

    to dimension a HSPA network by determining the maximum

    cell radius versus minimum required bit rate (guaranteed at a

    given cell coverage probability).

    For such abacuses, the specification of bit rate and

    required coverage probability is enough to find the maximum

    allowed cell radius by the help of both figures (Fig. 2 shows

    the complete range of bit rates: until 5.7 Mb/s, and Fig. 3

    extracts the part of abacuses whose bit rate doesnt exceed

    1 Mb/s for a better view).

    The discrete aspect of the abacuses translates the effect ofAdaptation in Modulation and Coding affecting with adiscrete manner the bit rate supported by the link (limited

    numbers of CQIs thus well determined Transport Block

    sizes).

    Note that at a given fixed coverage probability, themaximum radius doesnt exceed the maximum value

    corresponding to CQI = 1 (peripheral border of the cell

    offering the minimum offered bit rate).

    The higher the required coverage probability is, thesmaller the maximum cell radius (Dimensioning with

    stricter conditions).

    Figure 2. Abacus of the cell size versus offered bit rate (High capability

    terminals) for different coverage probability values

    Figure 3. Abacus of the cell size versus offered bit rate (Low capability

    terminals) for different coverage probability values

    Fig. 4 is similar to Fig. 2 except by taking the shadowing

    standard-deviation as parameter instead of the required area

    coverage probability. It shows dimensioning abacuses of

    coverage limited bit rate (bit rate guaranteed in 95% of the

    cell). Note in particular that the dimensioned radius is lower

    for higher shadowing standard-deviation values. This is

    effectively logical since the shadowing margin to include in

    the link budget increases with the standard-deviation.

    Fig. 5 summarizes entirely the abacuses of the threefigures, Fig. 2 to 4, while combining both the following

    parameters: coverage probability and shadowing standard-

    deviation. The same remarks can be extracted with a global

    vision of the impact of both parameters together (shadowing

    standard-deviation and area coverage probability). In

    particular, the smaller the coverage probability (case of 70%),

    the less the impact of shadowing standard-deviation is

    important on the dimensioned cell size (due to the impact of

    coverage probability on the shadowing margin). The abacuses

    of Fig. 6 allow to dimension the cell radius (coverage limited)

    198

  • 8/6/2019 Air Interface Dimension Ing Tec

    4/8

    versus required coverage probability while knowing the

    required bit rate and the shadowing standard-deviation.

    It is evident that the higher the required bit rate, thesmaller the dimensioned cell radius (because the CQI at

    the edge of the cell is important).

    The impact of the shadowing standard-deviation on thedimensioned cell size diminishes with the increase of the

    maximum guaranteed service bit rate. In other words,shadowing effect is more important with a lower bit rate

    service. This is due to the fact that the required CIR of a

    low bit rate service is below that of a higher bit rate, thus

    more sensitive to propagation channel variations due to

    shadowing.

    Figure 4. Abacus of the cell size versus offered bit rate (High capability

    terminals) for different shadowing standard-deviation values

    Figure 5. Abacus of the cell size versus offered bit rate (High capability

    terminals) for different coverage probability and shadowing standard-

    deviation values

    The more the coverage probability increases and

    approaches to 100%, the more the dimensioned radius

    decreases asymptotically (near the coverage probability of

    100%). This is due to the shadowing effect requiring an

    infinite margin to reach a coverage probability of 100% (not

    reached in practice).

    Figure 6. Abacus of the cell size versus the area coverage probability for

    different services (different nominal bit rates) and different shadowing

    standard-deviation values

    IV. SCHEDULING METHODS BASED DIMENSIONING(CAPACITY-LIMITED)

    After dealing with the coverage-limited case, we present in

    this section the capacity-limited dimensioning. Since the

    scheduling is the main bottleneck for this case study, our

    dimensioning methods are based on different scheduling

    techniques. The last sub-section is a suggestion of a new

    introduced scheduling method that can enhance dimensioning

    performance of HSPA based UMTS networks.

    A. Fair Resource scheduling based dimensioningHSPA capacity is limited either by the number of codes

    HS-PSCH (High Speed Physical Shared Channel) [1][5] or

    by the total node B power. The cell size in the first case (Rc) is

    the greatest radius R verifying the equality in the following

    condition of HSPA code limitation:

    15)()(

    21

    21

    221 00

    0

    + ++

    + kkki

    iii rRnrrn

    where ri = ri,min and ri+1 = ri+1,min = ri,max, ri and ri+1 denote

    respectively the lower and upper limits of the range of the

    sub-cell (ring) having a CQI equal to CQIi (same modulation,

    coding rate and number of physical shared channels or codes

    ni used thus having the same corresponding block size TBSi

    according to the adequate table of [10]), and such that

    10 +kCQI is the CQI of the ring including the border of the cell

    having the size R. The number 15 in the numerator of the

    second term of (7) refers to the number of the codes

    (physical shared channels) allocated to HSPA (the value of 15

    is taken here as an example of the maximum number allocated

    for HSDPA until the end of this paper). In the second case

    (node B power limited capacity), the sizeRp of the cell is such

    that:

    (7)

    199

  • 8/6/2019 Air Interface Dimension Ing Tec

    5/8

    10102 1010TotTX PP

    pR =

    where PTot is the total power of the node B. The capacity

    limited cell radius is: ),(min pccap RRR = .

    By assuming the dimensioning and the bit rate limited only

    by coverage and link quality but not by capacity, the cell

    radius R will depend on minimum bit rate Rmin of theconcerned service (by taking always the assumption of the

    Fair Resource as the used scheduling technique). The bit

    rateRmin refers to the Transport Block Size TBS0 as follows:

    TBS0 =i

    min { TBSi / TBSi tables [10]

    and TBSiRminTTIdelay }

    In this case, and by accomplishing a dimensioning without

    codes multiplexing, the cell radius R will be the distance r0

    referring to TBS0 or exactly the minimum radius of the

    internal CQI ring (core) corresponding to the maximum CQI

    that radio condition and terminal capability allow. In fact,

    TBS0 refers to a CQI0 (the maximum allowed value) [10] to

    which refers a minimum SINR value (SINRmin) computed as

    follows:

    )( 0min OffsetCQICQISINR ratio =

    From the last value, we can extract the maximum

    attenuation from SINR definition expression, and thus the

    corresponding radius r0. If we assume the total power

    transmitted by the node B constant (not depending on the

    number of mobiles served), then the computation of the cell

    radius can be made immediately. Yet in reality, the power

    transmitted depends on the traffic, and is proportional to the

    number of mobiles served in the cell (since individual transmit

    power is constant), hence the transmitted intracellular power is

    proportional to the number of active mobiles of the cell, so it

    depends on the cell radius. Therefore, for more accuracy and

    precision, we must apply an iterative algorithm until its

    convergence to the cell radius. The last parameter shouldnt

    exceed, in any case and whatever the service and its required

    bit rate, the value of the radius referring to the minimum

    allowed value of CQI (equal to 1).

    In contrast, by assuming the dimensioning is capacity-

    limited (either by number of codes allocated to HSPA if

    15)(2

    0 0>

    R

    drrrnd where n(r) is the number of codes

    referring to the CQI of a virtual mobile at a distance rto thenode B or by the total available power of the node B if

    10102 1010TotTX PP

    pR > ), or in other words assuming traffic

    density above some value, then the bit rateRu guaranteed per

    user (assuming always a uniform traffic and the use of the

    Fair Resource scheduling technique) can be written as

    follows:

    2

    2

    idelay

    capiu

    rTTI

    RTBSR

    =

    where TBSi is the Transport Block Size referring to the ring i

    of the CQI = i (at the border of the cell) having an outer radius

    ri+1 >Rcap (capacity limited cell size) in case of one service. In

    this last case, in order to guarantee a minimum bit rate Rmin at

    the border of the cell, the Transport Block Size TBS0 at the

    cell border should be given by:

    TBS0 =i

    min { TBSi / TBSi tables [10]

    and TBSiRmin 2cap

    delay

    R

    TTIri+1 }

    In this case, the cell radius can be concluded from TBS0 as

    for the previous case of coverage limited dimensioning; yet

    the expression (12) above ofTBS0 depends on the cell radius

    (through ri+1), then the planner should apply an iterative

    process or by dichotomy to converge to the exact cell radius

    or extract it from some mathematical formula referring exactly

    to the required bit rate Rmin. Thus, if2

    12min1 ++< i

    cap

    i rR

    TTIRTBS

    (where i is such that TBSi = TBS0), then the dimensioned cellsize Rdim is equal to the limit size of the CQI ring with TBS0

    (i.e. ri+1), elsemin

    1dim

    RTTI

    TBSRR icap

    = + (ri+1Rdim < ri+2).

    B. Fair Throughput scheduling based dimensioning

    This paragraph provides expressions of maximum bit rate

    per user ensured by Fair Throughput scheduling technique

    with and without consideration of codes multiplexing. Thus

    we can determine Fair Throughput dimensioning procedure

    with its analytical support.

    Assuming TTIdelay the Transmit Time Interval duration,and TBS1, TBS2, TBS3,, TBSi, are the respective

    Transport Block Sizes of each of the users in the cell

    according to their CQIs. Thus the maximum ensured bit rate

    by each of the users (without codes multiplexing) can be

    written as follows:

    =

    j jdelay

    ens

    TBSTTI

    R1

    1

    We can easily check that iTTI

    TBSR

    delay

    iens ; . In

    particular, Rens is always below or equal to the most limitingcoverage bit rate at the cell border.

    The bit rate in expression (13) is thus the minimum

    guaranteed bit rate independently of the number of codes and

    multi-codes available in HSPA (It refers to the minimum

    required number of codes always below the number of

    available HSPA shared codes assumed to be equal to 15).

    Otherwise, we can ensure a user bit rate (in Fair

    Throughput) above that given by (13) by using more OVSF

    (8)

    (9)

    (10)

    (11)

    (12)

    (13)

    200

  • 8/6/2019 Air Interface Dimension Ing Tec

    6/8

    (Orthogonal Variable Spreading Factor) HSPA codes (with

    codes multiplexing).

    By considering codes multiplexing, we establish that the

    maximum ensured bit rate per user can be written as follows:

    ( )

    =

    ii

    idelay

    FTens

    TBS

    nTTI

    R15

    where ni is the corresponding number of codes referring to the

    useri position within the cell (given by 3GPP standard: Third

    Generation Partnership Project[8]).

    In the uniform traffic case study, (14) becomes by

    reasoning on the different CQI rings of one cell with a

    uniform area density as follows:

    ( )

    =

    =

    +i

    iii

    idelay

    R

    delay

    FTens

    rrTBS

    nTTI

    drrrTBS

    rnTTI

    R

    )(

    15

    )(

    )(2

    15

    221

    0

    assuming that ri and ri+1 are the radii of the CQI ring borders

    (ri+1 = ri+1,min = ri,max). Hence, if minimum bit rate value Rmin to

    guarantee for the service is known, the maximum allowed area

    density of users max can be determined as follows:

    =

    =

    +

    i

    ii

    i

    idelay

    R

    delay

    rrTBS

    nRTTI

    drrrTBS

    rnRTTI

    )(

    15

    )(

    )(2

    15

    221min

    0min

    max

    The planner should therefore determine the CQI ring m

    referring to bit rate Rmin. So, the dimensioned cell size Rdim

    guaranteeing a minimum bit rate Rmin can be extracted from

    (15) as follows: