LINK PERFORMANCE EVALUATION FOR E

Document Sample
LINK PERFORMANCE EVALUATION FOR E Powered By Docstoc
					       LINK PERFORMANCE EVALUATION FOR EGPRS WITH MULTIPLE ANTENNAS
             W. C. Freitas Jr1 , F. R. P. Cavalcanti1 , C. C. Cavalcante2 , D. Zanatta Filho2 and A. L. F. de Almeida1
                1
                                                                                        a
                GTEL-UFC: Wireless Telecom Research Group, Federal University of Cear´ , Fortaleza, Brazil
                                                  URL:www.gtel.ufc.br
                                              {walter,rod,andre}@gtel.ufc.br
            2
              DSPCom: Digital Signal Processing for Comm. Lab., State University of Campinas, Campinas, Brazil
                                     URL:www.decom.fee.unicamp.br/ dspcom.ufc.br
                                        {charles,daniloz}@decom.fee.unicamp.br



Abstract - Transmission diversity schemes have recently                As internet traffic is expected to be asymmetric, with higher
emerged in wireless systems as an attractive solution in order      data rates on the forward link, the issue of transmit diversity
to mitigate fading effects. Space-time block codes (STBC)           becomes important. This importance arises because the link
with two antennas can provide similar order diversity as            quality control of EGPRS will translate into throughput gain
maximal-ratio receiver combining (MMRC). In this paper              any gain perceived at the link-level, such as a diversity gain.
transmit diversity scheme using STBC with two and four an-          Some possibilities of transmit diversity are then investigated
tennas is applied for the EDGE/EGPRS system and its results         in this work and their gains evaluated.
are evaluated in a interference-limited scenario. We jointly           In [1], Alamouti has proposed a simple transmission diver-
compare this strategy of multiple antennas with incremental         sity scheme using space-time block codes (STBC) with two
redundancy (IR), a technique for link quality control (LQC)         antennas. The obtained diversity order is similar to apply-
in EDGE.                                                            ing a maximal-ratio receiver combining (MMRC) with two
Keywords - EDGE, Space Time Block Codes, Incremental                antennas at the receiver. This scheme requires no bandwidth
Redundancy                                                          expansion, as redundancy is applied in space across multi-
                                                                    ple antennas. In [2], an extension to a scheme similar to
                      I. I NTRODUCTION                              Alamouti’s STBC for more than two antennas in transmis-
   The enhanced general packet radio service of EDGE                sion is proposed using the Theory of Orthogonal Design [3].
(EGPRS) combines multilevel modulation and an efficient              Alamouti’s scheme is indicated for flat fading channels, there-
link quality control in order to reach the high data rates de-      fore without intersymbol interference (ISI). A generalization
manded by third generation (3G) systems. With those char-           for channels with ISI is given in [4].
acteristics EGPRS arises as a natural evolutionary path for            In this work we study the performance of multiple antennas
two TDMA-based second generation (2G) systems, namely               applied in the context of the EDGE system. The structure uses
GSM and IS-136. In the physical layer of EGPRS there                STBC to provide transmission diversity with two and four an-
are nine modulation and coding schemes (MCSs), MCS-                 tennas. We also evaluate the performance of STBC with a
1 through MCS-9. Four of them use GMSK modulation                   more efficient LQC strategy such as IR. Firstly, in section II e
(MCS-1 through MCS-4) and five use the 8-PSK multilevel              describe the space-time codes and the techniques involved to
modulation (MCS-5 through MCS-9). As link quality con-              provide transmit diversity. In section III the EDGE/EGPRS
trol strategies, EGPRS uses link adaptation (LA) and incre-         simulator model is presented. After that, in section IV the
mental redundancy separately or jointly. Through this dy-           LQC strategy used in EDGE, incremental redundancy (IR), is
namic adaptation in agreement with the link quality, one may        explained. In section V the simulation results are presented
choose between transmission rates and protection to maxi-           and discussed. Finally in section VI we establish our conclu-
mize throughput.                                                    sions.
   The fading effect is one of the most important drawback
factors of the maximum data rates reached by wireless com-                              II. S PACE T IME C ODES
munication systems. To mitigate fading some diversity strate-          Space-time codes (STC) use channel coding techniques
gies are usually provided, among which we can highlight:            combined with multiple transmit antennas to increase data
time, frequency and space diversity. Currently, in 2G net-          rates over wireless channels. STC introduces temporal and
works, the most common strategies used for receive diversity        spatial correlation into signals transmitted from different an-
include: space diversity - multiple antennas in the base sta-       tennas, in order to provide diversity at the receiver, and coding
tions are employed to provide receive diversity; time diversity     gain over an uncoded system without sacrificing bandwidth
- channel coding with interleaving; and frequency diversity -       [1].
frequency hopping (FH) for TDMA systems and RAKE re-                   Two techniques widely used for STC are: space-time block
ceiver for CDMA systems.                                            codes (STBC) and space-time trellis codes (STTC). In STTC
when the number of antennas is fixed, the decoding complex-                                   Table 1
ity of STTC (measured by the number of trellis states at the                 Modulation and Coding Schemes in EGPRS.
decoder) increases exponentially as a function of the diversity
level and transmission rate. In addressing the issue of decod-                MCS           User       Code      Header       Blocks
ing complexity, Alamouti [1] discovered a remarkable STBC                                   Rate       Rate       Code         per
scheme for transmission with two antennas [5]. In this work                                [Kbps]                 Rate        20ms
the simulation results will only consider STBC.                         9(8-PSK)            59.2         1.0      0.36          2
                                                                        8(8-PSK)            54.4        0.92      0.36          2
A. Alamouti’s Space Time Block Codes (STBC)                             7(8-PSK)            44.8        0.76      0.36          2
                                                                        6(8-PSK)            29.6        0.49       1/3          1
   Alamouti proposed a simple scheme of transmission diver-
sity with two transmit antennas in which two symbols s1 and             5(8-PSK)            22.4        0.37       1/3          1
s2 are simultaneously transmitted by different antennas at a            4(GMSK)             17.6         1.0      0.51          1
given symbol period, where s1 is the signal transmitted by              3(GMSK)             14.8        0.85      0.51          1
antenna one and s2 is the signal transmitted by antenna two.            2(GMSK)             11.2        0.66      0.51          1
In the next symbol period, antenna one transmits −s∗ and                1(GMSK)              8.8        0.53      0.51          1
                                                       2
antenna two transmits s∗ . It is assumed that the channel is
                         1
constant during two periods of consecutive symbols. Con-
sidering this space-time coding, the received symbols in two      order to deal with this drawback some changes in the STBC
consecutive symbol periods are:                                   rule were made, as it will be shown in the next section.

          r1          s1     s2           h1         n1
                =                   ·          +           (1)    B. STBC for Channels with Intersymbol Interference
          r2         −s∗2    s∗
                              1           h2         n2
                                                                     Lindskog and Paulraj in [4] extended the considerations
where the channels samples h1 and h2 may be modelled by a
                                                                  made by Alamouti from symbol-wise to block-wise in order
complex multiplicative distortion and n1 and n2 are gaussian
                                                                  to treat channels with ISI. In this scheme a block of symbols
noise samples. This representation can be reorganized in a
                                                                  d[n] is divided in two sub-blocks d1 [n] and d2 [n] as well as
similar manner as:
                                                                  the transmission frame. In the first half of the frame, d1 [n]
          r1         h1      h2           s1         n1           is transmitted by antenna one and d2 [n] by antenna two. In
           ∗    =                   ·          +           (2)    the second half of the frame, d2 [n] time reversed, complex
          r2         h∗
                      2     −h∗1          s2         n∗
                                                      2
                                                                  conjugated1 and negated is transmitted by antenna one and
In a matrix equivalent form we have: r = Hs + n.                  d1 [n] time reversed and complex conjugated from antenna
   For detection, Alamouti proposes the multiplication of the     two. This can be represented by:
received signal vector r by matrix HH . Therefore the symbol
estimated can be obtained by ˆ = HH ·r = HH ·Hs+HH ·n.
                              s
It can be noted that H is orthogonal and HH is the matched                    r1            h1      h2            d1           n1
                                                                                     =                       ∗           +                 (4)
filter, so: HH · H = (|h1 |2 + |h2 |2 ) · I and ˆ is the vector
                                                s                              R
                                                                              r2            hR
                                                                                             2     −hR1           d2           nR
                                                                                                                                2
of matched filter output. Those observations imply that the
symbols s1 and s2 can be recovered from the matched filter        In this case the ordinary multiplication is replaced by convo-
output.                                                           lution. For detection, the same strategy used for the previous
   Tarokh, in [2], achieved, through orthogonal designs, sim-     scheme can be applied here.
plicity similar to Alamouti’s STBC for more than two anten-
nas for transmission. Using four antennas the received sym-                 Tail
                                                                                   58 symbols
                                                                                                     Training
                                                                                                                 58 symbols
                                                                                                                                    Tail
                                                                            Bits                    Sequence                        Bits
bols, in four consecutive symbol periods are:

                                                                               Fig. 1. Normal burst GSM.
  r1        s1        s2      s3        s4      h1     n1
 r2   −s∗          s∗     −s∗        s∗   h 2
                                                   n2 
           2        1        4        3     +                In EDGE/EGPRS each user payload (RLC radio block) is
 r3  =  −s3        s4      s1
                                             ·
                                        −s2   h3
                                                   n3 
                                                                  interleaved over two (MCS-1 up to MCS-6) or four bursts
  r4       −s∗4      −s∗3    s∗2        s∗
                                         1      h4     n4         (MCS-7 up to MCS-9), and each burst contains 116 user data
                                                          (3)     symbols (modulated with either GMSK or 8-PSK, see Table
The same detection strategy used for two antennas can be          1), 26 training symbols in the middle and 3 tail bits in the ex-
used in this case.                                                tremities. In [6], a new burst format scheme using two trans-
   This Alamouti’s scheme is indicated for channels without       mit antennas and compatible with the GSM burst format is
intersymbol interference (ISI), but due to the nature of the      proposed.
mobile radio channel, transmitter, receiver, pulse shape and
modulations present in the EDGE, ISI may be present. In            1 Time   reversed and complex conjugate is represented for R .
            Tx 1                                                                 in EGPRS specifications, the header and data are coded sep-
                                                                                 arately. At each iteration the first task of the simulator is to
                          − d 2R [n] L
                   Tail                   Training                  Tail
                   Bits                  Sequence    L   d1[n]      Bits         separate the RLC radio block in two parts: header and data
                                                                                 blocks. After that, each part is coded and punctured sepa-
       Tx




            Tx 2                                                                 rately in agreement with [7] and the interleaving and map-
                   Tail                Training                     Tail         ping of burst is made for subsequent transmissions in the air
                   Bits   d1R [n]   L Sequence L         d 2 [ n]   Bits
                                                                                 interface.


                                                                                              IV. I NCREMENTAL R EDUNDANCY
                   Fig. 2. New burst format.
                                                                                    As link quality control strategies, the EGPRS can use link
                                                                                 adaptation (LA) and incremental redundancy (IR) separately
   Due to ISI in the downlink channels, some “edge effects”                      or together. Each MCS has up to three puncturing patterns,
are introduced in the transmitted signals, and these effects                     P1 , P2 and P3 , with each representing the same RLC radio
need to be considered in the design of the new burst format.                     block, which can be used to decode an RLC radio block or
Let L denote the maximum delay of the two-downlink chan-                         can be combined together to provide coding gain (IR mode).
nels [6].                                                                        In [8] a description of LQC methods is made.
   In Figs 1 and 2, two burst formats are described. In the                         Using IR, the data block coded is divided in smaller sub-
Fig. 1 the normal GSM burst format is shown, while in Fig.                       blocks of the same size, through the puncturing process. Ini-
2 the new burst format suitable to two transmit antennas is                      tially, one sub-block containing little or no redundancy is
presented.                                                                       transmitted reaching thus one high rate of transmission since
   In the new format, the block contains 116 data symbols,                       the rate coding is sufficiently high, generally close to 1. If the
{d[n]115 } , which are divided in two sub-blocks and now
      n=0                                                                        data block is not received correctly more redundancy is sent
transmitted in two bursts. The sub-blocks are defined as fol-                    in the next retransmission with a different puncturing pattern.
low:                                                                             The erroneous blocks are stored, and combined with the sub-
                                                                                 sequent blocks until the successful decoding of the data block
               d1 [n] = d[n];       n = 0, ..., 57                               transmitted.
                                                                           (5)
               d2 [n] = d[n];       n = 58, ..., 115

      III. EDGE/EGPRS L INK L EVEL S IMULATOR                                                    V. P ERFORMANCE R ESULTS
   A EDGE/EGPRS link level simulator has been con-                                  We compare the performance of a single antenna and STBC
structed based on 3GPP specifications. The physical layer of                     with two and four uncorrelated antennas for MCS-1 and
EGPRS is simulated including: channel encoder, interleav-                        MCS-4. Some simplifications on the inner physical layer
ing, burst mapping modulator, pulse shaping, channel mod-                        are assumed. The BPSK modulation scheme replaces GMSK
elling, equalizer, de-interleaving and decoder. The simula-                      with perfect phase recovery in the receiver. Flat fading is also
tor is divided in two parts: inner core and outer core. The                      assumed and can be considered a good approximation for the
outer core encompasses the functions at the radio link control                   typical urban (TU) channel [9]. We are also assuming per-
(RLC) block level, while the inner core is responsible for the                   fect channel estimation in the receiver. When a more realistic
burst level. In this paper the outer core part is implemented                    channel estimation is considered, e.g. based on the training
following the detailed 3GPP specifications [7]. On the other                      sequence, some decrease in the performance is expected.
hand, some simplifications are assumed for the inner core as                         With perfect timing at the receiver and synchronized in-
discussed later.                                                                 terference, no further considerations about pulse shaping are
   The Jakes’ fading model is used for generating the chan-                      essential at this moment. We use the burst format displayed in
nel response. Its time variation and correlation depends on                      Fig. 3 when considering two transmit antennas with L = 0.
the mobile velocity. For the frequency hopping (FH) case in-                        The STBC scheme utilizing four transmit antennas is sim-
dependent channel samples are generated for each burst. For                      ilar to scheme for two antennas, in this case the data blocks
no frequency hopping (NoFH), the degree of channel corre-                        is divided in four sub-blocks d1 [n], d2 [n], d3 [n] and d4 [n].
lation, from burst to burst, depends on the velocity. For the                    The transmit matrix is show in 3 when using four antennas. A
interference-limited scenario, one time-aligned interferer is                    RLC radio block is mapped in four bursts when transmitting
assumed. The same assumptions regarding fading, modula-                          with four antennas following the fig. 3.
tion and velocity made for the desired user are extended for                        We assume that the scenario is interference-limited, hence
the interferer. The severity of the Doppler spread is controlled                 throughput in Kbps is plotted versus C/I in dB, with the noise
according to the selected mobility conditions.                                   power assumed to be 3dB below the interference power. In
   The basic unit of time in the simulator is an RLC radio                       this scenario the implementation of frequency hopping (FH)
block; each RLC radio block is referred to as iteration. As                      may be considered or not (NoFH). The throughput for each
                                                                                                                                            Interference Limited FH 3Km/h
             Tx 1                                                                                                        12
                    Tail     *                    Training         *              Tail
                    Bits
                           -d4[n]   -d3[n]   L   Sequence    L   -d2[n]   d1[n]   Bits

                                                                                                                         10

             Tx 2
                    Tail     *                    Training        *               Tail
                    Bits
                           -d3[n]   d4[n]    L   Sequence    L   d1[n]    d2[n]   Bits                                    8




                                                                                                     Throughput [Kbps]
        Tx




             Tx 3
                    Tail     *                  Training           *              Tail                                    6
                    Bits
                           d2[n] d1[n]       L Sequence L        -d4[n] d3[n]     Bits


                                                                                                                                                                           MCS−1,   1Tx−1Rx   (LA)
             Tx 4                                                                                                         4                                                MCS−1,   2Tx−1Rx   (LA)
                                                                                                                                                                           MCS−1,   2Tx−1Rx   (IR)
                    Tail    *                     Training         *              Tail
                    Bits
                           d1[n] -d2[n]      L   Sequence    L   d3[n] d4[n]      Bits
                                                                                                                                                                           MCS−1,
                                                                                                                                                                           MCS−1,
                                                                                                                                                                                    4Tx−1Rx
                                                                                                                                                                                    4Tx−1Rx
                                                                                                                                                                                              (LA)
                                                                                                                                                                                              (IR)

                                                                                                                          2




 Fig. 3. New burst format transmitting with four antennas.                                                                0
                                                                                                                              0       5    10                15
                                                                                                                                                          C/I [dB]
                                                                                                                                                                           20             25         30




                                                                                               Fig. 4. Throughput performance using STBC with transmit
MCS can be found from the BLER values by using the fol-                                        antennas.
lowing rule:
                                                                                                                                                Interference Limited FH 3Km/h
                                                                                                                          18



             Throughput = (1 − BLER) · Rmax                                              (6)                              16


                                                                                                                          14


where Rmax is the user data rate for a given modulation and                                                               12

coding scheme, e.g. 8.8 Kbps for MCS-1, see Table 1.
                                                                                                     Throughput [Kbps]
                                                                                                                          10


   For results the BLER is measured over 3000 RLC blocks                                                                      8

with two and four transmit antennas in the downlink. As LQC                                                                   6

the use of IR is also evaluated. It is assumed that the total                                                                                                              MCS−4,   1Tx−1Rx   (LA)
                                                                                                                              4                                            MCS−4,   2Tx−1Rx   (LA)
transmit power for STBC is the same as the one for a single                                                                                                                MCS−4,
                                                                                                                                                                           MCS−4,
                                                                                                                                                                                    2Tx−1Rx
                                                                                                                                                                                    4Tx−1Rx
                                                                                                                                                                                              (IR)
                                                                                                                                                                                              (LA)
                                                                                                                                                                           MCS−4,   4Tx−1Rx   (IR)
antenna. A block is said to be erroneous, if one or more the                                                                  2



following events occurs:                                                                                                      0
                                                                                                                                  0   5     10               15             20            25         30
                                                                                                                                                          C/I [dB]

  •   Cyclic redundancy code fail for the header;
  •   Cyclic redundancy code fail for the RLC data block.                                      Fig. 5. Throughput performance using STBC with transmit
   Figs. 4 to 7 show comparative results using STBC with                                       antennas.
two and four antennas where the use of IR can or cannot
be used. On Fig. 4 it is shown results for MCS-1 in a sce-
nario limited by interference when FH is used in low mobility                                                                             VI. C ONCLUSION
(3 Km/h). For high mobility (100 Km/h) similar results are
achieved since MCS-1 is well protected. Fig. 6 show result                                        The achieved results states that STBC could combat the
for NoFH in high mobility. Comparing Figs. 4 and 6 we see                                      fading effects, thus offering increased throughput when two
that transmit diversity provides better gain when FH is used.                                  transmit antennas are employed. As a general rule, trans-
This can be explained by the high protection found in MCS-1.                                   mit diversity provides higher relative gains for less protected
   In case of MCS-4, Figs. 5 and 7, the use of FH represent a                                  MCSs under bad channel conditions. Further improvement
loss of performance. This can be explained by not perceived                                    in performance can be obtained by using incremental redun-
temporal diversity for less protected MCS when FH is used,                                     dancy together with transmit diversity.
with FH, one single burst experiencing a poor channel state                                       For the use of STBC some changes in EDGE/EGPRS stan-
may be enough to produce an erroneous block. Use of IR                                         dards are needed. Among these changes we can highlight the
enhance the performance throughput for low quality channel                                     required two different training sequences transmitted by the
conditions.                                                                                    two antennas for channel estimation in the receiver and STBC
   Tables 2 to 5 summarize the throughput gains relative some                                  encoder.
reference scenarios. Transmit diversity is an important tech-                                     The results in this paper also show that most of the bene-
nique to mitigate the fading effects specially in less protected                               fit gained from transmit diversity is already realized with two
MCS. This can be proved by observation of these table re-                                      antennas, and that four transmit antennas may not ne neces-
sults for low SIR (6dB) where the fading effects are more                                      sary.
pronounced. The better performance is achieved for less pro-                                      The results presented in this paper are based on perfect
tected MCSs for most of scenarios, especially when the refer-                                  channel estimation. The absolute gains shown in Tables 2-
ence scenario considered is (1Tx-1Rx, FH 3Km/h), as shown                                      5 should be taken as indicative values of the potential gains
in Table 3.                                                                                    achievable with transmit diversity.
                         12
                                            Interference Limited NoFH 100Km/h
                                                                                                                                        Table 4
                                                                                                                  Throughput gains [%] for MCS-1 (relative to 1Tx-1Rx,
                         10                                                                                                       NoFH) at 100Km/h
                          8
                                                                                                                                   NoFH 2Tx-1Rx       NoFH 4Tx-1Rx
     Throughput [Kbps]




                          6
                                                                                                                                       SIR[dB]            SIR[dB]
                                                                                                                                     6       15         6       15
                          4
                                                                             MCS−4,   1Tx−1Rx   (LA)
                                                                                                                     MCS1(LA)      65.36    6.46      81.23    6.64
                                                                             MCS−4,   2Tx−1Rx   (LA)
                                                                             MCS−4,
                                                                             MCS−4,
                                                                                      2Tx−1Rx
                                                                                      4Tx−1Rx
                                                                                                (IR)
                                                                                                (LA)
                                                                                                                     MCS1(IR)      65.36    6.46      81.23    6.64
                                                                             MCS−4,   4Tx−1Rx   (IR)
                          2




                          0
                              0       5      10                15            20             25         30
                                                                                                                                        Table 5
                                                            C/I [dB]
                                                                                                                  Throughput gains [%] for MCS-4 (relative to 1Tx-1Rx,
                                                                                                                                  NoFH) at 100Km/h
Fig. 6. Throughput performance using STBC with transmit
antennas.                                                                                                                          NoFH 2Tx-1Rx       NoFH 4Tx-1Rx
                          18
                                                  Interference Limited FH 3Km/h                                                       SIR[dB]            SIR[dB]
                          16
                                                                                                                                     6      15          6      15
                                                                                                                     MCS4(LA)      325.54 52.11       612.23 55.02
                          14
                                                                                                                     MCS4(IR)      411.07 52.38       785.74 55.02
                          12
     Throughput [Kbps]




                          10


                              8
                                                                                                            tract.
                              6

                                                                             MCS−4,   1Tx−1Rx   (LA)
                              4                                              MCS−4,
                                                                             MCS−4,
                                                                             MCS−4,
                                                                                      2Tx−1Rx
                                                                                      2Tx−1Rx
                                                                                      4Tx−1Rx
                                                                                                (LA)
                                                                                                (IR)
                                                                                                (LA)
                                                                                                                                        R EFERENCES
                                                                             MCS−4,   4Tx−1Rx   (IR)
                              2
                                                                                                            [1]    S. Alamouti, ”A Simple Transmit Diversity Technique
                              0
                                  0   5      10                15             20            25         30          for Wireless Communications”, IEEE Journal of Selected
                                                            C/I [dB]
                                                                                                                   Areas in Communications, vol. 16, no. 8, pp. 1451-1458,
                                                                                                                   Oct.1998.
Fig. 7. Throughput performance using STBC with transmit
                                                                                                            [2]    V. Tarokh, H. Jafarkhani and A. R. Calderbank, ”Space-
antennas.
                                                                                                                   Time Block Codes from Orthogonal Designs”, IEEE
                         Table 2                                                                                   Transactions on Information Theory, vol. 45, no. 5, pp.
Throughput gains [%] for MCS-1 (relative to 1Tx-1Rx, FH)                                                           1456-1467, July 1999.
                        at 3Km/h                                                                            [3]    A. V. Geramita and J. Seberry, ”Orthogonal Designs,
                                                                                                                   Quadratic Forms and Hadamard Matrices”, Lectures
                                             FH 2Tx-1Rx                            FH 4Tx-1Rx                      Notes in Pure and Applied Mathematics, vol. 43. New
                                              SIR[dB]                               SIR[dB]                        York and Basel: Marcel Dekker, 1979.
                                              6      15                             6      15               [4]    E. Lindskog, A. Paulraj, ”A Transmit Diversity Scheme
                         MCS1(LA)           121.10 4.47                           132.94 4.47                      for Channels with Intersymbol Interference”, in Proc. Int.
                         MCS1(IR)           121.10 4.47                           132.94 4.47                      Conf. Communications, New Orleans, LA, pp. 307 - 311,
                                                                                                                   June 2000.
                         Table 3                                                                            [5]    A. F. Naguib, N. Seshadri and A. R. Calderbank., ”In-
Throughput gains [%] for MCS-4 (relative to 1Tx-1Rx, FH)                                                           creasing Data Rate over Wireless Channels”, IEEE Signal
                        at 3Km/h                                                                                   Processing Magazine, May 2000 pp.76-92.
                                                                                                            [6]    K. Zangi, D. Hui and J. F. Cheng, ”Physical-Layer Issues
                                            FH 2Tx-1Rx                              FH 4Tx-1Rx                     for Deploying Transmit Diversity in GPRS/EGPRS Net-
                                              SIR[dB]                                 SIR[dB]                      works”, in IEEE VTC Fall 2001, Atlantic City, NJ, USA.
                                             6        15                             6        15            [7]    3GPP, ”Channel Coding”, TS 45.003, v5.1.0, July 2001.
    MCS4(LA)                               151.92 51.81                            178.33 66.29             [8]    S. Eriksson et al., ”Comparison of Link Quality Control
    MCS4(IR)                              1280.77 57.36                           3510.36 66.46                    Strategies for Packet Data Services in EDGE”, in Pro-
                                                                                                                   ceedings of IEEE VTC’99.
                                                                                                            [9]    H. Olofsson, M. J. Almgren, C. Johansson, M. Hook
                 ACKNOWLEDGEMENTS                                                                                  and F. Kronestedt, ”Improved Interface between Link
  This work is supported by Ericsson Research-Brazilian                                                            Level and System Level Simulations Applied to GSM”,
Branch under the ERBB/UNI.33 Technical Cooperation Con-                                                            In Ericsson Radio System AB.

				
DOCUMENT INFO