Macro-diversity versus micro-diversity system capacity with realistic receiver RFFE model
I. Ouachani∗,∗∗ , P. Duhamel∗∗ , K. Gosse∗ , S. Rouquette-Léveil∗ and D. Bateman∗
* Centre de Recherche de Motorola - Paris-
Espace Technologique - 91193 Gif-sur-Yvette Cedex - FRANCE
**Laboratoire des Signaux et Systèmes
Plateau des Moulon, 3 rue Joliot-Curie - 91190 Gif sur Yvette - FRANCE
Abstract— This paper studies the evolution of capacity in the
context of macro-diversity antenna configurations, with respect
to system parameters such as the distance between mobile
devices. In this context, we redefine an accurate channel model
which involves a realistic model of the RF front-end (RFFE).
The impact of this improved model on system capacity is first
checked in the micro-diversity context, when comparing our
results with the model classically used in the literature. This
new model is shown to explain some phenomena oberved in
actual implementations. Then, in the macro-diversity context,
we concentrate on the usefulness of adding antennas, depending
on the distance to the transmitter or receiver. The obtained
results are strongly impacted by the RFFE model.
I. I NTRODUCTION
Wireless networking constitutes an important component
of future information technology applications. To improve
the reliability of communication over the wireless channels
and combat the fading, many papers have focused on the use
of diversity techniques. One of the most studied technique
is the use of multiple antennas at wireless transmitters and
receivers. It exploits the spatial micro-diversity created by
several antennas colocated on the same mobile device [1]
[2]. However, spatial diversity can also be exploited by joint
processing of signals transmitted or received by separated
devices, access points or terminals. This type of diversity is
known as macro-diversity.
Previous works addressing the macro-diversity [3]–[8]
have been published. The earliest works have focused on
the up-link in cellular networks in order to improve the
systems coverage and enlarge the cells size. The measure
of the link quality has been either (i) the instantaneous
or local mean signal-to-noise ratio (SNR) or (ii) signal-
to-interference (+ noise) ratio (SINR) [3] [4], or (iii) the
corresponding bit-error-rate (BER) [6] [7]. In this paper the
main objective is to assess the impact of system parameters
values, such as distance between devices, on the capacity
and the outage capacity of macro-diversity based system.
This point-to-multipoint communication link is met when
a given communicating terminal can transmit information
through several communication routes, either in decentral-
ized networks or in systems in which access points can share
received information.
In this context, it is necessary to redefine a proper channel
model adapted to the macro-diversity situation, in connection
with the propagation model, since the relative distance
between communicating devices plays an important role due
to path loss effects. This requires also to redefine realistic
constraints on emitting powers. In that purpose, we propose
a model of the receiver radio frequency front-end (RFFE),
enabling to specify relationships between receiver SNR
values and emitted/transmitted powers in various operating
zones.
In order to assess the impact of the realistic RFFE model
(presented in section II) and the related improved channel
model on the system capacity, we first address the case of
point-to-point multiple-transmit multiple-receive (MTMR)
communications (micro-diversity), and compare our results
to the ones obtained with classical MTMR channel models
[1] [2]. It is shown that the RFFE exhibits two main
operating modes, one of which results in an unusual channel
behavior. The impact of these operating modes is first
evaluated in a microdiversity scenario (section III), in terms
of the ergodic and outage capacities. Then, in section IV,
we concentrate on the macro-diversity context itself. The
objective is to evaluate the impact of increasing the number
of antenna (possibly at some distance of the other ones) on
the ergodic and outage capacities. This study thus provides
means of evaluating whether considering additional antennas
that are at some distance as part of a multi-antenna receiver
has a large potential impact or not.
II. M ODELLING THE M ULTIPLE T RANSMIT M ULTIPLE
R ECEIVE ANTENNA SYSTEM
This paper makes use of the following notations: boldface
lower symbols denote column vectors (e.g. x) while bold-
face capital symbols denote matrices (e.g. H). Whenever
required, subscripts emphasize their sizes: xM indicates a
size M × 1 vector; HM indicates an M × M matrix and
HM ×T has the size of M × T . Elements of vectors or
matrices are indexed by subscripts or double subscripts (e.g.
xk or hl,k ). IM denotes the size M identity matrix.
()∗ denotes complex conjugation; ()T denotes transposi-
tion and ()H stands for Hermitian transposition.
A. MTMR antenna system presentation
We focus on a MTMR system that employs nT trans-
mitting and nR receiving antenna elements. In the micro-
diversity context, the nT transmitting antennas are co-
located, as are the nR receiving antennas. In the macro-
diversity context, the transmitting antennas and the receiving
ones are dispatched on mobile terminals.
Assume we have Tx transmitting terminals and Rx
receiving terminals, and nTj and nRi (j = 1 . . . Tx ,
i = 1 . . . Rx ) denote the number of antennas co-
located on the j th transmitting terminal and ith receiv-
ing terminal respectively. This system is denoted as an
(nT1 , nT2 , ..., nTTx )x(nR1 , nR2 , ..., nRRx ) MTMR, and the
distance between the j th transmitting terminal and the ith
receiving one is denoted by Di,j . The input/output relation
of the MTMR system is described by the following equation:
y = Hx + b (1)
where x = [x1 x2 · · · xTx ]T is the transmitted vector; xi
denotes the nTi × 1 vector transmitted by the ith terminal.
b = [b1 b2 · · · bRx ]T denotes the additive white Gaus-
sian noise vector, which the covariance matrix cov{b} =
diag{cov{b1 }, cov{b2 }, · · · , cov{bRx }} with cov{bi } is
given by:
 2
N1 0 ··· 0

 0 N2 · · · 2 signal is filtered in order to remove adjacent band signals.
0 
cov{bi } =  ..

.. .. 

(2)
 . ··· . . 
0 · · · 0 Nn2R
i
where Ni2 , (i = 1...nR ) is the noise power introduced by
the ith received signal. The channel matrix is given by:
H = {Hi,j }i=1...Rx , j=1...Tx (3)
where Hi,j denotes the channel matrix containing the indi-
vidual channels from the j th transmitting antenna to the ith
receiving antenna.
In the micro-diversity situation, the system will be re-
ferred to as nT xnR MTMR; and the distance separating the
transmitting antennas to the receiving ones will be denoted
by D.
B. Channel fading model statistics
The propagation fast fading, seen by the different antenna
elements, is assumed to be independent Rayleigh flat fading.
Thus, the entries hk,l
i,j of matrix {Hi,j }i=1...Rx , j=1...Tx are
circularly-symmetric Gaussian with zero mean, independent
2
real and imaginary parts, having variance σi,j /2. The vari-
2 2
ance σi,j is given by the path loss σi,j = Ko (Di,j /do )−δ
where do = 1m is a reference distance, Di,j is the distance
separating the ith transmitter to j th receiver, δ is the path
loss exponent; Ko = (c/4πdo fc ) is the channel power gain
at the reference distance.
C. Model of the Radio-Frequency Receiver Front-End
Since we investigate the impact of macro-diversity sce-
narios on the overall system capacity, we have to redefine
appropriate channel models, in order to make explicit the
relative role of the emitted powers, of the pathloss, and of
the resulting SNR on the receiver branches introduced by the
RFFE. Classically, the RFFE is modelled only by additive
white Gausian noise (AWGN), which represents the thermal
noise of the components, a model which implicitly assumes
some behavior of the Automatic gain Control (AGC) inside
the receiver. We now investigate the receiver RFFE model
and study its impact on the capacity calculation.
AGC
Gr (Pr + N)
Pr
Other devices ADC
(N1, G1 ) (N2, G2 ) (N3, G3 )
(N,Gr )
Fig. 1. Simplified receiver scheme
N Gr
Pr Ps
+ *
Fig. 2. Receiver model scheme
1) Receiver description: Generally, the receiver RFFE is
mainly composed of three parts: in the first part, the received
Then, the amplifying block (second block, delimited by the
dashed rectangle √ in Fig. 1) multiplies the received signal
by some gain Gr , so that the signal dynamic range is
adapted to the ADC input range. The gain Gr is adjusted
by the Automatic Gain Control AGC device. And finally,
the ADC converts the analog signal to a digital one.
A generic AGC consists of a (Low Noise) Amplifier, with
a constant gain G1 and a noise figure N1 followed by a
variable gain amplifier with a gain G2 and introducing a
noise figure N2 . All other components are grouped in the
third block having a constant gain G3 , and introducing a
noise figure N3 .
This receiver RFFE model can be simplified into the
scheme plotted in Fig. 2, which is strictly equivalent. There,
the received signal is first affected by the additive thermal
noise; it is then amplified by the gain Gr = G1 G2 G3
adjusted by the AGC, and it is finally sent to the ADC. The
characteristics of the thermal noise and the amplifiers are
described below.
2) Thermal noise: The individual contributions of the
receiver components to the thermal noise can be gathered
into a single noise source at the input of the receiver, and
its power is given by:
N2 N3
N = KT B (N1 + + ) (4)
G1 G1 G2
where K is the Boltzmann constant, T is the temperature;
KT is -174 dBm/Hz at room temperature, and B is the
signal bandwidth.
3) Dynamic range: The dynamic range of the receiver
expressed in dB is the difference between the blocking
level and the receiver sensitivity. The blocking level is
the maximum power supported by the receiver electronic
components, here denoted by Prmax . The receiver sensitivity
is the signal power at the input of the receiver at which the
paquet error rate (PER) equals 10−1 , denoted by Prmin .
4) Amplifier gain: The RFFE global gain is given by
Gr = G1 G2 G3 . Because of individual components
limitations, the amplifiers are not always able to provide the
necessary gain to compensate for the channel fadings. In

fact, when the received power is lower than some threshold,
here denoted by PrAGC , the amplifying gain in the AGC is
bounded by Gmax r . SNRsat
To simplify the model, we assume that the minumum
of the dynamic range of Gr (i.e. Gmin r ) corresponds to
Prmax . Hence, the variation of the gain has two modes.
When Pr is in [Prmin , PrAGC ], the amplifying block gives the
maximum gain Gmax r . When the received power increases
in [PrAGC , Prmax ], the amplifying block gain Gr decreases
linearly till Pr equals the receiver blocking level Prmax . At
this stage, the gain Gr = Gmin r .This description introduces
three modes for the AGC: in a first mode, (saturation) the
gain is set to Gmax
r . In the nominal mode (where the AGC
plays its role: the signal at the entrance of the ADC has
its nominal value. Finally, when Pr > Prmax , the received
signal dynamic range is outside the operating range of the
ADC, thus introducing undesirable clipping effects. In the
sequel, we will not address this third mode. In Fig. 3(a),
only the two first modes are presented.
5) Receiver SNR: Given these models of the noise floor
and of the gains, the instantaneous SN R at the input of the
ADC reads:
Pr Pr
SN R = N = (5)
KT B(N1 + N
G1 +
2 N3
G1 G2 )
where Pr is the received power.
Define the received power Prsat corresponding to a gain
G2 , so that we have: G1 NG3sat = (N1 + G
sat N2
1
). This
2
value delimits two approximations of the SNR with different
behaviors:
• If Prmin ≤ Pr ≤ Prsat , the amplifier gain Gr is large
and the expression of the noise power simplifies to:
N = KT B (N1 + N2 /G1 ), a constant noise power.
Thus the SN R varies linearly with the received signal
power :
Pr
SN R = N2
(6)
KT B(N1 + G1 )
• If Prsat ≤ Pr ≤ Prmax : the gain Gr is small, and N1 +
N2 N3
G1 is negligible compared to G1 G2 . Consequently, the
thermal noise simplifies to: N = KT B G1N3G2 . Since
G2 is variable, the noise power is no more constant
and the SNR expression can be approximated by:
G 1 G 2 Pr
SN R = (7)
KT B N3
Since G2 and Pr are variable, deducing the SNR varia-
tion from its expression is not direct. In general, PrAGC ,
which is the threshold under which the amplifying gain
is maximum, is lower than Prsat . Generally, the system
is designed so that Prsat is as close as possible to Prmax ,
thus reducing the interval in which the noise varies with
Pr . Then, since the product G2 Pr is constant, (proof is
not given due to lack of space), we a have a constant
SN R for Pr ∈ [Prsat , Prmax ] .
(a) (b)
SNR Gr (dB)
Gmax
r
Gmin
r
3 dB Pr (dBm) Pr
Prmin Prsat Prmax Prmin
Fig. 3. (a)Amplifier gain variation versus the received power, (b)SNR
PrAGC Prmax
variation versus the received power
Considering the two SN R expressions above, the variation
of the SN R versus the received signal power Pr is plotted
on Fig. 3.(b)
From the explanations given above, the receiver dynamic
range can be divided into three intervals: [Prmin , PrAGC ],
[PrAGC , Prsat ] and [Prsat , Prmax ]. Since the performance
of the system is investigated in terms of capacity, which
depends only on the noise power, only two intervals
[Prsat , Prmax ] and [Prmin , Prsat ], denoted by I1 and I2 re-
spectively, are of interest to us. It is to point out that the
"nominal" zone for receiver processing is: [PrAGC , Prsat ],
since we are outside the saturation zone of the AGC and the
saturation zone of the SNR.
D. System performance measures
1) Ergodic capacity: The mutual information (in
b/s/Hz) of the macro-diversity spatial multiplexing system
under an average transmit power constraint is given by [9]:
I = log2 det InR + cov{b}−1 HΣHH



(8)
where Σ = E(XXH ). The capacity is defined in the ergodic
case as the maximum of the mutual information under an
average transmitter power constraint. Assume the channel is
not known to the transmitter. Telatar has shown in [1] that the
corresponding optimum power allocation is achieved when
the power is distributed evenly between components of X.
2) Outage capacity: A commonly used information theo-
retic performance measure is the outage capacity. The outage
is defined as the event that the maximum average mutual
information is inferior to the target spectral efficiency R
desired for transmission. Hence, the outage probability is
simply the probability that an outage event occurs:
pout = P (I < R) (9)
The outage probability is often chosen as pout
0 = 5%, and
the value of R verifying (9) is the outage capacity.
III. MTMR SYSTEMS EXPLOITING MICRO - DIVERSITY
In this section, we focus on a MTMR system that uses
nT transmitting and nR receiving antenna elements, in
a micro-diversity scenario. The aim here is to calculate
the system capacity, which depends, as shown in (8), on
channel coefficients and signal and noise variances on the
various receiver branches. This computation is based on the
 TABLE I
propagation model defined in section II-B, and on the signal
PARAMETERS OF A TYPICAL IEEE 802.11 A RECEIVER
and noise variances as given by the RFFE model. Prmin -85 dBm N -88 dBm
Under the classical assumptions made by Foschini in [2], PrAGC -68 dBm Gmax
r 65 dB
sat -48 dBm Gmin 17 dB
the SNR is the same on all receiver branches, and is Pr r
Prmax -20 dBm SN Rsat 40 dB
expressed as: PADC -3 dBm SN Rmin 3 dB
P
SNR = (10)
N
where P is the average received power on each branch. This −10

is the "standard" model. Here, the extension of the single −20

antenna RFFE model described in section II-C to the MTMR −30

SNR saturation
Gr = f(Pr): linear
case puts into question these assumptions, and leads to a

Rceived power in dBm

−40

−48 dBm
modified relation between the SNR at the output of the RFFE −50

block and the received power on a given receiver antenna. −60

−68 dBm
This new relationship also depends on the way the multi- −70 SNR=g(Pr): linear

antenna front-end is designed, in terms of number of analog −80

G = f(P ): linear
r r

SNR = g(Pr):linear −85 dBm

Gr = Gmax
RFFE blocks and in terms of AGC strategies. In any case, the −90
r

0 10 20 30 40 50 60 70 80 90

resulting SNR vs received power relation affects capacity, distance in meter

Fig. 4. Received power versus the distance between the transmitter and
and we compare below the results obtained with the standard the receiver (STSR case).
and the improved multi-antenna RFFE models.

A. Multiple-antenna receiver model

In a multiple-antenna receiver, three strategies can be
applied in the front-end design: (1) one RFFE block is
assigned for each antenna and the AGC of each branch is
communicating with the AGCs of the other ones, so that
the same gain is applied on all branches, (2) one RFFE
block is assigned for each antenna and the AGC of each
branch is independent of the AGCs of the other ones, thus
different gains are applied on the receiver branches, (3) a
single RFFE block is assigned to all antennas, the receiver
compares all the received powers and selects the highest
one. This case does not fully exploit diversity and will not
be further considered here.
Remembering the single antenna RFFE model presented
in subsection II-C, it is clear that the instantaneous noise
power depends of the received signal power, or, in other
words, of the channel fading coefficients. Thus, in the
MTMR setting, the final value of the SNRi received on the
ith branch depends on the AGC strategy. Due to lack of
space, this comparison is made only through simulations.
B. Simulation results
Consider again RFFE characteristics matching with the
physical layer specifications of the IEEE 802.11a standard,
as summarized in table I. The Rayleigh fading variances are
given by the path loss, which is characterized by the follow-
ing parameters chosen for a typical indoor channel without
shadowing: the path loss exponent is δ = 3.1, the operating
frequency is fc = 5.2GHz and Ko = 2.33 10−5 . The total
transmit power is fixed to Pe = 20 dBm for all systems.
B is assumed to be equal to 16.25MHz, corresponding to
typical WLAN bands (KT B = 72dB).
We now compute the ergodic capacity and the outage prob-
ability for the 2x2 system, in the micro-diversity context,
considering the two receiver RFFE models, which will be
compared to the standard system. The ergodic capacity and
outage probability are investigated versus the distance D
between the transmitter and the receiver. These results will
serve as a basis for studying the macro-diversity case.
We first investigate the variation of the received power
versus the distance D between one transmitter and one
receiver, in order to give the bounds of the zones correspond-
ing to the three signal power intervals: [-85 dBm, -68 dBm],
[-68 dBm, -48 dBm] and [-48 dBm,-20 dBm]. The received
power versus D is given in Fig. 4.
In summary, the "nominal" working region of the RFFE
block is where the distance between the transmitter and the
receiver is in [6m, 23m], when the receiver gain is below
Gmax and the SNR is varying linearly with the received
r
signal power.
In the following, we use U1 and U2 to denote the inter-
vals [1m, 6m] and [6m, 76m], respectively. Fig. 5 illustrates
the variation of the ergodic capacity versus the distance
separating the transmitter to the receiver; we note that the
ergodic capacities when using either communicating AGCs
or independent ones, are the same in all the receiver dynamic
range. Further, when we compare the ergodic capacity when
considering the modelled RFFE blocks to the standard
system, we note that: (i) they have the same performance
when D ∈ U1 , (ii) they have different behavior in the SNR
saturation zone, i.e. when D ∈ U2 . When D tends to zero,
the standard system capacity converges to infinity whereas
the system capacity when considering the RFFE models
saturates. This behavior is observed in real systems.
From Fig. 5, we remark that when D ∈ U1 , the 2x2
MTMR system outage capacity of the three models are
different. It is to note that although the ergodic capacity
is the same for the 2x2 MTMR system when using the
communicating AGCs and independent ones, their outage
capacities are different. Thus proving the precision brought
by the the outage capacity as a performence measure.
 2x2 MIMO context, Pe = 20dBm
40
standard system The transmitters are assumed synchronized with the receiver
RFFE model, independent AGCs
35 RFFE model, communicating AGCs
so that the signals arrive to the receiver at the same symbol
30
period.
Ergodic Capacity in bps/Hz
U2
25
till D = 78m
For sake of clarity, we illustrate the relationship between
20
U
1
the spatial positions of the transmitters and the total received
15 power by delimitting the receiver surrounding area with
10 respect to the zones involved in the RFFE model, i.e. when
5 the AGC saturates, when the SNR saturates, and in the
0
0 10 20 30 40 50 60
nominal AGC zone (linear SNR and linear AGC curve).
D in meter
This question is of importance in this MTSR case, since the
Fig. 5. Ergodic capacity, 2x2 MTMR micro-diverse system with various
2-antenna receiver RFFE models definition of the regions depend on the total received power
(hence on both distance). In the STMR case, the problem is
microdiverse 2X2 MIMO context, d=2m, Pe = 20dBm similar to the STSR case.
0.9 Due to lack of space, we omit the computations providing
0.8 the fronteers of the various zones, plotted in Fig. 7. It is
0.7
seen that 5 zones can be distinguished: when the couple of
Outage probability P(I < R)

0.6
distances (D1 , D2 ) is in one of the two shaded areas, the
0.5

0.4
transmitted signals are not treated by the receiver since the
0.3
received power is outside the dynamic range.
Standard system
0.2
out Independent AGCs
P = 5%
Communicating AGCs
0.1

0
10
Fig. 6. Outage capacity curves for the 2x2 MTMR micro-diverse systems
with various 2-antenna receiver models, for D = 2m.
To conclude, we can say that when the distance separating
the transmitter to the receiver is sufficiently high, both the
standard and the improved models provide almost the same
results. However, when the receiver is closer to the transmit-
ter, the receiver SNR saturates in the realistic model, thus
resulting in a more accurate estimate of the performances;
Thus, it is shown that the use of the classical model in this
region could result in an overestimation of the performance.
IV. MTMR SYSTEMS EXPLOITING MACRO - DIVERSITY
We now focus on macro-diversity systems, when the
antennas are dispatched on different mobile terminals. We
restrict our simulations to the case where each terminal
has a single antenna. We concentrate on scenarios where
the distances between all the receive (or transmit) mobiles
decrease so that we move smoothly from a macro-diversity
to a micro-diversity configuration. The system consisting
of Tx transmitting terminals and Rx receiving ones, in the
macro-diversity context is denoted by: the Tx × Rx macro-
diversity MTMR system.
In the following, we present the simulation results for
some scenarios, keeping the same simulation parameters as
considered in section III-B.
A. 2x1 MTSR employing macro-diversity: delimiting the
receiver surrounding area
In this scenario, the system is now made of 2 transmitters
and one receiver. The transmitters have the same total
transmit power. Furthermore, the distances separating the
receiver from the transmitters are simply denoted by Di .
B. The (1,1)x(1,1) MTMR system: the two receivers are
15 20 25 30
Spectral efficiency R in bps/Hz
35 40 45
located at the same place
We consider the scenario where the two receivers are
located at the same place and the two transmitters are mobile
around. Then we compare its performence to the 1x2 MTSR
system in order to study the impact on the ergodic capacity
of adding a second transmitter at a certain distance from the
first one.
In Fig. 8 and Fig. 9, the curve with stars is plotted for
(1,1)x(1,1) system where one transmitter is located at a fixed
distance D1 from the 2 receivers and the other one is moving
far away, being always farer from the two receivers than the
first transmitter.
Fig. 8 depicts the variation of the ergodic capacity when
D1 = 1m, while Fig. 9 is the corresponding curve for D1 =
57m. Clearly, when the distance of the second transmitter
goes to infinity, the capacity should converge to that of a
1 × 2 STMR system. This is clearly seen on Fig. 9, since
the convergence is fast, while on Fig. 9, it is seen that the
impact of the second transmitter extends to a much wider
zone.
f we look at the improvement brought by the macro-
diversity (adding the second transmitter), we can see that
the farer the fixed transmitter is the better is the increase in
the global system capacity.
If we look in more details at the near surrounding region
of the two receivers, we can note a surprising behavior of the
ergodic capacity. As it is shown in Fig. 10, starting from the
case the two transmitters are located at the same distance
D1 = D2 in U1 , then one of them starts moving closer
the two receivers, the system ergodic capacity decreases.
Roughly speaking, this can be explained by the fact that in
this region the SNR saturates. This seems to be due to the
fact that, when the transmitter is becoming nearer to the two
receivers, the received power increases consequently thus
increasing the introduced noise (since it is proportional to
 30
2x2: Rx1=Rx2, d1=1m
1x2 (indep. AGCs): Tx1=Tx2
the received power), which causes a decrease in the system 25

ergodic capacity.

Ergodic capacity in bps/Hz

20

V. C ONCLUSIONS 15

In this paper, the evolution of the capacity in both the 10

macro-diversity and micro-diversity contexts have been in- 5

vestigated, with respect to the system parameters such as the

0
distance between the mobile terminals. For both contexts, we 0 10 20 30 40
d in meter
2
50 60 70 80 90

have proposed a realistic channel model which involves an Fig. 8. Comparing the (1,1)x(1,1) system ergodic capacity, when the 2
receivers are in the same place, to the 1x(1,1) one for D1 = 1 m
accurate model of the receiver radio front-end.
In the micro-diversity context, it was seen that our results 6
2x2: Rx1=Rx2, d =57m
1
1x2 (indep. AGCs): Tx1=Tx2

are similar to those provided by the model classically used in 5.5

the litterature, when the distance separating the transmitter to 5

the receiver is higher than a given threshold. However, when 4.5

Ergodic capacity in bps/Hz

4
the distance between the transmitter and the receiver tends to
3.5
zero, the modelled MTMR system ergodic capacity saturates
3

whereas the standard MTMR system one tends to infinity.

2.5

This new model allows to predict a behavior observed in 2

actual implementations. 1.5

55 60 65 70 75 80 85 90
In the macro-diversity situation, it was shown that the d in meter
2

Fig. 9. Comparing the (1,1)x(1,1) system ergodic capacity, when the 2

coverage area of the STSR system is strongly increased by
adding a mobile transmitter. Furthermore, adding a mobile
transmitter to the 1x(1,1) STMR system, composed of one
fixed transmitter and two colocated receivers, has different
impact on the system ergodic capacity, depending on the
distances between the transmitters and the receivers. From
a practical point of view, the good news are that macro-
diversity is proportionnally more useful in difficult situations
(when the receivers and the transmitters are far away) than
when they are closer. The SNR saturation zone was also
shown to be subject to somewhat unexpected phenomena
(bringing one antenna closer decreases capacity).
Finally, it was shown that adding antennas significantly
increases the MTMR system performence in most macro-
diversity scenarios, and that the new RFFE model explain
some phenomena observed in actual implementations.
MISO 2x1 macro−diverse system
Outside the receiver dynamic range
100
80
receivers are in the same place, to the (1,1)x1 one for D1 = 57 m
[3] R.C. Bernhardt. Macroscopic diversity in frequency reuse radio
systems. IEEE Journal on Selected Areas in Communications,
5:862–870, June 1987.
[4] L.-C. Wang, G. L. Stüber, and C.-T. Lea. Effects of Ricean
fading and branch correlation on a local-mean-based macro-
diversity cellular system. IEEE Transactions on Vehicular
Technology, 48:429–436, March 1999.
[5] A. M. D. Turkmani. Performence evaluation of a composite
microscopic plus macroscopic diversity system. In IEE Pro-
ceedings I, Communication, Speech, Vision, volume 138, pages
15–20, February 1991.
[6] Zang J and V. A. Aalo. Effect of macrodiversity on average-
error probabilities in a Ricean fading channel with correlated
log-normal shadowing. IEEE Transactions on Communica-
tions, 49:1503–1507, January 2001.
[7] A. A. Abu-Dayya and N. C. Beaulieu. micro- and macro-
diversity MDPSK on shadowed frequency-selective channels.
IEEE Transactions on Communications, 43:2334–2343, August
P = −85dBm
r
1995.
P = −68dBm
r
P = −48dBm
[8] S. Mukherjee and D. Avidor. Effect od microdiversity and
r

60
SNR = g(Pr): linear
P =−20dBm
r correlated macrodiversity on outages in a cellular system. IEEE
D1

G = Gmax
Transactions on Wireless Communications, 2(1):50–58, January
r r
40 2003.
G = f(P ):linear
[9] T. Cover and J. Thomas. Elements of Information Theory.
r r
20
Wiley, 1991.
sat
0 SNR = SNR

0 20 40 60 80 100 26
D1 = 1m
D2
D1 = 3m
Fig. 7. Delimitation of the receiver surrounding area (MTSR case). 25 D1 = 6m

24
Ergodic capacity in bps/Hz

23

22

R EFERENCES 21

20

[1] I.E. Telatar. Capacity of Multi-antenna Gaussian Channels. 19

Technical report, AT & T Bell Labs, 1995.
18
[2] G.J. Foschini and M.J. Gans. On Limits of Wireless Com-
munications in a Fading Environment when Using Multiple 17
0 2 4 6
D2
8 10 12

Antennas. Wireless Personal Communications, 6:311–335, Fig. 10. Capacity behavior in the surrounding region of the two receivers.
1998.