Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
6, JUNE 2014
3299
AbstractDistributed antenna systems (DAS) represent a suitable method for extending multiple-operator radio coverage into
indoor premises, tunnels, etc. With the advent of 4G (and beyond)
mobile communications systems and of multiple inputmultiple
output (MIMO) transmission techniques, a common problem for
wireless operators is how to upgrade existing DAS systems to the
MIMO technology in a cost-effective way. In this paper novel
indoor solutions based on the combination of DAS and MIMO
transmission techniques (Interleaved-MIMO DAS solutions) are
proposed, and their performance is investigated in realistic cases
with the aid of measurements, ray tracing simulations, and LTE
link-level simulations.
Index TermsIndoor radio communication, MIMO systems,
mobile communication, optical fiber cables, radio propagation,
ray tracing.
I. INTRODUCTION
0018-926X 2014 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
3300
IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 62, NO. 6, JUNE 2014
DAS architecture to an interleaved MIMO DAS solutions requires almost no additional deployment costs as no modification
of the existing infrastructure (antennas, cables, cable runways
etc.) is necessary.
A drawback with respect to c-MIMO DAS is the nonuniform
SNR of the different MIMO branches at the mobile terminal.
Although recent standards such as LTE can cope with up to
1215 dB of power imbalance between different branches at the
mobile receiver, i-MIMO DAS implementation in real environments cannot guarantee such a figure over the whole service
area.
The evaluation of the consequences of this problem is object
of investigation in this paper.
The i-MIMO concept, introduced for the first time in [10],
is further developed, theorized and evaluated in a systematic
way in the present paper using an ad-hoc measurement setup,
ray tracing (RT) propagation simulation and LTE system simulation. The scheme and characteristics of a possible, practical
implementation of i-MIMO DAS are described in Section II.
In Section III, the measurements performed using a suitable
MIMO setup in two different environments are described. In the
same section an advanced RT simulator [11] is validated against
measurements for the first time in terms of important MIMO
parameters such as the condition number and the power-imbalance. Once validated, the RT tool is then used to feed propagation data into an LTE link level simulator [12] to determine
the performance of 2 2 i-MIMO DAS in a number of realistic
environments (Section IV). Results show that i-MIMO DAS solutions can actually approach the performance of c-MIMO DAS
at a much lower cost. Performance assessments of higher order
i-MIMO DAS and general planning guidelines are outlined in
Section V. Conclusions are drawn in Section VI.
From a general point of view, a MIMO system with transmitting antennas and receiving antennas can be described by
the channel matrix, i.e., a
matrix of channel impulse
responses [1].
It is well known that MIMO capacity depends on both the
SNR and the singular values of the channel matrix. Instead of
the latter, a more synthetic parameter has been considered in
this paper, i.e., the condition number (CN), which is strictly related to channel capacity [14]. CN is defined as the imbalance
between the singular values of , or between the eigenvalues
of
(where the superscript stands for matrix transposition
and conjugation). In other terms
(1)
and
are the maximum and the minimum eigenwhere
value of the matrix, respectively. Especially for 2 2 MIMO
systems, CN can be considered a good indicator of the potential multiplexing gain. From a practical point of view CN values
should not be too high (e.g. lower than 15 dB) to fully exploit
the spatial multiplexing mode recently introduced in the latest
4G standards (LTE/LTE-A). In addition to the CN, SNR is also
of crucial importance.
In the present work we show that another important performance parameter specific for i-MIMO DAS deployments is the
3301
VITUCCI et al.: INTERLEAVED-MIMO DAS FOR INDOOR RADIO COVERAGE: CONCEPT AND PERFORMANCE ASSESSMENT
Fig. 1. Example of (a) c-MIMO F-DAS and (b) i-MIMO F-DAS deployments for a 2
Fig. 2. The 2
and
are estimated through the down link reference signals (RS) over
a 5 MHz channel.
Two extensive measurement campaigns have been carried out
with the above described setup in different indoor scenarios.
The first scenario is Villa Griffone, the historical villa in the
outskirts of Bologna (Italy) were young scientist Guglielmo
Marconi carried out his first experiments. Its an 18th century
building, with thick stone walls and can, therefore, be considered a hostile environment for radio propagation, representative
of historic buildings of many European cities. Measurements
in this scenario have been carried out at the frequency of
1935 MHz. Three different MIMO DAS deployments have
been considered.
2 2 c-MIMO DAS, i.e., co-located MIMO with a single,
2-branch MIMO RAU. Tx elements are spaced by
.
2 2 i-MIMO 2-DAS
2 2 i-MIMO 4-DAS
At the Rx side, arrays of two omni-directional antennas with
a
spacing were used. A more detailed description of this
measurement campaign can be found in [10].
A second measurement campaign at a different frequency
(858 MHz) and with similar MIMO DAS deployments was
carried out within a different building (the Commscope-Italy
building) featuring a typical modern indoor office environment, with internal walls made of plasterboard and, therefore,
quasi-transparent to radio waves. The RAU locations and the
Rx positions (routes B, C, H) used in this second measurement
campaign are displayed in Fig. 3 on the plan of the building.
Regarding the Tx antenna configuration, the considered
MIMO DAS deployments and the related active antennas
locations are listed below (0 and 1 are the identifiers of the
two Tx MIMO branches):
c-MIMO 1-DAS with 2 antenna elements located in Tx0
and Tx1 (
antenna spacing)
i-MIMO 2-DAS (RAUs located in Tx0 and Tx1)
i-MIMO 4-DAS (RAUs located in Tx0, Tx1 , Tx0 and
Tx1 )
3302
IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 62, NO. 6, JUNE 2014
Fig. 3. Plan of the office building with transmitting locations and measurement routes.
Fig. 4. Average Condition Number and Power Imbalance along a corridor for:
(a) 2x2 i-MIMO 2-DAS deployment and (b) 2x2 i-MIMO 4-DAS deployment.
dB-value local averages are computed on the 4x4 point grid centered on each
measurement position.
poor result is probably due to the presence of metallic objectsa coffee machine and a built-in wardrobenot included
in the digitized map.
VITUCCI et al.: INTERLEAVED-MIMO DAS FOR INDOOR RADIO COVERAGE: CONCEPT AND PERFORMANCE ASSESSMENT
3303
TABLE II
TX-SIDE CORRELATION COEFFICIENTSi-MIMO 2-DAS CASE
antennas are spaced of several wavelengths or when dual-polarized antennas are used, both PI and small-scale fading correlation can affect the MIMO performance: that is why in [19]
an empirical model for CN evaluation is proposed, based on
the knowledge of PI and channel correlations at Tx side (often
called Tx antenna correlations for simplicity).
The complex correlation coefficient at Tx side can be defined
as
Fig. 5. Route H with 2
2 i-MIMO 4-DASMeasured versus simulated
CDFs of Condition Number in positions 1 and 4.
TABLE I
RT SIMULATIONSPREDICTION ERRORS
It is worth noticing that the deterministic prediction of coherent parameters such as the CN is a challenging task since
a realistic reproduction of the amplitude and phase distributions of the multipath components is required. In this context,
the adoption of hybrid models based on integration of deterministic ray models and statistical diffuse scattering models is
a key asset, as recently highlighted by a number of investigations [15][18]. The diffuse scattering model described in [15],
embedded into the RT simulator adopted here, is an example
of such hybrid technique: as shown in [16][18], this approach
leads to a sensible improvement in the prediction of multidimensional channel parameters.
The results in Fig. 4 highlight a clear correlation between the
CN and the PI. This fact is rather intuitive as, similarly to a low
CN, a low PI between different branches received through different propagation links is necessary to achieve good capacity in
distributed MIMO. Conversely, in co-located MIMO with small
antenna spacing, were only one propagation link exists, the CN
only depends on small-scale fading correlation, and, therefore,
the PI has little significance. In some intermediate cases, when
(3)
which is the correlation between channels originating from
transmit antennas and and arriving at receive antenna .
However, one of the peculiarities of distributed MIMO deployments is precisely to obtain negligible channel correlations
at the Tx side, because of the very high spacing between the
antenna elements. This is confirmed by our measurement campaigns (as shown in Table II for the 2 2 i-MIMO 2-DAS deployment) and, therefore, the PI alone can be considered a good
performance indicator in i-MIMO systems.
Since MIMO capacity strongly depends on the SNR value,
we can conclude that the SNR and the PI are likely the most important parameters for i-MIMO DAS performance. Differently
from the CN, the PI can be estimated just on the base of simple
path loss models, such as [20].
This aspect can facilitate the development of planning tools
and the investigation of effective deployment strategies for distributed i-MIMO systems without resorting to complicated RT
algorithms.
IV. INTERLEAVED-MIMO DAS PERFORMANCE ANALYSIS
In order to investigate the performance of the proposed
i-MIMO DAS solution in a realistic case, a complete LTE link
level simulation has been carried out combining RT prediction
and the LTE simulator developed at the Technical University
of Vienna [12].
By default the LTE simulator dynamically generates different
MIMO matrixes at each simulation run, exploiting the powerdelay profiles (PDP) provided by several, embedded standardized channel models (e.g., AWGN, ITU-R Pedestrian, Vehicular and Indoor models, WINNER, etc.) and taking into account
some additional characteristics such as the type of fading (block
fading, fast fading) and channel correlations.
In this paper the channel matrices were generated differently;
namely directly through RT simulations and fed into the LTE
simulator, so as to get a more realistic performance evaluation
of the measured scenarios (see the next section).
Link level simulations show that the performance (i.e., the
throughput) of an i-MIMO DAS system mainly depends on the
3304
IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 62, NO. 6, JUNE 2014
TABLE III
SETTINGS OF THE RT SIMULATIONS
(4)
where
is the Frobenius norm of the channel matrix (see,
also, [13, p. 74]), and
is the MIMO order. Using such
VITUCCI et al.: INTERLEAVED-MIMO DAS FOR INDOOR RADIO COVERAGE: CONCEPT AND PERFORMANCE ASSESSMENT
3305
TABLE V
SETTINGS OF THE LTE LINK LEVEL SIMULATOR
TABLE IV
SETTINGS OF THE MATLAB PROCEDURE
normalization in a 2
and, therefore
was considered (20 MHz), and the LTE transmit mode was
set to Closed loop spatial multiplexing (i.e., Transmit Mode
4 according to 3GPP TS 36.211), which means that the Base
Station adjusts, besides the modulation and coding scheme,
also the MIMO transmission parameters such as the number of
independent transmitted streams and the pre-coding applied to
them, as a function of the channel conditions reported by the
Rx mobile terminal (User Equipment, UE in the LTE standard).
Summing up, for each position of the UE within the considered
environment the complete simulation process provides the
physical layer throughput, the bit error rate (BER), the CN
and the PI. For a more detailed description of the simulation
procedure, see [21].
The procedure described above has been used to evaluate the
performance of i-MIMO DAS versus c-MIMO DAS, assuming
that the DAS system is hooked up to a 2 2 MIMO-LTE Base
Station (referred to as eNodeB in the LTE standard). In particular, four different RAU locations have been chosen (P1, P2,
P3, and P4 of Fig. 3), and two different deployments have been
simulated.
Case #1 with all antenna locations (P1, P2, P3, P4) active. For this deployment, the performance of c-MIMO
and i-MIMO 4-DAS configurations are analyzed and compared, assuming that both Tx branches are assigned to
each RAU location in the first configuration, and only one
branch to each RAU location in the second configuration.
Case #2 with only two antenna locations (P1 and P3) active. For this deployment, the performance of c-MIMO
and i-MIMO 2-DAS configurations are analyzed and compared, similarly to case #1.
B. Analysis of the Simulation Results
The physical layer throughput as function of the SNR averaged over all the Rx routes for the considered deployments are
reported in Fig. 8 and Fig. 9, for both carrier frequencies and for
four Tx power values.
The comparison between the solid and the dashed traces at
the lower frequency [Figs. 8(a) and 9(a)], clearly shows that the
c-MIMO DAS and the i-MIMO DAS solutions reach almost
the same throughput despite the lower cost of the interleaved
solution which requires half the number of antennas and cables
(see Fig. 1).
3306
IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 62, NO. 6, JUNE 2014
Fig. 10. LTE throughput as a function of SNR and Power Imbalance for the
i-MIMO DAS configuration.
Fig. 8. Average LTE Throughput as a function of SNR for deployment #1
(c-MIMO 4-DAS versus i-MIMO 4-DAS) at (a) 750 MHz and (b) 2680 MHz.
Only a slight throughput degradation is visible in the deployment case #2 [Fig. 9(a)], for which the radio coverage is worse
due to the lower number of Tx antenna locations. It can be also
verified that the throughput approaches the saturation level of
147 Mbps for both deployment cases at the SNR level of 40
dB. Concerning the higher frequency [Figs. 8(b) and 9(b)], for
both deployment cases a 1015 Mbps throughput advantage for
the c-MIMO DAS solution in the intermediate SNR range of
2535 dBs can be seen. The performance gap gets narrower as
the SNR approaches the saturation level of 40 dB. The higher
propagation losses with respect to the 750-MHz carrier, and the
consequent higher power imbalances, could be brought in to explain the performance degradation of the interleaved solution
at 2680 MHz. However, all the throughput plots are congruent
since they are computed as function of the SNR.
In order to investigate more deeply the influence of both coverage and PI on the physical layer throughput, Fig. 10 shows a
multidimensional plot of the throughput obtained by combining
the results of the interleaved deployments #1 and #2 at both carrier frequencies.
For each pair of SNR and PI values, the throughput is displayed in a color-coded format: different throughput zones
can be identified in Fig. 10, according to the color.
It is evident that increasing SNRs and decreasing PIs produceon the averagehigher channel throughput.
This further testifies that in addition to the SNR the PI between the MIMO branches is also of crucial importance for the
correct deployment of an LTE i-MIMO DAS, as already discussed in the previous sections.
If these requirements cannot be satisfied (green/blue points in
Fig. 10), diversity schemes can be, however, adopted to improve
the BER value.
Moreover, it should be also noted that these results have a
general validity for LTE systems operating in single user mode
and in the considered, reference environments. Of course, other
systems adopting different MIMO schemes or operating in
very particular propagation environments might show different
behaviors.
V. I-MIMO PLANNING PRINCIPLES
A. High-Order i-MIMO Deployments
In this section the deployment of i-MIMO DAS of arbitrary
order in large indoor environments is studied. If n is the MIMO
order at the base-station side and
the number of RAUs,
then we have M/n repetitions of the n MIMO branch-signals
which must be properly interleaved over the environment to
both optimize radio coverage and minimize power imbalance. It
is evident that in a uniform, linear coverage deployment along a
VITUCCI et al.: INTERLEAVED-MIMO DAS FOR INDOOR RADIO COVERAGE: CONCEPT AND PERFORMANCE ASSESSMENT
Fig. 11. Uniform, 2D 4 4 i-MIMO DAS coverage. Left: wrong, noninterleaved coverage solution. Center: interleaved coverage solution with a 4-branch
cluster highlighted in black. Right: 1-shift interleaved coverage solution. Each
color/letter corresponds to a different MIMO branch-signal.
corridor different branches must simply be alternated to minimize imbalance. In a 2D coverage case, however, and even
more so in a 3D case, the best deployment scheme is not always trivial.
For example in a four-branch case, the solution on the left
in Fig. 11 is clearly unbalanced and, therefore, should be discarded. In order to limit the spatial power imbalance, it is intuitive that the four branches should be arranged in a square
cell cluster, highlighted in black in Fig. 11. Of course, different clusters can be laid over the service area in different
ways; for instance, they can be simply aligned to each other
(Fig. 11-center), or a 1-cell down-shift can be introduced between adjacent columns (or rows) of clusters (Fig. 11-right). The
shifted solution might yield different imbalance and, therefore,
performance levels, but first simulation results do not seem to
confirm this effect.
Even in the best deployment, the received power from a
given branch peaks in the vicinity of the corresponding RAU
location and, therefore, the power imbalance also peaks. Moreover, high-order MIMO deployment solutions corresponding
to big clusters will necessarily yield greater spacing between
different-branch RAUs and, therefore, larger imbalance, especially in propagation environments where path loss increases
rapidly with distance.
B. Planning Method Outline
Power imbalance can be expressed through the parameter PI
defined in (2) only if
. For higher order i-MIMO DAS
systems (i.e.,
), the overall power-imbalance level at a
given Rx location can be expressed through the following standard deviation parameter:
(5)
is the local averagemade over all Rx antenna
where
elementsof the i-th branch power.
is, of course, the total
power received from all the RAUs radiating the same i-th
MIMO-branch signal and
is the average received power.
represents the standard deviation of the set
and represents in our opinion the best metric
to measure power imbalance when
. According to
preliminary simulations performed using the LTE link-level
simulator presented in Section IV,
, together with the
average SNR computed as
(6)
3307
3308
IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 62, NO. 6, JUNE 2014
REFERENCES
[1] A. J. Paulraj, D. A. Gore, R. U. Nabar, and H. Blcskei, An overview
of MIMO communicationsA key to gigabit wireless, Proc. IEEE,
vol. 92, no. 2, Feb. 2004.
[2] T. Haustein, V. Kumar, and T. Wirth et al., Measurements of multi-antenna gains using a 3GPP-LTE air-interface in typical indoor and outdoor scenarios, presented at the European Wireless 2008 (EW2008),
Prague, Jun. 2008.
[3] S. Catreux, V. Erceg, D. Gesbert, and R. W. Heath, Jr., Adaptive
modulation and MIMO coding for broadband wireless data networks,
IEEE Commun. Mag., vol. 40, no. 6, pp. 108115, Jun. 2002.
[4] A. Valenzuela, A. Rustako, and R. Roman, Distribute antennas for
indoor radio communications, IEEE Trans. Communications, vol.
COM-35, no. 12, pp. 12451251, Dec. 1987.
[5] M. Fabbri and P. Faccin, Radio over fiber technologies and systems:
New opportunities, in Proc. 9th Int. Conf. Transparent Optical NetworksICTON 2007, Jul. 15, 2007, pp. 230233.
[6] R. Ibernon-Fernandez, J.-M. Molina-Garcia-Pardo, and L. Juan-Llacer,
Comparison between measurements and simulations of conventional
and distributed MIMO system, IEEE Antennas Wireless Propag. Lett.,
vol. 7, 2008.
[7] M. Alatossava, A. Taparugssanagorn, V.-M. Holappa, and J. Ylitalo,
Measurement based capacity of distributed MIMO antenna system
in urban microcellular environment at 5.25 GHz, in VTC Spring
2008, IEEE 67th Vehicular Technology Conf., May 1114, 2008, pp.
430434.
[8] B. K. Lau; M.A. Jensen, J. Medbo, and J. Furuskog, Single and multiuser cooperative MIMO in a measured urban macrocellular environment, IEEE Trans. Antennas Propag., vol. 60, no. 2, pp. 624, 632,
Feb. 2012.
[9] V. Jungnickel, S. Jeckel, L. Thiele, L. Jiang, U. Kruger, A. Brylka,
and C. von Helmot, Capacity measurements in a cooperative MIMO
network, IEEE Trans. Veh. Tech., vol. 58, no. 5, pp. 23922405, Jun.
2009.
[10] L. Tarlazzi, P. Faccin, E. M. Vitucci, F. Fuschini, and V. Degli-Esposti,
Characterization of an interleaved F-DAS MIMO indoor propagation
channel, presented at the 6th Loughborough Antennas & Propagation
Conf. 2010 (LAPC 2010), Loughborough, U.K., Nov. 89, 2010.
[11] V. Degli Esposti, D. Guiducci, A. deMarsi, P. Azzi, and F. Fuschini,
An advanced field prediction model including diffuse scattering,
IEEE Trans. Antennas Propag., vol. 52, no. 7, pp. 17171728, Jul.
2004.
[12] C. Mehlfhrer, M. Wrulich, J. C. Ikuno, D. Bosanska, and M. Rupp,
Simulating the long term evolution physical layer, presented at the
17th Eur. Signal Processing Conf. (EUSIPCO 2009), Glasgow, Scotland, Aug. 2009 [Online]. Available: http://publik.tuwien.ac.at/files/
PubDat_175708.pdf
[13] J. Capmany and D. Novak, Microwave photonics combines two
worlds, Nature Photon., vol. 1, pp. 319330, 2007.
[14] C. Oestges and B. Clerckx, MIMO Wireless Communications: From
Real-Word Propagation to Space-Time Code Design. Oxford, U.K.:
Academic (Elsevier), 2007.
[15] V. Degli-Esposti, F. Fuschini, E. M. Vitucci, and G. Falciasecca, Measurement and modelling of scattering from buildings, IEEE Trans. Antennas Propag., vol. 55, no. 1, pp. 143153, Jan. 2007.
[16] V. Degli-Esposti, V. M. Kolmonen, E. M. Vitucci, and P. Vainikainen,
Analysis and modeling on co- and cross-polarized urban radio propagation for dual-polarized MIMO wireless systems, IEEE Trans. Antennas Propag., vol. 59, no. 11, pp. 42474256, Nov. 2011.
[17] F. Mani, F. Quitin, and C. Oestges, Directional spreads of dense multipath components in indoor environments: Experimental validation of
a ray-tracing approach, IEEE Trans. Antennas Propag., vol. 60, no. 7,
pp. 33893396, Jul. 2012.
[18] E. M. Vitucci, F. Fuschini, and V. Degli-Esposti, MIMO channel characterization through ray tracing simulation, presented at the 1st Eur.
Conf. Antennas and Propagation (EuCAP 2006), Nice, France, Nov.
610, 2006.
[19] S. Nanba, Y. Hirota, and Y. Kishi, Channel capacity estimation for
2 2 MIMO considering received power imbalance and antenna correlation, presented at the 6th Eur. Conf. Antennas and Propagation
(EuCAP 2012), Prague, Czech Republic, Mar. 2630, 2012.
VITUCCI et al.: INTERLEAVED-MIMO DAS FOR INDOOR RADIO COVERAGE: CONCEPT AND PERFORMANCE ASSESSMENT
Pier Faccin received the M.Sc. degree in electronic engineering from the University of Bologna,
Bologna, Italy, in the 1986.
He was a fellow Researcher at Fondazione
Guglielmo Marconi, Bologna, working on high
power modal converters for corrugated circular
waveguides. He then joined the Italian Navy Research Laboratory Giancarlo Vallauri, Livorno,
Italy, and he was involved in radar cross section
measurements of aero-naval moving targets. He
continued his career at Alcatel-Telettra, Chieti, Italy,
being involved in the R&D Laboratory dealing with military spread-spectrum
radio-communications. Upon spending some years as Hardware Designer in
the industry of professional musical keyboards, in the 1998, he joined Tekmar
SistemiAn Andrew Company, today CommScope Italy, Faenza, as Core
Technology Manager. He has been in charge of the Optical Research Department within the CommScopes Distributed Coverage and Capacity (DCCS)
Business Unit, i.e., the Group division which develops, manufactures and
install optical product such as repeaters and distribution systems that provide
radio coverage and capacity solution for wireless networks. Challenged by the
development of new applications of Radio over Fiber (RoF) technology in the
fiber distributed antenna systems (F-DAS) field of application, he introduced
the wavelength multiplexing technology inside the worlds first F-DAS commercially available. The keynote technical theme of such implementation was
the usage of the Erbium Doped Fiber Amplifier (EDFA) for optical budget
restoring. Unstoppable technology enthusiast, together with his team, later on
he pioneered and successfully introduced, the concept of interleaved multiple
inputmultiple output (i-MIMO) for a smart positioning of the antennas in
a MIMO system. More recently, he has been in charge for standardization
activities of wireless repeaters inside the 3GPP and ETSI standard organizations. Aimed with passion for technical research, outside of his engineering
activities, he continuously maintained relationships with the scientific world.
He has been invited as a speaker to several conferences at international level
and, with his team, he has been involved in the 7th European Framework
Programs ICT-ALPHA Project. A talented researcher and inventor, he has
authored and coauthored more than 40 conferences and journal papers, four
patents and patent applications on optical and wireless communications. Today
he is working as freelance telecommunication professional.
3309