3D Beamforming Trials with an Active Antenna

Array
Johannes Koppenborg, Hardy Halbauer, Stephan Saur, Cornelis Hoek
Alcatel-Lucent Bell Labs
Stuttgart, Germany
{Johannes.Koppenborg, Hardy.Halbauer, Stephan.Saur, Cornelis Hoek}@alcatel-lucent.com

Keywords- cellualar mobile radio; vertical beam steering;

downtilt adaptation; 3D beamforming

I.

INTRODUCTION

Beamforming and MIMO schemes with multiple antennas

at the transmitter are widely used to form the horizontal beam
pattern of mobile wireless base stations (eNodeB). Target is to
improve the signal strength at the receiver or to enable
increased capacity due to spatial multiplexing. Also beam
coordination schemes to minimize inter-cell interference
through appropriate scheduling of the individual terminals
(UE) on appropriate beams or precoding vectors are feasible.
For this the measured interference and scheduling decisions of
the adjacent cells need to be taken into account. This became
feasible with the extensions of the Third Generation
Partnership Project (3GPP) Long Term Evolution Advanced
(LTE-A) standard, enabling message exchanges between the
eNodeBs through the so called X2 interface. However, the
eNodeB antennas usually have a fixed vertical antenna pattern
and are mounted with a fixed downtilt to statically optimize
cell coverage. The beam pattern adaptation is achieved through
multiple antenna elements arranged horizontally and fed with
digitally preprocessed data streams. In consequence the
capabilities of interference avoidance and throughput
optimization are also restricted to the horizontal plane.
With 3D beamforming we introduce the capability to adapt
the beam pattern individually for each UE also in vertical
direction. This allows exploiting an additional degree of

freedom for beam coordination and serving of multiple users,

leading to an additional increase of throughput and coverage.
In this paper the basic principle of 3D beamforming will be
described and an overview on related theoretical work will be
given. As a major scope of this paper practical implementations
and trial setups for proof-of-concept will be shown. The
presented lab and field trials will verify the potential of 3D
beamforming in combination with practically implemented
algorithms in real propagation environment.
II.

PRINCIPLE OF 3D BEAMFORMING

A. Overview
Adaptation of the vertical beam pattern in addition to the
horizontally applied multi-antenna scheme is the key element
for the extension towards 3D beamforming. But whereas in
horizontal direction usually an angle between -60 to +60
relative to the eNodeB boresight must be covered, the situation
in vertical direction is different. For the example of an inter-site
distance (ISD) of 500 m and a eNodeB height of 30 m, a
simple geometrical consideration (Figure 1) shows that the
minimum downtilt at cell edge is 6, and 95% of the downtilts
needed to cover the cell area are below 20.
1
CDF (P downtilt < x-axis)

Abstract-- Beamforming techniques for mobile wireless

communication systems like LTE using multiple antenna arrays
for transmission and reception to increase the signal-to-noiseand-interference ratio (SINR) are state of the art. The increased
SINR is not only due to a larger gain in the direction of the
desired user, but also due to a better control of the spatial
distribution of interference in the cell. To further enhance the
system performance not only the horizontal, but also the vertical
dimension can be exploited for beam pattern adaptation, thus
giving an additional degree of freedom for interference avoidance
among adjacent cells. This horizontal and vertical beam pattern
adaptation is also referred to as 3D beamforming in the
following. This paper describes investigations of the potential of
3D beamforming with lab and field trial setups and provides
initial performance results.

0,8
0,6
0,4
0,2
0
0

10

20
Downtilt []

Figure 1. CDF of downtilts in a multicell system

(ISD = 500m, eNodeB height = 30 m)

30

This indicates that the adjustment range of the vertical

beam pattern in a typical macrocell deployment is much less
than in horizontal direction, and it will be even smaller for
lower eNodeB heights. Further, the vertical beam usually is
much smaller than the horizontal beam, the half-powerbeamwidth (HPBW) is typically in the 5 to 10 range. In
combination with UE specific vertical beam adaptation this
requires a higher accuracy of the downtilt adjustment, but also
helps to improve vertical beam separation.
One possible way to built antennas with narrow vertical
beam is to vertically arrange a high number of transmitting
elements and feed them with the same signal with a proper
phase shift between the transmitting elements. The generic rule
for the resulting beam angle and the phase shift is given by

,
2 d

= arcsin

where is the beam angle with respect to boresight direction,

d is the distance between the transmitting elements,
represents the phase shift between adjacent transmitting
elements and is the wavelength of the RF carrier. The
individual antenna element can be realized as vertically or
horizontally polarized dipole or as group of two cross-polarized
orthogonal dipoles.
boresight direction

d
direction of beam with
downtilt

antenna elements
RF transmission
signal

Figure 2. Principle of beamforming antenna

Antennas with fixed downtilt usually realize the required

phase shift between the transmitting elements with a passive
feeder network. To achieve a dynamic adaptation, as required
for 3D beamforming, this approach is not suitable. For each
transmitting element each frequency subband intended for a
specific UE must be available with its specific phase shift. This
can be achieved e.g. with individual baseband processing for
each transmitting element, applying different phase shifts for
different subbands according to the selected downtilts. New
upcoming hardware concepts using individual active antenna
elements now enable the realization of such advanced 3D
beamforming schemes and allow practical verification of the
theoretically investigated benefits.
B. Simulation
The concept to operate a cellular communication system
with dynamic adaptation of the vertical beam on a per UE basis

to reduce interference, and the option to combine it with an

interference coordination scheme was firstly published in the
scope of the ARTIST4G project [1]. In the following, different
operation modes for 3D beamforming have been proposed [2],
[3]. They are either based on different fixed downtilts, where
each downtilt value is appropriate for a certain region in the
cell, e.g. a steep downtilt for UEs close to the eNodeB, or
based on exact steering of the vertical main lobe towards the
location of the UE in order to maximize the received signal
power. The system parameters have been optimized with
respect to spectral efficiency and cell edge user throughput for
different cell sizes and HPBWs. [2], [3] also investigate the
option to combine vertical with horizontal beamforming. It has
been shown that even without any kind of additional
coordination a significant reduction of interference can be
achieved.
In [4], [5] a distributed scheduling scheme for beam
coordination was proposed. A simplified approach based on a
predefined strategy for the assignment of radio resources, also
referred to as implicit coordination, works without any
exchange of scheduling information among neighboring
eNodeBs. The key result of this analysis is that gains resulting
from 3D beamforming and interference coordination are almost
independent effects and are additive up to a certain amount. It
is also important to note that both spectral efficiency and cell
edge user throughput can be improved simultaneously with this
approach. Depending on the configuration of the system,
emphasis can be placed on one or the other performance
metric. Overall, we can achieve around 20% gain in spectral
efficiency and around 60% in cell edge user throughput.
The drawback of these results derived by pure simulation
analysis is the uncertainty with respect to the channel model.
Usual models like the 3GPP Spatial Channel Model Extended
(SCME) include multiple rays solely in horizontal direction. In
the vertical plane, a line of sight relation between eNodeB and
UE has to be assumed, i.e. effects like diffraction at roof tops
are not considered. Also construction based impairments of the
antenna hardware are not fully included in the simulation
models. Hence, we have initiated additional lab and field trials.
They aim at an assessment whether the predicted performance
gain derived from simulations can be achieved as well in
practice.
III.

LAB AND FIELD TRIALS WITH 3D BEAMFORMING

A. Description of Measurement Setup

An enhanced 3D beamforming prototype has been built
with an eNodeB emulator, an active antenna array (AAA) and
test UEs as shown in Figure 3. The AAA consists of 8 antenna
elements with 8 cross-polarized dipoles arranged in vertical
direction. This means the polarization of the beams is 45 as
used in LTE networks, and the used frequency is 2.6 GHz with
20 MHz bandwidth. This provides a static sector pattern in
horizontal direction, whereas in vertical direction a fully
flexible beam pattern per resource block can be formed
towards the UE location.
Test signals with 8 data streams with dedicated pilots per
antenna element are sent from the eNodeB emulator to the test

UEs via the active antenna array using vertical beam steering.
The received signals at the test UEs are analyzed and the
optimum beamforming weights for each subcarrier are reported
via a fixed link or a generic wireless link to the eNodeB and
applied for the following time interval. This setup allows lab
and field trials with quasi-real-time channel feedback in the
uplink and is therefore suitable for proof-of-concept field trials
covering different deployment scenarios and algorithms
designed to feed back optimum beam pattern weights.
20 MHz@2.6 GHz

eNodeB
Emulator

CPRI

(Baseband)

UE 0
UE 1

Uplink feedback

Figure 3. Principle view of 3D beamforming Set up

B. Lab Trials with 3D Beamforming

In a first step the behavior and performance of 3D
beamforming with the fully adaptive algorithm described
above has been tested in an indoor lab environment. The
antenna array was mounted on a rack, with a height of 1.8 m
for the center of the antenna array. Two test UEs (UE 0 and UE
1) are vertically mounted on two stages with a separation of
only 20 cm in a distance of 4 m from the antenna array. The
test UEs can be moved +/- 80 cm in vertical direction. This
means that the minimum distance of the two UEs is 20 cm, if
the vertical position is equal.

where the received signal power of both UEs is shown. Both

test UEs are operating with 20 MHz bandwidth and just to
visualize the excellent separation of the wanted and unwanted
signal the representation for each UE has been limited to 10
MHz. On the lower 10 MHz part of the spectrum the situation
is shown as seen for UE 0 (blue curve), and on the higher 10
MHz part of the spectrum the situation for UE 1 (green curve)
is shown.
In this test the suppression of the unwanted test UE is more
than 35 dB. Similar results can still be achieved when the two
UEs are mounted close together with a distance of only 20 cm
or even in the case of non-line-of-sight (NLOS) propagation
conditions. In each case the adaptive algorithm takes advantage
from the reflections from walls and surfaces in the room to
achieve excellent performance.
C. Field Trials with 3D Beamforming
In a second step proof-of-concept outdoor field trials were
conducted for a macro cell environment with high buildings.
For these trials the adaptive antenna array was mounted in a
height of 7 m at a wall of a building. The beams were steered in
direction of a straight street with presumably almost line-ofsight (LOS) propagation conditions, see Figure 5.

Garbage
Container,
Metal fences
and trees

2
AAA mounted on
wall of building
(height over
street level 7 m)

Figure 5. Overview of field trial scenario (Source: Open Street View)

The two test UEs where located in a test van and connected
with two receive antennas mounted on the roof of the van with
a separation of 50 cm.

Figure 4. Received Power for two UEs by using a zero forcing algorithm

For a typical arrangement with a vertical separation of only

30 cm for the two test UEs a very good separation of the two
beams can be achieved. This result is illustrated in Figure 4,

Figure 6. Separation of the wanted and unwanted signal for LOS condition

For a first test with LOS condition the antennas where

adjusted in the driving direction of the test van at position 1 in
Figure 5, which was parked in a distance of approximately 20
m from the eNodeB antenna. It could be expected that a strong
LOS component in the beams in such a straight direction
between eNodeB antenna and van is present. But even for this
LOS condition there are additional multipath components due
to reflections as visible in the frequency response ripples
shown in Figure 6. The signal of the two UEs could be
separated reliably. The suppression of the unwanted test UE is
depending on the frequency and ranges between 10 and 15 dB
as shown in figure 6.

the second test UE was mounted on street level about 5 meters

away at a position with a strong LOS component. As can be
seen in Figure 8, the suppression of the unwanted signal is
about 15 dB for UE 1 with the strong LOS component and
below 5 dB for the mobile with NLOS condition. A reason for
this asymmetry is that the received signal of the mobile with
the LOS component (UE 1) is stronger than the signal of the
UE with NLOS condition (UE 0). This behavior indicates that
the adaptive algorithm is working as reliable as before, but due
to the different power levels the achieved advantage is smaller.

Figure 9. Separation of the two signals for non line of sight condition for one
UE and line of sight to the other with beam reflected from a wall

Figure 7. Separation of the two signals for non line of sight condition

In a second scenario the test van was parked in a distance of

about 30 m from the eNodeB antenna array (position 2 in
Figure 5). It is ,blocked by metal fences, metal garbage
container and trees. Thus, there wasnt a strong LOS
component present any longer. Again the two receive antennas
on the roof of the test van had a separation of 50 cm. As
shown in Figure 7, the suppression of the unwanted signal is
varying between 0 and 20 dB depending on the frequency. In
situations like this, system improvements are still feasible, but
would require frequency selective scheduling in addition.

Finally, in a fourth scenario, the transmitting antenna array

at the eNodeB has been rotated, so that the boresight direction
of the horizontal pattern pointed towards a high building
nearby. In this scenario we expected strong reflections in the
direction of the two test UEs. The arrangement of the receive
antennas at the test UEs was kept as in the third scenario. The
modified propagation conditions resulted in an improved
performance for both UEs with a suppression of the unwanted
signal of about 15 dB as shown in Figure 9.
The above presented trials in a typical outdoor environment
show that quite high signal suppression can be achieved for
fully adaptive algorithms. The suppression depends on the
surrounding like metal fences, reflections from buildings and
so on. Additional performance improvements are possible by
using frequency selective scheduling. The best potential for
signal separation is seen in scenarios with a high amount of
reflections.
IV.

CONCLUSION AND OUTLOOK

Simulations have shown that the performance of cellular

mobile communication systems can be significantly improved
by adapting the vertical dimension of the antenna pattern at the
eNodeB individually for each UE according to its location.
With first lab and field trials we provide evidence of the
expected gains in real indoor and outdoor deployments with
both LOS and NLOS propagation conditions.
Figure 8. Separation of the two signals for non line of sight condition for
one UE and line of sight to the other

In a third scenario the test van stayed at the same position

(point 2 in Figure 5) as before with the NLOS condition, but
with only one antenna on the roof of the van. The antenna of

The conducted lab and field trials with adaptive vertical

beamforming algorithm have shown significant improvements
of the system performance with respect to the capability to
separate two signals sharing the same radio resources by taking
advantage of reflections from building, walls and other strong
reflectors.

The trials have been done with all impairments of a real

wireless system like phase jitter, time synchronization and
power amplifier distortion. Two UEs could be separated even if
the channels for the both UEs were very similar in a LOS
scenario. Consequently, the improved SINR values allow
higher throughput for the addressed UEs and an overall
increased system capacity
The trial set-up will be enhanced with multiple cell
arrangements for testing coordinated multipoint (CoMP) and
network MIMO algorithms. Enhanced feedback compression
algorithms will be studied to reduce the uplink signaling load.

REFERENCES
[1]

[2]

[3]

[4]
[5]

ACKNOWLEDGMENT
The research leading to these results has received funding
from the European Commissions seventh framework
programme FP7-ICT-2009 under grant agreement n 247223
also referred to as ARTIST4G.

[6]

ARTIST4G consortium, D1.1 Definitions and architecture

requirements for supporting interference avoidance techniques, project
report, December, 2010.
ARTIST4G consortium, D1.2 Innovative advanced signal
processing algorithms for interference avoidance, project report,
December 2010.
S. Saur, H. Halbauer, Exploring the Vertical Dimension of Dynamic
Beam Steering, 8th International Workshop on Multi-Carrier Systems
and Solutions, May 3-4, 2011, Herrsching, Germany.
ARTIST4G consortium, D1.3 Innovative advanced signal processing
algorithms for interference avoidance, project report, March, 2011.
H. Halbauer, S. Saur, J. Koppenborg, C. Hoek, Interference Avoidance
with Dynamic Vertical Beamsteering in Real Deployments, 4G Mobile
Radio Access Networks Workshop at IEEE Wireless Communications
and Networking Conference (WCNC 2012), April 1, Paris France.
ARTIST4G consortium, D6.2 Laboratory and field trial results
connected to the first set of innovations, project report, September,
2011.