SEMI-DIRECTIONAL SMALL ANTENNA DESIGN
FOR UWB MULTIMEDIA TERMINALS
Christophe Roblin, Member, IEEE, Alain Sibille, Senior Member, IEEE, and Serge Bories
ENSTA, Paris, France – roblin@ensta.fr
Abstract—. The design and results of a new UWB printed
dipole antenna with a dielectric lens are presented.
Index Terms—UWB antennas, small antennas, printed
antennas.
I.
INTRODUCTION
T
HE design of small size and low cost UWB (Ultra
Wide Band) antennas, notably for mass product
consumer applications, is nowadays one of the
challenges for the “antennists” involved in the UWB
communications domain. The targeted applications
extend i.e. from LDR (Low Data Rate) systems such as
sensor networks, to HDR (High Data Rate) terminals –
fixed or nomadic – for Multimedia applications for
BAN/PAN (Body / Personal Area Network), Home
entertainment or WLAN.
For most of these applications (which are, it is
worthwhile to recall it, essentially full duplex), omnidirectional antennas are for obvious reasons the first a
priori choice. On the other hand, because of the severe
constraints compelled by the regulation authorities
(FCC i.e.), notably as regards the transmitted power
(precisely the EIRP), the link budget is of most concern,
even for LOS (Line Of Sight) short range
communications. It’s the reason why the improvement
of the link budget, even of a few dBs (typically 3 to 6)
is of main importance : the transmitted power being
restricted, the only way to proceed at the RF HW
(Hardware) level is to act at the receiver side, either by
reducing the noise figure or by introducing antenna
gain (or, of course, both). The latter solution should be
in general more cost effective because the UWB VLNA
(Very Low Noise Amplifiers) are not cheap.
Semi-directional antennas may be a very effective
approach in order to mitigate link budget while
preserving link robustness. In many scenarios there will
be LOS or quasi LOS between the transmitter and the
receiver, meaning that most of the channel energy will
be contained in the direct paths. Obviously maximizing
the antenna gain in the direction of this path will be
highly beneficial to the link budget. A further
consolidated argument stems from experiments on the
obstruction of the direct path by a human at 5 GHz [1],
resulting in about 10-15 dB extra attenuation. The
received level can be described by a two-path model
diffracted by the edges of the body. Clearly, provided
the person intercepted the main path well within the
antenna lobe, we can expect a good proportionality of
the signal to the antenna gain from such a two path
model.
However in many of the contemplated scenarios, we
cannot force the antenna(s) to have its main lobe be
properly aligned on this direct path. Consequently, since
gain is associated with directivity and directivity is
associated with angular filtering, there is a risk to lose
robustness whenever the main path shifts too far from
this main lobe. The consequence of that is a tradeoff.
Although the antenna may rotate and it is difficult to
ensure it will be properly aligned, in many instances it
is sensible to think it will be aligned “roughly”, i.e. in
the right half space or even in the right quarter of space.
Therefore a reasonable choice would be to specify the
antenna to be semi-directional, i.e. little gain on its back
side, and a main lobe covering about 120° in half power
azimuth beam width.
A simple design approach is to use a quasi omnidirectional planar radiator as will be described in
section II-A, and to add a dielectric lens in order to
redirect the major fraction of the radiation on the
substrate side of the circuit (§ II-B). Such techniques are
well-known for quasi-optical antennas, and present the
advantage to start from an existing design, and to be
relatively simple in terms of production feasibility in
particular through moulding of the lens.
II. ANTENNA DESIGN AND RESULTS
A. Quasi omni-directional planar dipole design
The design is based on several solutions developed in
our lab. for the DVB-T (with proper frequency scaling)
within the European IST project CONLUENT [2, 3]
(during which, notably, we developed cpw-fed elliptical
planar dipole UWB antennas) and on the work
published in [4].
The adopted geometry is a wide planar dipole fed with a
coplanar waveguide (milled over the bottom part of the
dipole) : this provide a first size reduction with respect
to a thin dipole. The use of a substrate of intermediate
permittivity (RO3006® with εr = 6.15 and thickness h =
1.27 mm) offers a further significant size reduction. The
"optimized" dimensions obtained after many
simulations with the electromagnetic software WIPL®
are (fig. 1) : L = 31 mm, W = 21.64 mm, Lb = 12.1 mm,
Lt = 10.4 mm, Ls = 8.5 mm, the size of the central gap
between poles being Lg = 0.6 mm and Wg = 5 mm ; for
the 50 Ω CPW feeding line, wcpw = 1.2 mm and gcpw =
0.22 mm.
partial Eθ gain on boresight varies from 0.9 to 3.3 dB in
the same band. The cross component Eϕ level is very
low for the whole bandwidth for ϕ = 180 or 0 deg, and,
in any case, lower than -20 dB over 3-6 GHz.
L
W
W
z
Fig. 4 : gain versus frequency
y
Lb
Ls
Lt
Fig. 1 : sketch of the 1st antenna
Note that the choice of the substrate was only dictated
for demonstration purpose, considering that many other
simulations have demonstrated that cheaper substrate
with relative permittivity in the range 4-5 (i.e. FR4)
provide as well interesting results.
Fig. 5 : measured return loss
The measured input bandwidth is typically [3-10.5]
GHz for a VSWR < 2.5 and [3.3-5.2] for a VSWR < 2.
Fig. 2 : simulated return loss
B. Semi-directional planar dipole-lens design
The antenna parameters have been “re-optimized” with
a dielectric lens (of various shapes) added. Its
permittivity should be close to the substrate one (εr =
6.15 was used for both in the simulations). Eventually,
the adopted shape is a half prolate ellipsoid (semi axis :
a = b = 11 mm, a < c = 25 mm).
The obtained characteristics seem adequate with respect
to wished performance, with an improved gain of ~5.7
dBi in the main lobe at 5 GHz (3-4 dB gain increase on
average).
a
c
Fig. 3 : Gain in azimuth (θ = 90 deg – xy plane)
The input bandwidth (VSWR < 1.92) is 2.9 – 10.5 GHz
(fig. 2). The FTBR (Front to back ratio) is less than 0.5
dB over the band 3-8 GHz (fig. 3 & 4). The ordinary
Fig. 6 : 3D view of the antenna with lens (εlens = 6.15)
Fig. 7 : simulated return loss (antenna with lens εlens = 6.15)
Fig. 11 : measured return loss (antenna with lens εlens = 2.33)
Fig. 8 : Simulated front and back gains (antenna with lens)
Fig. 12 : measured effective gain in azimuth over [3-8] GHz
(Dipole with dielectric lens εlens = 2.33).
5.17
dBi
Fig. 9 : Simulated 3D gain pattern at 5 GHz (antenna with
lens)
The following results are concerned with a “variant” of
the preceding with a plexyglass lens (εlens = 2.3) instead
of the chosen material (Nylon-type), not yet available.
Although this results in a significant increase of the lens
size, an optimized prototype was fabricated to validate
the principle.
Fig. 13 : Dipole with lens (εlens = 2.33) ; measured gain versus
frequency : effective gain (front : plain blue – back : red
circles) & ordinary gain (front : green dashed-dotted).
Fig. 10 : Simulated gain versus frequency (εlens = 2.33).
As can be seen (fig. 11), the antenna with lens offers a
gain bandwidth which is closer to [3.5-8] GHz than to
the targeted [3-6] GHz. This obvious discrepancy with
the simulation, especially over the low frequency part,
is due on one hand (but probably slightly) to the fact
that the simulation is loss free, and on the other hand to
the difficulty of modeling of the feeding part and, in the
same time, of the measurement particularly in the case
of small antennas. Actually, small antennas without a
“real” screening ground plane are difficult to design,
simulate and characterize because, particularly for the
lower frequencies, they radiate “as a whole”, the close
environment being strongly involved, notably the
feeding cable.
Nevertheless, it can be observed (fig. 13) that, apart
from this “frequency shift”, the achieved gain and front
to back ratio (FTBR ~= 6 dB on average) are fully
satisfactory.
III.
ANTENNA COMPARISONS AND DEMONSTRATION
WITH THE ULTRAWAVES PLATFORM
Several tests of antennas performance have been carried
out with the UWB communication platform developed
in the Ultrawaves IST project [5]. The involved
antennas were commercial ones (Skycross and others
not presented here) and prototypes designed at ENSTA
and ENST Paris (in collaboration). A part of the results
is summarized in the following table.
The acronyms uses are :
• SKY : Skycross (commercial) antenna
• SKYG : Skycross antenna with a small home
made added ground plane
• FPF : F-probe fed planar triangular patch
(ENST Paris)
• BIC : shaped small bicones [6] (ENSTA)
• PDDL : Planar Dipole Dielectric Lens
presented antenna (ENSTA)
IV. CONCLUSION
The realized antenna gives satisfactory results, in
particular as regards the tests performed with an UWB
platform in realistic situations.
The results concerning the targeted antenna (with a
smaller lens of higher permittivity) will be presented at
the conference.
ACKNOWLEGMENTS
This work has been partly funded by the European
Union within the framework of the IST
ULTRAWAVES project [5]. We address our thanks to
all partners and particularly to Bram van der Wal from
Philips Consumer Electronics (Eindhoven) for his
welcome to his lab. and as well Domenico Porcino for
his contribution to fruitful discussions.
REFERENCES
[1]
[2]
[3]
[4]
[5]
UWB
Antennas
SKY-FPF
Data Rate Range
comment
(Mb/s)
(m)
30
> 9.5 LOS ; OBST (body) :
~OK @ 3 m
SKY-PDDL
30
brief body obst ==> a few err
+ slight blocking (@ 3m)
BIC-PDDL
30
brief body obst. ==> a few err
+ slight blocking (@ 3m)
FPF-PDDL
30
> 9.5 LOS. < 7 m when obstructed
by a body (small errors)
SKY-SKY
48
6.5 LOS1 Æ episodic blocking &
Err ; hand obst. Æ loss of Rx
SKYG-SKYG
48
6.8 LOS Æ episodic blocking ;
hand obst. Æ No loss of Rx
SKY - BIC
48
6 LOS2 Æ episodic blocking &
Err ; hand Æ loss of Rx
SKYG-FPF
48
> 9.5 LOS
SKYG-PDDL3
48
> 8.4 LOS (~OK @ ~9m); hand Æ
No loss of Rx
FPF-FPF
48
> 9.5 LOS ; Obstructed by a body :
~< 5 m (small errors)
FPF - PDDL
48
> 9.5 LOS ; Obstructed by a body :
~< 5 m (small errors)
Table 1 : range results in the various antenna
configurations and data rates
1
Loss of link rather often
Loss of link episodically ; sensitive to proximity body perturbation
3
Planar Dipole with Dielectric Lens (ENSTA)
2
[6]
Jonas Medbo, Jan-Erik Berg and Fredrik Harrysson, Temporal
Radio Channel Variations with Stationary Terminal, COST 273,
TD(04) 183, Duisburg, 20-22 sept 2004 (also presented at
VTC2004-Fall in Los Angeles, September 26-29, 2004)
A. Guena, D. Zapparata, A. Sibille and G. Pousset, Mobile
diversity reception of DVB-T signals using roof or window
antennas, COST 273 TD(04) 014, Athens, Jan. 2004.
CONFLUENT (IST-2001-38402) project Deliverable D2.3,
Nov. 2003.
Y. Kim and D.H. Kwon, “CPW-fed planar ultra wideband
antenna having a frequency band notch function”, Electr. Lett.
Vol. 40 n°7, 2004.
A. Sibille, Ch. Roblin, S. Bories, X. Begaud and A.-C. Lepage,
Prototypes and analysis tools for ultra wide band antennas,
ULTRAWAVES (IST-2001-35189) project deliverable D6.1,
Nov. 2004.
H. Ghannoum, S. Bories, Ch. Roblin & A. Sibille, “UltraWideband slightly distorting bicone antenna for UWB channel
measurements”, to be presented at IWAT 2005, Singapour, Mars
2005.