Proceedings of the 40th European Microwave Conference

Metamaterial matching of high-permittivity coatings

for 79 GHz radar sensors
Frerk Fitzek #1 , Ralph H. Rasshofer #2 , Erwin M. Biebl ∗3
#
BMW Group Research and Technology
Hanauer Str. 46, 80992 Munich
1
[email protected]
2
[email protected]
∗
Technische Universität München - Fachgebiet Höchstfrequenztechnik
Arcisstr. 21, 80333 Munich
3
[email protected]

Abstract— Future automotive radar sensors in the 79 GHz Metamaterials (MTM) could be a solution in this case
regime are intended to be integrated behind painted, plastic to match the bumper environment to free space. In order
fascia. This integration has to be carefully studied, concern- to accomplish this task, a suitable metamaterial has been
ing reflections coming from the bumper in line of sight of
the sensor. To reduce these reflections, impedance matching investigated in detail in [1] and [2]. How to match this kind
of multilayer structures with high permittivities (e.g. metallic of structure (fishnet) to the multilayered bumper is the aim of
paints) is mandatory. Besides conventional matching techniques, this paper. The mathematical model for stratified media allows
metamaterials are a possibility to realize a matching layer. Based direct comparison between theory and measurements for plane
on an adopted parameter extraction algorithm for multilayer and homogene materials [3]. Frequency selective surfaces are
dielectrics, a bandwidth evaluation of a possible metamaterial
design is shown. The measurement results show that for a used in [4] for a anti-ice radome. This publication has a similar
bandwidth up to 3.8 GHz the metamaterial outperforms the intention, but does not take additional paintings on top of the
benchmark. radome into account. Metallic paints, in particular, influence
the electromagnetic fields, because of their high permittivity.
I. I NTRODUCTION AND R ELATED W ORK
II. M ETALLIC PAINTS
Integration of 79 GHz sensors (77 to 81 GHz allocated The electromagnetic behaviour of stratified media depends
by the European Union replacing the 24 GHz ultrawideband on their index of refraction and therefore on their relative
radar sensors) into modern cars represents a challenging task permittivity and relative permeability. Metallic paints feature
since both electromagnetic and car body design constraints small particles of metal (e.g. aluminium), in which the elec-
have to be met. This problem is even more challenging tromagnetic field induces currents. This polarization effect
when different sensors have to be integrated, due to the results in an increased relative permittivity. The configuration
fact that hidden integration of the sensors behind a plastic of standard finishing is usually first a primer layer to planish
bumper is often required. As is clear from both theoretical the surface of the bumper. The second layer is the base paint
investigations and from previous experience with 76 GHz radar which might contain metal particles. The clear coat finilizes
systems, operation of mm-wave radar sensors behind painted, the paint on top of the base paint. In this paper we summarize
plastic fascia will lead to performance degradation if matching the behaviour of these three "paint" layers in one layer (called
aspects are not considered. In the millimetre wave regime, the in the following just "‘paint"’) as depicted in Figure 1. The
electromagnetic field propagates quasi-optically, which might general extraction and matching procedure is also valid for a
result in high reflection and loss factors for obstacles or media detailed layer assembly, even for multiple paintings.
in the line-of-sight. This is the case if the radar sensor is to
be integrated in the bumper or behind other fascia of the car. A. Measurements of metallic paints on substrate
The measurement setup consists of a vector network ana-
lyzer and includes corrugated horn antennas to generate plane
waves right after the aperture of the antenna. In fact, a Gaus-
sian beam will be forming while having the painted substrate
exactly positioned at its waist. Since the waist width is much
smaller than the prototype dimensions (15 cm×15 cm), the
wave probes a quasi-infinite layer which is one assumption
for the simulation. Additionally a time-domain gating is per-
formed to exclude multiple reflections coming from the mea-
Fig. 1. Assembly of MTM for metallic paint matching surement setup. The measured S21meas includes the effects of

978-2-87487-016-3 © 2010 EuMA 1401 28-30 September 2010, Paris, France

the whole setup and hence is normalized to the S21FS element
 
obtained from the same measurement setup without the painted
substrate (free-space measurement). Similarly, the measured 
S11meas is normalized to the SPEC
11 element obtained from the

UHIOHFWLRQDQGWUDQVPLVVLRQG%

same measurement setup with a “PEC layer”. Therefore, the
scattering parameters characterizing the painted substrate can 

be calculated as [5], 
S21meas
S21 = (1) 
S21FS
S11meas 
6PHDVXUHPHQW
S11 = (2)
S11PEC 
6PHDVXUHPHQW
6PRGHO
Due to reciprocity S12 is equal to S21. The reflections S22 and 6PRGHO
 
     
S11 differ due to the losses in the media (see Figure 2). S11 IUHTXHQF\*+]
describes the reflection of the wave coming from the sensor
and hence might disturb the sensor. The reflection S22 is not
Fig. 3. Measurement and model data from 76 - 81 GHz
relevant for the radar system since it never reaches the sensor.

the start parameters for an optimization algorithm, which tries

UHIOHFWLRQDQGWUDQVPLVVLRQG%

  to minimize the magnitude, phase and thickness errors. The

thickness is hereby also a degree of freedom whereas it is fixed
6PHDVXUHPHQW

6PHDVXUHPHQW (d2,meas = 0.12 mm) for the analytical parameter extraction.
6PHDVXUHPHQW
 6PRGHO The general procedure to extract the material parameters for
6PRGHO multilayered media is to measure and extract only the first
6PRGHO
 
       layer with the decribed method. Afterwards the measurement
IUHTXHQF\*+] of the two layer structure is done. With a simple S-to-T
 
transformation, the material parameters of the second layer
 can be isolated. The measurement for the two layer structure
SKDVHGHJ

 is summarized in
 
 S11 S12
Smeas = , (3)
  S21 S22
      
IUHTXHQF\*+]
which results in the corresponding Tmeas . With the known
model parameters of the first layer the scattering parameter
Fig. 2. Measurement and model data from 60 - 90 GHz can be defined as
S11substrate S12substrate
 
Figure 3 shows the measured and simulated S21 and S11. Ssubstrate = . (4)
mod
S21substrate S22substrate
To get the proper model parameters, an extraction method has
to be derived, which is described in the next section. The corresponding T-matrix is Tsubstrate
mod . The T-matrix of the
paint (Tpaint
meas ) is calculated as
B. Parameter extraction for metallic paints
 −1
The parameter extraction algorithm is based on [6]. With Tpaint
meas = Tsubstrate
mod · Tmeas (5)
this algorithm the normalized impedance z and the index
of refraction n are calculated directly from the s-parameters. which is only a virtual measurement because it is not done in
Important to mention is the sign convention of e+jwt which reality. Nevertheless the S-parameter can be calculated out of
is used here. For this assumption the imaginary part of the Tpaint
meas as
index of refraction has to be negative for a passive medium.
S11paint S12paint
 
The imaginary parts of the relative permittivity and perme- Spaint = (6)
meas
S21paint S22paint
ability have to result in this negative imaginary part of the
index of refraction. This results in a decaying real part of These s-parameters are again used for the above described
the wave solution. The advantage of this algorithm is the extraction procedure. It is important to mention that possible
dispersive extraction of εr and µr for a known thickness errors in determing the parameters of the first layer result in
of the material if the measurement inaccuracies are small. erroneous determination of the parameter values of the sec-
Due to non perfect measurement, the extracted parameters ond layer. Also the measurement inaccuracies are completely
are erroneous. Nevertheless, the parameter range for the real comprised in the extracted parameters of the second layer.
solution is defined. The median of the values are used as But for phase accurate measurements, this extraction method

1402
delivers good estimates for the real solution. Figure 4 shows goal can be described as
the extracted parameters of the second layer (paint).  
MTM substrate paint 1 0
T ·T ·T = (7)
    0 1

which results in
QRUPDOL]HGLPSHGDQFH



LQGH[RIUHIUDFWLRQ


  
1 0 −1


TMTM = · Tsubstrate · Tpaint (8)
 0 1
 UHDOSDUW  UHDOSDUW

 
LPDJLQDU\SDUW
 
LPDJLQDU\SDUW
The resulting SMTM cannot be realized with passive strucures
             
IUHTXHQF\*+] IUHTXHQF\*+] due to the losses. Nevertheless it is possible to calculate
  
analytically the optimal solution for one given frequency.
 Finding the optimum for more than one frequency and for

UHODWLYHSHUPHDELOLW\
UHODWLYHSHUPLWWLYLW\

an easy manufacturing process is non-trivial, which leads to


 full-wave simulation tools where parameter ranges can be

predefined (e.g. the thickness of the matching layer). Based
 UHDOSDUW  UHDOSDUW on former studies [2] the MTM fishnet structure was designed
LPDJLQDU\SDUW LPDJLQDU\SDUW
 
       

      to exhibit a negative index of refraction around 79 GHz. This
IUHTXHQF\*+] IUHTXHQF\*+]
structure is now used for matching the complete structure
(MTM and painted bumper) to freespace. Therefore the pa-
Fig. 4. Parameter extraction
rameters of the structure wh , wv , lv and lh are optimized to
exhibit low reflection for a given goal. The optimization goal
As decribed before, it is assumed that the original three is to minimize the maximum reflection and to maximize the
layers (primer, base paint, clear coat) can be described in the minimum transmission at the band borders 76 and 81 GHz.
RF domain as one layer (called paint) with the same behaviour.
A. Simulation Results
C. Model parameters The simulation setup uses periodic boundary conditions
Because the measurement inaccuracies directly affect the (PEC and PMC) to insure infinite reproducibility of the cell
extracted parameters, the frequency selective parameters are in the desired directions, leading to the simulation of an
erroneous. Hence, assumptions for the real parameters have equivalent infinite MTM layer. To calculate the s-parameters
to be made, which depends on the polarization effects of the of the structure, the transient solver from CST Microwave
high permittivity coatings. Considering a non-magnetic, non- Studio [8] is used.
dispersive material, an optimization algorithm minimizing the
least-sqare error of magnitude and phase of the wohle setup
and the thickness of the paint layer (degrees of freedom: εr
and d) results in εr,2 = 11.5 − 0.26j with d2 = 0.12 mm.
As mentioned before the measured thickness of layer two
d2,meas was also 0.12 mm. The extracted model parameters are
summarized in Table I and are further used for the simulation.

Fig. 5. Simulation setup

TABLE I
M ODEL PARAMETERS The simulated results are depicted in Figure 6 in comparison
Layer d εr to the benchmark (s-parameter of the simulation are deembed-
Substrate 3.01 mm 2.86 - 0.06j ded and unnormalized). It is clearly shown that the matching
Paint 0.12 mm 11.47 - 0.26j can be improved. The worst case reflection within the whole
5 GHz band is worse than that of the benchmark. On the other
hand at 79.5 GHz the reflection could be decreased by over
III. M ETAMATERIAL M ATCHING 20 dB.

In [7] the authors showed that for metallic paints with B. Measurement Results
higher permittivities (condition depends on the permittivity of The measured S21 and S11 are depicted in Figure 7.
the substrate) the thickness optimization (integrated matching In comparison to the simulation results obtained with the
layer) is the optimal solution considering low-cost matching given parameters the resonance frequency is shifted by ap-
possibilities. This matching is the benchmark for further proximately 500 MHz to 80 GHz. With certain tolerances for
matching designs and is shown in Figure 2 and 3. In terms the parameters within the tolerance range (see Table II) the
of T-matrices of the whole setup (see Figure 1) the matching resonance frequency shifts to 79.5 GHz in the simulation (see

1403
 0  

-5 

UHIOHFWLRQXQGWUDQVPLVVLRQG%
reflection and transmission / dB

% *+]
-10 

% *+]
-15 

% *+]
-20 S21 w/o MTM 
S11 w/o MTM
S21 with MTM PHDVXUHG6ZLWK070
S11 with MTM PHDVXUHG6ZLWK070
-25  
76 77 78 79 80 81 82 83        
frequency / GHz IUHTXHQF\*+]

Fig. 6. Painted bumper with (simulated) and w/o MTM matching (measured) Fig. 7. Measurement results of the complete structure (MTM with bumper)

TABLE II
PARAMETER VALUES FOR THE DIFFERENT MATCHING DESIGNS .

Parameter MTM Matching Design Tolerances

wv 1 mm 0.50 mm ± 0.02 mm
lv 2 mm 2.02 mm ± 0.02 mm
wh 2.4 mm 2.25 mm ± 0.02 mm
lh 1.6 mm 0.34 mm ± 0.02 mm
t 254 µm 254 µm -
εr,s 2.2 2.2 -
tan δ 0.0009 0.0009 -
tCu 17 µm 17 µm - Fig. 8. Manufactured MTM and matching design (15 cm × 15 cm)

ACKNOWLEDGMENTS
Figure 6). The mechanical tolerances have the main influence
on the resonance frequency. Additionally, the simulation re- The authors wish to thank the German Federal Ministry of
sults are sensitive to the losses of the media, which strongly Education and Research (BMBF) for the funding of the RoCC
depends on the accuracy of the measured phases. However, joint project and Continental AG, namely Dr. J. Massen and
the reflections of the measurement are below -10 dB for a U. Möller for supporting them with measurement equipment.
bandwidth of 3.8 GHz, below -15 dB for a bandwidth of 2 GHz R EFERENCES
and even below -20 dB for a bandwidth of 1 GHz. This shows
[1] M. Kafesaki, I. Tsiapa, N. Katsarakis, T. Koschny, C. M. Soukoulis, and
that metamaterials can improve the matching of such bumper E. N. Economou, “Left-handed metamaterials: The fishnet structure and
assemblies for bandwidths up to 3.8 GHz. its variations,” Phys. Rev. B, vol. 75, pp. pp. 235 114–(1–9), 2007.
[2] F. Fitzek, Z. Abou-Chahine, R. H. Rasshofer, and E. M. Biebl, “Auto-
motive radome design - fishnet structure for 79 GHz,” in Proc. German
IV. C ONCLUSIONS Microwave Conference (GeMIC), 2010.
[3] S. K. Kurt Altenburg, Wellenausbreitung in geschichteten Medien bei
senkrechtem Einfall und die Anwendung auf Leitungstheorie, elektrische
This paper showed that metamaterials can be used for Wellen, Optik, Akustik, Wellenmechanik sowie mechanische und elek-
multilayer radome matching especially for high permittivity trische Vierpolketten, B.-A. Deutsche Akademie der Wissenschaften,
Laboratorium für Kunststoffe, Ed. Annalen der Physik, 1953, vol. 448,
coatings like metallic paints. To extract the material parameters no. 1-5.
of paints, an adopted extraction method is shown. With the [4] G. Gafford, P. H. Gebert, W. S. Arceneaux, and R. D. Akins, “Anti-ice
benchmark from former studies, a bandwidth evaluation of radome,” US Patent 5 528 249, 1996.
[5] A.Elhawil, L. Zhang, J. Stiens, C. D. Tandt, N. A. Gotzen, G. V.
the metamaterial is shown for a standard bumper assembly Assche, and R. Vounckx, “A quasi-optical free-space method for dielectric
consisting of a substrate material and some paint layers (stan- constant characterisation of polymer materials in the mm-wave band,” in
dard finishing / worst case painting with highest permittivity). Symposium of IEEE-LEOS-Benelux, 2007, pp. 187 – 190.
[6] X. Chen, T. M. Grzegorczyk, B.-I. Wu, J. Pacheco, Jr., and J. A.
Concerning the worst case reflection within the band from Kong, “Robust method to retrieve the constitutive effective parameters
76 GHz - 81 GHz, the benchmark shows better results because of metamaterials,” Physical Review E, vol. 70, pp. 016 608–7, 2004.
the metamaterial suffers from the highly resonant behaviour. [7] F. Fitzek, R. H. Rasshofer, and E. M. Biebl, “Comparison of matching
layers for automotive radome design,” in Advances in Radio Science
Nevertheless, for constrained bandwidth requirements (e.g. up (ARS), 2010.
to 3.8 GHz) the maximum reflection can be further reduced in [8] Computer Simulation Technology GmbH, “CST Studio Suite 2010,”
comparison to the benchmark, even below -20 dB. http://www.cst.com, 2010.

1404