176
IEEE ANTENNAS AND WIRELESS PROPAGATION LETTERS, VOL. 1, 2002
Spiral and Dipole Antennas for Indoor MIMO-Systems C. Waldschmidt, Student Member, IEEE, T. Fügen, Student Member, IEEE, and W. Wiesbeck, Fellow, IEEE
Abstract—Influences and characteristics of antennas are usually neglected in multiple-input multiple-output (MIMO) system simulations, but are important for the modeling of realistic transmission channels. This paper shows how antennas influence MIMO channels and presents a comparison of spiral and dipole antennas in MIMO systems. The comparison is based first on measurements and second on a stochastic channel model for MIMO channels. It is shown that the pattern of the single antenna elements strongly influences the MIMO capacity. Index Terms—Channel capacity, channel measurements, channel modelling, multiple-input multiple-output (MIMO) antennas, MIMO measurements.
I. INTRODUCTION
R
ECENT studies have shown the potential of exploiting the spatial properties of multiple input multiple output (MIMO) channels by utilizing several transmit and receive antennas. Antennas, usually modeled as isotropic radiators or neglected in MIMO system simulations, modify the transmission channel in reality. Antennas that radiate omnidirectional in the azimuth plane are considered to be the best solution for MIMO systems, since they offer the largest number of degrees of freedom to exploit the spatial properties of the channel. But especially in indoor scenarios channels have many propagation paths outside the azimuth plane and using them can possibly increase diversity of the channel. An isotropic radiator would illuminate the whole environment but is physically impossible. Thus the question arises, what antennas do perform best in indoor scenarios. Therefore, a measurement campaign with different antennas has been performed showing the influence of the antennas on the MIMO indoor channel. Additionally, an extended stochastic channel model, that takes an accurate modeling of the antennas into account, has been verified with the measurements. II. MEASUREMENTS Two different types of omnidirectional antennas were chosen for the measurements. First, half wavelength dipole antennas were used since they are commonly used as standard or reference antennas. This type of antenna radiates omnidirectionally in the azimuth plane and is linearly polarized. The elevation pat. Second, two-arm spiral antennas tern is described by were examined. They are circularly polarized and have an exManuscript received October 24, 2002; revised December 4, 2002. The authors are with the Institut für Höchstfrequenztechnik und Elektronik (IHE), Universität Karlsruhe (TH), Germany (e-mail:
[email protected]). Digital Object Identifier 10.1109/LAWP.2002.807787
tremely large bandwidth, only limited by the feeding network and the geometrical size of the antenna [1]. The antennas are robust against polarization mismatching due to arbitrary orientations of transmitter and receiver because of the circular polarization. Spiral antennas mainly radiate omnidirectionally 45 from the azimuth plane (see Fig. 1) and the gain is 5 to 6 dBi. The comparison of these two types of antennas allows to draw conclusions on the multipath richness outside the azimuth plane, since the spiral antenna mainly radiates 45 and the dipole mainly 0 from the azimuth plane. The measurement system consists out of a two channel vector network analyzer, an amplifier and coaxial switches. The time for one measurement is less than 0.1 s, which is essential to eliminate any time variance of the channel during the measurement. The minimum signal level is 15 dB above the noise of the measurement system. The measurements were performed in an office building, with concrete ceilings and concrete and wood covered walls. The average office size is 5 5 m. All measurements were done during night, when there were no persons inside the building, in order to reduce the time variance. The antennas were mounted onto a plastic board in linear arrangements. The whole measurement setup was placed at 20 different positions inside the building (no line-of-sight situation) and at each position measurements with , , and linear antenna arrangements with spacings of were carried out. The transmit and receive array consisted of four antennas each. The distance between transmitter and receiver was between 8 and 20 m. All measurements were done at 2 GHz. III. SIMULATION MODEL The channel model used for comparison is an extended version of the model described in [2]. This stochastic threedimensional double-directional channel model is based on ray-tracing simulations and measurement campaigns, done in the same building as the MIMO measurements described above. In addition to the channel impulse response of the multipath signal the stochastic channel model delivers the angle of departure of all paths at the transmitter and the angle of arrival of all paths at the receiver for the SISO (system with one transmit and receive antenna) case. With this knowledge the MIMO channel matrix can be estimated, as long as, first, the antenna spacings do not extend several wavelengths thus the same plane waves impinge at all antenna elements. Second, the distances between transmitter and receiver and any obstacles in the channel are large enough to assume plane waves impinging at the arrays. With these assumptions the MIMO channel matrix results from the coherent addition of the impinging waves at the
1536-1225/02$17.00 © 2002 IEEE
WALDSCHMIDT et al.: SPIRAL AND DIPOLE ANTENNAS FOR INDOOR MIMO-SYSTEMS
177
Fig. 1. Elevation radiation pattern and azimuth radiation pattern of spiral antenna at an elevation angle of 85. The null in the azimuth plane is explicable by the influence of the feeding network of the antenna.
different antenna positions. The channel impulse responses or in the flat fading case the channel coefficient are derived by placing virtual antennas in the vicinity of the SISO transmit and receive antennas. The phase difference of the incident plane waves at the different antenna positions is described by , illustrated in Fig. 2, which is a function of the angle of and the antenna positions arrival of the waves and
(1) The difference of the amplitude of the plane waves at the positions of the virtual antennas is neglected, in other words the amplitude is assumed to be constant for all antenna positions. The result is an estimation of the channel coefficient for all virtual receive and transmit antennas
Fig. 2. Plane wave impinging at the array. The phase difference of the plane waves at the different antenna positions is '.
1
(2) is the attenuation of the th path in the SISO where is its phase. The power azimuth profile channel and of the stochastic channel model is composed out of several Laplacian functions, each modeling a cluster of scatterers, see profile is also modeled by Laplacian Fig. 3. The elevation functions for paths with a short delay time , merging into a sinusoidal function for paths with a long delay time. With this extended channel model simulations where performed with regard to the antennas. The three dimensional radiation pattern of the antennas was calculated with a standard simulation tool based on the method of moments [3] and compared with the measured pattern.
9
Fig. 3. Azimuth and elevation # angle of departure and arrival in the stochastic channel model, that is based on ray-tracing simulations and measurements.
IV. RESULTS The instantanious capacity of MIMO transmission channels in the presence of spatially uncolored noise can be calculated by [4] (3)
where denotes the identity matrix, is the mean signal-tonoise ratio (SNR) of all receive antennas and is the minimum number of receive and transmit antennas. is the normalized channel matrix, whose elements are the channel impulse responses between the single antenna elements. denotes conjugate complex transpose. In order to assess the
178
Fig. 4.
IEEE ANTENNAS AND WIRELESS PROPAGATION LETTERS, VOL. 1, 2002
Comparison of measurements and simulations.
channel, usually a fixed SNR (see e.g., [5]) and a channel mais used. Additionally, trix, normalized with equally distributed mean attenuations of the elements of are assumed. But if comparing different antennas in the same scenario, equally distributed attenuations neglect the gain of single antenna elements. Besides, for a fixed SNR and a normalized channel matrix any interrelation between the mean attenuation and the correlation properties of , that strongly influence the is not normalized, that means the capacity, is neglected. If path loss and the gain of single antenna elements are included in , (3) can be written as (4) This formulation allows a comparison of simulations and is the total transmit power, which is equally measurements. distributed among all transmit antennas. The mean path loss for the simulations equals the mean path loss of all measurements with the dipole antennas. The reflection coefficient of both types of antennas is below 14 dB to eliminate errors due to mismatching. The result of the measurements and simulations is presented in Fig. 4. Due to the relatively small number of measurements (total 300) the 10% outage capacity has no significance, thus the mean value is given. For the simulations (2000 Monte Carlo simulations) both the mean and 10% outage capacity are given. It is clearly visible that the capacity decreases for small antenna spacings, since the elements of are strongly correlated. Basically two the channel matrix factors determine the capacity of a MIMO system. First, the
path loss and second the multipath richness. Dipole antennas lead to a slightly lower path loss than spiral antennas. On the other hand, for a fixed SNR spiral antennas slightly outperform dipoles, thus the multipath richness is higher. There are some antenna arrangements in which the high gain of the spiral antennas leads to lower attenuation of the channel, but for arbitrary orientations of the antennas that does not hold. V. CONCLUSION The comparison of spiral and half-wavelength dipole antennas in indoor scenarios has shown that concentrating the radiated energy into the azimuth plane is reasonable, though many propagation paths outside that plane exist. The extended stochastic model agrees reasonable well with the measurements and allows a comparison of other antenna-types and antenna arrangements. REFERENCES [1] E. Gschwendtner and W. Wiesbeck, “Multi-Service dual-mode spiral antenna for conformal integration into vehicle roofs,” in Proc. IEEE Antennas and Propagation, Salt Lake City, UT, July 2000, pp. 1532–1535. [2] T. Zwick, C. Fischer, and W. Wiesbeck, “A stochastic multipath channel model including path directions for indoor environments,” IEEE J. Select. Areas Commun., vol. 20, pp. 1178–1192, Aug 2002. [3] EMSS—EM Software and Systems, FEKO, Stellenbosch, South Africa, 2000. [4] G. J. Foschini and M. J. Gans, “On limits of wireless communications in a fading environment when using multiple antennas,” Wireless Personal Communications, vol. 6, no. 3, pp. 311–335, March 1998. [5] M. Jensen, “Characteristics of measured 4 4 and 10 10 MIMO wireless channel data at 2.4 GHz,” in Proc. IEEE Antennas and Propagation Society, vol. 3, Boston, MA, Mar. 2001, pp. 95–98.
2
2
