Performance of lens antennas in wireless indoor millimeter-wave ...

732

IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 47, NO. 6, JUNE 1999

Performance of Lens Antennas in Wireless Indoor Millimeter-Wave Applications Carlos A. Fernandes, Member, IEEE, and José G. Fernandes, Member, IEEE

Abstract—Dielectric lens antennas can be designed to produce highly shaped beams that significantly improve the system performance in emerging wireless indoor millimeter-wave systems. A lens configuration is analyzed in this paper that produces a circularly symmetric cell with uniform spatial power distribution, fairly sharp boundaries, and scalable cell radius. The last characteristic is used to control the reflections at sidewalls. A hemispherical coverage lens antenna is designed for the mobile terminal (MT) to ensure relatively free movement. The impact of these antennas is analyzed in terms of cell coverage and channel time dispersion, considering the effect of cell radius scaling, and MT antenna tilting. Measurements and simulations show that the proposed lens antennas outperform common solutions based on pyramidal horns or biconics. Index Terms—Lens antennas, millimeter-waves, shaped beams, wireless applications.

I. INTRODUCTION

T

HE wide acceptance of mobile services by the users and strong competition in the field has led to an explosive growth in cellular and mobile telephony in the last ten years. The increasing need for larger bandwidths to support broadband services has been a major driving force pushing the development of mobile/wireless broad-band systems, aiming to extend to the mobile users the wide range of services available in the broad-band integrated services digital network (B-ISDN). Due to the saturation of the lower part of the frequency spectrum, this type of system will operate in the millimeter-wave band. It has been shown that, at these frequencies, the antennas have a significant impact on the characteristics of the multipath radio channel and, therefore, they play a key role in system performance [1]–[3]. The power available from solid-state devices is limited and, thus, it should be uniformly distributed over the cell while keeping the channel time dispersion at low values since it directly impacts on the achievable carrier bit rate (CBR). Adaptive equalization is unavoidable in high bitrate systems, but it cannot cope with large time dispersion values often encountered in typical scenarios. Hence, some directivity is required both at the base and mobile station Manuscript received December 1, 1998; revised February 4, 1999. This work was supported in part by the Funda¸ca˜ o para a Ciência e a Technologia under Project PBIC/C/TIT/2501/95. C. A. Fernandes is with the Instituto de Telecomunica¸co˜ es-Pólo Lisboa, 1049-001 Lisbon, Portugal, and is also with the Instituto Superior Técnico, 1049-001 Lisbon, Portugal. J. G. Fernandes is with the Instituto de Telecomunica¸coes-Pólo Aveiro, 3810 Aveiro, Portugal, and is also with Universidade de Aveiro, 3810 Aveiro, Portugal. Publisher Item Identifier S 0018-9480(99)04280-5.

antennas to favor the link budget and to cooperate with the equalizer in the mitigation of the channel time dispersion. However, the antenna directive requirements must not entail a restriction of terminal mobility. In this paper, we analyze the impact of using a novel and quite inexpensive antenna configuration for indoor scenarios that is able to cope with the above requirements. The basic idea is to use a shaped dielectric lens antenna at the base station (BST) hanging from the ceiling, which produces a ) type of radiation pattern in the elevation secant squared ( plane in order to compensate for the free-space attenuation at each direction. The BST antenna is paired with an hemispherical pattern lens in the mobile terminal (MT) so that the average received power remains reasonably constant for all positions of the mobile or portable terminal within the cell. The radiation patterns of the BST and MT lens antennas are circular symmetric. The proposed lens combination further provides very sharp cell boundaries with negligible radiation outside the cell limits. patterns is that cell A remarkable characteristic of dimensions are scaled to the antenna height. This provides a simple means to control the illumination of the walls at the edges of the cell to maintain an adequate compromise between multipath effects and the need for alternative paths in case of line-of-sight (LOS) blockage. II. LENS ANTENNA The general design principles for axisymmetric amplitudeshaping dielectric lenses are addressed in [4]. In the present case, the lens is fed by the aperture of a circular metallic waveguide, which is embedded in the lens body. The lens surface is conveniently shaped to transform the feed aperture radiation pattern into the desired output beam. This antenna configuration is quite flexible, allowing the design for different target patterns from secant squared to hemispherical type, with linear or circular polarization. The lens antenna design involves a two-step procedure based on geometrical and physical optics. Specific software tools were developed, which allow a reliable calculation of the lens profile and prediction of the corresponding radiation ) characteristics. Plexiglas material ( was used in all prototypes described throughout, although other very low-loss commercially available materials ( ) with the same permittivity were tested for these type of lenses, giving similar radiation patterns, but higher gain. For linear polarization, the BST lens is excited by the mode of the circular waveguide, producing a circular

0018–9480/99$10.00  1999 IEEE

FERNANDES AND FERNANDES: LENS ANTENNAS IN WIRELESS INDOOR MMW APPLICATIONS

(a)

733

(b)

(a)

(c) Fig. 1. Elevation power patterns of shaped dielectric lens antennas fed by : ). (a) BST antenna circular waveguide, measured at 62.5 GHz (" 2 target pattern for pattern superimposed on ideal 01 excitation. (b) BST antenna pattern for 11 mode excitation. (c) MT pattern for 11 mode excitation.

TE

sec

r = 2 53

TM

TE

symmetric cell with a vertically polarized field ( component is dominant). Fig. 1(a) shows the corresponding elevation pattern measured at 62.5 GHz, superimposed on an target pattern. The lens diameter is 66 mm. Ground direction , and the horizon to . The corresponds to , which radiation is intentionally limited up to corresponds to 11-m-diameter cell when the height difference ) is . between BST and MT antennas ( characteristic, The measured pattern is close to the ideal , due to the intrinsic limitation of linear poexcept near larization in the axial direction of circular symmetric radiation patterns. Prototypes using circular polarization do not show this limitation. In this case, the lens is excited by the mode of the circular waveguide with circular polarization, producing a circular symmetric cell with circular polarization illumination. The resulting BST elevation pattern is presented in Fig. 1(b), showing a good coincidence with the desired characteristic for . -mode excitation was used for the MT The same antenna, although seeking a different lens output radiation pattern. The desired elevation characteristic in this case is flat , and no radiation outside this interval top up to in order to discriminate possible reflection paths. This flattop characteristic and the pattern circular symmetry favor free movement of the mobile within the cell limits, and even some tilting of the MT. Fig. 1(c) shows the measured pattern of

(b) Fig. 2. Received power measured along selected paths in test room A for h : m. (a) h : m. (b) h m.

m = 15

1 = 05

1 =1

a 66-mm lens that matches the above specification. For the corresponds to the ceiling direction and MT antenna to the horizon. Notice the steep fall of radiation in all the patterns of Fig. 1, which significantly for prevents multipath pick up. III. CELL COVERAGE A. Measurements Versus Simulation Results Continuous wave (CW) measurements were performed in 8.8 m 4.0 m room to evaluate the proposed a 10.9 m antenna configuration (room A). Ceiling and sidewalls are made of concrete and the floor is covered with 1-in-thick ceramic tiles. An anechoic chamber with laminated wood external walls, tables, shelves, and equipment racks are the main objects in this room (see Fig. 2). Different lens antenna combinations were used in the BST and MT. Fig. 2 shows the BST antenna location and the measured received power distribution along 12 linear paths taken in the uncluttered part of the room for antenna height differences

734


Fig. 5. Side view of the antenna arrangement with the BST at two different heights showing the LOS ray and first-order reflected rays on lateral walls. Fig. 3. NRP along path A in test room A for m and h m.

h

Fig. 4. NRP along path B in test room A for m and h : m.

h

1 =1

1 = 10

m = 1 5 m, with 1 = 0 5 :

h

:

m = 1 5 m, with 1 = 0 5 :

h

:

Fig. 6. Simulated NRP distribution over room B for h : m.

1 = 15

m and m. Circular polarization lenses were used in both the BST and MT. The circular symmetric nature of the cell and its sharp boundary are quite apparent. The received power level is reasonably constant within about m), and falls a 3.7-m-diameter region in Fig. 2(a) ( off rapidly outside this region. The cell diameter is doubled in m), showing the scalable nature of the cell. Fig. 2(b) ( The theoretical cell diameter comes directly from the simple geometry of the antenna configuration (see Fig. 5) and is given by (1) is the fall-off direction of the radiation pattern where for the present lenses). A quantitative measure ( of this behavior is given in Figs. 3 and 4, which correspond, respectively, to paths A and B, marked in Fig. 2. Figs. 3 and 4 present the received power normalized to the transmitter power level (NRP). Simulation results obtained with a ray tracing tool, described in [5], are superimposed on the measured data, showing a satisfactory agreement. Despite the omnidirectional characteristic of BST and MT antennas, fading depth within the cell is negligible due to weak illumination of the walls for values, as illustrated in Fig. 5. low A more detailed observation of the received power level in Figs. 3 and 4 shows that the flat top is not horizontal. This is due to a 3 mislevelling of the BST lens. This fortuitous tilting was taken into consideration in the ray-tracing tool simulations. It is stressed that a small tilt of the BST antenna

h

m = 15 m :

and

yields a visible deformation of the cell boundary. Taking (1), it can be easily shown that an error of a single degree, e.g., to , gives a cell diameter of 8 m instead of from m. This suggests that highly shaped 7.4 m for lenses must be carefully aligned for perfect cell symmetry. The power level difference of about 5–6 dB between the m and flat top parts of the curves obtained for m is mainly due to the factor of two in the direct ray since the reflected components are path length, i.e., significantly attenuated by the antennas radiation pattern. B. Further Simulation Results Further simulations were performed in an empty, but otherwise similar, room, called room B, for different BST antenna heights ( ) and MT antenna tilting angles ( ) to demonstrate the performance of this antenna configuration, as well as the allowed user movement freedom. The simulation uses measured antenna radiation pattern data, and the MT antenna height ( ) is kept constant at 1.5 m. Fig. 6 shows the NRP m. The cell is wider than distribution calculated for in Fig. 2(b) by a factor of 1.5, as expected from (1). Previous conclusions about the uniformity of power distribution and sharp cell boundary apply again. The corresponding cumulative distribution function (CDF) curve in Fig. 7 reveals that the NRP is above 73 dB in 90% m. This value is still well of the room area for above the 95-dB NRP threshold specified for a 64-Mb/s


735

Fig. 7. CDF of the NRP and SDW parameters for several hb values in room B (in meters).

Fig. 8. CDF of the NRP and SDW parameters for several hb values in room B (in meters). LOS ray is suppressed.

Fig. 9. CDF of the NRP and SDW parameters for several tilting angles m of the MT antenna in room B for hm

MBS system operating with 22-dBm transmitted power [6]. The impact of the present antennas on system performance is m, the NRP distribution detailed in Section IV. For tends to be uniform all over the room since the coverage circle defined by the BST antenna radiation pattern covers all room area and the mean value decreases due to the increases, the channel time higher path loss. Moreover, as dispersion [represented here by the sliding delay window (SDW) parameter, containing 90% of the energy of the channel impulse response SDW 90%] also increases because the MT picks up more and more significant multipath components mainly reflected from the sidewalls (see Figs. 5 and 7). This is an important characteristic of the proposed antenna configuration since alternative paths can be provided in a controlled way through to cope with possible LOS blockage, as demonstrated in Fig. 8 and sketched in Fig. 5. The comparison

= 1:5 m and 1h = 2 m.

of Fig. 8 with Fig. 7 leads to the conclusion that, for higher values of , the NRP degradation is only 5 dB in more than 80% of the room area due to LOS blockage, while for lower values of , the degradation can be more than 25 dB, which is very significant. In what concerns the SDW, the degradation is negligible for higher values of , while it is very significant for lower ones. It is stressed, however, that this example is very pessimistic since the LOS is blocked for all positions in the room, which, in practice, is a very unlikely situation. In any case, the conclusion is as follows: to cope with LOS blockage, the BST antenna must be placed high enough to avoid significant link degradation, provided the system is able to mitigate the resulting higher channel time dispersion. Fig. 9 depicts the CDF of the NRP and SDW obtained in , with m. The room B for several tilting angles , but it worsens NRP degradation is not significant for

736

Fig. 10.


CDF of the NRP and SDW parameters for several

hb values in room C (in meters).

for higher angles in about 50% of the room area where the MT antenna main lobe fails to contain the direction of the BST antenna. The degradation is not very significant in the case of non-LOS since this component was already not available: 2 dB at 10% probability for the NRP and 10 ns at 90% probability for the SDW. The tilting freedom of the MT was also studied for , showing a relatively small degradation due to the presence of several multipath components: 1 dB in 10% of the room area for NRP, and 3 ns at 90% probability for SDW, considering values up to 20 . Similar simulations in identical conditions were carried out in a medium reflective room, named room C, 24-m long by 11-m wide (see [5] for its description) in order to evaluate the applicability of this antenna configuration to larger rooms. For was adopted for such dimensions, a fall-off angle both the BST and MT lens antennas in order to enlarge the cell dimensions without a significant increase in BST antenna height. The CDF curves of NRP and SDW, depicted in Fig. 10, above show that the power distribution is quite uniform for 3 m, with power levels similar to those obtained in room B, . On while an incomplete coverage is obtained for the other hand, the SDW is larger due to larger dimensions of the room, since the propagation delay imposed by the longer multipath components are directly related with larger dimension of the room—80 ns in 80% of the cases in room C compared to the 36 ns obtained in room B. This is confirmed by the ratio 80 ns/36 ns and the room length ratio 24 m/11 m. The performance of the antenna configurations under study was also evaluated in rooms with even larger dimensions. The behavior is quite similar to the one observed in rooms B and C: the NRP level decreases as the antennas beamwidth increases, essentially due to lower gain, and the channel time dispersion increases with the room dimensions. This suggests that the proposed antenna configuration is more suited for small- and medium-size rooms, but not so advantageous for large environments if very high bit rates are required, as will be discussed in the next section. IV. IMPACT

ON

SYSTEM PERFORMANCE

Considering a system with transceiver specifications used in [1] and [2], i.e., a decision feedback equalizer (DFE) with seven taps in the forward and feedback filters with a Ts/2 (Ts:

symbol time) tap spacing in the receiver, and using the method described in [1] to estimate the maximum CBR, it results that a minimum CBR of 175 Mb/s ( ns) can be achieved ns) with the in room B and 100 Mb/s in room C ( proposed antennas, even in a non-LOS condition, provided value is chosen. that the adequate In a LOS situation and with low MT antenna tilting, the CBR can be much higher since the channel time dispersion is near zero for low antenna heights. On the other hand, a significantly high bit rate can be achieved even without equalization. Using the rule-of-thumb that the root mean square (rms) delay spread (DS) cannot exceed 10% of Ts [7], a symbol rate of 10 Msymbol/s can be achieved in a LOS up to 12 for m in room B and condition with 4 Msymbol/s in room C. For the sake of comparison, taking the values of DS ( 32 ns) and DW ( 100 ns) parameters obtained with biconic horn antennas at the BST and MT in a room with 7-m wide (see dimensions similar to room B [11-m long [5])] one concludes that, for biconics, the maximum CBR results in about 70 Mb/s using the equalizer, and 3 Msymbol/s without equalization, which are significantly lower values. Furthermore, this configuration also limits the tilting freedom of the MT in the vertical plane. Also, taking the values of DS ( 45 ns) and DW ( 150 ns) parameters obtained with biconic horn antennas at the BST and MT in room C (see [2] and [3]), gives 45 Mb/s and 2.2 Msymbol/s with and without equalization, respectively. Another antenna configuration was previously studied where two pyramidal horns were located in the two narrower walls of the room, pointing toward each other and tilted downwards to each cover two-thirds of the room. This configuration yielded DS ( 30 ns) and DW ( 70 ns), resulting in a 100 Mb/s and 3.3 Msymbol/s with and without equalization, respectively (see [1]–[3]). This particular configuration gives results similar to those obtained with the proposed lenses, but it is more expensive to implement since it requires two BST’s and eventually restricts more the MT freedom of movement. This highlights the performance improvement obtained with the proposed antenna configuration relatively to conventional biconic–biconic and horn–biconic configurations. An antenna configuration that also aims at constant received power characteristic was recently presented [8], but using a different antenna technology. The BST, located near the room


ceiling, uses a single-element microstrip patch antenna with four parasitic elements to produce a down-looking shaped broadbeam with a 120 beamwidth. The MT uses a 64-element pencil-beam patch antenna with a beamwidth of 7.5 . Bit error rate (BER) measurements were conducted with this antenna 9.9 m configuration in a room with dimensions of 14.5 m 2.6 m with the BST located at 1.8 m height and the MT at 0.7 m. One-way transmission was made at 155 Mb/s using an eight-phase-shift keying (PSK) modulation scheme, a five-tap transversal equalizer, and a forward-error-correction technique. A BER lower than 10 was obtained in a circular area of 5-m radius centered just below the BST. These results obtained for a very directive link are similar to those described for room B, despite the fact that lens antenna gain is sacrificed by the deliberate choice of circular symmetric radiation patterns for both BST and MT in favor of complete terminal mobility. The solution presented in [8] severely restricts the movement freedom, which is only acceptable for low-mobility terminals. This adds weight to our antenna proposal for the implementation of wireless broadband systems in small- and medium-size rooms. For large rooms, multiple BST’s can be used giving multiple cells, but with a controlled channel time dispersion, which is essential to support very high bit rates. A portable MBS terminal integrating similar lens antennas was built [9] and the results of demonstrations will be reported elsewhere in the near future. V. CONCLUSIONS The proposed shaped dielectric lens antennas were shown to produce cells with uniform power distribution including sharp boundaries and little restriction of MT movement freedom. The antenna configuration is flexible enough to also accommodate nonsymmetric cell layouts with linear or circular polarization. BST antenna height controls the cell radius and, consequently, the amount of sidewall illumination which directly impacts on the channel time dispersion and alternative paths in case of LOS blockage. Experimental and simulation results indicate that these antennas outperform conventional solutions based on horns and biconics in terms of achievable CBR. Although antenna tilting and LOS blocking degrade the NRP parameter, in many cases, there is enough margin for compensation of its effects by an automatic adjustment of the transmitter power. Other lens combinations and different polarizations were tested leading to similar results. The low fabrication tolerance of the lenses [4] and the possibility of molding, makes this an attractive solution for simple and inexpensive mass production. REFERENCES [1] J. Fernandes, “Transmission capacity of a broad-band wireless radio link,” in Proc. IEEE Radio Wireless Conf. Colorado Springs, CO, 1998, p. 225.

737

[2] J. Fernandes et al., “Performance evaluation of mm-wave wide-band digital radio transmission,” in Proc. IEEE 2nd Symp. Commun. Veh. Technol., Benelux, Belgium, 1994, p. 95. [3] J. Fernandes, O. Afonso, and J. Neves, “Impact of antenna set-up and arrays on mobile radio systems,” in Proc. ICUPC’95, Japan, 1995, p. 387. [4] C. A. Fernandes, V. Brankovic, S. Zimmermann, M. Filipe, and L. Anunciada, “Dielectric lens antennas for wireless broadband communications,” Wireless Personal Commun. J., vol. 10, no. 1, pp. 19–32, June 1999. [5] J. Fernandes, P. Smulders, and J. Neves, “mm-wave indoor radio channel modeling vs. measurements,” Wireless Personal Commun. J., vol. 1, no. 3, pp. 211–219, 1995. [6] J. Fernandes and C. Fernandes, “Impact of shaped lens antennas on MBS systems,” in Proc. 9th Int. Personal, Indoor, Mobile Radio Commun. Symp., Boston, MA, 1998, p. 744. [7] J. C.-I. Chuang, “The effects of time delay spread on portable radio communication channels with digital modulation,” IEEE J. Select. Areas Commun., vol. SAC-5, pp. 879–889, June 1987. [8] A. Kato, T. Manabe, T. Ihara, and M. Fujise, “Development and evaluation on the millimeter-wave indoor wireless LAN demonstrators,” in Proc. 9th Int. Symp. Personal, Indoor, Mobile Radio Commun., Boston, MA, 1998, p. 28. [9] A. Plattner, B. Byzery, C. Fernandes, and T. Karttaavi, “A compact, portable 40 GHz transceiver for the mobile broadband system,” in Proc. ACTS Mobile Commun. Summit, Rhodes, Greece, 1998, p. 843.

Carlos A. Fernandes (S’86–M’89) received the Licenciado, M.Sc., and Ph.D. degrees in electrical and computer engineering from the Instituto Superior Técnico (IST), Technical University of Lisbon, Lisbon, Portugal, in 1980, 1985, and 1990, respectively. In 1980, he joined the IST and, since 1993, has been an Associate Professor in the Department of Electrical and Computer Engineering in the areas of microwaves, radio-wave propagation, and antennas. Also since 1993, he has been a Senior Researcher at the Instituto de Telecomunica¸co˜ es, Lisbon, Portugal, where he has been the Leader of the antenna activity for national and European Projects such as the mobile broad-band system RACE 2067-MBS and the system for advanced mobile broad-band applications ACTS AC204-SAMBA. He has co-authored a book and several technical papers in international journals and conference proceedings in the areas of antennas and radiowave propagation modeling. His current research interests are in the areas of dielectric antennas for millimeter-wave applications and propagation modeling for mobile communication systems.

José G. Fernandes (S’93–M’97) received the fiveyear course and Ph.D. degrees from the University of Aveiro, Aveiro, Portugal, in 1990 and 1997, respectively, both in electrical engineering. Following graduation, he joined the Department of Electrical and Telecommunications Engineering, University of Aveiro, first as a Researcher in microwave electronics and then in radio propagation channel modeling for mobile broad-band communications, where he is currently a Professor. In 1994, he joined the Instituto de Telecomunica¸co˜ es, Aveiro, Portugal, where he performs most of his research work. He has been participating in some national and European research projects, such as the mobile broad-band system RACE 2067-MBS, the microwave optical duplex antenna link RACE 2005-MODAL, and the system for advanced mobile broad-band applications ACTS AC204-SAMBA, as well as in some COST actions. He has published over 30 technical papers. His research interests include radio propagation channel modeling, adaptive antennas, and adaptive techniques to improve transmission performance of high-speed systems.