MEASUREMENT OF THE INTERFERENCE REJECTION CAPABILITY OF SMART ANTENNAS ON MOBILE TELEPHONES Christian Braun, Martin Nilsson y and Ross D. Murch z Allgon Mobile Communications AB, P.O. Box 500, S-184 25, Åkersberga, Sweden y Allgon System AB, P.O. Box 541, S-183 25 Täby, Sweden y also with Dep. of Signals, Sensors and Systems, Royal Institute of Technology, Stockholm, Sweden z Hong Kong University of Science and Technology, Clear Water Bay, Kowloon, Hong Kong e-mail: christian.braun@allgon.se, y martin.nilsson@allgon.se, z eermurch@ee.ust.hk Abstract We are reporting investigations into the interference rejection capabilities of a dual antenna system on a mobile handset prototype for the 1800 MHz band. The performance evaluation is done by using a simulated incoming eld on the handset antennas and by downlink measurements in an indoor environment. Complex signal samples are used in order to investigate the interference rejection gain with dierent diversity combining schemes and with one and two interferers present. I. INTRODUCTION The fast growth of mobile telephone users has created a need for higher capacity in the cellular networks. However, the capacity is to a large extent limited by co-channel interference. Hence, any means of reducing the interference would increase the capacity [1, 2]. Furthermore, the future demands on higher data rates over the radio channel will create a need for a better channel quality [1, 3]. One way ofovercoming the capacity problem on the downlink is to use antenna diversity at the handset. It is well known that antenna diversity has the capability of reducing multipath fading and cochannel interference. Previous research in this area has concentrated on methods for estimating the diversity gain available from diversity antenna systems on the handset in a noise limited environment [46] or reducing the complexity of the diversity combining algorithms [2,7,8]. However, to our knowledge no study has yet been dedicated to measuring the interference rejection capability of an antenna diversity system on the handset. In this study we are investigating the interference rejection capabilities of handsets with antenna diversity systems. Two handsets prototypes with dualantenna systems are evaluated for diversity combining, using sampled RF antenna signals. Three different combining schemes are evaluated, Selection Combining (), Equal Gain Combining () and Interference Rejection Combining (). Both simulated and measured data are used in the evaluation, and the results are compared. The simulations are based on a model for the incoming eld, in which the 3D complex radiation patterns from the handset prototypes are used in order to generate the antenna signals. The measurements are done in an indoor environment where the handset antenna signals are sampled. The results show that the simulated and measured data give similar results. outperforms the other combining schemes but its performance moderates when more than one interferer is present. and are relatively insensitive to the number of interferers, and the dierence between these and is not dramatic when two interferers are present. We believe pre-detection combining schemes based on or could be a cost eective method to reduce the downlink interference in cellular systems. II. MODEL In this section, we describe a model for the signal received by handset antennas in a scattered eld and diversity combining employed on this model. Received Antenna Signal In the environments where mobile telephones are frequently used, there is seldom a line-of-sight between the base station and the mobile station. The mobile telephone sees incoming eld from many dierent
directions. If the instantaneous incoming eld is described by the electrical eld component, it can be written as: E (;t)=e(;t)^ = r (;t) e,j (;t)^ E ' (;t)=e'(;t) ^' = r ' (;t) e,j'(;t) ^'; (1) where is the space angle (; ') and ^ and ^' are the two orthogonal polarizations of the incoming eld. The time dependence describes the non-stationary behaviour of the eld, which is due to changes in the environment or to the mobile station moving around. The antennas on the mobile telephone are described by their radiation patterns. Since we are dealing only with two-branch diversity, we will dene two antenna ports; port A and port B. The antennas are characterized by the antenna vectors, a and b: a() = a() ^a() b() = b() ^b(); (2) where a() is the directional antenna gain and phase for antenna A, and ^a() is the polarization for the specic direction. The signals at the antenna ports, x A and x B, can be seen as the scalar product of the incoming eld and the antenna vector, integrated over all : x A (t) = = x B (t) = = he ; ai + he ' ; ai d [a Eh^; ^ai + a E'h ^'; ^ai] d [he ; bi + he ' ; bi] d [b Eh^;^bi + b E'h ^'; ^bi] d; (3) where the - and t-dependence of the integrands is understood. The antennas signals (3) will be complex, Gaussian variables when (1) describes a multipath scenario. The signal received by the handset will contain not only the desired signal, but also interfering signals and noise. The interfering signals we shall consider are co-channel interferers, i.e. signals from nearby basestations using the same frequency as the desired signal. Since these basestations are located at dierent positions, the desired and interfering signals will be independent from each other. Hence, the output vector x(t) containing the total output signal on the two antenna ports is: x(t) = x A (t) x B (t) = s(t) + IX k=1 i k (t) +n(t): (4) Here s(t) is the signal vector from the desired basestation on the two antenna ports. i k (t) is the signal vector from interferer k and and I is the number of interferers present. The noise vector n(t) is assumed to be equally distributed zero mean white Gaussian noise. Diversity Combining We will investigate three dierent diversity combining schemes: Interference Rejection Combining (), Equal Gain Combing () and Selection Combining (). For more detailed descriptions of these schemes, see e.g. [9, 10]. The diversity combining scheme can be represented byavector, here denoted w, and the resulting signal can be written as: y(t) =w(t) x(t): (5) Using this notation, the output Signal to Interference plus Noise Ratio (SINR) is expressed as: (t) = IP k=1 jw(t)s(t)j 2 jw(t)i k (t)j 2 + jw(t)n(t)j 2 : (6) The diversity combining schemes discussed below operate so that they maximize the channel quality, here represented by the instantaneous SINR (6). Combining Schemes Interference Rejection Combining,, is a linear combination of the antenna signals which employs complex weights on each antenna branch. The complex combining vector w is given in [9] as: w irc = R,1 xx r xd; (7) where R xx is the covariance matrix for the received signals and r xd is the projection of the received signals onto the desired signal. Equal Gain Combining,, adds the antenna signals with a phaseshift employed on one of the branches. The combining vector can be written as: w egc (t) = e,j (t) 1 ; (8)
The evaluation was done by employing diversity combining on complex signal samples. By simulating an incoming eld and using the 3D radiation patterns from the handset antennas, the received antennas signal could be generated. To validate the simulation, measurements of the received signal on the handset antennas in an indoor environmentwere performed. Simulations Figure 1: Schematic illustration of the handset prototypes, where antennas A and B are indicated in the gure. where is the phase shift, in this case employed on branch A,chosen so that the output SINR is maximized. Selection Combining,, simply chooses the branch which is currently has the best SINR. The combining vector is: III. w sc (t) = ( 1 0 0 1 if SINR A > SINR B : (9) otherwise DUAL ANTENNA HANDSETS Two handset prototypes with two-branch antenna systems were used, using one external and one inbuilt antenna each (Fig. 1). Prototype 1 consisted of a quarterwave monopole and a shorted patch antenna on a Printed Circuit Board (PCB), prototype 2 consisted of a monopole and a planar meander antenna on a similar PCB. The distance between the antenna feeding points was about 20 mm for both prototypes. The diversity evaluation was made at 1800 MHz, where the return loss was better than 10dB for each antenna. The coupling between the antenna ports was below -17dB and below -6dB for prototypes 1 and 2, respectively. The 3D radiation patterns from the antennas on the prototype were measured in an anechoic chamber. The measurements were done with a phantom head and hand in order to simulate a real user scenario. In [6], the 3D radiation pattern measurements are further described. IV. PERFORMANCE EVALUA- TION In the simulation part, complex antenna signals were generated using the model discussed in Section II. As a model for the incoming eld, we used a three dimensional extension of Clarke's model [11], in which the amplitude, r, and phase,, had the following distributions: r ;' (;t)=exp(, (, m) 2 ) (10) 2 2 ;' (;t)=rand(0:2): (11) In order to model an indoor environment, the constants were set to: m = 70 and = 20. The angular resolution of the incoming eld was 5 for both and ', corresponding to the resolution of the measured 3D radiation patterns of the handset antennas. As can be seen from (10) and (11), the eld was xed with no time dependence. However, by letting the handset (actually the radiation patterns of the handset antennas) move along a straight line in the incoming eld, the antenna signals (3) experienced typical time varying signals with Rayleigh fading statistics. In order to model the case with interfering signals present, an interfering eld was set up independently from the desired eld. The amplitude distribution was was the same as for the desired eld, with an oset corresponding to mean Signal-to-Interference Ratio (SIR). The phase was generated independent from the phase of the desired eld. In this way, the desired and the interfering signals had an uncorrelated behaviour. The simulation generated 3x5120 complex signal samples (one desired and two interfering signals). These were stored and further processed for diversity combining. Measurements In order to validate the simulation data, and to extend the investigation, downlink channel measurements were performed in an indoor environment. These were done so that they would correspond to a Rayleigh fading, interference limited scenario, having similar environment parameters as used in the simulation. The measurements were done with a
was calculated from the signal samples. The signals deriving from the simulation showed a fairly constant mean level, whereas the measurements showed a large part of long-term fading due to dierent path loss in dierent parts of the room. Therefore, in order to calculate the short-term correlation from the measured signal, a demeaning had to be performed. This was done according to [5] with a demeaning window of 1 s, corresponding to a displacement of about 1 m. Figure 2: Overwiev of the measurement area. The measurement routes A-J are marked. Network Analyzer in the external mode where it performed a three-port measurement. The RF signal was fed out to one port, connected to a transmitting dipole antenna, and complex signal samples were performed simultaneously at the two other ports, connected to the two antennas on the handset prototype. Hence, the received complex signals at the two antennas were sampled simultaneously. The output power from the transmitting antenna, which was positioned behind a metallic screen so as to avoid line-of-sight in the measurements, was about 0 dbm. The handset antennas were connected to the Network Analyzer by coaxial cables, on which 12 cm long ferrite chains were put, close to the handset, to prevent currents from owing on the cables. The handsets were held in talk position, by a real person, who moved along a number of dierent paths in the room (see Fig. 2). For each prototype, 10 measurement runs were taken. The measurements were done at 1800 MHz with an IF bandwidth of 100Hz. Since the bandwidth was small, the received signal was well above the noise oor. 100 samples/s/channel during 10 s for each route gave a grand total of 10x1000 samples for each handset antenna. Signals from dierent routes were used as desired and interfering signals, respectively (5 routes each). A second interferer was created by taking the interfering routes backwards. Data Processing The mean signal levels from each antenna and the envelope correlation between the antenna signals The desired and interfering signal samples were then added to form a total output signal at each antenna port according to (4). The interfering signal levels were adjusted so that SIR 1 = 15dB, SIR 2 = 20dB. Moreover, a noise oor was introduced, SNR =25dB. Diversity combining was then employed so that the resulting SINR (6) would be maximized. V. RESULTS Table 1 shows the mean signal levels from each antenna, P A and P B, the dierence, P, and the envelope correlation between the antenna signals, e. The simulated data shows P of about 1dB for both prototypes. e is very low for prototype 1 and slightly higher for prototype 2. The measured data has a generally much lower mean signal level, which is due to the specic path loss in the room in which the measurements were performed. The measured P is higher than the simulated for prototype 1 whereas prototype 2 shows a lower dierence. The measured correlation is still low for prototype 1 and higher for prototype 2. Table 1: Simulated (S) and Measured (M) Mean Signal Levels and Correlation S M P A P B P e Prot. 1-6.1 db -7.1 db 1.0 db 0.01 Prot. 2-5.4 db -6.5 db 1.1 db 0.13 Prot. 1-57.4 db -60.0 db 2.6 db 0.06 Prot. 2-63.9 db -64.4 db 0.5 db 0.26 The SINR improvements from using the diversity combining schemes are shown in Table 2. The Table shows both the simulated and measured values, and the cases with one and two interferers. The values shown are the SINR improvements compared to the antenna having the best SINR, which is equivalent
to the diversity gain. CDF for prototype 1, SNR = 25 db and SIR1 = 15 db, SIR2 = 20 db 0 10 Table 2: SINR improvements in db at the 1% outage level for prototypes 1 and 2, for one (I) and two (II) interferers. Simulation Measurement P.1 I 11 13 23 7 10 24 II 10 12 15 8 10 16 P.2 I 10 13 22 9 12 26 II 10 12 14 9 12 16 Probability that For prototype 1 the resulting SINRs are presented as Cumulative Density Functions (CDF's). Due to space limitations, the plots for prototype 2 are not shown. The CDF's from the two prototypes had a similar behaviour, and the quantitative results are shown in Table 2. Figs. 3-4 show the results from the simulation and Figs. 5-6 from the measurements. Probability that 10 0 CDF for prototype 1, SNR = 25 db and SIR1 = 15 db Figure 3: CDF from simulation on prototype 1, one interferer. The CDF's from the simulation show a quite good Rayleigh distribution, whereas the measured values do not. The mean SINRs on the antennas are equal for simulations and measurements (about 4dB), but the median values dier. It is likely that the environment in which the measurements were performed did not give an ideal multipath propagation. Nevertheless, the SINR improvement from using the diversity combining schemes on the measured values is close to the simulated ones. It can clearly be seen that Figure 4: CDF from simulation on prototype 1, two interferers. is able to cancel one interfering signal almost completely, whereas the simpler combining schemes have a more moderate performance. With two interferers however, the dierence gets smaller. The measured dierence in mean signal levels on prototype 1 seemed to aect the diversity performance negatively for and. These schemes are more sensitive to signal level dierences on the antenna branches than, whose performance is not aected. This is because the complex weights also aect the antenna signal amplitude, which can compensate for initial signal level dierences. VI. DIUSSION From the results we nd that there is a signicant amount of interfence rejection gain available on the handset even with a simple combining scheme such as or. outperforms the others, but it may require two separate receiver chains because the combining most likely would be performed at baseband level. Moreover, when there is more than one interfering signal present, the dierences between the and the others can be only a few db's at the 1% outage level. In order to be attractive for a mass market, the diversity combining for mobile telephones has to be low-cost and low-complex. We believe this can be done with the simple combining schemes presented in this paper.
10 0 CDF for prototype 1, SNR = 25 db and SIR1 = 15 db CDF for prototype 1, SNR = 25 db and SIR1 = 15 db, SIR2 = 20 db 0 10 Probability that Probability that Figure 5: CDF from measurement on prototype 1, one interferer. REFERENCES (1) E. Nikula and M. Doetsch, Wideband TDMA based radio interface for high bit rate packet access, in VTC'98, (Ottawa, Canada), pp. 845849, IEEE, May 1998. (2) A. Carter, D. Hilborn, N. Secord, and A. Abu- Dayya, A diversity co-channel interference canceller for AMPS cellular systems, in VTC'98, (Ottawa, Canada), pp. 2630, IEEE, May 1998. (3) T. Frankkila, S. Labonte, and R. Ramesh, The evolution of IS-136: Voice service, in VTC'98, (Ottawa, Canada), pp. 835839, IEEE, May 1998. (4) M. A. Jensen and Y. Rahmat-Samii, Performance analysis of antennas for hand-held transcievers using FDTD, IEEE Trans. on Antennas and Propagation, vol. 42, pp. 11061113, August 1994. (5) M. LeFevre, M. A. Jensen, and M. D. Rice, Indoor measurement of handset dual-antenna diversity performance, in VTC'97, (Pheonix, A), IEEE, May 1997. (6) C. Braun, G. Engblom, and C. Beckmam, Antenna diversity for mobile telephones, in AP-S'98, (Atlanta, GA), pp. 22202223, IEEE, July 1998. (7) P. Wong and D. Cox, Low complexity diversity combining algorithms and circuit architectures for co-channel interference cancellation and frequency selective fading mitigation, IEEE Trans. on Communications, vol. 44, pp. 11071116, September 1996. (8) C.-N. hang, W. Lam, and C. Ling, A low complexity antenna diversity reciever suitable for TDMA handset implementation, in VTC'97, (Pheonix, A), IEEE, May 1997. Figure 6: CDF from measurement on prototype 1, two interferers. (9) R. Compton, Adaptive Antennas -Concepts and Performance. New Jersey: Prentice Hall, 1988. (10) M. Schwartz, W. R. Bennet, and S. Stein, Communication Systems and Techniques. New York, NY: McGraw-Hill, 1965. (11) R. Vaughan and J. Andersen, Antenna diversity in mobile communications, IEEE Trans. on Vehicular Technology, vol. 36, pp. 149172, November 1987.