Effects of Antenna Polarization and Beam Pattern on Multipath Delay Spread and Path Loss in Indoor Obstructed Wireless Channels

Transcription

1 ,~ YO 17 t' 7 mrnm I Effects of Antenna Polarization and Beam Pattern on Multipath Delay Spread and Path Loss in Indoor Obstructed Wireless Channels C. M. Peter Ho and Theodore S. Rappapm1 Mobile and Portable Radio Research Group Bradley Departme!lt of Electrical Engineering ABSTRACT The effects of different combinations of antenna patterns and polarizations on path loss and rms delay spread in indoor obstructed channels at 2.45 G Hz are treated in this paper. Based on measurement results, and assuming the main beams of the antennas are lined up on boresight, it appears that using vertically polanzed (VPJ antennas at both the transmitter and the receiver can provide smaller delay spread and smaller path loss compared to those of other antenna combinations. The correlation between path loss and rms delay spread of different antenna combinations is presented. It IS found that path loss is uncorrelated with rms delay spread when the antennas are omnidirectional. However, strong correlations are found when the antennas are directional and both polarizations of the transmitter and receiver are vertically polarized (VP) or horizonta: ' olarized (HP). 1. INTRODUCTION Wireless communications is one of the faste' _,wing segment oi the telecommunications industry. Before the end of this century. wireless communication will probably be able to provide ubiquuous voice and data communications through Personal Communication Networks (PCN) [1]. Two fundamental requirements of the future PC~ systems are: i) low portable power. and ii) high data rate. The first requirement is desired because a lightweight portable unit is more convenient for the user. Assuming a fixed coverage range; the transmitter po;wer required depends on the path loss of the environment. The second requirement of the PCN system is limited by multipath dela) spread. Multipa~h delay spread can be quantified by m1s delay spread (crnru;) and maximum excess delay (10 db down) which were formally defined in [2]. The results of (3] show that a circular polarized (CPl directional antenna is able to provide smaller rms delay spread when the transmitting and receiving antennas are aligned, and there exists a direct line-of-sight between the antennas. This observation sug,!!ests that some antenna polarizations and patterns may provide smaller delay spread and smaller path loss than those of other antenna combmauons in a pan 1cular channel. Hence, by appropriately Implementing the proper antenna combinations in different channels, one may improve the quality of transmission. In a line-of-sight (LOS) channel. the power budget of the system can definitely be improved by using antennas wjth h1gher gains while the rrns delay spread can be reduced by using CP antennas. The effects of various obstacles on the selectton of antenna polanz.ations and patterns are not treated in any previous paper. If the mdoor channel 1s obstructed, the power budget of the communication lmk may not necessarily be improved by Just increasing the gains of the antennas. Also, obstructed channels can cause severe depo lanza!lon to the transmitted wave, and hence co-polanzed antennas may not necessarily This work is sponsored by Teledyne Microwave Monol1th1c Inc. prov1de a higher received power than cross-polarized antennas. In this paper, the effects of antenna polarization. and antenna pattern on the rms delay spread and path loss in an obstructed (OBS) indoor environment are presented. From our results. it appears that the use of vertically polarized di.rect10na\ antennas at both the transmitter and the receiver will provide low delay spread and high received power in indoor obstructed channels. 2. EXPERI~1ENTAL DESIGN A II of the measurements reported in this paper were carried out in Teledyne's two-floored office building in California. On the second floor. where all the measurements were made, there are many soft partitions which make individual office cubicles in a large open area, and some concrete walls which form extra rooms at the peripheral area of the build1ng. The offices consist of standard office furniture and equipments such as computers and copying machines. The receiver was mounted on a cart and was moved to different locations while the transmitter was tixed within a sof:-?artltioned office. For all results presented in this paper. there was 11: direct line-of-sight between the transmitter and the receiver at each measurement location. Typical obstacles between the transmitter and the rece1ver were soft partitions and office furniture. The measurement setup was very similar to the one described in [5]. A pulse havmg an absolute width of 33 ns (rms width 25 ns) was generated t:-y the H P 8()82A pulse generator. The pulse was then mixed With a 2 AS G Hz earner. The modulated signal was amplified and transmitted by the antenna. A relatively simple square law detector was used a: the: ;~.ce1ver. Various antenna combinations were tested in our measurements. Antennas used in the measurements were: DISC:t)ne L:near p,1iarized Omnidirectional Antenna (YP or HP) lga1r: =:.db Linear Polanzed D1rec11onal Corner Reflector (YP or.hp) (Gam= i 2 db.l Circular Polanzed DirectiOnal Hehcal Antenna (Gain= 12 db) Conference on Unviersal Personal Communication. Dallas, TX, Dept. 30h 19_ ~7803-0!J91 ~419L/UUUU-0092 S3.00 ~ :992 IEEE Since the antennas With different polarizations and patterns were used at the transmitter and the rece1ver, a total of twenty-five (5x5) antenna combinations were tested at each location in this measurement program. Measurements were performed ar SIX different locations and T-R separation ranges from 15m to 45m. Data from one measurement location was discarded due to an apparent error in the recorded attenuator settings of the' me8surement. Although the data are limited for broad based conclusions, the preliminary results and evaluation crjterton of antenna performance provide some direction for future research m indoor wneiess system design. Antenna heights of both the

2 transmitter and receiver for all the results in this paper were 1.8 m above the ground. The main beams of the directional antennas were lined up on boresight as if a line of sight path existed. For each antenna combination at each location, the receiver was moved over a distance of 1 meter where 10 snapshots of power profiles were recorded. 3.1 DELAY SPREAD RESULTS Table 1 gives delay spread results of some antenna combinations at measurement location M where typical results were observed. Column one of the table shows some of the antenna combinations used, where the corresponding antenna code is shown in Table 2. The half power beamwidths (HPBW) of the antennas are shown in Table 2 as well. The first letter in column one of Table 1 corresponds to the antenna used at the transmitter while the second letter corresponds to the antenna used at the receiver. All values in Table 1, except A( ), are in nanoseconds. arms is the average rms delay spread of the ten profiles measured along a 1 meter track for each antenna combination at location M. The average was computed from the instantaneous rms delay spread values of each of the ten profiles. ).(arms> is a measure of how far arms of each antenna combination is from the average rms delay spread computed over all antenna combinations at location M, and is given by: where A. (rms- delay- spread) = IL- 1Lau a all (EQ 1) A.(rms-delay-spread) =normalized distance of averaged rms delay spread from the average over all antenna combinations at location M I! = average rms delay spread of a particular antewa combination at location M!!an = average of averaged rms delay spread of all an: :::nna combinations at location M a, 11 = standard deviation of averaged rrns delay spread of all antenna combinations at location M Table 1: Delay Spread Results of Some Antenna Combinations atlocationm T R separation:24m combination from the average largest maximum excess delay over all antenna combinations and is obtained in a similar fashion as MannJ MED in column six is the averaged maximum excess delay (10 db down) of all profiles for each antenna combination. A.(MED) in the table is a. normalized distance computed from column six via. a similar method as A.(arms> and A.(Max MED). If a. number in third, fifth and seventh column is negative, then it is below the ensemble average of co.lumns two, four, or six respectively. For good channels (with small intersymbol interference), we want rms delay spread and maximum excess delay (10 db down) to be as small as possible and hence the numbers in columns three, five and seven to be as negative as possible. Table 2: Antenna Code Antenna. Combination Antenna Code HPBW HPBW azimuth plane vertical plane omni VP b omni 600 omni HP c 60 omni directional VP d 50 NA directional HP e NA 50 directional CP f To find good antenna combinations, we want to determine if ther.e are any antenna combinations which. cause columns three, ~ve, and seven of Table 1 to be consistently negative while keeping path loss at a. minimum, and therefore we need to look at results from all locations. 3.2 ANALYSIS OF DELAY SPREAD RESULTS RMS delay spreads for all measurement locations and all antenna. combinations range from 20 ns to 70 ns. The spread of data is not very large. One explanation for this is that the distance between the transmitter and receiver is not too large (maximum T-R separation = 45 m). Maximum excess delay (10 db down) ranges from 50 ns to over 200 ns. i 1 Antenna arms /..(arms> Max. /..(Max MED I..(MED) Code MED MED) bb cc cf dd ee ff O.J. maximum minimum average standard deviation Max MED in the fourth column is the largest maximum excess delay (10 db down) of the ten profiles for each antenna combination at location M. Hence, column four is the worst case maximum excess delay (10 db down) of all profiles of each antenna combination. ).(Max MED) is 1the normalized distance of Max MED of each antenna It turns out that a few antenna pairs have small delay spread (i.e. A.(arm 5 )< 0.3 and A.(Max MED)<-0.3 and /..(MED)<-0.3) at all locations. They are antenna. combinations bf, ce, cf, dd, and fd. The rms delay spread and maximum excess delay (10 db down) of cf is not always the minimum at any location, but this antenna. combination provides delay spreads which are consistently close to the minimum at each location. Hence, from our results, it seems that an omnidirectionaf HP transmitting antenna and a directional CP receiving antenna. is the antenna combination that provides the most consistency in minimizing delay spread for all locations. From our limited data, it is clear that some antenna combinations work consistently better than others at mitigating delay spread. There is another interesting observation from our delay spread results. When a transmitter uses an omnidirectional antenna. and a. receiver uses a. directional antenna point-ed towards the transmitter (no matter what type of polarization) always (except for one antenna pair at one location), there is lower delay spread than when the transmitter uses the same directional antenna pointed at the receiver and the receiver uses the same omnidirectional antenna at the same measurement location. This suggests that the channel is not symmetrical. That is, for our experiments, scatterers tend to be closer to the receiver than the transmitter. 0093

3 Rr==r 2m 'WJ!N"M'YF 'I M =s. Regarding the performance of co-polarized antenna pairs, we have made the following observations. CP/CP (directional helical antennas at the transmitter and the receiver) and VP/VP (directional corner reflector at the transmitter and the receiver) can offer smaller delay spread than HP/HP (directional comer reflector at transmitter and receiver). In general, co-polarized directional antenna pairs provide smaller delay spread than co-polarized omnidirectional antenna pairs. Moreover, cross-polarized (omnidirectional or directional) antenna pairs give higher delay spread than the other antenna pairs. 4.1 COMPUTATION OF PATH LOSS In the past, path loss was calculated by referring the po.,.,..er received at a particular T-R separation distance R to the power received under free space conditions at a close distance, say, 10 wavelengths [2] or 1 meter [4]. In our experiments, multiple antennas were tested and there are some difficulties in. finding a good reference. For example, if we want to find the path loss of cross-polarized antenna pairs, it is hard to separate the effects of the antenna patterns, the depolarization caused by the channel and the polarization mismatch of the antennas. Similar difficulties are encountered when the path loss of co-polarized antenna pairs is computed. Thus, we suggest to incorporate the polarization mismatch factor into the path loss, and path loss will be calculated by referring the actual received power at the receiver antenna terminal to the received power of a back-to-back calibration. Hence, path loss is given by are approximately one standard deviation below the average path loss over all antenna combinations. Table 3: Path Loss Results of Some Antenna Combinations at Location M PLrree space(isotropic):68db T-R separation:24m (Highly Obatructed) Antenna Code PL' (db) I..(PL') PL(dB) A.(PL) bb be Q.6 cc 72 -o.s cf dd ee ff 73 -Q maximum minimum average standard deviation Effects of Beam Pattern on Path Loss PL(dB) = PWR(calib) -PWR(R) +Gt+Gr (EQ 2) where Gt and Gr are the gains of antennas, PWR(cn.','h)=power received ofback-to-back calibration, and PWR(R)=power rece:ved at particular T R separation distance R. In addition, we define system path loss PL' as follows. PL' (db) = PWR (calib) - PWR (R) (EQ 3) In defining PL', we are viewing the channel and the antennas as a whole system. That is, we absorb the antenna gains and polarization loss into our variable PL' which is a function of antenna combinations and distances. In fact, this definition matches with intuition. Our goal is to have as much power received as possible with a fixed transmitter power. It does not matter what kind of depolarization mechanism occurs in the channel as long as the received power is high. Wide-band techniques presented in [2] are used to compute total power as the area under the power delay profile. Path loss results at location M of some antenna combinations, calculated by equation. (2) and (3), are shown in Table ANALYSIS OF PATH LOSS RESULTS In general, the system path loss for most of the antenna combinations is greater than path loss of an isotropic radiator in free space separated by the same T-R separation distance. This is reasonable because the signal undergoes attenuation through obstacles and hence much energy is lost during propagation despite the gain offered by the. directional antenna. Antenna combinations dd and ee, which use directional antenna of the same polarization at both the transmitter and receiver, provide consistently low path loss for all locations. Quantitatively, the path loss values of these two antenna combinations Table 4 summarizes some comparisons between the power received for co-polarized antenna pairs, the antenna patterns of which may be omnidirectional or directional. The first two cases in Table 4 consider the effect of the receiving antenna pattern (with a fixed transmitting antenna) on the received power, while the last two cases in the same table consider the effect of the transmitting antenna pattern (with a fixed receiving antenna) on the received power. Let's look at an example in case i) of the table. Antenna combination #1 is bb, and antenna combination #2 is bd for vertically polarized antennas in case i) of the table. The differences between the received powers of bb and bd were calculated for all locations. An average value of PWR(bd)- PWR(bb) over all measurement locations is computed and is shown in the fourth column of the table (3.0 db). A similar average value is computed for HP antennas for case i) and is shown in the fourth column of the table ( 4.3 db). An overall average (both HP and VP), based on the values in column four of case i), is computed for case i) and is shown in the fifth column of the table (3.8 db). All averaging in this table are performed in linear scale. In fact, similar average values are found when averaging in db scale for case i) and ii) because the individual values are fairly close. When the transmitting antenna is fixed, i.e. case i) and ii), the signals arriving at the receiving antenna will be the same for antenna combination # 1 and #2 because the same transmitter antenna is used for antenna combination # 1 and #2, and comparisons are performed at the same measurement location. If signals arrive only in the main beam of the receiving antenna, the difference between the receiving power for antenna combinations #1 and #2 in case i) will be roughly equal to the difference of the gains of directional and omnidirectional antennas, i.e. 10 db. On the other hand, we expect less difference between the two antenna systems if the signals arrive from all directions. The overall averaged difference of received power between antenna combination #1 and #2 is 3.8 db for case i), and 5.2 db for case ii). Our results suggest that a directional transmitting antenna localizes the signals to be arrived within the main beam of the receiver by only a few decibels, when

4 nw aw trzrr ttrr nnar,.t rrts PP"S rnrr---.-mnsz r compared with an omnidirectional transmitting antenna, in indoor obstructed channels. Table 4: Comparisons Between Co-Polarized Antenna Pairs with Different Beam Patterns Antenna Antenna PWR(#2)-PWR(#l) overall Combination Combination (db) average #1 #2 (db) i) OMNI/OMNI OMNI/DIR VP: bb VP: bd HP: cc HP: ce 4.3 ii) DIR/OMNI DIR!DIR VP: db VP: dd HP: ec HP: ee 5.6 iii) OMNI/OMNI DIR}OMNI VP: bb VP: db HP: cc HP: ec 2.5 iv) OMNI/DIR DIR/DIR VP: bd VP: dd HP: ce HP: ee 3.6 At location B, the power received with bb is better than that of bd. Therefore, the gain of the directional transmitti: _ antenna does not always help to improve the power received at e receiver. Similar observations are found at other locations. On _, verage, directional receiving antennas pointed on boresight on the h\ othetical LOS path can provide more power at the receiver than omr: _ irectional antennas using the same co-polarized transmitting antenna, cut the improvement (4.5 db) is less than the difference of gains between the omnidirectional antenna and the directional antenna. The effect of the antenna patterns of the transmitting antenna on a fixed receiving antenna is studied for case iii) and iv). The power improvement of using a directional antenna at the transmitter over an omni-directional antenna at the transmitter is about 4 db if co-polarized antennas are used at the receiver. From the above experiments, the power improvement using a directional antenna at the transmitter over an omni-directional antenna at the transmitter, when using a co-polarized receiving antenna, is about the same (4 db) as the improvement offered by a directional antenna at the receiver over an omni-directional antenna at the receiver, when a co-polarized transmitter antenna is used Effects of Polarization on Path Loss When studying co-polarized directional antenna pairs, dd and ee provide consistently low path loss at all measurement locations. On the other hand, ff gives fairly high path loss at two measurement locations. This may due to the fact that the sense of polarization of a CP wave ts changed for every reflection and hence some reflected components could not be detected by the CP directional receiver. When the transmitter is omnidirectional (HP or VP), the CP directional antenna almost always provides the highest.power at the receiver, when pointed towards the transmitter on the hypothetical line 'of-sight path. Intuitively, we might expect co-polarized directional antennas at the receiver to provide the highest received power. However, since the channels are of obstructed topography, depolarization of the signal will be high [5]. We may conclude the average polarization stat-~ of the transmitted signal, generated by a linearly polarized omnidirectional antenna, is moved from the equator (HP/VP) on the Poincare' sphere [6] to the poles (CP), after going through the channel. When the transmitter is a linearly polarized directional antenna, a receiver antenna with the same beam pattern and polarization as the transmitter always provides the highest received power. The received power of a linearly polarized directional transmitting antenna with a CP directional receiving antenna is, on average, 3 db smaller than the highest received power of all antenna combinations with the same transmitting antenna. The 3 db difference agrees with the theoretical polarization discrimination between LP and CP, assuming no depolarization occurs in the channel. From the above observations, it seems that the depolarization of the signal generated by linearly directional antennas through the channel is smaller than omni-directional antennas. More measurements and analysis are needed to determine a physical explanation for this. The average difference between path loss of ff and fd (or fe) is small and is, on average, less than 1 db. Assuming no depolarization, the difference between the received power of fd and ff would be around 3 db since the gains of f and d are the same. The above observation shows that depolarization of a CP signal is fairly large. This can be explained physically by the fact that a CP signal changes its sense after each reflection. The results also agree with the results mentioned previously that the system path loss of ff is higher than dd or ee. The average cross-pol discrimination (XPD, defined as the ratio of the signal level at the output of a receiving antenna that is nominally copolarized to the output of receiving antenna of the same gain but orthogonally polarized to the transmitting antenna) of the HP and VP waves are 4.5 db and 4.6 db respectively. XPD for directional antennas are higher than that ofomnidirectional antennas. This agrees with the analysis performed above since small depolarization of a signal corresponds to a large XPD value. The overall average and standard deviation of XPD of all measurements are 4.5 db and 2.8 db. The overall average is slightly higher than reported in [3] due to fact that directional antennas were employed in our measurements at both the transmitter and the receiver. 5. PATH LOSS VS. RMS DELAY SPREAD Linear regression analysis is done for path loss vs. rms delay spread for different antenna combinations. The correlation coefficient, R, of PL' vs. crrms of different antenna combinations is also evaluated. For copolarized omnidirectional antenna, PL' is uncorrelated with rrns delay spread. This result is similar to that of [2J. Antenna combination with a HP directional transmitter provides some correlation between PL' and rms delay spread for all receiving antennas. In general, directional antenna pairs are more likely to provide high correlation between PL' and rms delay spread. Antenna combination dd has the highest correlation (R 2 =0.63) between path loss and rms delay spread among all antenna combinations. If R 2 is larger than 0.50, at least half of the relation between rms delay spread and path loss can be described by the linear regression model[?]. Antenna combinations dd and ee have correlation coefficients higher than Other antenna combinations, that have some correlation (R larger than 0.50) between PL' and rms delay spread, include ce, de, eb, ec, ed, ee, and fe. For all other antenna combinations, PL' appears to be uncorrelated with rms delay spread. Antenna combinations that have high correlation between PL' and rms delay spread are desired for system design because both power and 0095

5 delay spread are important system parameters in the future PCN ~ystems. If the correlation between PL' and rms delay spread is high, then two parameters of the linear regression model, y-intercept and slope, can describe the relation between PL' and rms delay spread fairly well, 6. CONCLUSION The effects of different antenna combinations with different antenna patterns and polarizations on path loss and delay spread are investigated in this paper. Antenna combination cf, which uses an omnidirectional transmitting antenna with a directional CP receiving antenna, provides the lowest rms delay spread a.nd the lowest maximum excess delay (10 db down) among all antenna combinations. Unfortunately, the system path loss experienced by cf is fairly large at. two of the five locations. Antenna combinations dd and ee are able to provide consistently high received power at all measurement locations. Hence, the results of our experiment show that using vertically polarized directional antennas at both the transmitter and the receiver can give us a relatively low delay spread and low path loss. Moreover, the fact that a high correlation exists between the path loss and rms delay spread for this antenna combination is an advantage for system design of future PCN system. However, there is one disadvantage of this antenna combination. During the experiment, we have lined up the main beams of the transmitter and receiver manually. For practical implementation, an additional adaptive mechanism is needed to line up the beams if the antennas are directional. Table 5 summarizes some key results of this paper. Depolarization of the signal generated by omnidirectional antennas is generally higher than that generated by directional antennas. Therefore, for most cases, a CP receiving antenna is able to give the lowest path loss among all receiving antennas when the transmitting antenna is omni-directional. Antenna combination ff (CP directional antenna at both the transmitter and the receiver) which proved to give small delay spread in LOS channels [3], does not give the lowest delay spread among all combinations in our measurements. The path loss of ff is not reduced consistently well either. Due to fact that twenty-five antenna combinations were tested at each location, we were unable to make measurements at many locations. The data base we obtained is not large enough to make an statistically significant assessment of the relation between distance and PL' for different antenna combinations. Nevertheless, the results give us some direction for further research of antenna design for future PCN systems. Our results may provide some preliminary guidelines for antenna design. By considering the same transmitting antenna, the probability of two or more receiving antennas of different polarizations and beam patterns to have rms delay spread lower than a particular value can be evaluated. Table 5: Summary of Antenna Combinations that have small delay spread or small path loss or high correlation between PU and arms at all measurement locations Small Small Path High Carr. Delay Loss Btw.PL' Spread and arms bf ce ce de Antenna cf dd dd ee eb ed Combinations fd ef fe ACKNOWLEDGEMENT The authors would like to thank Kurt Schaubach, Mike Keitz, and Kenneth Blackard ofmprg for helping to conduct the experiment, and Steve Ludvik of Teledyne for supporting the work and arranging the measurements. REFERENCES 1. T. S. Rappaport, "The Wireless Revolution", IEEE Comm. Magazine, Nov. 1991, pp T. S. Rappaport, "Characterization of UHF Multipath Channels in Factory Buildings," IEEE Trans. Ant. & Propag., Aug 1989, ppl D. A. Hawbaker, "Indoor Wide Band Radio Propagation Measurements and Models at 1.3 GHz and 4.0 GHz," Master Thesis, Virginia Tech, May D. Devasirvatham, "A Comparison of Delay Spread Measurements within Two Dissimilar Office Buildings,'' Proc. of 1986 Int. Conf. on Comm., vol T. S. Rappaport and D. A. Hawbaker, "Wide Band Microwave Propagation Parameters using Circular and Linear Polarized Antennas for Indoor Wireless Channels," IEEE Trans. Comm., val 40, no 2, Feb 1992, pp M. Born and E. Wolf, "Principles of Optics," Pergamon Press, J. S. Milton and J. C. Arnold, "Introduction to Probability and Statistics," McdrawHill, dd