8/13/2019 00245316

1/4

PROPAGATION INTO AND WITHIN BUILDINGS AT 900,1800 A N D 2300 M H z

A. F. de TOLEDO and A. M. D. TURKMANI

The University Of Liverpool, Department of Electrical Engineeri ng andElectronics, Brownlow Hill P 0. Box 147, Liverpool L69 3BX, UK.

ABSTRACTInvestigations of propagation into, and within,buildings at 900, J800 and 2300 MHz have beenundertaken. The composite Rayleigh-plus-log-normaldistribution successfully modelled the measured cu-mulative distributions of all measurements. The av-erage measured penetration loss at ground floor levelwas found to be 14.2, 13.4 and 12.8 dB respectivelya t 900, 1800 and 2300 MHz. The rate of change ofpenetration loss with height was -1.4 dB per floor.The ra te of cha nge of mean signal level for signalstravelling within buildings was, on average, 8.5 dBper floor. The path loss attenuation factor, n, thatbest modelled the within-building measurement wasfound to be 5.3, 5.5 and 6.0 respectively for 900, 1800and 2300 MHz.

1. IntroductionThe dramatic changes in the nature of mobile radiotechnology in the past decade have had a very clearimpact on the general communications environment. Therelatively unsophisticated technology of the early 1970 shas given way to much more advanced components andequipment, and the demand for mobile communicationservices has led to the development of new systems thatoperate in higher frequency bands. The move towards'personal communications', i.e. communicati on withhand-held rather th an vehicle-borne equipment, ha s ledto the realisation that not enough is known about radiopropagation either into or within buildings. In this con-text, into is used to identify the scenario where a con-ventional base station on a hilltop site or atop a highbuilding communicates with a radio receiver that is in-side another building. On the other hand within is usedto identify the case when both transmitter and receiverar e inside the same building.This paper contains the results of an investigation ofradio propagation into and within buildings at 900, 1800

and 2300 MHz. Measurement s of th e average signalstrength and th e signal variability have been made usingbuildings within the University of Liverpool precinct.These tests were intended, inter alia, to establishwhether the Rayleigh-plus-lognormal statistical modeltha t successfully describes th e signal received a t street-level outside buildings still applied, or whether it wouldbe necessary to devise an alternat ive model. The signalvariability is discussed in both into and within buildingmeasurements. The building loss factor, which is in-cluded in the vehicular mobile model to account for theinc:ease in atte nuat ion of th e received signal observedwhen the mobile is moved from outside a building toinside, is discussed in the into building measurements.The variation in signal strength as a function of dis-tance is described in the within building measurements.

2 Experimental procedureOne readily accessible property of the signal transmittedover a mobile radio propagation path is the variationof its envelope as t he position of th e mobile terminal i schanged. It is well established that the signal statisticscan be modelled as a combination of a small-scalequasi-stationary process (multipath), superimposed on alarge-scal e process (shadowing). The scat ter ing modelwhich describes the local, i.e. small area statistics, fol-lows a Rayleigh distribution, and the local mean (i.e.the signal strength averaged over the Rayleigh fading)is log-normallly distributed.'For a given distance from the transmitter, the differencebetween the mean signal strength on one floor of thebuilding and the mean signal strength measured in thestreets, immediately outside, can be defined as the pen-etration loss for that floor.2 One purpose of the currentpropagation study has been to determine the buildingloss and its variation in a form that can be combinedwith vehicular mobile studies to give an estimate ofcoverage into buildings.

60-7803-0673-2192 3.00 1992 IEEE

_ _ _ ~ 7-

8/13/2019 00245316

2/4

It has been also shown tha t the local mean varies withthe distance between the base station and the receiverand, is given by a law of the form

mean-signal-strength ntercept 10 log,&d ) 1)

The obtained values for the path loss attenuation factor,n, are given in the within buildings measurements.The method used to obtain the results was to transmita continuous wave (CW) signal from a fixed base station;this was received and recorded for subsequent analysisby mounting a receiver and data-logging system (DLS)on a trolley which could be moved around within thebuilding concerned. Spatial sampling was facilitated byattaching a slotted disc to a fifth wheel on the trolley.In order to process the data collected in a particularroom in the building, each sample was normalised by t heaverage signal strength within that room. Thenormalised data for each room was then collated to f o ha d ata file consisting of the fast-fading component only.The distribution of this fast-fading signal describes thesmall scale signal variations.The large scale signal distribution was determined bytesting the departure (in dB) of the average signalstrength of each room from the average signal strengthfor the whole building. The difference between the av-erage of the measurements carried out in all rooms ofone particular floor and the average of the measure-ments at street level was treated as the building pene-tration loss for that floor.

3. Into b uilding measurementsMeasurements of the received signal strength wereundertaken within buildings in the University ofLiverpool precinct. The buildings were Blocks A and Bof the Department of Electrical Engineering & Elec-tronics and the Departments of Computer Science andLife Sciences, these buildings being at 180, 240, 300 and350 m, respectively from the transmitter. No line-of-sightexisted between Electrical Engineering Block B and thetransmitter. Partial line-of-sight existed, however to thethree other buildings. A general description of eachbuilding was given in reference 4. The transmitter waslocated on the roof of the Mechanical Engineeringbuilding, at a height of 40 m.Twelve experiments, four for each frequency setting,were conducted. The cumulative distribution of thesmall scale signal variations and of the large scale sig-

nal variations for 1800 MHz were discussed in references4 and 5. Nevertheless the significant conclusions, con-sidering all three frequency settings, are:1. The small scale variations are Rayleigh distributed.2. The large scale variations are log-normally distrib-

uted with a standard deviation related to the condi-tion of transmission. The standard deviation werefound to be 8.09 7 60 and 7.56 dB for 900, lS00 and2300 MHz respectively, i.e. the values of standarddeviation decreased slightly as the carrier frequencyincreased. Examination of the results also shows avalue for the standard deviation of around 5 dBwhen no line-of-sight existed, whereas, for a partialline-of-sight condition, the standard deviation in-creased to approximately 8.5 dB.

References 4 and 5 describe in some detail the buildingpenetration loss for 1800 MHz. The average penetrationloss at ground floor level was found to be around 14.2,13.4 and 12.8dB respectively at 900, 1800 and 2300 MHz.It has also been found that the penetration loss de-creases with height at a rate of 1.4 dB per floor in av-erage -1.38, 1.36 and -1.50dB for 900, 1800 and 2300MHz). It was noticed however, that the penetration lossincreased for floor levels higher than the sixth floor,with a rate of change of 0.2 dB per floor in average 0.45,0.36 nd -0.22 B). These anomalies, which had also beenreported in the literature, were attributed to the relativeposition of base station, measured and obstructingbuildings or other physical structures. It should befinally noted that, as reported in reference 2, he pene-tration loss decreased slightly as the frequency oftransmission was increased. This is in contrast to thewell known fact that path loss increases with the trans-mission frequency. In free space propagation, for exam-ple, the path loss increases by 6 dB when the frequencyof transmission is doubled. Therefore, as far as propa-gation into buildings is concerned, by increasing thefrequency of transmission, some of the additional pathloss can be compensated by lower building penetrationloss values?

4. Within building measurementsPersonal Communication Network (PCN) systems areexpected to utilise a small cell structur e in order to meetthe very large demand expected for mobile radio ser-vices. In dense urban areas (e.g. the city of London),microcells, or even picocells, may be used. In some situ-

634

8/13/2019 00245316

3/4

ations one large building alone can comprise a microcelland in this case the base station will be situated withinthe building. An understanding of the propagationmechanism within buildings is therefore very importantand essential.

Twenty two experiments for 1806 and 2300 MHz, and sixexperiments for 900 MHz have been conducted withinthe four buildings of the University precinct. In orderto determine the transmitter location which gives thebest si gnal coverage, six different rooms in th e ElectricalEngine ering Block A, and thre e in the Computer Sciencewere selected as base station for 1800 and 2300 MHz.The small-scale signal variations, at the t hree frequen-cies, were very closely represented by a Rayleigh dis-tributi on an d overall, the transmission conditions hardlyaffected this distribution. Although, the results suggestthat the large-scale signal variations can also be mod-elled by a log-normal distribution, the standard devi-ati on s ar e high: 16.0 dB. 16.5 dB and 17.8 dB respectivelyfor 900, 1800 and 2300 MHz. The val ues a re high, mainlybecause the mean signal strength on th e floor where thetransmitter is situated is very high, while the mean sig-nal levels on floors further away from the transmitterare much lower.In order to compare the dependence of the median (50%)signal level on the position of the transmitter inside thesame building, the cumulative distribution of the large-scale signal variations, for the six experiments in theElectrical Engineering Block A, and for the three ex-periments in t he Computer Science, were plotted and theresults for 1800 MHz are shown in reference 4. Similarresults were found for the 2300 MHz setting (see Fig. 1);considering, for example, the Electrical EngineeringBlock A, it ca n be observed that, when the transmitterwas placed in a very large room, which occupies thecentre of th e building between the fifth and th e sixthfloor, the median signal level increased by about 15 dBcompared with th e value measured when t he transmitterwas in the foyer. When the transmitter was located ina very large room, with a large window area, near thecent re of the building, the median was 7.5 dB higherthan when it was located in a small room, with no win-dows, in th e centre of the same floor. Although, th e verylarge room a t the to p of the building yielded the highestmedian signal strength, the large room in the middle ofthe building, with large window area, provided the bestcoverage: the minimum mean signal value of its distrib-ution was found between 7 dB and 10 dB higher than theminima of the other distributions.

Room: 403 - oom: 306Room: 6CR - oom: 302Room: Foyer - oom: 602

ken sicpl, dBCom par i s on o f the cumulative dis tr ibution fitnc-

t i o n of l ar ge -s c a le s igna l var ia t ions for d ine r e n ttr ans m i t ter lo c a t ions w i th in E le c tr i c a l E ng ine e r ingB l o c k A . at 2300 M H z

The floor mean signal levels were calculated and thevalues plotted in a graph which shows the gradient oftheir variation from floor to floor. Fig. 2 gives, for ex-ample, the results for 2300 MHz. Examination of thosegraphs, for each frequency, reveals two different slopes:6.5, 6.1 and 6.7 dB per floor, respectively for 900, 1800and 2300 MHz, when considering floors below that onwhich the transmitter was located and, similarly, -10.5,-10.4 and -10.8 dB per floor, when considering the meas-urements conducted on floors above the transmissionlocation. The results may be attributed again to therelatively different positions of base station and ob-structions around the lower and upper floors. However,changing the frequency, has scarcely affected the valueobtained i n eithe r event, i.e. receiver below or above th etransmitter. Therefore, th e global average rate ofchange of the mean signal level calculated, was 8.5 dBper floor.

0 . 6.7 dE per Floor -10.8 dB per Floor1 0

-20-30

g -405060

s 70-80

..Y

d

.

Y

\ t /

8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8F l w r

F i p r e 2. Normalised l loor mean s ignal levcl aga ins t num-bc r o r floors w p a r a t i n g thr r e c e ive r and tr ans m i t trr

6 5

8/13/2019 00245316

4/4

The regression analysis applied to the scatter diagramof the mean signal strengths, calculated for each roomin one particular experiment, yields the path loss atten-uation factor n, defined in eqn. (l), This factor describesthe association between the mean signal strengths anddistances from the transmitter. The correlation coefficientof the scat ter diagram also in dicates the degree of asso-ciation between the two variables, i.e. a correlation coefficient close to 1 indicates a strong association, whichmeans that knowing one variable helps in predicting theother, whereas a correlation coefficient close to 0 indi-cates a weak association, and information about onevariable does not give much information about theother.The path loss attenuation factors were found equal to5.3, 5.5 and 6.0 respectively for 900, 1800 and 2300 MHz,i.e. increasing slightly with frequency. The correlationcoefficients have been determined around 0.7.

5. ConclusionsIt has been confirmed that the Rayleigh plus lognormalmodel is adequate to describe the signal envelope vari-ations, the standard deviation of the lognormal compo-nent being between 5 and 8.5 dB depending on thetransmission conditions and frequency. The averagepenetration loss at ground floor level was found to be14.2 dB at 900 MHz, 13.4 dB at 1800 MHz and 12.8 dBa t 2300 MHz. As the receiver was moved to the upperfloors of a building, the penetration loss decreased at arate of about 1.4 dB per floor.Anomalies, which had alsobeen reported i n the literatu re, were found at levelsabove the 6th floor where the loss again started to in-crease.The Rayleigh plus lognormal statistical model can alsobe used to model propagation totally within buildingsalthough the standard deviation of the lognormal com-ponent is much higher at about 16.8 dB. Coveragewithin buildings is highly variable depending on the lo-cation of the transmitter and the number o obstructionsbetween it and the receiver. Although difficult to drawfirm conclusions, it has been found th at t he coverage ofa building can be maximised by locating the transmitternear the centre of the building and in as large a roomas possible.

mately 8.5 dB per floor. The signal attenuates quiterapidly and according to regression analysis the best-titregression line has a path loss attenuation factor, n,equal to 5.3, 5.5 and 6.0 for 900 1800-and2300 MHz re-spectively. This is somewhat higher tha n values reported in the literature, which are themselves quitevariable, and is an example of the difficulties involvedin arriving at definitive conclusions in a highly-variableenvironment. Nevertheless, sufficient results have beenobtained to justify the conclusions drawn, and the results should be useful to system designers an d frequencyplanners.

6 AcknowledgmentThe work was undertaken under the terms of a contractawarded by DTI Radiocommunications Agency, UnitedKingdom.

7.1.

2.

3.

4.

5.

The rat e of change of mean signal level for propagationtotally within a building has been measured at approxi-

636

ReferencesJAKES, W. C.: 'Microwave mobile communications ,( John Wiley and Sons, New York, 1974 )RICE, L. P.: 'Radio transmission int o buildings a t 35and 150 MHz', Bell System Tech. J., 1959, 38, (l) , pp.COX, D . C., MURRAY, R. R. and NORRIS, A. W.:'Measurements of 800 MHz radio transmission intobuildings with metalic walls', Bell System Tech. .TURKMANI, A. M. D., TOLEDO, A.. F., 'Radiotransmission at 1800 MHz into, and within, multi-story buildings', IEE Proc. Par t 1, 138, No. 6, De-cember 1991, pp. 577-584.TURKMANI, A. M. D. PARSONS, J. D., andTOLEDO, A. F.: 'Radio propa ation into buildingsat 1.8 GHz , COST 231, Outpute bocume nt COST 231,TD( iO)117.

197-210.

1983, 62 9), pp. 2695-2717.