Estimating Coverage Of Radio Transmission Into And Within Buildings By Linear Prediction Filter For

Estimating Coverage of Radio Transmission into and within Buildings for Line of sight visibility between two points in terrain by Linear Prediction Filter Jamal Fathi Abu Hasna [email protected] Electrical & Electronics Engineering Department, Near East University, Cyprus, Turkey via Mersin-10, KKTC

Keywords Signal Strength, Network Measurement Report (NMR), Multi-path, Base Transceiver Station (BTS), Linear Prediction Filter. Abstract Investigations of propagation into and within buildings at 900, and 1800 MHz have been undertaken, using buildings in the Near East University of Cyprus The emphasis of this article is the modeling of radio transmission into buildings that uses the measured penetration loss values in order to adjust the propagation models developed for the outside areas, and the modeling of radio transmission within buildings, starting with the simple distance-power law, then applying Linear Prediction filter to the received signals.

surrounding buildings. Several researchers have studied the problem of receiving radio signals inside buildings and model it as the distance dependency of the path loss when the mobile is outside a building, plus a building loss factor. The building loss factor is included in the model to account for the increase in attenuation of the received signal observed when the mobile is moved from outside a building to inside. This model was first proposed by Rice [9] in 1959, also Antonio Fischer De Toledo [4] in 1998, and has been used in most subsequent investigations. In addition to penetration loss, system designers are also interested in learning about the received signal variability and the effects of building height, conditions of transmission, construction materials, and frequency of operation. Several research activities that deal with these aspects have been reported in the literature [1, 3, 9–11].

I. Introduction A physical understanding and consequent mathematical modeling of the radio propagation inside buildings is very important because it facilitates a more accurate prediction of system performance and provides the mechanism to test and evaluate methods for mitigating the deleterious effects caused by the radio channel in such environments. This article reports the results of narrowband measurements into buildings at 900, and 1800 MHz, with the transmitter located on the roof of one building and the receiver located in a different building. It also describes measurements that have been undertaken with both transmitter and receiver situated within the same building. These latter measurements are classified as propagation within buildings. The experiments were conducted in order to determine statistics related to the random variation of a continuous wave (CW) signal received in indoor environments [1–3]. Empirical models which allow the path loss between the transmitting and receiving antennas to be predicted have also been developed and are presented in this article. This additional loss will depend on a large number of factors with various degrees of importance. Among them are the transmission frequencies, the distance between the transmitter and receiver, the building construction material, and the nature of the

II. Measurement Results a) Into Building Experiments The tests were undertaken using a fixed base station transmitter and a mobile receiver. The signal transmitted from the base station was received using a purpose-built data logging system, which was moved around the building. The base station consisted of a CW transmitter feeding a collinear antenna, raised clear of local obstructions. It produced an effective radiated power (ERP) of 29 dBm at 900 MHz and 24 dBm at 1.8 GHz. A vertically polarized omni-directional antenna was also used at the mobile, with a ground plane 1.65 m above the floor. Each sample of data collected in a particular room in the building was normalized by the average signal strength within that room. The normalized data for each room were then collated to form a data file consisting of fast fading only. The distribution of this component describes the small-scale signal variations. The local mean was estimated by averaging the signal strength over 200 samples symmetrically adjacent to every point (i.e., the process known as moving average). The large-scale signal distribution was determined by testing the departure (in decibels) of the average signal strength of each room from the average signal strength for the whole building. The inside average signal strength was determined for each room of the buildings measured. Outside signal strength

was measured at street level around the perimeter of the building, along the closest available path to the building’s outside walls. Fifteen tests were conducted in the Near University of Cyprus in order to assess the effect of transmission condition on signals propagating into buildings at 900, 1800 MHz in different buildings of the university to examine the values of penetration loss at ground-floor level at 1800 MHz only. The two distinct regions where penetration loss measurements, at ground-floor level, took place can be characterized as a highly built-up area (i.e., the Medical Department) and a medium built-up area (i.e., the Electrical & Electronics and Computer Engineering Department). Two different locations for the transmitter were selected for each area. For the first set of trials (i.e., the Medical Department), the transmitter was located on the roof of the Medical Department (TX1), which is approximately 5 floors high. For the second set of measurements in the Engineering Department which is 4 floors high including Innovation and Information Technologies, as shown in Fig. 1, the transmitter was set up on the roof of the Engineering Department (TX2), which is approximately 20 m high. Fifteen experiments were conducted for each transmitter location. The 10 buildings selected in the university precinct for the 20 penetration loss tests are described in Table 1, and their relative position appears in Fig. 1. 1800 MHz are described in Table 1; the relative position of the buildings is presented in Fig. 1. Electrical Engineering, four floors, 20 m high, floor area of 600m2, steelframed construction with offices, large laboratories, lecture and research rooms, 2×3 glass windows, looking to the Medical Department, and large glass entrance. Offices on all floors of all buildings were crowded with typical office furniture and the teaching or research laboratories contained experimental equipment according to the specialized demand of each area of study, and The Near East University Techno Park which is the complex that will host the joint venture between Near East University and IBM (NEU Innovation and Information Technologies Centre). The center’s primary objective is to carry out research, development and innovation. The Innovation and Information Technologies Center is the only one found in the region and this includes the following regions collectively. Eastern Europe, the Middle East, Central Asia and Northern Africa. The new super computer with its advance platform is ranked 76th in the world, 13th within certain nations and is ranked 10th amongst the other universities in the world. b) Penetration Loss at Ground-Floor Level The mean signal levels outside and inside buildings at ground floor level and the mean value of penetration loss for the two experiments carried out in the

university are presented in Tables 2 and 3. 14

13

5 7

8 4 9

3

2 10 1

12

TX2

TX1 6 11 15

Figure 1. Relative position of the building tested for penetration loss in the university.

Table 1. General Description of the Engineering Department buildings in the University Building

Address

Front of Building

Ground Floor

Innovation and Information Technologies Centre

E E & C Engineering

Front Glass Wall

Computer Engineering

CE

Front Glass Wall

Electrical Engineering

EE

Front Glass Wall

Mechanical Engineering

ME

Front Glass Wall

Large open area, 4 U shape large corridors, Under ground floor, 5 annex structures, 3 labs, and 5 lecture halls. Large open area, 1 U shape large corridor, 28 rooms, including labs and research rooms. Large open area, 1 U shape large corridor, 28 rooms, including labs and research rooms. Large open area, 1 U shape large corridor, 28 rooms, including labs and research rooms.

In the experiments conducted in the university, the average values of penetration loss at ground-floor level were found to be significantly different (20.39 dB and 17.4 dB) for the two transmitter locations. This difference (i.e., approximately 2.99 dB) was due to the Internet receiver

dish mounted on the roof of the Engineering department. The two sets of fields trials yielded an average value of penetration loss equal to 18.8957 dB. In the university there were important changes in the relative position of the transmitter concerning the measured buildings, were changed.

However, it is necessary to remember that the validity of most outdoor propagation models, such as those of Okumura [4], Hata [16] and Ibrahim [17, 18], have been developed for large cells, whereas for personal communication the suitable cell diameter is often less than 700 m. Table 2. General Description of the Medical Department buildings in the University Building

Ad

Front of Building

Ground Floor

Reception and First Aid Department

M C

Front Glass Wall

Large open area, 2 L shapes large corridors, annexes structures, 18 rooms, and 7 Offices.

General Surgery Department

M C

Front Glass Wall

Pregnancy Department

M C

Front Glass Wall

Offices

M C

Front Glass Wall

Large open area, 2 U shape large corridors, 18 rooms, including labs and Operation rooms. Large open area, 2 T shape large corridors, 18 rooms, including labs, Operation rooms, and First aid room. Large open area, 1 U shape large corridor, 28 rooms, including labs and research rooms.

Table 3. Penetration Loss in the University Rx Loca Innov Cafete Comp EE ME Dek Secr Aver

Out dBm -88.72 -78.82 -72.29 -85.35 -68.01 -66.05 -70.14 -75.62

Grou dBm

Pen. Loss

Out dBm

Grou dBm

Pen. Loss

-101.35 -100.63 -99.27 -102.09 -91.84 -90.07 -86.87 -96.01

12.63 21.81 26.98 16.74 23.83 24.02 16.73 20.39

-91.37 -83.21 -86.21 -86.21 -77.54 -74.37 -85.26 -83.45

-104.4 -109.24 -105.97 -107.06 -92.27 91.42 95.82 -100.88

13.03 25.87 19.76 20.85 14.73 17.05 10.56 17.4

Therefore, those models cannot be fully trusted when used for the indoor environment without further investigations. In addition, predicting first the signal outside the building of interest and then, from that result, determining the signals inside the building leads to an inevitable reduction in accuracy. Therefore, prediction of the path loss for radio transmissions into buildings may be more accurate if it has been undertaken directly and not merely as an extension of outdoor propagation models. A similar approach was adopted by Barry and Williamson [3] to analyze measurements undertaken in New Zealand at 851 MHz.

Toledo and Turkmani [19] have obtained direct modeling of propagation into buildings at 900, 1800, and 2300 MHz. The latter have performed the prediction using information based on what might be termed incomplete data. Such information may have an element of uncertainty and a risk of being incorrect. Using appropriate statistical techniques, it was possible to generalize from a given set of data to a more broadly applicable statement, and for that purpose specific and rigorous techniques have been applied in order to estimate the degree of uncertainty. Details of the statistical techniques applied can be found in [19] and a full description in [1]. Propagation into (and within) buildings involves a more complex multipath structure than that of the outdoor land-mobile radio channel, which is dependent on path length, effective base station antenna height, and the environment local to the mobile. In addition to these variables, indoor propagation is also affected by other empirically observed variables such as building structure and layout of the rooms. After collating all the survey measurements in the university precinct buildings and investigating the relationships between a large number of variables, the best of all results, for the into building case, was obtained when three variables were present in the regression equations: the logarithm of the distance, d, the logarithm of the floor area, Af, and the number of building sides seen by the transmitter on each floor of the building housing the receiver, SQ The resulting models for the path loss, at 900, and 1800 MHz, were found to be ^ Y i,900 = –37.7 + 40.0log10d + 17.6log10Af– 27.5SQ (1) Yi,1800 = –27.9 + 40.0log10d + 23.3 log10Af– 20.9SQ (2) with the root mean square errors (RMSEs) equal to 2.07, and 1.96 dB, respectively. III. Linear Prediction Filter Coefficients The signal data as the output of an autoregressive process driven by white noise. Use the last 4096 samples of the AR process output to avoid start-up transients. The prediction error, e(n), can be viewed as the output of the prediction error filter A(z) shown Fig 2, where H(z) is the optimal linear predictor, x(n) is the input signal, and p is the order of the prediction filter polynomial, a = [1 a(2) ... a(p+1)], p = length(x)-1, estimate for each column in the rows of matrix a and a row vector of prediction error variances g. x(n)

-1

-2

-p

H(z)=-a(2)Z –a(3)Z - ..a(n+1)Z

 x (-n)

e(n)

+ A(Z)

Figure 2. Prediction Error Block Diagram The autocorrelation method of autoregressive (AR) modeling to find the filter coefficients. The generated filter might not model the process exactly even if the data sequence is truly an AR process of the correct order. This

is because the autocorrelation method implicitly windows the data, that is, it assumes that signal samples beyond the length of x are 0.

is biased. The MSE for the biased predictor is

 , the value of MSE for the biased predictor can be determined. For complete prediction, ˆ   0 , that means p  pˆ for when E    E  p  p

Xa  b  x(1) 0  0   x(2) x(1)     ,  x(2) 0    X   x( m)   x(1)  0 x (m)  x (2)           0  0 x (m )   

2

2

2 

1  1  0 a (2)    ,   b  a         0 a ( p  1) 

obtaining zero error. V. Line of Sight Visibility between two Points in Terrain To compute the mutual visibility between two points on a displayed digital elevation map. The current object if it is a regular matrix map or the first regular matrix map found on the current axes. The map's zdata is used for the profile. The color data is used in the absence of data in z. The two points are selected by clicking on the map. The result is displayed in a new figure. Markers indicate visible and obscured points along the profile. The profile is shown in a Cartesian coordinate system with the origin at the observer's location. The displayed z coordinates accounts for the elevation of the terrain and the curvature of the body. The elevations are provided as a regular matrix map containing elevations in units of meters. The two points are provided as vectors of latitudes and longitudes in units of degrees. The resulting logical variable vis is equal to one when the two points are visible to each other, and zero when the line of sight is obscured by terrain. As shown in Fig. 4. The prediction error is approximately white Gaussian noise, as expected for a third-order AR input process.

and m is the length of x. Solving the least squares problem via the normal equations

X H Xa  X H b leads to the Yule-Walker equations

 r (1) r (2)  r ( p )      r (2) r (1)  r (2)     r (2)     r ( p)  r (2) r (1) 



 

E  2  E p  pˆ

 a(2)   a(3)         a( p  1) 

 r (2)   r (3)         r ( p  1) 

Original Signal vs. LPC Estimate 6 Original Signal LPC Estimate

where r = [r(1) r(2) ... r(p+1)] is an autocorrelation estimate for x computed using xcorr. The Yule-Walker equations are solved in o(p2) flops by the LevinsonDurbin algorithm [20] as shown in Fig. 3.

Amplitude

4 2 0 -2 -4

IV. Prediction of Channel Power

0

10

20

30

40 50 60 70 Sample Number Autocorrelation of the Prediction Error

80

90

100

3000

4000

5000

1

N

ˆi  based on the predicted channel power: p

 hˆ

2

ij

j 1

j=1 and picks antenna i corresponding to maximum power gain with Known pilot symbols are transmitted ˆ max  max  i  N t pˆ i . The from each antenna p

 

 

2 h,

average channel power gain is E pi  N r  the average predicted power gain

and is

E pi   N r  h2  N r rwH Rw1 rw . Note the average



2

H

1

value of the error E   N r ( k  rw R w rw while predicting power is not zero, thus the power prediction

Normalized Value

At instant k, the receiver selects a transmit antenna 0.5

0

-0.5 -5000 -4000 -3000 -2000 -1000

0 Lags

1000

2000

Figure 3. Autocorrelation of the Prediction Error Signal.

0 Terrain Visible Obscured Observer Line of Sight

Vertical Distance from Observer

-1000

-2000

-3000

-4000

-5000

0

1000

2000 3000 4000 5000 Horizontal Distance from Observer

6000

7000

Figure 4. The visibility between two points on the peaks map. VI. Conclusion This article contains the results of investigation and modeling of radio propagation at 900, and 1800 MHz for the into and within building scenarios. Measurements of signal strength and signal variability have been made using buildings within the University of Near East in Cyprus. The article also includes linear prediction filter for Line of sight visibility between two points in terrain. It is interesting to observe that the Barry and Williamson models, derived from experiments carried out in Auckland, New Zealand, yielded slightly worse values of RMSEs: 3.9 dB for the line of sight case, and 7.2 dB for the obstructed path case [3], and also ANTONIO FISCHER DE Toledo models carried out in the University of Liverpool of RMSEs: 2.4, and 2.2 dB. Thus, predicting propagation path loss using the models presented in this article should produce more precise results. VII. References [1] A. F. Toledo, “Narrowband Characterization of Radio Transmissions into and within Buildings at 900, 1800, and 2300 MHz,” Ph.D. thesis, Dept. Elec. Eng. and Electron., Univ. of Liverpool, U.K., May 1992. [2] A. M. D. Turkmani, “The Mobile Radio Channel,” Ch. 3, Personal Communication Systems and Technologies, J. Gardiner and B. West, Eds., Artech House, 1995. [3] P. J. Barry and A. G. Williamson, “Statistical Model for UHF Radio-wave Signals within Externally Illuminated Multistory Buildings,” IEE Proc. —Part I, Aug. 1991, vol. 138, no. 4, pp. 307–18. [4] Y. Okumura et al., “Field Strength and Its

Variability in VHF and UHF Land Mobile Radio Service,” Rev. Elec. Commun. Lab., 1968, 16, pp. 825–73. [5] K. Allsebrook and J. D. Parsons, “Mobile Radio Propagation in British Cities at Frequencies in the VHF and UHF Bands,” IEEE Trans., 1977, vol. VT-26, no. 4, pp. 313–23. [6] D. C. Cox, R. R. Murray, and A. W. Norris, “Measurements of 800 MHz Radio Transmission into Buildings with Metallic Walls,” Bell Sys. Tech. J., 1983, 62, no. 9, pp. 2695–2717. [7] A. M. D. Turkmani and A. F. Toledo, “Radio Transmission at 1800 MHz into and within Multistory Buildings,” IEE Proc. — Part I, vol. 138, no. 6, Dec. 1991, pp. 577–584. [8] A. F. Toledo and A. M. D. Turkmani, “Propagation into and within Buildings at 900, 1800 and 2300 MHz,” Proc. 42nd IEEE VTC, Denver, CO, May 1992, pp. 633–36. [9] L. P. Rice, “Radio Transmission into Buildings at 35 and 150 MHz,” Bell Sys. Tech. J., 1959, 38, no. 1, pp. 197– 210. [10] J. M. Durante, “Building Penetration Loss at 900 MHz,” Proc. IEEE VTC., 1973, pp. 1–7. [11] A. M. D. Turkmani, J. D. Parsons, and D. G. Lewis, “Measurement of Building Penetration Loss on Radio Signals at 441, 900 and 1400 MHz,” J. IERE, vol. 58, no. 6 (supp.), 1988, pp. S169–74. [12] S. E. Alexander, “Radio Propagation within Buildings at 900 MHz,” Elect. Lett., 18, no. 21, 1982, pp. 913–14. [13] A. J. Motley and J. M. P. Keenan, “Personal Communication Radio Coverage in Buildings at 900 MHz and 1700 MHz,” Elect. Lett., 9 June 1988, 24, no. 12, pp. 763–64. [14] D. C. Cox, “Universal Digital Portable Radio Communications,” Proc. IEEE, 1987, 75, no. 4, pp. 436– 77. [15] T. S. Rappaport, “Wireless Communications – Principles and Practice,” IEEE/Prentice Hall, 1996, pp. 102–10. [16] M. Hata, “Empirical Formula for Propagation Loss in Land Mobile Radio Services,” IEEE Trans., 1980, VT-29, no. 3, pp. 317–25. [17] M. F. Ibrahim and J. D. Parsons, “Signal Strength Prediction in Built-Up Areas, Part 1: Median Signal Strength,” Proc. IEE, Part F, 130, no. 5, 1983, pp. 377–84. [18] J. D. Parsons, The Mobile Radio Propagation Channel, Pentech Press, 1992. [19] A. M. D. Turkmani and A. F. Toledo, “Modeling of Radio Transmissions into and within Multistory Buildings at 900, 1800 and 2300 MHz,” IEE Proc. — Part I, vol. 140, no. 6, Dec. 1993, pp. 462–70. [20] Jackson, L.B., Digital Filters and Signal Publishers, 1989. pp. 255-257.Processing, Second.