Comparison Of Propagation Models For Fixed Wimax System

Comparison of Propagation Models for Fixed WiMAX System based on IEEE 802.16-2004 Amarasinghe K.C.∗ , Peiris K.G.A.B. ∗ , Thelisinghe L.A.D.M.D. ∗ , Warnakulasuriya G.M. ∗ , and Samarasinghe A.T.L.K.† ∗ Department

of Electronic and Telecommunication Engineering, University of Moratuwa, Moratuwa, Sri Lanka Email: {050016, 050304, 050455, 050469}@ent.mrt.ac.lk † Department of Electronic and Telecommunication Engineering, University of Moratuwa, Moratuwa, Sri Lanka Email: [email protected]

Abstract— The study of empirical propagation models for mobile channels has been done extensively, but the applicability of those models to a Fixed Wireless Access (FWA) System is not appropriately tested. The candidate models include Hata Okumura, COST 231 Hata, COST-Walfisch-Ikegami and Erceg. From those models Erceg model seems to be the most suitable. In this paper field measurements are taken for the 3.5GHz Fixed WiMAX network in Katubedda, Sri Lanka. Those are used to validate the applicability of the above mentioned propagation models in a sub urban environment in Sri Lanka.

I. I NTRODUCTION The knowledge of Propagation conditions is mandatory for the development and optimization of FWA systems. Propagation models help to understand the interferences in the network, which results in developing a well structured network with better quality. Those can be classified mainly into two extremes, i.e. Fully empirical models and Deterministic models. There are some models which have the characteristics of both types. Those are known as Semi-empirical models. Empirical models are based on practically measured data. Since few parameters are used, these models are simple but not very accurate. In this paper we consider three models which are categorized as empirical models for macrocellular environment. These include Hata-Okumura model, COST231 Hata model and Erceg model. On the other hand, deterministic models are very accurate, but these need high computational power and large number of site specific details. Those estimate propagation of radio waves analytically. In this paper we are not considering any model that is classified as a deterministic model. Some of the examples include Ray Tracing and Ikegami model. As mentioned earlier, semi-empirical models are based on both empirical data and deterministic aspects. COST231 Walfisch-Ikegami model is categorized as a semiempirical model and this will be considered in the next sections of this paper. All these models estimate the mean path loss based on parameters such as antenna heights of the transmitter and receiver, distance between them, etc... These models have been extensively validated for mobile networks, but applicability of those to the FWA systems has not been tested at a large scale. In this paper we are comparing the field measurements taken at Katubedda, Sri Lanka with the predictions made by the

propagation path loss models. Some models that are discussed here, have been tested earlier in Belgium [1] [2], UK [3], Croatia [4] and Norway [5] [6] [7]. II. DATA MEASUREMENT METHODOLOGY We used a commercially available IEEE 802.16d network in Sri Lanka, which belongs to Dialog Broadband Networks (DBN) Private Limited, for the field measurements. Under the sponsorship of DBN we gathered Fixed WiMAX signal parameters in Katubedda area in Sri Lanka. The network has three Base Transceiver Systems (BTS) in Katubedda area with approximately 400 customers. The arrangement includes a BTS with sectoral antennas, J34216V01120N of Laird Technologies, in three sectors. This is a high performance, WiMAX/Broad Band wireless, 3.5 GHz operating, vertical polarization, 1200 antenna with 16 dBi gain. We considered two BTSs in our field measurements. First one which is known as Katubedda BTS was 42m above the ground level and the other one which is known as Kaldemulla BTS was 28m above the ground level. The channel bandwidth (BW) of the Fixed WiMAX system can be varied between 1.25MHz to 28MHz [8]. This network used 1.75MHz and 3.5MHz channel BWs. The Subscriber Unit (SU) is a BreezeMAX broadband data Customer Premises Equipment (CPE) with 17 dBi gain. It supports BPSK, QPSK, 16QAM and 64QAM modulation schemes. We constructed a tripod, to carry the CPE and take the measurements easily at 3m, 4m and 5m CPE heights. A pickup truck was used to carry the tripod with the CPE. A dedicated software was developed to acquire signal parameters automatically. This software detects the correct sector of the base station and it automatically sends the signal to the CPE to rotate to the base station’s direction. Then the rotated CPE will be locked to the BTS’s frequency and the measurements will be taken. The measured data is stored in a text file inside a separate folder with the site name. Those parameters include SNR and RSSI of Up Link and Down Link radio channels. A Global Positioning System (GPS) was used to accurately take the location coordinates. All measurements were taken within a 100m to 1.5km distance range from the BTS. The upper boundary of the distance is set by the coverage area of the BTSs, since the antennas in the BTSs are tilted in such

a way that the coverage area is limited to a 1.5km radius. This terrain has moderate to heavy tree density with buildings having about 10m height on average. The Katubedda area can be categorized as a suburban environment in Sri Lanka. Approximately 200 field measurements were taken at 5m height for the Katubedda BTS. Also around 100 measurements were taken at 3m and 4m heights for the Katubedda BTS. For the Kaldemulla BTS around 150 measurements were taken at different locations at 5m CPE height. III. PROPAGATION PATH LOSS MODELS Least Square (LS) regression was taken as the basis for the comparison of the models. Equation (1) gives the standard equation for the LS regression. In that d refers to the distance between BTS and CPE, d0 refers to the reference point at far field of the base station antenna which is considered as 100m and n refers to the path loss exponent. P L(d) = P L(d0 ) + 10nlog10 (d/d0 )

(1)

Following empirical models were compared against the LS line to verify the validity of those models to the data measured.

C. COST Walfisch-Ikegami model This is the COST 231 proposed Walfisch and Ikigami combined model [11]. This gives a better path loss prediction. Characteristics of urban environment such as, height of buildings (hroof ) in m, width of roads (w) in m, building separation (b) in m, and road orientation with respect to the direct radio path (ϕ). In our analysis we have used 10m for hroof , 12m for w, 20m for b and 630 for ϕ. The model has separate equations for Line of Sight (LOS) and Non LOS (NLOS) conditions. Equation (5) gives the equation for NLOS conditions, which we used in our analysis. P LN LOS (dB) = LF S + Lrts (wr , f, ∆hm , ϕ) + LM SD (∆ht , ht , d, f, bs ) (5) LF S gives free space loss, which is defined in (8), Lrts gives the Roof-to-street loss (9), and LM SD is the multiple screen diffraction loss (11). ∆hm is given by (6) and ∆ht is given by (7). ∆hm = ht − hroof (6) ∆ht = ht − hroof

A. Hata-Okumura model This model is best suited for large cell coverage (distances up to 100 km) and it can extrapolate predictions in the 150 1500 MHz band. Also this is the widely used model for most of the signal strength predictions in macrocellular environment [9], [10]. Although, its frequency band is outside the band of Fixed WiMAX, its simplicity has made it to be used widely in propagation predictions. The path loss equation is give by (2). P Lurban (dB) = 69.55 + 26.16log10 (fc ) − 13.82log10 (ht ) − a(hm ) + [44.9 − 6.55log10 (ht )]log10 (d) (2) fc is the operating frequency in MHz, ht and hm are the BTS antenna height and the CPE height in m, d is the distance from BTS to CPE in km and a(hm ) is the correction factor for mobile unit antenna height in dB (3). a(hm ) = 3.2(log10 (11.75hm ))2 − 4.97

(3)

B. COST-231 Hata model COST 231 project is the development of the outdoor propagation models for application in urban areas at higher frequencies. It has extended the earlier Hata-Okumura model to support frequencies ranging from 1500 MHz up to 2000 MHz [11]. The main advantage is that it contains corrections for urban, suburban and rural (flat) environments. The basic equation for path loss in dB is given by (4), P Lurban (dB) = 46.3 + 33.9log10 (fc ) − 13.82log10 (ht ) − a(hm ) + [44.9 − 6.55log10 (ht )]log10 (d) + Cm (4) The symbols have the same meaning as in the Hata-Okumura model in III-A. Cm is defined as 0 dB for medium and sub urban areas with moderate tree densities and 3 dB for Metropotitan centers.

(7)

where ht gives base station height (m) and hroof gives the height of the building (m). LF S = 32.4 + 20log10 (d) + 20log10 (fc ) Lrts = −8.8 + 10log10 (fc ) + 20log10 (∆hm ) − 10log10 (w) + Lori

(8)

(9)

In the above equation hm gives the CPE height in m and Lori is the street orientation function which depends on ϕ. We used the function (10) for this. Lori = 4.0 − 0.114(ϕ − 55) since 550 ≤ ϕ ≤ 900

(10)

LM SD = Lbsh + ka + kd log10 (d) + kf log10 (f ) − 9log10 (b) (11) In equation (11), Lbsh is given by (12), Ka is 54, Kd is 18, and Kf is given by (13). Lbsh = −18 ∗ log10 (1 + ∆ht )

(12)

Kf = −4 + 0.7((fc /925) − 1)

(13)

D. Erceg model This was developed by Erceg et al. and the experimental data were taken in several suburban areas in New Jersey and around Seattle, Chicago, Atlanta, and Dallas. The base antenna heights were in the range from 12 to 79 m [12]. This has categorized three different terrain categories. The maximum path loss category is hilly terrain with moderate-to-heavy tree densities (Category A), the minimum path loss category is mostly flat terrain with light tree densities (Category C) and the middle category can be characterized as either mostly flat terrain with moderate-to-heavy tree densities, or hilly terrain

with light tree densities (Type B). This model is recommended by IEEE 802.16 Broadband Wireless Access Working Group [8]. The Path Loss in dB is given by equation (14), P L = A + 10γlog10 (d/d0) + s

; d ≥ d0

(14)

where, A gives decibel path loss at distance d0 (15), γ gives path loss exponent (16) and s is the shadowing component given by (17). A = 20log10 (4πd0 /λ) (15) In this λ gives the wavelength in m. γ = (a − bhb + c/hb ) + χσγ

(16)

The parameter hb is the base station antenna height in meters (80m ≥ hb ≥ 10m), χ is a zero-mean Gaussian variable of unit standard deviation N[0,1] and a, b, c and σγ are constants for each terrain category given by Table I. s = yσ

(17)

TABLE I N UMERICAL VALUES OF THE E RCEG M ODEL PARAMETERS Model Parameter

Terrain Type A

Terrain Type B

Terrain Type C

a b (in m−1 ) c (in m) σγ µσ σσ

4.6 0.0075

4.0 0.0065

3.6 0.0050

1 2.6

17.1

20.0

0.57 10.6 2.3

0.75 9.6 3.0

0.59 8.2 1.6

IV. COMPARISON WITH MEASUREMENTS FOR THE DOWN LINK A. Path loss analysis For the ease of presentation we have used 5m and 3m CPE heights in this paper. Fig.1 gives the scatter plot of the measured path losses with the path loss plots from standard propagation path loss models. In Table II the corresponding error statistics in terms of the mean prediction error, µ, and the standard deviation of the prediction errors, σ, are given for each model plotted in Fig.1. The prediction errors are calculated as the error between practical data collected and the points in path loss lines of the propagation models. From the statistical analysis in Table II we can see that the Erceg B model, i.e. the Flat/Moderate-to-Heavy Tree density terrain has closer µ and σ values to the Least Square Regression Fit. Fig.2 gives the scatter plot of the measured path losses along with the path loss plots for standard propagation path loss models for reciever height of 3m. The corresponding error statistics are given by the Table III. As can be seen from the table, the best fit at 3m CPE height is the Free Space Loss Model. This is due to the nulls obtained in that receiver level. By analyzing the plots it is seen that the most suited model is the Hata model. From the Erceg Model’s terrain types, the

Fig. 1. Comparison of the measured Path Loss with Propagation Models at 5m CPE height for the Down Link TABLE II E RROR STATISTICS OF M ODEL P REDICTIONS COMPARED WITH THE R EGRESSION FIT AT 5 M CPE HEIGHT FOR THE D OWN L INK Model Least Square Erceg A Erceg B Erceg C Free Space Loss Hata W-I COST231 Hata

µ

σ

0 2.9767 0.9180 8.0789 22.8252 7.5525 39.5640 12.6101

23.031 25.5477 24.9519 25.8940 32.5112 25.2179 46.3027 27.1453

most suited type is Erceg Type C, i.e. Flat/Light Tree Density. It is due to the fact that such a lower CPE height sees lesser trees compared to the 5m CPE height. As the CPE height varies, the terrain which is seen by the CPE also varies. This is a issue faced only in Fixed Wireless networks. B. RSSI analysis Fig.3 shows the Down Link(DL) measured Receiver Signal Strength Indicator(RSSI) scatter plots at 5m CPE height with the RSSI calculated using the standard propagation path loss models. Fig.4 shows the DL RSSI scatter plot at 3m CPE height with the values from propagation path loss models. Those figures shows that all signal strength measurements drawn in scatter plots are better than the W-I model for Path Loss in this suburban environments. An improvement in the RSSI can be seen with the increase in the CPE height in certain distances well as a reduction also take when increasing height compared to lower height levels due to the effects of foliage. Fig.5 shows the Cumulative Distribution Function(CDF) of the RSSI at 3m, 4m and 5m CPE heights. The values at 3m and 4m (median value of -65 and-64.1 dBm) are closer together than those at 5 m (median value -84dBm). The received power

Fig. 2. Comparison of the measured Path Loss with Propagation Models at 3m CPE height for the Down Link

Fig. 3. Comparison of the measured RSSI at 5m CPE height with Propagation Models for the Down Link

TABLE III E RROR STATISTICS OF M ODEL P REDICTIONS COMPARED WITH THE R EGRESSION FIT AT 3 M CPE HEIGHT FOR THE D OWN L INK Model LS Erceg A Erceg B Erceg C Free Space Loss Hata W-I COST231 Hata

µ

σ

0 20.4487 16.3785 11.1126 9.8841 23.1383 56.1258 28.2658

33.5043 43.0340 40.9087 38.8335 37.1924 43.9185 67.2199 46.7561

at 5 m is considerably higher than at lower receiver heights. C. SNR analysis Fig.6 and Fig.7 show the SNR as a function of RSSI for all DL measurements carried out at 5m and 3m CPE heights respectively. The Signal to Noise Ratio (SNR) is a better measure for the actual operating conditions of the receiver than the RSSI value. This is due to fact that SNR value takes into account interference and noise conditions in addition to signal strength. On the other hand, the SNR and RSSI values should be closely correlated. Equation (18) shows the theoretical relationship between the SNR and RSSI [5]. SN R =

34 (1 +

5.2∗1013 1/9 (102+RSSI)9 )

(18)

For the low RSSI values there seems to be a linear relationship between the RSSI and SNR values as expected, where a 1 dB increase in SNR gives a 1 dB increase in the RSSI value. For higher RSSI values, the SNR approaches a limit mainly caused by saturation effects in the receiver.

Fig. 4. Comparison of the measured RSSI at 3m CPE height with Propagation Models for the Down Link

The low SNR values for relatively good RSSI values might be caused by interference. In many situations it can be useful to have a mathematical expression for estimating the SNR as a function of the RSSI. V. COMPARISON WITH MEASUREMENTS FOR THE UP LINK In the previous sections the analysis for the DL has been carried out. In this section the analysis for the UL will be done. A. Path loss analysis Fig.8 and Fig.9 show the comparison of the path loss models results with the field measurements at 5m and 3m CPE height

Fig. 5. Comparison of the cumulative distribution curve for the three CPE heights

Fig. 7. SNR plot of measured data and theoretical results at 3m CPE height for the Down Link

Fig. 6. SNR plot of measured data and theoretical results at 5m CPE height for the Down Link Fig. 8. Comparison of the Path Loss with the Propagation Models at 5m CPE height for the Up Link

respectively. B. RSSI analysis Fig.10 and Fig.11 show the comparison of the path loss models results with the field measurements at 5m and 3m CPE height respectively. C. SNR analysis Fig.12 show the SNR as a function of RSSI for all UL measurements carried out at 5m height. As can be seen from the graph related to UL, it is evident that the UL also gives the same results as the DL. Interferences in the both channels closely related. Hence, same conclusions as in the previous section which is for the DL case are valid in the UL as well. VI. CONCLUSION In this paper propagation measurements for FWA network operating based on IEEE 802.16d at 3.5 GHz are analyzed

and discussed. A statistical path loss model for a suburban Sri Lankan environment is proposed. Different receiver heights are analyzed to evaluate the effect of the CPE height differences for the Fixed WiMAX network. The path loss exponent highly depends upon the receiver height. ACKNOWLEDGMENT All the results mentioned in this paper were gathered during a project funded by Dialog Broadband Networks (Pvt)Ltd., Sri Lanka. All the authors like to thank Dialog Broadband Networks for their funding and support.

Fig. 9. Comparison of the Path Loss with the Propagation Models at 3m CPE height for the Up Link

Fig. 10. Comparison of the RSSI with the theoretical values at 5m CPE height for the Up Link

Fig. 11. Comparison of the RSSI with the theoretical values at 3m CPE height for the Up Link

Fig. 12. SNR plot of measured data and theoretical results at 5m CPE height for the Up Link

R EFERENCES [1] W. Joseph, L. Roelens, and L. Martens, “Path Loss Model for Wireless Applications at 3500 MHz,” in IEEE Antennas and Propagation Society International Syposium with USNC/URSI National Radio Science and AMEREM Meetings, 2006, pp 4751 - 4754 [2] W. Joseph and L. Martens, “Performance Evaluation of Broadband Fixed Wireless System based on IEEE 802.16,” in IEEE Wireless Communications and Networking Conference, 2006, pp 978 - 983 [3] V.S. Abhayawardhana, I.J. Wassell, D. Crosby, M.P. Sellars, and M.G. Brown, “Comparison of Empirical Propagation Path Loss Models for Fixed Wireless Access Systems,” in IEEE 61st Vehicular Technology Conference, 2005, pp 73-77 [4] J. Milanovic, S. Rimac-Drlje, and K. Bejuk, “Comparison of Propagation Models Accuracy for WiMAX on 3.5 GHz,” in IEEE International Conference on Electronics, Circuits and Systems, 2007, pp 111-114 [5] P. Grnsund, O. Grndalen, T. Breivik, and P. Engelstad, “Fixed WiMAX Field Trial Measurements and the derivation of a Path Loss Model,” in M-CSC, 2007. [6] O. Grndalen, P. Grnsund, T. Breivik, and P. Engelstaad, “Fixed WiMAX Field Trial and Analyses,” Mobile Summit, 2007.

[7] P. Grnsund, P. Engelstaad, T. Johnsen, and T. Skeie, “The physical performance and path loss in a fixed WiMAX deployment,” in Proceedings of the International Conference on Communication and Mobile Computing, 2007, pp 439 -444 [8] IEEE Std. 802.16-2004, IEEE Standard for Local and Metropolitan area networks, “Part 16: Air interface for fixed broadband wireless access systems”, 2004. [9] M. Hata, “Empirical formula for propagation loss in land mobile radio services,” IEEE Transactions on Vehicular Technology, vol. VT-29, pp. 317325, September 1981. [10] Y. Okumura, “Field strength and its variability in VHF and UHF land-mobile radio-services,” Review of the Electrical Communications Laboratory, vol. 16, September-October 1968. [11] COST 231 Final Report, Digital Mobile Radio Towards Future Generation Systems, Brussels: COST Telecom Secretariat, 1999. [12] Erceg Vinko, et al., “An Empirically Based Path Loss Model for Wireless Channels in Suburban Environments”, IEEE Journal on selected areas in communications, vol. 17, no. 7, pp.1205-1211, July 1999.