Site-Specific Validation of ITU Indoor Path Loss Model at 2.4 GHz

Site-Specific Validation of ITU Indoor Path Loss Model at 2.4 GHz Theofilos Chrysikos (1), Giannis Georgopoulos (1) and Stavros Kotsopoulos (1) (1) Wireless Telecommunications Laboratory Department of Electrical & Computer Engineering University of Patras 26500 Greece txrysiko@ece.upatras.gr, igewrgopoy@ece.upatras.gr, kotsop@ece.upatras.gr Abstract This paper presents a site-specific validation of the ITU indoor path loss model at 2.4 GHz. Based on measurements acquired in a recent experiment for a WLAN indoor office environment, we are able to accomplish a numerical adjustment of the Site-general ITU model to the specific measured data that reflect the intrinsic characteristics of the complex indoor topology, thus validating the Site-Specific ITU model at 2.4 GHz. For the first time, we are in position to provide values for the model s parameters that concern specifically this frequency band, which is of utmost importance for wireless networks, mostly for WLAN channels. This marks a major step towards an even greater degree of precision and reliability in the usage of indoor RF models for link budget design in modern wireless communications. 1. Introduction A plethora of path loss models have been developed in order to calculate the average path loss (in dbm) and therefore, by knowledge of the transmitted power, the average received power at a certain distance, for the given frequency of the operating system under study. Path loss models can be distinguished in two basic categories: the empirical models, whose mathematical formula, often complex, is derived out of experimental data and site measurements, and the semi-empirical models that combine theoretical work and real-case measurements [1]. The fundamental path loss model is the Free Space Model that predicts an inverse square dependence of the average received power versus distance between transmitter and receiver. While other path loss models adopt the inverse-square law for outdoor scenarios, extensive research has developed a significant number of modified power-law models that try to take into account the complex nature of real life wireless channels. The indoor propagation channel is a much more complex case than the outdoor environment, due to the increased number of obstacles, whose dimensions are close to the wavelength of the propagated electromagnetic wave, where the presence of multiple types of walls and floors add to the complexity of the calculations. As a result, various path loss models have been developed to describe the indoor channel and its multiplicative effects that cause the attenuation of a transmitted signal [3]. What matters most is not only the calculation of the average path loss (enabling the calculation of the average received power) but the thorough, to the best degree, description and prediction, on behalf of the path loss model, of all the different attenuation factors and their combined impact on the propagated wave and its power levels [3]. In a recent work [4] we examined two different indoor scenarios: a residence in a building consisting of home apartments in downtown Patras, Greece, and an office, the Wireless Telecommunications Laboratory of the Department of Electrical and Computer Engineering of the University of Patras. The Laboratory s premises are located on the second floor of a three floor building of the Department of Electrical and Computer Engineering. In both buildings, a WLAN was considered, where an Access Point (AP WLAN router) was in each case the fixed transmitter, and the receiver was a laptop equipped with a WLAN card and the NETSTUMBLER 0.40 software for proper reception and processing of the measured data. Both transmitter and receiver antennas were omnidirectional. Measurements were taken in various distances between transmitter and receiver, including both LOS and NLOS scenarios. Same floor measurements as well as multiple floor measurements were acquired, enabling us to perform an analytical comparison of the theoretical values 978-1-4244-4439-7/09/$25.00 2009 IEEE

predicted by the RF models and the measured data. In this way, we were able to validate the most important existing path loss models for the specific frequency of 2.4 GHz. Based on this background work, we are now able to provide mathematical corrections to the ITU model and thus validate this model for 2.4 GHz and particularly for the office environment. Due to insufficient number of multiple floor measurements in the home scenario (which is caused by our inability to have access within the residences of different floors of the apartment building), our adjustments are limited to the office scenario. However, this by no means diminishes the significance of our work, since this is by no means a typical office scheme, as it will be explained in the following section. 2. Site and Method Description The exact topology of the building where the Wireless Telecommunications Laboratory is hosted can be seen in the Appendix Section of the present paper, featuring the second floor of the building, where the premises of the WTL can be found, including a main room, several offices scattered around this large room, a public corridor, the lab area and the WC rooms. The transmitter is a fixed AP of the WLAN (router) located on the external wall of the large room. The notion of the standard office environment does not stand: both the topology of the second floor (same-floor measurements) and the topologies of the other two floors (where multiple-floor measurements were taken) demonstrate a radical shift of the stereotype office scheme. This way we are able to witness the direct impact of a site-specific environment and its very particular intrinsic characteristics on the propagation of the signal and its average signal strength, as well as incorporate these phenomena dynamically in our validation, by comparison to the theoretical values predicted by the model and the mathematical correction of its parameters. A total of 22 measurements were taken for the single floor measurements, 10 measurements for the one-floor difference (fixed transmitter on the second floor, moving receiver on the third floor of the building) and 7 measurements in the auditoriums and public hall of the ground floor marking a two-floor difference between the transmitter and the moving receiver. In each case, the predicted theoretical value of the ITU model is compared to the measured one, for the same exact location. The relative error (%) is calculated and then compared to the relative error of the adjusted model for the same location, thus proving the positive contribution of our work. 3. Site-General ITU indoor RF model The ITU site-general model for path loss prediction in an indoor propagation environment is given by [5]: L 20log f Nlog d Lf n 28 db (1) Where N is the distance power decay index, f is the frequency in MHz, d is the distance in meters (d > 1m), Lf(n) is the floor penetration loss factor and n is the number of floors between the transmitter and the receiver. The power decay index and the floor penetration loss factor are specified for a number of frequencies and types of indoor environment by tables provided in [5]. For a zero value of the floor penetration loss factor, thus considering the transmitter and the receiver on the same floor level, and for N=20, the ITU model is identical to the free space model for the outdoor environment. The indoor channel study requires N=18 for a LOS path between transmitter and receiver (a path loss exponent equal to 1.8). Propagation around corners or through walls, requires N = 40. In the case of long paths, the reflected path(s) may interfere, resulting again in N = 40. However, no exact specifications existed so far for 2.4 GHz, despite its importance as the operating frequency of WLANs in an indoor propagation environment. The closest specifications concern the frequency zone of 1.8 2 GHz which provided the values for our original estimations of the ITU model, before its site-specific adjustment and validation from our own experimental measurements. 4. Measurements and Validation A. Same floor measurements Initially the single floor scenario is examined. In that case, Lf(n)=0. At 2.4 GHz and for distances between the transmitter and the receiver that reach up to 16 m, the ITU model is mathematically expressed by (N=28 since NLOS paths have to be taken into consideration): 39.9 28 log (2) Since the transmitted power is considered to be equal to 17 dbm, the average received power is given by: 23 28log (3) For distances beyond 16 m, the power decay index equals N=38 (increasing impact of the distance-caused attenuation on the average signal strength). Thus the ITU model and the average received power in this case are respectively given by:

39.9 38 log (4) 23 38log (5) Table 1. Measured and ITU- predicted values of average received power (same floor scenario) Location T-R Pr ITU Error % (m) (dbm) (dbm) A 8-48 -47.3 1,45833 B 11.5-55 -51.6 6,18182 C 13-53 -53 0 D 15-54 -54.7 1,296296 E 5-45 -41.8 7,11111 F 13-51 -53 3,921569 G 20-66 -69 4,545455 H 21-71 -70 1,40845 I 18-71 -67.7 4,64789 J 16-59 -65.7 11,35593 K - - L - - M 20-65 -69 6,153846 N 22-75 -71 5,33333 O 20-74 -69 6,75676 P 10.5-55 -50.5 8,18182 Q 10.5-41 -50.5 23,17073 R 10-40 -50 25 S 10-40 -50 25 T 16.5-63 -66 4,761905 U 17-68 -66.7 1,91176 V 11-49 -51.1 4,285714 W 15-54 -54.7 1,296296 X 16.5-59 -63 6,779661 Figure 1. Response of the ITU model compared to the measured values (same floor measurements) As it can be seen, the ITU model with the varying power decay index corresponds rather well as compared to the measured data. The total of the 22 measurements provide an average error of 7.299 %. This however, is not a fair approach, as the average error is influenced heavily by the measurements in the locations Q, R, S, where a strong LOS path is present. In that case, as seen in [4], only the Free Space Model and the Multi-Wall-Floor model can predict reliably the actual signal strength. Should we decide to ignore these three measurements for the sake of an approach where the OLOS scenario is more fairly represented, the average error equals 4.599 %. The ITU model for same-floor measurements provides, therefore, on general OLOS terms, an average error less than 5% (compared to the actual measured values), approximately equal to 4.6%. If the LOS measurements are also taken into account, the

average error increases to a 7.3% that remains however within satisfactory levels, even if it adopts a less fair approach as to the number of OLOS measured locations compared to the three LOS locations. 23 38 log 4 1.375 = 23 38 log 5.5 (9) B. Multiple floor measurements In the multiple floor scheme, the floor penetration loss factor is 15 + 4(n - 1), where n is the number of floors penetrated. Thus the original ITU model is: 39.9 38 log 15 4 1 (6) Figure. 4. Response of ITU model-2 By proposing a radically new adjustment of the ITU model, the ITU model-3 is derived: 23 28 log 7.5 1 (10) Figure. 2. Response of original ITU model For Lf(n) = 4n, which is the value for the home environment, the average path loss and average received power (ITU model-1) are given by: 39.9 38 log 4 (7) 23 38 log 4 (8) Figure. 5. Response of ITU model-3 Figure. 3. Response of ITU model-1 By adjusting the model with the use of a numerical corrective factor (1.375) the ITU model-2 is derived: All the predicted values of all the suggested ITU models and their respective errors compared to the measured data are presented in Table 2. For one-floor difference between transmitter and receiver, the best performance belongs to the ITU-1 model (4n). For two-floor difference between the transmitter and the receiver, the best performance is observed by the ITU-3 model where Lf(n) = 7.5(n+1). The ITU-2 model (5.5n) provides the in-between model, with an

average prediction (and error) with less deviation between the one-floor and the two-floor scenarios (an intermediate solution). When n=1 and there is one floor penetrated by the transmitted signal in an indoor scenario at 2.4 GHz (WLAN), the ITU-1 model will be applied. If n=2, then the ITU-3 model will be applied. For an application of a moving user shifting between floors, where n=1 and n=2 at varying periods of time, then the ITU-2 model shall be applied. Table 2. Measured and ITU-predicted values of average received power (multiple floors scenario) Location T-R distance Pr (dbm) ITU 1 Lf(n)= Error-1 % ITU 2 Lf(n)= Error - 2 % ITU 3 Lf(n)= Error - 3 % (m) 4n 4*1.375*n 7.5*(n+1) A (1-f) 5.8-60 -55.2-8 -56.7 5,5-59.3 1,16 B 11.4-67 -66.1 1,34-67.6 0,89-67.5 0,74 C 10.9-62 -65.3 5,32-66.8 7,74-67.05 8,14 D 10.4-58 -64.6 11,37-66.1 13,96-66.48 14,62 E 11.4-67 -66.1 1,34-67.6 0,89-67.5 0,74 F 20.2-77 -75.3 2,2-76.8 0,25-74.55 3,18 G 17.7-77 -73.1 5,06-74.6 3,11-72.9 5,32 H 17.7-75 -73.1 2,53-74.6 0,53-72.9 2,8 I 16.7-73 -72.2 1,09-73.7 0,95-72.2 1,09 J 17.7-76 -73.1 3,81-74.6 1,84-72.9 4,07 A (2-f) 9.6-78 -67.3 13,71-70.4 9,74-73 6,41 B 15.2-76 -74.7 1,71-77.7 2,23-78.4 3,15 C 17.9-76 -77.3 1,71-80.3 5,65-80.5 5,92 D 18.9-81 -78.2 3,45-81.2 0,24-81.24 0,29 E 24.2-81 -82.2 1,48-85.2 5,18-84.2 3,95 F 12.9-76 -72 5,26-75 1,31-76.6 0,78 G 17.5-81 -77 4,93-80 1,23-80.31 0,85 5. Conclusions The Site-Specific validation of the ITU indoor RF model at 2.4 GHz manifests that, for Lf(n)=0, the power decay index is N=28 for distances up to 16 m, and for greater distances N=38. The average error is 7.299 % when taking into consideration the LOSdominated measurements, and 4.599 % for usual OLOS locations. The Multi-Wall-Floor predicts with an average error of 2.406 %. Table 3. Validation of ITU model at 2.4 GHz Distance (m) Power decay index (N) Site-specific ITU indoor model at 2.4 GHZ 1 < d < 16 28 39.9 28 log d > 16 38 39.9 38 log For multiple floor scenarios, the power decay index equals N=38. As far as the floor penetration loss factor is concerned, there are three adjustments that need to be taken into account at 2.4 GHz. If n=1, that is if the transmitter and the receiver are separated by one floor, then Lf(n) = 4n and the ITU model average path loss prediction is given by Eq. 7. If n=2, that is if the transmitter and the receiver are separated by two floors, then Lf(n) = 5.5n and the ITU average path loss prediction is given by: 39.9 38 log 5.5 (11) Finally, if there is a more complex motion scenario of the receiver, where n does not remain constant but variates between n=1 and n=2, then Lf(n) = 7.5 (n+1) and the ITU predicted path loss is expressed by: 39.9 38 log 7.5 1 (12) Table 4 depicts the average error between each ITU suggested model and the measured values. Thus we

can confirm which numerical adjustment of the floor penetration loss factor is more suitable. The MWF model predicts with an average error of 2.019 %. Table 4. Evaluation of ITU suggested adjustments Average 4n 5.5n 7.5(n+1) Error (%) 1-floor 2,606 3,566 4,186 difference 2-floor 4,607 3,654 3,05 difference multiple floor 3,607 3,602 3,618 [4] T. Chrysikos, G. Georgopoulos, K. Birkos and S. Kotsopoulos, Wireless Channel Characterization: On the validation issues of indoor RF models at 2.4 GHz, First Panhellenic Conference on Electronics and Telecommunications (PACET), Patras, Greece, March 20-22, 2009. [5] J. Seybold, Introduction to RF Propagation, Wiley Interscience, 2005. Appendix Second floor of WTL (Same floor measurements) 6. Future work Immediate future work consists of studying the site-specific estimation of shadowing effects, most notably the standard shadowing deviation (db), thus allowing us a dynamic re-evaluation of the Log- Distance path loss model and the coverage percentage. This way, we shall be able to reflect on the room s topology the varying deviation that describes the random shadowing effects, thus establishing a geodynamic depiction of the randomness of shadowing and its impact on the coverage percentage in comparison to a theoretical value, i.e. 95%. Secondly, the choice of the best value as the average received signal strength must be re-approached on a different basis, that will take into account the worst case scenario as well (worst of all instantaneous signal strengths to be chosen as the average signal strength for a predetermined time period) and for a median value case study that will adopt a third, in-between approach. The comparisons of all relative errors of all indoor RF models for these three different scenarios will help further validate and evaluate the existing path loss models in a variable scenario of sub-optimal case studies. Thirdly, the indoor WLAN channel and the topology in question shall be depicted electronically so that the electromagnetic wave propagation over the wireless channel shall be simulated with ray tracing software to provide a benchmark that can be applied to predict reliably the path loss in such environments. 7. References [1] J. D. Parsons, The Mobile Radio Propagation Channel, Wiley Interscience, 2000. [2] A. Goldsmith, Wireless Communications, Stanford University, 2005. [3] S. Kotsopoulos and G. Karagiannidis, Mobile Communication, Papasotiriou SA Publication, 1997.