Sensitivity of Aggregate UWB Interference Models to their Parameters

Transcription

1 Sensitivity of Aggregate UWB Interference Models to their Parameters Werner Sörgel 1, Michael Baldauf 1, Marwan Younis 1, and Werner Wiesbeck 1 1 Institut für Höchstfrequenztechnik und Elektronik, Universität Karlsruhe (TH), Kaiserstr. 12, Karlsruhe, Germany, Werner.Wiesbeck@ihe.uka.de Abstract This paper is concerned with the calculation of the aggregate interference power from ultra-wideband devices received by a point to point communication system. The source based model for the aggregate interference power and its parameters are presented together with results for different sets of the parameter values. The results show that the applied wave propagation model has a very strong influence. The paper treats the free space propagation model, the log-distance pathloss model with different propagation exponents and an adapted Hata model. The differences in interference power between free space propagation and the Hata model for urban propagation are up to 22 db. 1 Introduction Low power ultra-wideband (UWB) technology offers new and promising possibilities for high data rate communications and high resolution sensors. The key issue for the deployment of such systems is the coexistence of UWB systems with narrowband systems located in the same frequency range. Therefore the maximum transmit power spectral density (PSD) has to be regulated in a way that no harmful interference occurs. The United States FCC has issued an upper limit of dbm/mhz for the frequency range of GHz [FCC02]. In Europe, the regulation process and the related investigations are still ongoing and among others the coexistence with point-to-point communication and fixed wireless access systems is of very high importance. For the assessment of the interference by multiple UWB devices theoretical models for the aggregate interference PSD at the victim receiver are used [Mit04, ECC04]. The present paper introduces an interference model that allows for an accurate calculation of the interference and that is convenient to use for a wide range of parameter sets. The calculated examples show the sensitivity of the aggregate interference to wave propagation effects. 2 Model Geometry It is assumed that the aggregate interference power is the incoherent sum of the received interference powers generated by all UWB devices within the victim s scope. The contributions from different interferers are incoherent since the single sources are not synchronized. The aggregate interference PSD at the victim receiver is determined by several factors: the transmit power of a single UWB device, which is given as the effective isotropic radiated PSD (EIRPSD) p Tx. The EIRPSD is an upper bound for the radiated power and assumes that the maximum antenna gain of the UWB device is valid for all directions. Then the spatial distribution of the UWB devices around the victim receiver is important. As long as no further knowledge about a specific scenario is available a uniform distribution is a fair assumption. Due to the uncertainty of the exact positions of the concentrated 201

2 sources their power is smeared out in the regarded area or volume, yielding an average for a certain device density. In the present case it is assumed that the UWB devices are distributed uniformly in the plane below the victim with a constant density 2D. The victim is located in the height H Tx above the ground as shown in Fig. 1. The resulting average two-dimensional EIRPSD in W/(Hz m 2 ) is p Tx,av = a 2D p Tx. (1) The activity factor a accounts for the fact that only a fraction a of the deployed UWB devices is active at the same time. A very crucial parameter is the path-loss L p between the distributed sources and the victim antenna. Together with the effective antenna area A w,rx (,) of the victim the path loss determines the propagation attenuation depending on the distance R to the regarded source point. Fig. 1: Model geometry for aggregate interference calculations. 40 ITU R F.699 ITU R F.1245 isotropic 30 gain in dbi off boresight angle α in deg Fig. 2: Envelope antenna radiation patterns for the victim (f =4 GHz and dish diameter 3.7 m) for ITU-R F.699 (peak envelope) and ITU-R F.1245 (average sidelobe level), as reference an ideal isotropic receiving antenna is shown. 202

3 The effective antenna area A w,rx is calculated from the antenna gain G Rx : A W,Rx (,) = 2 4 G Rx (,). (2) The victim is assumed to be a point-to-point communication system operating at 4.0 GHz with a 3.7 m diameter dish antenna and a maximum antenna gain of G Rx = 41 dbi. These values present a conservative scenario since the path loss is increasing with frequency and the 4 GHz band is located the in the lower part of the UWB frequency range. The antenna characteristic G Rx (,) is calculated from the rotational symmetric radiation pattern by a coordinate transformation. The radiation pattern of the dish is given as G Rx () with the argument denoting the off-boresight angle. The boresight direction of the dish = 0 coincides with = 90 and = 0 in the global coordinate system, i.e. there is no downtilt applied. For G Rx () two ITU recommendations are available: ITU-R F.699 defines a reference radiation pattern, i.e. the peak envelope of the pattern including the peaks of the sidelobes and ITU-R F defines an average radiation pattern, which gives a mean sidelobe level. Fig. 2 shows the radiation patterns for the victim at f = 4 GHz with a dish diameter of 3.7 m for both ITU recommendations. Additionally the radiation pattern for an ideal isotropic antenna is shown. 3 Path Loss Models In the following the different path loss models are presented: the log-distance path loss model (single slope), which includes the important special case of free space propagation, the dual slope log-distance path loss model and an adapted Hata Model, which has been adapted from earlier co-existence studies. 3.1 Log-Distance Path Loss Model (Single Slope) A basic empiric channel model is the log-distance path-loss model with the intercept distance and a single slope given by the path loss exponent n 2 1 = L p,n 4 R n L p,n db = L 0 db +10n log 10 R. (3) For a path loss exponent of n=2 the free space condition is preserved. For the single slope model is the distance at which the free space path loss value is assumed independently of n: L p,n ( ) = L p,2 ( ). The free space wavelength is denoted by and the distance between transmitter and receiver is R. This model is the basis for various more detailed wave propagation models like Dual- and Multi-slope models, which introduce different path loss exponents for different range intervals. Also empiric models based on extended measurements like Okumura-Hata, use the log-distance path loss assumption [Sar03]. Equation (4) yields the received interference PSD for a 100% active, single interferer:,s = p Tx G Rx (,)/L p.n = p Tx A W,Rx (,) 4 2 R n. (4) Applying (4) to the model geometry in Fig. 1 the aggregate interference PSD at the victim receiver is the two-dimensional integral of all incident interference power: = p Tx,av R A W,Rx (,) 0 cos( ) 2 tan( ) dd (5) cos 2 ( ) = min = 0 n HRx The minimum elevation angle min is determined by: tan( min ) = D / H RX. The integral (5) is solved numerically. The victim antenna height is set to = 20 m. This is the lowest height that allows a 203

4 clear first order Fresnel zone for an assumed link distance of 20 km over plane ground. This value is very conservative regarding the interference calculation and most often larger are encountered in the real world. However this value is taken as a reference for comparison with other studies like [ECC04]. The UWB devices emit an EIRPSD of p Tx = dbm/mhz. For all graphical results given the activity factor is a = 100% and the interferer density is set to 2D = 1000/km 2 for easy comparison with other studies [ECC04, Hir04]. The path loss exponent n is a crucial parameter. It strongly depends on the scenario (urban, sub-urban and rural). There exist many measurements for mobile communication systems in the frequency range of 1 2 GHz [Sar04] yielding n especially dependent on the height. There are also reports available for the frequency range of 5 6 GHz e.g. [Xio02] yielding n = 3.3 for line of sight (LOS) conditions in rural areas. The coexistence studies given in [ECC04] apply n = 2, assuming free space propagation conditions with a constant correction of -5 db for shadowing effects. This does not take into account that the shadowing effects increase with distance. For increasing path loss exponents (which are especially applicable for urban scenarios with high interferer density) drastic differences to the pessimistic calculation (n = 2) occur. Assuming a single slope model with n = 3.3, = 10 m and D = 15 km the difference is >28 db. Furthermore the integral (5) diverges with n = 2 for D. This is not consistent with the observation that the interference power at the victim antenna is relatively low despite of all the spurious emissions from electronic devices around it. However the basic log distance path loss model with a constant n may be inaccurate for small D. Therefore it is used mainly as reference for the comparison with other models (see Fig. 8). 3.2 Dual Slope Path Loss Model In order to account for the LOS conditions in the direct vicinity of the victim antenna the dual slope model [Sar03] introduces free space propagation conditions (n 1 = 2) for distances 0<R< and for R> the larger path loss exponent n 2 is used: n = nr ( )= 2 R. (6) n 2 R > For sub-urban scenarios the path loss exponent n 2 = 3.3 is suggested for R based on the literature [Sar03, Xio02]. For the calculations the intercept point is set to = 100 m in accordance with other coexistence studies like [ERC98]. With these assumptions ( = 100 m, = 20 m, n 2 =3.3, 3.7 m dish antenna, pattern acc. to ITU-R F.699) the aggregate interference increases compared to the single slope model with n = 3.3 and = 10 m. However the difference between the pure free space propagation and the sub-urban dual-slope model is still 15 db for D = 15 km. The aggregate interference scales linearly with a 2D. Therefore assuming a typical UWB device activity factor of a = 1 % [Hir04] and a conservatively estimated suburban device density 2D = 1000/km 2 [ECC04] the reference values are lowered by 20 db. The UWB devices are deployed mainly indoor. The indoor emissions are attenuated by the wall insertion loss (typically db). Depending on the percentage of indoor devices this yields an additional over all indoor loss of 7 10 db. Due to its moderate complexity the dual slope model is very well suited for investigating the sensitivity of the aggregate interference regarding the parameters,, n 2, and different antenna patterns. The choice of is relevant regarding the estimation of the aggregate interference emitted in the neighborhood of the base station. This determines the bias of the different curves in Fig. 3. The difference between = 100 m and = 200 m is 3.3 db additional interference for n 2 = 3.3. The intercept distance is strongly correlated to the distance to the neighboring buildings and it is therefore assumed that = 100 m is a good and conservative choice for typical urban and sub-urban environments. 204

5 = 50m = 100m = 150m = 200m Fig. 3: Aggregate interference power spectral density for the at the victim receiver for a variation of the intercept distance (fixed parameters: 3.7 m dish antenna and radiation pattern acc. ITU-R F.699, dual slope model with n 2 = 3.3, = 20 m, 2D = 1000 UWB devices/km 2 and activity factor a = 100 %) = 20m = 40m = 60m = 80m = 100m Fig. 4: Aggregate interference power spectral density for the at the victim receiver for a variation of the victim s height (fixed parameters: 3.7 m dish antenna and radiation pattern acc. ITU-R F.699, dual slope model with n 2 = 3.3, = 100 m, 2D = 1000 UWB devices/km 2 and activity factor a = 100 %). 205

6 n = n = n = n = n = n 2 = Fig. 5: Aggregate interference power spectral density for the at the victim receiver for a variation of the path loss exponent n 2 (fixed parameters: 3.7 m dish antenna and radiation pattern acc. ITU-R F.699, dual slope model with = 20 m, = 100 m, 2D = 1000 UWB devices/km 2 and activity factor a = 100 %) isotropic ITU R F.699 ITU R F Fig. 6: Aggregate interference power spectral density for the at the victim receiver for a variation of the victim s antenna pattern (fixed parameters: dual slope model with n 2 = 3.3, = 100 m, 2D = 1000 UWB devices/km 2 and activity factor a = 100 %). 206

7 n = 2.0 n = 2.5 n = 3.0 n = 3.3 n = 3.5 n = interferer distance D in km Fig. 7: Interference power spectral density,s at the victim receiver generated by single interferer moving along the ground in main beam direction of the dish (fixed parameters: 3.7 m dish antenna and radiation pattern acc. ITU-R F.699, = 20 m, = 100m). The height influences the distance at which the antenna mainlobe hit the ground. Additionally the interference power collected by the victim s antenna sidelobes is reduced due to the larger distance. As can be seen from Fig. 4 an extension of the victim antenna height from = 20 m to = 100m yields a reduction of interference power by 8 db for the applied dual slope model with n 2 =3.3. It is expected that for free space propagation the reduction is somewhat stronger, because the interference contributions from the main lobe are weighted higher. This is in accordance with a reduction of 12 db for n 2 = 2 reported in [Hir04]. As mentioned above the path loss exponent n 2 is the crucial parameter in (6), accounting for shadowing effects that grow stronger with the distance. As can be seen from Fig. 5 the gradient of the aggregate interference at a fixed area radius D is decreasing with n 2. Therefore the difference of the graphs between n 2 = 3 and n 2 = 4 is 5.2 db for D = 15 km whereas the difference between n 2 = 2 and n 2 = 3 is 13 db. In Fig. 6 the different antenna patterns are compared (n 2 = 3.3, = 20 m): the ITU-R F1245 pattern with the lower average side lobe level yields 2.1 db less aggregate interference power (for D =15 km). The values are compared with an ideal isotropic receiving antenna with antenna gain 0 dbi. It is notable that the isotropic antenna collects roughly the same interference power like the highly directive ones. This is mainly due to the nearly isotropic sidelobe level, while the interference is incident from all azimuth angles. For this reason the tendencies of the presented results can be carefully applied also to other services like fixed wireless access (FWA) with less directive antennas. Up to now only the aggregate interference case has been investigated. The single interferer may be located within the footprint of the victim s antenna mainlobe, which could have a more disturbing effect then the spatially distributed two-dimensional EIRPSD. Therefore the interference power due to a single interferer moving along the ground for the direction = 0 is calculated for different n 2 ( = 20 m). The results are presented in Fig. 7, which utilizes a logarithmic scale for the interferer distance D. For free space propagation with n 2 = 2 the maximum received interference occurs not in the direct vicinity of the victim but for a distance of D = 1.5 km. However with increasing shadowing 207

8 (increasing n) the local maximum is reached at the intercept point of the model, which becomes clearly visible. 3.3 Hata Path Loss Model The phenomenon that the attenuation due to shadowing increases with distance is well acknowledged in conservative coexistence studies like [ERC98], which utilizes a Hata model (ITU-R P.529-2) for urban wave propagation. This model cannot be applied directly for the present case at 4 GHz, which is the frequency of the regarded victim since it is valid only for frequencies up to 2 GHz and base station heights down to 30 m. However it is generally agreed that the path loss increases with frequency. Therefore a conservative estimate can be calculated from the wave propagation model applied in [ERC98] conservatively evaluating all frequency dependent empiric terms for 2 GHz (while leaving the free space attenuation at 4 GHz). If base station heights < 30 m were applied this would result in lower path loss. Therefore the basestation height is kept at = 30 m. The height of the UWB devices is set to H Tx =1.5 m. This results in the following adaptation of the Hata model given in [ERC98]: L P,Hata (R) db 4000MHz R log log 10 1 MHz 1km R = log 10 1km 2000MHz 30m log log 10 1MHz 1m log 30m R 10 log 1m 10 1km free space for R < 0.1 km interp. for 0.1 km < R <1 km (7) Hata model for R > 1 km As can be seen from Fig. 8 the application of the Hata model yields an aggregate interference that remains constant for D > 1 km. This emphasizes the fact that in dense urban environments with low base station heights especially the conditions in the vicinity of the base station determine the aggregate interference. The difference to the free space propagation yields 22 db. Together with an activity factor of a = 1 % and a typical metropolitan population density for 2D = 3000/km 2 [Hir04] this is further lowered by 15 db. Again the model does not account for the indoor to outdoor wall insertion loss HATA DS n=3.3 SS n=2 SS n= Fig. 8: Numerical results for the aggregate interference PSD at the generic victim receiver with 3.7 m dish antenna and radiation pattern acc. ITU-R F.699 for 2D = 1000 UWB devices/km 2 and a = 100 %: free space propagation (dotted line, SS n=2), dual slope model with n 1 = 2, n 2 = 3.3, = 100 m (dashed line, DS n = 3.3), adapted Hata model (solid line), single slope model with n = 3.3 and = 10 m (dash-dot-line, SS n=3.3). 208

9 4 Conclusions The results presented in this paper emphasize the importance of conservative but realistic and fair wave propagation modeling for the investigation of the coexistence of ultra wideband devices with the narrowband services in the same frequency band. Different propagation models are investigated. Based on the literature it can be stated that pure free space propagation is a very unlikely phenomenon for urban and sub-urban scenarios for large areas. Many measurement campaigns provide the knowledge that the shadowing effects are increasing with the distance. The sensitivity of the dual slope log-distance path loss model has been investigated regarding the model parameters intercept distance, victim antenna height, antenna pattern G Rx (,) and path loss exponent n. It has been shown that the most sensitive parameter of the model is the path loss exponent n followed by the antenna height, and the intercept distance, which closely correlates to the distance of the surrounding buildings. A good choice for the intercept distance is = 100 m. The antenna pattern plays only a minor role; even an isotropic antenna pattern yields similar interference powers as highly directive victim antennas. For suburban scenarios a dual slope log-distance path loss model with path loss exponent n 2 = 3.3 is considered to be appropriate. It yields an interference power, which is 15 db lower than for the free space propagation model. Together with additional losses (wall attenuation, db applicable for about 80% of the devices) and considerations about the appropriate device density 2D according the population density [Hir04] the aggregate interference power of the UWB devices (emitting with dbm/mhz acc. [FCC02]) is most unlikely to exceed the protection criteria of -127 dbm/mhz [ECC04] for a point-to-point communication link at 4 GHz. This holds also for the dense urban environment where the yet conservative Hata model is applicable, which yields a 22 db lower aggregate interference power at the victim receiver than the free space propagation model. References [FCC02] Federal Communications Commission (FCC), Revision of Part 15 of the Commission s Rules Regarding Ultra Wideband Transmission Systems, First Report and Order, ET Docket , FCC 02-48, April 2002 [Mitt04] M. Mittelbach, C. Muller, D. Ferger, A. Finger, Study of coexistence between UWB and narrowband cellular systems, Ultra Wideband Systems 2004, Joint with Conference on Ultrawideband Systems and Technologies, Joint UWBST & IWUWBS International Workshop on, pp 40-44, May 2004 [ECC04] Electronic Communications Committee (ECC) Draft ECC Report on The Protection Requirements of Radiocommunication Systems Below 10.6 GHz from generic UWB Applications, Draft ECC Report 64, November 2004, Download: [ERC98] European Radiocommunications Committee (ERC), Compatibility Analysis Regarding Possible Sharing Between the UIC System and Radio Microphones in the Frequency Ranges MHz and MHz, ERC Report 62, Siofok, May 1998 [Sar03] T.K. Sarkar, J. Zhong, K. Kim, A. Medouri, M. Salazar-Palma, A survey of various propagation models for mobile communication, Antennas and Propagation Magazine, IEEE, vol. 45, no. 3, pp , June 2003 [Xio02] Z. Xiongwen, J. Kivinen, P. Vainikainen, K. Skog, Propagation characteristics for wideband outdoor mobile communications at 5.3 GHz, IEEE Journal on Selected Areas in Communications, vol. 20, no. 3, pp , April 2002 [Hir04] W. Hirt, C. Politano, PULSERS Comments, Comments by Members of the European Integrated Project PULSERS to Draft ECC Report 64, Dec (internal communication) 209