Transactions on the Built Environment vol 34, 1998 WIT Press, ISSN

Transcription

1 Experimental validation of propagation models for radiocommunications applications in industrial environments M. V. Castro, A. Seoane P., F. P. Fontan, J. Pereda Dpt. of Communications Technologies. University of Vigo Campus Universitario. E Vigo. Spain uvigo. es Abstract In this article, an overview is presented of the results obtained within a Project developed at Ansaldo (Naples, Italy) in the framework of the European Union's Human Capital and Mobility-Access to Large Installations Programme (E.U. Contract No. CHGE-CT European Commission-DGXII). The project aimed to characterise the radio propagation channel in an industrial environment. The way to do that consisted fundamentally of two phases: gathering a comprehensive set of measurements and a propagation modelling review. A deep knowledge of the limitations caused for an industrial environment will facilitate the setting up of the so convenient radio communications systems within industrial areas such as analogue trunked systems according to the MPT 1327 (400 MHz) standard, cordless systems according to the DECT (1800 MHz) standard or digital trunked systems, especially designed for voice and data applications, based on TETRA (400 and 900 MHz) standard 1 Introduction Radio technologies are of special interest in industrial environments for personal communications and data transmission. Data and voice transmission allow the setting-up of several services, such as sensor and alarm monitoring, telecommand, SOS services, workers communications to Control Centres, etc. A deep knowledge of the impairments caused by factors such as metallic structures, impulse/ignition noise, type of

2 806 Computers in Railways environment (open, urban, rural,...), etc. to radiocommunication system performance is vital for the definition and setting-up of optimum implementations for any voice or data radio application. The presented work intended to characterise the radio propagation channel by gathering a comprehensive set of narrow-band measurements to support propagation modelling efforts. The work concentrates on data and voice communication. Typical industrial propagation scenarios were simulated by studying links between an elevated transmitter point and scattered measurement points along the streets between the different factory buildings at the ANSALDO facility in Naples. The results gathered in this article may be used for the planning of outdoors voice and data communication systems in factory like areas. The measurements were carried out at the 450 MHz frequency band allocated in European countries for commercial Private Mobile Radio (PMR) systems. One such system will be installed at the ANSALDO Trasporti (Naples) plant. The pan-european digital Trunked standard TETRA (Trans-European trunked Radio) was selected for installation. However, at the moment the TETRA standard equipment is not available and, thus, an alternative temporary system will be installed based on the analogue trunked standard MPT A comprehensive measurement campaign was carried out in order to assess the suitability of the most common propagation models available for the required application (industrial communications). Measurements were carried out at the MHz frequency band which is close to one of the bands that will be allocated to TETRA networks. Three models were extensively analysed: the Okumura-Hata, the multi-knifeedge diffraction and the COST 231 model. A detailed statistical study of the prediction errors was carried out. The obtained results may be readily used in the planning stages of future TETRA installations. Other measurements being carried out within the same framework try to investigate the coverage of factory buildings from transmitters located outdoors, indoor coverage and the assessment of the magnitude of impulse and ignition noise in factories and in the railroad environment. 2 Selection of test frequencies In order to characterise the propagation characteristics at the 450 MHz UHF band, a test frequency needed to be identified. This frequency should not be used by other systems in operation in the vicinity of the study area. Also, in order to reduce possible co-channel and adjacent-

3 Computers in Railways 807 channel interference a maximum transmit power of 1.5 W was used in transmissions carried out later during a measurement campaign described in Section 3. In order to find free frequencies in the vicinity of the test area long observation periods were required given the low usage rate at these PMR frequency bands. The scanning system described below was set-up to continuously record any activity in the bands of interest for several days in order to produce reliable estimations of the spectral usage in the area. The MHz band was split into 2 MHz sub-bands in order to facilitate spectrum utilisation recordings. A computerised spectrum scanning system based on the HP 8546-A EMI receiver was implemented. A general layout of the system is presented in Figure 1. Figure 2 presents measured instantaneous spectra for one of the subbands considered. Finally, a working sub-band between MHz and MHz was selected to carry out the test transmissions. Specifically, a MHz carrier frequency was chosen to perform the measurement campaign described later. Figure 1: General block diagram of the spectrum scanning system used Figure 2: Instantaneous measured spectra at one of the 2 MHz study sub-bands. 3 Measurement Campaign In order to carry out a propagation measurement campaign to validate the three proposed propagation models a HP 8546-A receiver was located on an elevated position with respect to the Ansaldo facility (Test Room Building). A half wave dipole was set on a 2 m mast and connected to the receiving equipment using a low loss coaxial cable (L^ < 2 db). A 6 db attenuator was used to reduce the transmitted power. A 5-35 w VHF/UHF transmitter (Yaesu FT-5200) was used. The transmitter was placed on a car and fed directly by the car battery. A quarter wave monopole was placed on the car roof (acting as the ground plane) by

4 808 Computers in Railways means of a magnetic base. A diagram representing this set-up is shown in Figure 3. A large number of measurements were taken throughout the Ansaldo facility streets. Local measurements were averaged in order to produce representative received power levels by removing the fast signal variations due to multipath fading effects. The measurements were associated with measurement points and recorded on maps of the test area for further study (Figures 4-5). Figure 3: Measurement equipment. Figure 4: Measurement points. H h Figure 5: Measurement Points. Figure 6: Histogram of measured values. 4 Propagation Models In this section the propagation models used for validation with experimental data are briefly described. Okumura-Hata model This model is of an empirical nature [1], [2]. It was developed after carrying out a large set of measurements in different mobile communications bands. The basic path loss formula corresponds to urban areas. However, a set of correction factors are available to adapt the

5 Computers in Railways 809 model to other environments such as suburban, rural, etc. The formulas provided by the model are the following Avrw, = <%. JJ> /ogf/; - 7J.&2 /ogfa, ; - of A J + f J J /og^ ;; /ogw where/ is the transmit frequency in MHz ht is the transmit antenna effective height in m with (30 < ht < 200 m) hm is the receive antenna effective height in m with (7 < h^ < 10 m) for medium-small cities the following expression for a(h J must be used af Aj = ^7.7 /ogf/; - 0.7; ^ - for large cities expression for a(h J is / For suburban areas L urban must be corrected in the following way For rural areas L urban must be corrected in the following way LR^I = Lu*an ~ ^8 (log(f)f log( f ) Multi-knife-edge diffraction The multiple knife-edge model [3] is used to describe blockage effects by buildings. Figure 7 represents a radio path undergoing multiple diffractions at building rooftops. In case of having a visibility path, free space losses are considered unless insufficient clearance is detected, in which case diffraction losses are also considered. A visibility path (LOS: line-of-sight) is illustrated in Figure 8. Observations reported throughout the mobile propagation literature (for example [3]) report greater losses than those of free space propagation for mobile environments. This is due to the use of omni-directional antennas located near the ground (1-1.5 m). These excess losses reported may be due to specular reflections on the ground partially cancelling the direct ray. Test results carried out taking into account ground reflections are also reported here. The expression of the total received electric field in the case of reflected ray existence is the following: Where p is the reflection coefficient and Ad is the direct ray-reflected ray path difference. Figures 9 and 10 represent Non-LOS and LOS paths with ground reflections. COST 231 model This model is also suited for urban area propagation predictions [4]. Figure 1 1 presents the geometrical aspects of the model as well as its

6 810 Computers in Railways Figure 7: Multiple knife-edge diffraction path. Figure 8: LOS path Figure 9: LOS with ground reflection Figure 10: Non-LOS path with ground reflection input parameters. The models has been specially developed for the 900 and 1800 MHz bands however, its applicability to 450 MHz predictions is assessed in this work. The following expressions are given in the model A, ~ Ly + J^rts + ^msd Lbf are the free space losses given by the expression Zy = 32.4 J og(/V + 20 fog ^ with/in MHz and d in km. L^ are the diffraction losses at the building in the street where the mobile is located. These losses follow the expression 4,, = /ogf W Zogf /; ogM/% ; + I_ if Lrts ^ 0 then L^s - 0 must be used. Lon follows the expression Lmsd are the losses corresponding to the diffraction effects caused the building roofs along the transmission trajectory. 4w = L*H +k«+kj log(d) + k, log(f) - 9 log(b) where L^ = -18 log (1 + Ah,), if Ahe < 0, L^ = 0

7 Computers in Railways d_ 0.5 /or /or zlag > 0 / > 0. j < 0. J for Ahg < f 1 If or small size cities '925 ^ /or large cities the validity range of the model is the following 800 < f< < he < 50 m, 1 < h < 3 m, 0.02 < d < 5 km MHz 5 Experimental Validation Results In this section measured and predicted received powers using the models briefly described in the previous section are compared and detailed error statistics are given. Prediction errors are defined following the expression that is, if errors are positive, this means the computed path losses are lower that the real ones and if negative that the computed path losses are greater than the real ones. Figures 12 and 13 present all measured received powers classified according to their path length for LOS and Non-LOS conditions. A total of 36 LOS and 1 10 Non-LOS paths were analysed. ^frrn n n n Figure 11: COST 231 model. Path geometry.

8 812 Computers in Railways Measured data (36 points') Measured data (I 10 points) Distance m Figure 12: Received Powers (LOS conditions) Distance m Figure 13: receiver Powers (non-los conditions) The Okumura-Hata model results were compared with measurements by defining three different path length ranges. The maximum path length was approximately 450 m. Measured paths were classified in one single distance range (from 0 to 450 m), two distance intervals (from 0 to 225 m and from 225 to 450 m) and three distance ranges (from 0 to 150 m, from 150 to 300 m and from 300 to 450 m. Table I summarises the observed error statistics. Both LOS and Non-LOS were considered in these comparisons. For the computation of the results presented in Table I the Hata expression for rural areas was used. Table I. Okumura-Hata model error statistics (143 LOS and Non-LOS paths considered) Conditions 1 dist. range 2 dist. range 3 dist. range mean (db) std (db) min. (db) max. (db) As it can be observed, large standard deviation values of the prediction errors have been observed. The Okumura-Hata model produces very general information on the average received power at a given distance. However, if more detailed, path-to-path information is required, other models must be used. The COST 231 (Walfish-Ikegami) model was designed for the 900 and 1800 MHz bands and for urban areas. Although, in principle, not suited for the case under study, the model was compared with measured values. In Table II a summary of the observed error statistics is given. As it can be observed the application of the model for the environment under consideration can not be recommended.

9 Computers in Railways 813 Table II. COST 231 model error statistics (143 LOS and Non-LOS paths considered) Conditions LOS and NLOS LOS paths Non-LOS paths mean (db) std (db) min.(db) max.(db) Finally, an analysis is presented of the performance of the multi-knifeedge diffraction model with and without ground reflections. In these calculations the actual path lengths of both the direct and reflected rays were used to compute the phase term in equation (Eq.l). Table III summarises the prediction error statistics for this model and for Non- LOS paths. Table III. Error Statistics for multi-knife-edge diffraction with and without ground reflected rays. Non-LOS paths. Conditions multiple diffrac. Diffr. and Refl. 2 db refl. loss mean (db) std (db) min. (db) max. (db) Table IV. Error Statistics for Free space losses with and without ground reflected rays. LOS paths. Conditions mean (db) std (db) min. (db) max. (db) Free space losses Direct + Reflected 7 db refl. losses As for the LOS paths, free space losses and eventual insufficient clearance diffraction losses were considered. In order to improve predictions a ground reflected ray was considered both taking into account direct and reflected ray path differences and a constant 2 db reflected ray loss. In Table IV a summary of model error for LOS paths is given. As it can be observed from the results just presented the most suitable propagation model of the ones studied in this work is the multi-knifeedge diffraction model with a fixed 7 db loses due to the reflected ray for Non-LOS paths. As for LOS paths the free space attenuated direct ray with a 2 db attenuated reflection ray model produces the best overall

10 814 Computers in Railways statistics. The fixed reflection loss values have been derived empirically and more detailed investigation is needed to adequately quantify reflected ray effects on the overall received signal power. 6.- CONCLUSIONS In this paper the results of a comprehensive propagation study carried out at the Ansaldo Trasporti facility in Naples are presented. The objective of the study presented here was to evaluate the prediction error statistics of different propagation models for the 450 MHz UHF band. The knowledge gathered with this work will allow the adequate coverage planning of communication systems in industrial environments. The studies presented were carried out using locally free channels in the PMR band ranging from MHz. A spectrum scanning system was developed in order to determine what test frequencies could be used without upsetting existing communication systems in the area. Further, in order not to interfere (co-channel and adjacent channel) with other radio systems in the vicinity the transmit power was limited to 1.5 w. References [1] Okumura et al. Field strength variability in VHF and UHF land mobile service Rev. Elec.Comm.Lab., Sep-Oct 1968 [2] M.Hata Empirical formula for propagation loss in land mobile radio services IEEE Trans. Veh.Tech., August 1980 [3] J.D.Parsons The mobile radio propagation channel Pentech Press, 1992 [4] Urban transmission loss models for mobile radio in the 900 and 1800 MHz bands COST 231 TD(90)119 Rev. 2. The Hague. The Netherlands, Sept. 1991

11 Section 12: Pantograph and Catenary