Use of CLARIFY ® for RF Coverage Analysis and Propagation Model Optimization in GSM Networks

Abstract Propagation model tuning is a fundamental part of everyday GSM cellular engineering practice. The model tuning is usually accomplished through elaborate and costly tests based on CW measurements. This paper evaluates alternatives to CW testing where measurements are collected using traditional GSM scanners and PCTEL’s CLARIFY Interference Management System. The results of the analysis reveal that CLARIFY receiver provides a viable alternative for CW tests in many practical situations., Traditional GSM scanners are affected by the co-channel and adjacent channel interference and therefore their use should be limited to cases of relatively low frequency reuse.
A PCTEL Technical Paper

Adapted For Distribution - CL092009

the measurements are collected using existing commercial cellular network operating in the 850 MHz frequency band. The measured data is then used to determine the parameters for an optimized RF propagation model for a given morphological classification. Drive test methodology The RSL measurements are taken in two typical GSM network environments: suburban environment and dense urban environment. To achieve a high quality result for the modle tuning effort. That usually leads to drive testing of one test site at the time and model tuning for even the smallest cellular market may take days to accomplish. Additionally. one realizes that the RF propagation for the majority of the cells in the market is not tested in the process. one serving sector is selected for the study. it is of utmost importance that engineers have access to an accurate set of RF models. For the rest of the sites. 2. the level of improvement for the entire market is difficult to assess. the CW channel is transmitted without any co-channel or adjacent channel interference. Both environments are characterized with a relatively flat terrain. It is easy to see that the process of model tuning that is based on extensive CW testing is cumbersome and costly. one may get PCTEL Technical Paper – CL092009 Page 1 of 7 . Section 2 describes the experimental procedure used for data collection. The outline of the paper is provided as follows. To allow comparisons between the instruments a CW transmitter is set up on the selected sector. The reuse in the urban area network is N=30. even though the accuracy of the models is generally improved. The network in the suburban area uses frequency plan with the reuse of N=15 on the BCCH layer. including drive-test methodology. GSM base station and equipment setup. referred to in the industry as scanning receivers. The frequency selected for the CW transmitter is within the network guard band and therefore. it is critical that empirical path loss measurements are performance with high precision. Network planning. the parameters of the optimized models are applied across the board to all the cells in accordance with their morphology classification. Typically dedicated radios. both the GSM scanner and the CLARIFY receiver measure Broadcast Control Channel (BCCH) of the selected sector under the live network conditions. use of traditional GSM scanners or use of CLARIFY® high dynamic range receivers [1]. despite all of the efforts. Instead. In each environment. Therefore. are needed for optimal results. all alternative systems allow data collection on a live network without any special equipment set-up requirements. Furthermore. it is unused throughout the network. Introduction In the operation and maintenance of GSM networks. a regular reuse of the BCCH channels is not maintained. Over the past few years several alternatives to CW testing became possible. Section 3 presents the analysis of the obtained measured data. capital investment planning or automated cell planning processes depend heavily on the outputs of the RF propagation modeling tools. high sensitivity field equipment. they allow simultaneous measurements of all cells without any disruption in the system’s normal operation. Measurement procedure The data presented in this paper were collected using commercial measurement equipment and using processes generally embraced in the standard engineering practice. radio signal RF propagation modeling tools are widely used to accomplish many significant RF engineering tasks. The selected sectors are presented in light blue color. frequency planning. the accuracy depends on the similarity of the site’s RF propagation environment to one of the representative morphologies selected for the study. The details of the measurement procedure are provided as follows. the measurements are recorded simultaneously by each measurement device. Unlike the CW receiver. 2. In order to meet Lee sampling criteria [3]. Also. In theory. Strictly speaking. phone based systems do not offer a viable and cost effective alternative to CW testing. Each site under the test needs to be set up separately and the frequency plan needs to be modified to accommodate the CW test frequency. In common engineering practice. the maximum vehicle speed is maintained to accommodate the slowest RSL collecting device. The cellular market can comprise much such morphology. During the drive test. A rigorous analysis reported in [2] shows that the RSL measurements obtained by the phone based devices are not sufficiently accurate and repeatable. a group of test sites is selected to represent the morphology within a given cellular market. For each of the selected sites a Continuous Wave (CW) transmitter is mounted and detailed path loss measurements are performed. For that reason. for RF propagation modeling purposes. One may consider use of phone based measurement devices. each comprised of a distinct subset of test sites. The signals are sampled in radial and crossing routes within the beam-width of the transmitting antenna.1. The drive test routes for the suburban and urban areas are illustrated as red traces in Figs. As such. 1-2. the accuracy of the RF propagation models is achieved through careful integration of path loss measurements. The goal of this paper is to evaluate if CW based measurements can be replaced by measurements obtained using traditional GSM scanners or CLARIFY receivers. one can only guarantee that the accuracy is achieved for the sites that are in the selected test site group. The path loss measurements are collected using a process called model tuning. From the figures. Both networks deploy “ad-hoc” frequency plan and therefore. optimization. most RF propagation models are determined on the basis of qualitative assessment of cells’ RF propagation morphology. The routes are selected to capture the full dynamic range of signal strength of the transmitter. In this process.1. As a result. while observations and conclusions are then outlined in Section 4. The drive test routes are chosen in a manner with good engineering practices associated with RF propagation model tuning recommendations. In the end.

Drive-test equipment set-up 3. TABLE I.36 42. 3. As seen. The CW receiver has a 30 kHz band with a sensitivity of -122 dBm. as well as decode BSIC (Base Station Identification Code) if the C/I (Carrier to Interference ratio) value of the surveyed BCCH channel is greater than –2 dB. Parameter Tx centerline (ft) BASE STATIONS SET UP Suburban sector Urban sector 278 120 BCCH CW BCCH CW ARFCN 145 180(*) 144 170(*) EiRP (dB) 42. FL. Urban drive test area (Orlando. Selected sector setup At the selected sectors. Data analysis The primary goal of the data analysis is to establish the level of difference between CW. Suburban drive test area (Melbourne. while the urban sector is mounted on a side of a tall building in a city core area. if the C/I value is above -18 dB. maintaining similar path loss conditions. Further details of the setup are given in TABLE I. USA) Figure 3. The system is equipped with external GPS and RF antennas with the same characteristics.36 40. Figure 2. One may notice that there is a difference in the EiRP values between CW and BCCH signals of about 2dB in favor of the last one. scanner and CLARIFY PCTEL Technical Paper – CL092009 Page 2 of 7 . the drive test system contains a CW receiver.96 (*) Channels 170 and 180 are in the guard bands of the two systems 2. Due to the sophisticated signal processing techniques. FL. Equipment setup The equipment setup used for data collection is presented in Fig.better idea on the cell density within the two environments.2.96 40. USA) 2. Figure 1. a GSM scanner. and the CLARIFY high dynamic range receiver. Typical separations between sites in suburban area are about 35 miles. Every drive measurement system contains a laptop with appropriate measurement software for automatic data collection and location data association. the CW transmission shares the same antenna system used for the GSM cell. CLARIFY can measure and associate RSL signal to a specific sector. The scanner can measure and report the RSL. The suburban sector is on a self standing cell tower. while in urban area the distances are reduced below one mile. This difference is taken into consideration in the post-processing of the data and in the path loss calculations.3.

a bin is 30 m by 30 m square. The MSA size values for both environments are summarized in TABLE II. The differences between the results for the two bin sizes are negligible and for the sake of brevity only 30 m results are reported. of a great interest are the sizes of the area over which the instruments collects data. Some common bin sizes are 30 m. TABLE II.measurements. Representative MSAs obtained from measured data for the three devices are presented in Figs. or the overall statistical behavior of the data.. As seen. Such a tight reuse of frequencies decreases C/I which in turn directly impacts the capability of both GSM tools in taking the path loss measurements. of common bins with CW MSA in bins MSA relative to CW No. That indicates that in cases of a low to moderate frequency reuse. The comparisons may be made on two principle levels. within this relatively small area there are four cochannel BCCH reuses and there are seven adjacent BCCH assignments.57% 1802 CW scanner 8133 n/a n/a 4534 n/a n/a Measurable sector area (MSA) is defined as the size of the area in which the signal from the selected sector can be measured using a given tool. MSA can be expressed in either number of bins with measurements. This is good news since the areas of low to moderate reuse are typically in rural and suburban environments where the cells are of larger sizes. and N i is the total number of samples in i -th bin. the level of the data scattering about general trends. where RSLi is the averaged RSL in i -th bin. while CLARIFY MSA is shown with a green trace. the binning process is performed in accordance with RSLi = 1 Ni ∑ RSL j =1 Ni j (1) assignments are highlighted in Fig 5. 4-5. one may compare measurements themselves and determine how data collection results differ. On the first level. For example. it can be seen that a very good match exists between CW measurements and the measurements obtained through CLARIFY. In this study. The cells with co-channel and adjacent channel BCCH PCTEL Technical Paper – CL092009 Urban 3. Drive Device Parameter MSA in bins MSA relative to CW No. capability of taking measurements for multiple cells at the same time results in large cost savings. the scanner’s MSA is illustrated with a blue trace. Co-channel and adjacent reuse sectors are highlighted in red and yellow respectively. One should remember that in this paper. This is a standard process used in practice and it refers to spatial averaging of the individual RSL measurements over a small geographical area called bin. the analysis is done with bin sizes of 30 m and 50 m. the CW’s MSA is presented with a red trace. In these environments. On the other hand. from the standpoint of MSA size.68% 5032 1031 22. For example. one may use the data to optimize RF propagation models and then compare how close the parameters of the resulting models are. The binning process tends to eliminate impact of the fast fading. For the traditional GSM scanner the overlap is significantly smaller. all collected measurements are binned. The reason is predominantly due to the very tight frequency reuse deployed in the urban area under the test.71% 7703 2021 44. so the conversion between the bin count an the area size is straightforwrad. An assumption is made that the CW measurements are de facto benchmark and the analysis compares the measurements of the alternative devices against the ones obtained using CW. RSL j is the RSL of the signal expressed in dBm found in i -th bin. the comparison may be made between the outcomes that result from the data application.1. or the physical size of the area with measurements. both the scanner and CLARIFY exhibit significantly smaller MSA relative to CW. As seen. of common bins with CW MSA RESULTS GSM scanner 5342 65. Before analyses. the GSM scanner may be used only in parts of the network with a low frequency reuse. The table also contains the total number of bins common to both CW and the compared GSM tool. Measurable Sector Area (MSA) Suburban Page 3 of 7 . The size of the bin is usually determined by the terrain resolution used in the RF propagation modeling tool.74% 1031 CLARIFY receiver 7703 94. In the Figs. the MSA of CLARIFY receiver seems to be quite comparable to that of CW in the case of suburban environment. Therefore. In the case of urban environment. In the case of suburban environment. On the second level. CLARIFY receiver is a viable substitute for the CW measurement set. In this study. the limited dynamic range of the GSM scanner results in significant reduction of the MSA. existence of bias between the instruments. 50 m and 100m.

in whose case the result manifests as missing sample points or incorrect readings. These degradations may have considerable impact on the model tuning process during the RF propagation modeling. there are more sample points and also more sample with correct readings. However. The middle (in both horizontal and vertical layout) traces illustrate the CLARIFY receiver’s path loss data. the variations of PL values are in range of 2 dB and 4 dB for the CLARIFY receiver and the scanner respectively. Drive PATH LOSS RESULTS GSM scanner -0. Figure 5. in the vicinity of co-channel and adjacent channel Figure 6. The impact is more evident in the case of traditional scanners. However. For the suburban drive. However. owing to its high dynamic range receiver and its higher tolerance to co-channel and adjacent channel interference.47 4. the mean values of the differences between scanner and CW as well as the difference between the CLARIFY receiver and CW are both very close to zero.12 CLARIFY receiver 0. dev. The RSL measurements and the EiRP values from TABLE I are used to determine the path losses. PL measurements of the scanner are significantly degraded due to the impact of co-channel and adjacent channel interferers. In the case of CLARIFY receiver. the performance of both the traditional scanner and CLARIFY receiver is affected.26 Device Parameter Mean (cw – device) [dB] St. contains the principal results of the path loss (PL) statistical analysis. (cw – device) [dB] PCTEL Technical Paper – CL092009 Suburban Page 4 of 7 . For each common bin. TABLE III. Path loss analysis A representative portion of the RSL measurements using the three tools is presented in Figure 6. The CW RSL measurements are presented with the lowest (in horizontal layout) and left (in vertical layout) traces. From Fig. An example of the path loss measurements using different tools TABLE III. almost zero PL difference in case of the CLARIFY receiver is preserved. MSAs for the urban area for the sector in blue 3. The scanner path loss measurements are presented with the top trace in horizontal layout) and right (in vertical layout) traces.33 2.interferers. For the urban environment. 6.2. The traces are offset for easier representation. the pair wise differences between the path loss measurements are obtain using ∆ CW −tool = PathLoss CW [dB] − PathLoss tool [dB] (2) Figure 4. one may readily observe high agreement between RSLs collected by the CW and CLARIFY receiver. MSAs for the suburban area for the sector in blue where tool can be either the scanner or the CLARIFY receiver. This implies that all three tools express very similar PL calculations on average.

for different confidence levels and number of samples [4].7 CDF developed with CW PL data CDF developed with scanner PL data F(x) Figure 8. It is important to notice that the statistic does not depend on the form of F (x ) . is found using Figure 9. a statistic.83 D(n ) = max X FCW (x ) − Ftool (x ) (3) Figure 7. PDF/CDF for PL difference between the CW and the CLARIFY receiver for the suburban drive 0. On the other hand.9 0. The test is structured as follows.3. Empirical CDF 1 0. the Kolmogorov-Smirnov test (K-S test) of goodness-of-fit is used.5 0. which is used as the benchmark.Urban Mean (cw – device) [dB] St. PDF/CDF for PL differences between the CW and the GSM scanner for the suburban drive where FCW (x ) represents CDF developed using the CW data. The null hypothesis for the K-S test is that both data sets are drawn from the same continuous distribution. the empirical cdf functions are constructed. the sampling distribution of D(n ) is presented in tables.4 0. and Ftool (x ) corresponds to CDF found for either scanner or CLARIFY receiver data. The null hypothesis is accepted for GSM scanner data only in the suburban environment.42 3. the PL differences seem to exhibit largely a lognormal character with means and standard deviations as reported in TABLE III.6 0. In the first step.3 0.2 0. 80 90 100 110 PL [dB] 120 130 140 3. In this report. the almost overlapping CDF functions are constructed using the CW and the CLARIFY receiver data. as well as cumulative distribution functions are presented in Figs. An example of empirical CDF functions for the CW receiver and the GSM scanner PCTEL Technical Paper – CL092009 Page 5 of 7 . In other words. 9-10. 7-8. Usually. 9.73 -0.8 0. The K-S test is a good test to identify which tool. denoted by D(n ) .1 0 70 Typical normalized histograms of the differences between the PL measurements. D(n ) equals to the largest absolute deviation between two functions when all values of x are considered. Formally. provides PL measurements comparable to CW tool. From the shape of the curves. From Fig. one can easily observe a significant difference between CDF functions formed using the CW and the scanner’s data. In the second step. denoted by n . the hypothesis is accepted if the test is significant at the 95% level. Typical CDF functions obtained from the three tools are illustrated in Figs. The statistic D(n ) can be computed with respect to different confidence intervals. (cw – device) [dB] -5. but only on the sample size. dev. if any. The alternative hypothesis is that they are drawn from different continuous distributions. the K-S test passes the null hypothesis for the CLARIFY receiver in both suburban and urban environments. 7-8). Empirical CDF comparison (K-S test) In order to compare the CDF functions (as presented on Fig.69 3.

the application of the CLARIFY receiver data in model tuning leads to models that are very close to the ones developed on the basis of the CW data collection. • The average difference between the path loss measurements between the CW tool and the CLARIFY receiver is negligible (< 1 dB).5 0. A side-by-side comparison of the measurements collected by the three device types was performed and the findings may be summarized as follows. Dev. On the other hand.1 0 6.8 0. Model tuning The path loss measurements collected by the three instruments are used to perform the model tuning for the selected sector. despite differences in MSAs the models developed using CW and CLARIFY data are almost identical in both suburban and urban environments.4 -38.1 GSM scanner -65.4 0. • The MSA of CLARIFY receiver is quite close to the MSA of the CW receiver in cases of low to moderate frequency reuse (N > 15).7 0. (measured and predicted) [dBm] PMO RESULTS CW scanner -64. (measured and predicted) [dBm] Optimized intercept [dBm] Optimized slope [dB/decade] Mean (measured and predicted) [dBm] St. The optimistic model is a result of inability of regular GSM scanners to deal with the co-channel and adjacent channel interference.1 CLARIFY receiver -64.2 -37. • The CLARIFY receiver data leads to virtually identical RF propagation models as the ones developed using the CW measurements. • If the GSM scanner data is used for model tuning.2 -61. the resulting models are the over-predicting path loss.8 -40.Empirical CDF 1 0.4 The models obtained using the GSM scanner data seems to be quite optimistic when compared to the CW based model. the optimistic nature of the GSM scanner based model may result in a considerably higher one mile intercept or in a considerably lower slope.1 0 80 CDF developed with CW PL data CDF developed with Receiver PL data In a given scenario. This is especially the case in networks with low to moderate frequency reuse factor (N > 15). An example of empirical CDF functions for the CW receiver and the CLARIFY receiver 3.6 F(x) 0. The results of the study reported in this paper indicate that the CLARIFY receivers with high dynamic range (C/I >-18 dB). • As per Kolmogorov-Smirnov test of goodness-of-fit. 4.8 0 7.3 0. the MSA of the CLARIFY receiver is reduced. The bias depends on the frequency plan and resulting amount of co-channel and adjacent channel interference.6 -38.5 0 6.4 -35.6 -66. represent a viable practical alternative to CW testing. Therefore. GSM scanner exhibits good fit with the CW test only in a light frequency reuse environment. • Due to frequency reuse interference. A simple Lee macroscopic RF propagation model [3] is selected and the measurements are used to determine the optimum values for the slope and the intercept parameters of the model. Drive Device Parameter Optimized intercept [dBm] Optimized slope [dB/decade] Mean (measured and predicted) [dBm] St.4 -40.5 0 6. Dev. the statistics of CW data are in very good agreement with the CLARIFY receiver data for both environments.9 0 6. the GSM scanner shows a noticeable bias towards underestimating the path loss. TABLE IV.5 dB of each other in both cases. Observations and conclusions This paper considers the feasibility of using GSM scanners and CLARIFY receivers as substitutes for CW-based test systems. Even in the areas of high frequency reuse.2 0. The results of the model tuning are presented in TABLE IV. In urban areas of high frequency reuse. 90 100 110 PL [dB] 120 130 140 Figure 10. • • The MSA of the GSM scanner is affected by the frequency reuse and is considerably smaller than the MSA for the CW receiver.9 0.4.8 0 5. In contrast. the estimates of the RF propagation model parameters obtained from the CLARIFY receiver’s data seem to be quite close to the ones obtained from the CW measurements. the measurements obtained by regular GSM scanners seem to be quite sensitive to the cochannel and adjacent channel interference and they may approximate CW measurements only in limited scenarios when the frequency reuse is low. As seen. it seems that even though the size of the MSA is affected by the frequency plan.5 -66. PCTEL Technical Paper – CL092009 Urban Suburban Page 6 of 7 . The slope and intercept values are within 0.

Ivica Kostanic (Florida Institute of Technology.” in proceedings of ISWPC 2007. USA) Greg Evans (at&t wireless. 2 Edition. Mijatovic. 3rd Ed. January 10-13. Dickey. Kostanic. Inc. June 22-25 2008. Neter. Whitmore. Wasserman. Data analysis was performed with data post processing platform Gladiator from QualiTest Technologies. Allyn and Bacon. “Repeatability of Received Signal Level Measurements in GSM Cellular Networks. References N.C. from PCTEL. Lee.Y. Melbourne. Authors Nenad Mijatovic. I. [2] I. Inc. Originally presented at EW2008. 2008. Orlando. McGrawHill.] PCTEL Technical Paper – CL092009 Page 7 of 7 .” in proceedings of CCNC 2008. S. Wireless and Cellular Communication. ©2008. [1] 7. G. A. Kostanic. Acknowledgments Authors would like to express a sincere appreciation to Mr. W. MD. USA) [Adapted From “Use of Scanning Receivers for RF Coverage Analysis and RF propagation Model Optimization in GSM Networks” Mijatovic. FL. as well as PCTEL Inc. 1982. Applied Statistics. 2005. Mijatovic. Czech Republic. Boston. FL. Germantown. N. nd [4] J. 6. The authors are grateful to ATT Wireless for allowing use of their network. Kostanic & Evans. “Comparison of Receive Signal Level Measurement Techniques in GSM Cellular Networks. San Juan. RF Solutions Group. Puerto Rico (2007) [3] W. Prague.5. and Envision Wireless for providing exceptional support and tools. Dale Bass.