Comparison of Receive Signal Level Measurement Techniques in GSM Cellular Networks

Comparison of Receive Signal Level Measurement Techniques in GSM Cellular Networks Nenad Mijatovic *, Ivica Kostanic * and Sergey Dickey + * Florida Institute of Technology, Melbourne, FL, USA nmijatov@fit.edu, kostanic@fit.edu + PCTEL RF Solutions, Germantown, MD, USA sergey.dickey@pctel.com Abstract The Received Signal Level (RSL) is one of the fundamental measurements performed in cellular system operation and maintenance. Network RF optimization, coverage verification, and quality of service estimation rely on RSL measurements collected from the field using one of commercially available drive testing systems. This paper presents a comparative study of a series of RSL measurements in the GSM 1900MHz frequency band. The measurements were collected with three different drive test systems used in common engineering practice. Data analysis shows significant difference between the measurement systems. The differences were observed both in the size of the area from which the measurements may be collected as well as in the actual measured values. Computer simulation and theoretical analysis were used to verify the result of the measurements and extrapolate their validity to a set of more general cases. 1. Introduction In everyday practice, RF engineers use drive-test systems as primary tools for determining the performance of cellular networks. There are several types of commercial GSM drive test tools. They may be classified as the phone-based, regular scanning receivers, and high dynamic range scanning receivers. Each of these devices measures many network parameters. Among other things, all of them measure the RSL of the BCCH (Broadcast Control Channel) channels. However, they perform the measurement in a different manner, with different sampling frequencies, and with varying robustness with the respect to co-channel and adjacent channel interference [1]. Therefore, it is of a vital importance to understand the implication of these differences on the cellular engineering practice. The most important difference between the measurement devices is their dynamic range. The dynamic range is defined as the range of Carrier to Interference (C/I) values for which a device produces valid measurements. From a practical standpoint it is important to understand the impact of the dynamic range on the overall usability of the measurement system. In this paper, the impact is characterized and observed using two criteria. The first criterion is the size of the area from which the measurement may be collected in a typical GSM network implementing frequency reuse. The second criterion is the difference between the actual RSL readings among different measurement devices. The outline of the paper is as follows. Section 2 describes the measurement setup and drive-test methodology; Section 3 presents the data analyses. Finally, some conclusions are drawn in Section 4. 2. Measurement procedure The equipment set-up and drive-test methodology applied in the data collection are arranged to emulate a typical drive-test process used by RF field engineers. -based, scanner-based, and high dynamic range scanning receiver drive-test systems are placed in a standard SUV as presented in Fig. 1. 2.1. Equipment setup The phone-based system contains a measurement phone, an external GPS receiver, and a laptop with appropriate measurement software. The laptop is used to automatically log the phone measurements. Each measurement sample is associated with the time and position data obtained from the GPS receiver. In the setup, the measurement phone is attached to the passenger seat belt and it uses an integrated omnidirectional antenna with a gain of 0 dbi. During every drive, position of the phone tends to be the same. Legend Drive Test set-up Count Description Symbol 1 1 Scanner 1 Receiver 1 GPS 2 GPS antenna 2 RF antenna 3 Laptop computer -based drive test system Scanner-based drive test system Receiver-based drive test system Figure 1. Drive-test equipment set-up

For the phone, the list of the scanned BCCH channels is a function of the serving cell and its neighbor list. Specifically, the phone always measures and reports the BCCH RSL of the serving cell along with up to six strongest cells from its neighbor list. Therefore, throughout the drive tests, the list of BCCH channels scanned by the phone changes in a dynamic fashion. According to the phone specifications, if the C/I ratio of the BCCH channel is above 8dB, the phone-based system is capable of decoding BSIC (Base Station Identification Code). Using the BCCH/BSIC combination the RSL measurements may be easily associated with the serving sector or appropriate neighbor. The scanner-based system includes a scanner with an integrated GPS receiver, a data collection laptop, and external RF and GPS antennas. The scanner-based system uses more sophisticated signal processing techniques than the phone and it can measure and report the RSL if the C/I value of the surveyed BCCH channel is above 0 db. Above C/I of 0 db, the BSIC value is decoded as well. The scanner performs measurements for all present BCCH channels in a given area. The receiver-based system collects the data using a high dynamic scanning receiver with a built-in GPS receiver, a laptop with appropriate collecting software, and external RF and GPS antennas. The receiver-based system uses advanced signal processing techniques and it is capable of measuring RSL of the BCCH if the C/I value is above -18 db. Similar to the scanner, the receiver collects data on all BCCH channels as well. Both scanner s and receiver s RF and GPS antennas are mounted on the roof of the vehicle. The antennas used for the two devices are the same. Following test repeatability, antennas are placed on the same positions. Therefore, both antenna systems are identical, equalizing the total losses in the scanner and receiver signal paths. The scanner and receiver are set up to measure up to 15 BCCH channels in the 1900MHz-frequency band. 2.2. Drive test methodology The RSL measurements are taken in the Melbourne and Palm Bay areas of Brevard County, FL, USA over three consecutive days. The area under survey is in a suburban environment with a relatively flat terrain and covers approximately 255 sq. mi. The buildings in the area do not have more than two or three stories. All major streets in the area are selected for the network survey. The same routes are driven at approximately the same time of the day maintaining drive-test repeatability. During the drive test, the measurements are recorded simultaneously by each measurement device. The area is serviced by 33 GSM cells where each cell is implemented as a 120 degree sector. The survey is performed on a live operational network with a reuse of the BCCH frequencies that follows an ad hoc frequency plan having a cluster size of N = 15. Therefore, the scanner- and receiver-based systems constantly scan all available channels defined with the network configuration. 3. Data analysis The focus of the data post-processing is placed on the application side. Instead of analyses of how different instruments collect RSL measurements, the instruments are compared against two major criteria. The first criterion is the size of the area in which the measurement may be collected using one tool in a typical GSM network implementing frequency reuse - The Measurable Sector Area (MSA) is determined for each of the sectors. The second criterion is the difference between the actual RSL readings, which is defined for each sector as the difference between averaged RSL levels obtained using different measurement tools. 3.1. MSA analysis The MSA is observed using three types of analysis. The first one utilizes the RSL measurement data, and determines the MSA values from the drive-test itself. The second one simulates drive-test layout in the area and calculates predicted MSA values. The third one extrapolates the MSA calculations to scenarios of an ideal hexagonal network layout with different frequency reuse factors. RSL [dbm] -50-55 -60-65 -70-75 -80-85 -90-95 Scanner Receiver -100 400 405 410 415 420 425 430 435 440 445 450 Bin number Figure 2. RSL profile for different drive test tools Although, there are 33 sectors in the test area, only a group of 27 sectors is analyzed. The remaining five sectors are not part of the analysis due to relatively small number of measurements points for the phone-based system. These sectors are located on the edges of the

area and their antennas are pointed in the opposite directions from the roads selected for testing. Therefore, as a result, a total of 81 individual BCCH coverage plots are obtained from each measurement device across three measurement days. A representative portion of the averaged RSL measurements collected from the three tools is presented in Fig. 2. As expected, the measurements coming from the receiver and scanner show a very good agreement. The phone RSL readings are consistently lower due to the vehicle penetration loss. The trend shown in Fig. 2 is consistent across all data in the entire data set. 3.1.1. MSA analysis using measured RSL data There are four steps in determination of the MSA from the measured data. In the first step, the phone, scanner and receiver data are filtered by BCCH/BSIC combination. The filtering is performed to link each sector with measurements collected with three tools. In the second step, for every sector, for every measurement device and for every drive-test, data is binned. The bin size has a small impact on the overall calculation. The analyses performed for the bin sizes of 50 m and 100 m yield to a relatively close results for both the size of the MSA and the difference between the received signal level measurements. However, in our analysis, 50 m bin size is used. The average RSL values in every bin are calculated using logarithmic binning. In the third step, the MSA values for a given sector are calculated as a total number of bins measured with the tool. Finally, the MSAs between the tools are compared. Typical MSAs using the measured data for the three devices are presented in Fig. 3. The figure shown MSAs obtained from data collected on the first day for the beta sector (highlighted in light blue) of the cell site placed in the middle of the Fig. 3. On the Figure, the phone s MSA is presented with the green trace, the scanner s MSA is illustrated with the green and blue traces, while the receiver s MSA is shown with the green, blue, and red traces. Co-channel and adjacent channel interferers are presented as red and yellow sectors respectively. Analyzing MSA values for all sectors, the receiver exhibits the best performances compare to the phone and scanner systems. As it can be observed, the phone-based MSA is relatively small when compared to those obtained by the scanner and by the receiver. With respect to the receiver, scanner s MSAs is about 50-80% in size, while the phone s MSA is only at 20-40% level. For all the data collected in this study, the average observed values for the phone and the scanner s MSA are about 35% and 65% respectively. It is possible to somewhat extend the phone-based MSA by connecting the external antenna on the roof of the vehicle. In that case, the RSL measured by the phone is expected to increase by approximately 11dB, according to the values found in [2]. However, the phone populates its serving and neighbor list according to the c1 and c2 algorithms [3]. As a result, at any given time only the serving BCCH and up to six neighbor cell BCCH channels with highest C/I values are reported. 3.1.2. MSA analysis using drive test simulations In order to verify the measured MSA, drive test modeling is performed using computer simulation. Sample simulation results are presented in Fig.4. In the simulation, it is assumed that the MSA depends on a threshold of the drive-test instrument. The threshold is defined as minimum C/I value for which the tool is capable of measuring the RSL while decoding the BSIC. Thresholds of 8 db, 0 db, and -18 db are used for the phone, scanner, and receiver respectively. Scanner Receiver Figure 3. MSA analysis using measured RSL data Scanner Receiver Figure 4. MSA analysis using drive-test computer simulation

The drive test modeling calculates received signal levels and C/I values in the bins along the drive route. In the simulation, the parameters of the live network, such as cell site location and channel plan are used. As an example of the channel plan, co-channel and adjacent channel interferers of the server (in light blue) are highlighted as red and yellow sectors respectively on Fig. 3 and 4. For any given sector, the predicted MSA simulation is performed as follows. The RF propagation models is obtained and optimized using the receiver s measured data. In this analysis, the simple Lee s propagation slope-intercept model is used [5]. For every sampling point (bin), the received signal levels for each sector are determined using either optimized Lee s or default sub-urban slope-intercept RF propagation model. For every sector, utilizing the network channel plan, the C/I ratios are calculated in every bin. The interference is calculated as a linear sum of both co-channel and adjacent channel RSL levels. Additionally, an 18 db of adjacent channel rejection attenuation is applied on the interference coming from adjacent channels [4]. In other words, the C/ I values in every bin for every tool are obtained using: N I [ ] 2 C / I = C dbm 10 log + I i σ (1) N i= 1 where C represents RSL of the serving cell, I represents i RSL of the i-the interferer expressed in mw, represent the total number of interferers, and σ 2 Ν represents the average power of the narrow band thermal noise expressed in mw. Figures 3 and 4 may be used for a head-to-head comparison of the MSAs obtained from measurements and simulations. As one may observe that in both cases, phone based measurements can be found only in a relatively small area within the main antenna beam. However, the receiver based system allows data collection in a wider area, and even in vicinity of both co channel and adjacent channel sectors. This implies that the receiver is able to measure the RSL on channels with very low C/I values and closely approach the case when the RSL measurement are performed on a clear channel (i.e., without co- and adjacent channel interference). The curve in Fig. 5 shows the percentage of MSA for the given sector as a function of the dynamic range of the instrument (C/I value). In the figure, case of 100% corresponds to clear channel measurements. N I Percentage [%] 100 90 80 70 60 50 40 30 20 10 C/I = -18dB (85%) C/I = 0dB (63%) C/I = 8dB (41%) 0-60 -40-20 0 20 40 60 C/I values Figure 5. MSA as a function of instrument dynamic range From the curve, one may estimate the size of the MSA for given instrument relative to the size of the clear channel MSA. As seen, the receiver achieves 85% ratio, while the phone reaches only about 41%. 3.1.3. MSA in an ideal hexagonal cell layout The simulation methodology is extended to a regular hexagonal cellular system layout. Frequency reuse factors of 3, 4 and 7 are evaluated. The analysis is performed as follows. Network of 175 sectors is used. In the simulation, all sectors use the same antenna patterns, ERP, and antenna heights. The default Lee s suburban slope-intercept RF propagation model is used to predict the RSL level from each sector in each 50 m bin, and (1) is used to calculate C/ I values. For every frequency reuse factor, the MSAs for three device specific thresholds are obtained (8dB for the phone, 0dB for the scanner and -18dB for the receiver). Figure 6 shows an example plot of the MSA areas for the three devices obtained for the frequency reuse factor of N = 7. In the figure, the observed sector for which the MSAs are calculated is placed in the center of the area. The red diamonds represent co-channel sectors; and green diamonds represent non interfering sectors. The area where phone is capable of measuring the BCCH channel is presented in red. Scanner s MSA is comprises of both red and blue, while the receiver s MSA contains red, blue and green areas. Similar plots are obtained for reuse factors N=3 and N=4. With increasing frequency reuse factor, the level of interference decreases and the MSA for each instrument becomes larger. However, it is observed, the reuse factor has no effect on relative MSA ratios between different tools. That implies that the relative MSA sizes of different measurement tools always stay the same regardless of the frequency plan.

50 C/I ranges for frequency reuse factor 7 0.35 Mean = 14.12dB Stdev =5.18dB Count=858 40 0.3 Distance in km 30 20 10 0-10 -20 Relative frequency of occurence 0.25 0.2 0.15 0.1-30 0.05-40 -50-50 0 50 Distance in km Figure 6. MSAs for frequency reuse factor 7 If one observes the limited area around the cell site similar in size to the one used in drive-test simulation and appropriate frequency reuse plan, the scanner s and phone s MSAs relative to the receiver s MSA become about 72.6 % and 31.2 %. Both results are consistent to the MSA values obtained from the measurement and the drive-test simulation, as summarized in Table 1. TABLE I. SUMMARIZED MSA VALUES Measured values Modeled values Ideal case N=4 Scanne r Scanne r 44.21 80.71 50.44 80.94 31.00 76 Scanne r All MSA values are calculated relative to the receiver s MSA 3.2. Differences between RSL measurements To determine the extent to the differences between the RSL measurements collected using different drive-test systems, the statistical analysis is performed as follows. For each bin and for each test tool, the three averaged RSL values are calculated for a given sector (BCCH/BSIC combination). In each bin, the averaged RSL values for a given sector are compared. For every sector, every tool, and every day, the mean and the standard deviation of the difference between the average RSL values are calculated. Only common bins contained readings collected by all the instruments are considered in this analysis. The average RSL readings obtained by the scanner and the receiver are within the range of 1dB from each other. The measurements from the phone are on average 10 db lower due to the vehicle penetration loss. Figure 7 shows a typical histogram of the differences between the RSL measurements obtained by the phone and the receiver for the same sector observed in previous analyses. 0-5 0 5 10 15 20 25 30 35 Difference between Receiver & [db] Figure 7. Histogram of difference between the receiver and phone RSL measuremetns 4. Observations and conclusions The comparison study of the RSL measurements obtained from various data collection tools is presented in this paper. The measurement tools considered are phone based, scanner based and high dynamic range receiver based. Out of all examined data collection tools, the high dynamic range receiver has the largest MSA. Relative to the receiver, the percentage for scanner s MSAs is about 50-80% in size, while the phone s MSA is only 20-40%. For the data collected in this study, the average observed values for the phone and the scanner s MSA are about 35% and 65% respectively. A very small MSA makes the phone based measurements unsuitable for the tasks of propagation modeling and even RF optimization. On the other hand, the size of the MSA for the receiver is relatively large and comparable to the one obtained through clear channel CW measurements. The study demonstrated a relatively good agreement between measured values and computer simulations. A simple slope-intercept propagation model may be applied to predict MSA only in areas with similar clutter profile. However, in large number of cases, optimizing the slope and intercept may not to be sufficient to capture majority of signal variations. In cases, where the receiver data is used for propagation modeling, the selected model needs to have more then two degrees of freedom. Along with the slope and intercept, other parameters of the model such as clutter weights and obstruction may be considered for optimization as well. In cases of sectors with large number of RSL points, the size of MSA is in a relatively good agreement with results obtained from the theoretical hexagon layout. In general, from the study of the ideal system, it may be observed that a large benefit in terms of MSA size results from the increase of a dynamic range of the

receiver. In the ideal case, taking the entire area into account, the size of the scanner area relative to the receiver is only 22%, while this ratio for the phone based measurements is as low as 10%. However, these numbers are not observed in the measurements, due to a relatively small area under the test. Nevertheless, a high dynamic range of the receiver makes its MSA much larger than the one obtained by either the scanner or the phone. It is shown both theoretically and through measurements that the relative size of MSA areas between various measurement devices does not change as a function of frequency reuse. Therefore, the ratio between the sizes of the MSAs for different tools will be preserved regardless the applied frequency plan. The average RSL readings obtained by the scanner and the receiver are in the range of 1 db from each other. The measurements from the phone are on average 10 db lower due to the vehicle penetration loss. 5. Acknowledgements Authors would like to express a sincere appreciation to Mr. Dale Bass, from PCTEL and Mr. Bob Joslin, Mr. Scott Clay and Mr. Greg Akin from EVWI. Special thanks to Ms. Samira Noel, Mr. Rande Nicolls and Mr. Steve Vest for their contributions to this research. 6. References [1] S. Dickey, Novel technique for co-channel interference measurements in cellular networks, in proceedings of WNCG Symposium, October 22-24, 2003. [2] I. Kostanic, C. Hall, J. McCarthy, Measurements of the Vehicle Penetration Loss Characteristics at 800MHz, in proceedings of IEEE VTC 1998. [3] A. Mehrotra, GSM System Enginnering, Artech House, 1997. [4] J.D. Parsons, The Mobile Radio Propagation Channel, 2 nd Edition, John Wiley and Sons, New York, 1997.