A generalized formulation of the protection ratio applicable to frequency coordination in digital radio relay networks

Transcription

1 RADIO SCIENCE, VOL. 42,, doi: /2006rs003470, 2007 A generalized formulation of the protection ratio applicable to frequency coordination in digital radio relay networks Kyoung Whoan Suh 1 Received 27 January 2006; revised 7 June 2006; accepted 20 October 2006; published 28 February [1] This paper proposes an efficient and comprehensive algorithm for computing the protection ratio and illustrates some results applicable to the initial planning of frequency coordination for fixed wireless networks. A net filter discrimination depending upon transmitter spectrum mask and overall receiver filter characteristic is also examined to see the effect of adjacent channel interferences. Numerical simulations for cochannel and adjacent channel protection ratios are performed for the 6.2 GHz frequency band, including transmitter spectrum mask and receiver filter response. According to results for 64-QAM (quadrature amplitude modulation) and 60 km at bit error ratio 10 6, fade margin and cochannel protection ratio are 41.1 and 74.9 db, respectively. In addition, it is shown that the net filter discrimination for 30 MHz channel bandwidth provides 26.5 db at the first adjacent channel, which yields adjacent channel protection ratio of 48.4 db. The proposed method gives an easy and systematic method to compute the protection ratio and can be applied to frequency coordination in fixed wireless networks up to the millimeter wave band. Citation: Suh, K. W. (2007), A generalized formulation of the protection ratio applicable to frequency coordination in digital radio relay networks, Radio Sci., 42,, doi: /2006rs Introduction [2] In order to provide spectrum resources needed for upcoming wireless businesses such as a system beyond IMT-2000, next generation broadcasting, u sense network, telematics, and public safety and disaster relief, the Ministry of Information and Communication in Korea has been studying a plan for frequency movement and its rearrangement below 6 GHz, especially for some radio relay frequency bands. Those bands have been mainly used for a long-haul high-capacity transmission of voice, video, and data information. However, demand for those applications is rapidly decreasing during the last decade because of the optical transmission system and its nationwide deployment. Hence, to make proper frequency coordination for new service systems in conjunction with existing wireless networks such as radio relay, cellular, satellite, and radar, the protection criteria for interference analysis must be performed in advance to meet good 1 Electronics Engineering, Kangnam University, Kyeonggi-Do, South Korea. Copyright 2007 by the American Geophysical Union /07/2006RS quality of service recommended by the International Telecommunication Union [Suh, 2004]. [3] Regarding a basic method of frequency coordination in the wireless network, some criteria based upon the concept of a protection ratio have been adopted with a generic interference management methodology. The protection ratio defines a minimum relative power ratio of wanted to unwanted signals in the interfered receiving system. In general, two interference types such as cochannel and adjacent channel ones are considered as unwanted signals [Townsend, 1988]. The cochannel protection ratio relevant to cochannel interference is expressed as a function of S/N (signal-to-noise ratio) of modulation scheme, interference-to-noise ratio, multiple interference allowance, fade margin of multipath or rain attenuation [European Telecommunications Standards Institute (ETSI), 2005; International Telecommunication Union (ITU), 2004a, Radiocommunications Agency, 2003]. On the other hand, the adjacent channel protection ratio associated with adjacent channel interference includes the net filter discrimination as well as variables of the cochannel protection ratio. The net filter discrimination depends entirely upon the transmitting spectrum mask and the overall receiving filter responses. [4] Since the mid-1990s, two main studies relevant to protection from interference in the fixed wireless network 1of10

2 Figure 1. Cochannel interference. have been undertaken to define a maximum interference limit and a predictable protection ratio based upon a fade margin [National Spectrum Managers Association, 1992; Australian Communications Authority, 2003]. The former, performed by the Federal Communications Commission (FCC), European Telecommunications Standards Institute (ETSI), and Radiocommunications Agency (RA), has illustrated practical applications based upon system parameters such as effective isotropical radiated power, noise figure, occupied channel bandwidth, transmission capacity, and signal degradation due to interference. The latter, conducted by the Australian Communications Authority (ACA) has shown the initial frequency coordination method based upon the protection ratio by using the fade margin including interference limit, and S/N of modulation scheme was not taken into consideration as a variable even if it actually varied from modulation schemes [Suh, 2004]. [5] In this paper, in order to provide a comprehensive solution for the protection ratio, we suggest an efficient and systematic algorithm for the protection ratio calculation needed for initial planning of frequency coordination in radio relay networks. Moreover, the net filter discrimination associated with transmitter spectrum mask and overall receiver filter characteristic is examined to obtain the adjacent channel protection ratio. To show calculation procedure and practical applications, some computations for cochannel and adjacent channel protection ratios are performed in the actual radio relay frequency band of 6.2 GHz including transmitter spectrum mask and receiver filter responses, and several results including 38 GHz millimeter wave band are illustrated here as a guidance of initial planning for frequency coordination. 2. Interference and Fade Margin 2.1. Interference Type [6] For the radio relay network, there are basically two types of interference affecting the transmission quality 2of10 [ETSI, 2005]. When the center frequency of the interference signal coincides with that of the desired signal, this is called cochannel interference as shown in Figure 1. It is depicted that transmitter at A site with frequency f c is designed to transmit the wanted signal toward receiver at B site, but its signal may be overreached at receiver of C site where it is originally designed to receive only the signal from transmitter at B site, with the same frequency. This type of interference may usually occur in the network topology of star or grid. [7] On the other hand, the interference caused by an adjacent channel is called the adjacent channel one as depicted in Figure 2, which occasionally causes some problems in network design. As an example for receiver at D site, the signal with frequency f c transmitted by transmitter at A site may interfere with the wanted signal with frequency f c + Df coming from transmitter at C site. To reduce the interference from adjacent channel, the frequency separation Df between wanted and unwanted signals may be more important, including the overall filter discrimination in the interfered receiver Fade Margin [8] Two main factors that cause the wanted signal to fade are multipath clear air fading and flat fading due to rainfall. First, let us consider the dispersive fade margin caused by multipath fading in the radio relay link as shown in Figure 3 [Rummler, 1979]. It is well known that multipath fading in the atmosphere is more frequent at night and in the first morning hours, and it is seldom felt at midday or during periods of intense rain. [9] The probability that the received power p is less than or equal to p 0, at the frequency f (GHz), in the worst month, is given by [ITU, 2004b; Freeman, 1997] Pðp p 0 Þ ¼ Kd 3:6 f 0:89 1 þ e p 1:4 p p n ð1þ where p n is a received power at no fading, and je p j = 1 d jh r h t j is the absolute value of the path inclination in

3 Figure 2. Adjacent channel interference. milliradians where d is path length in km, f is frequency in GHz, and h t and h r are altitudes of the transmitter and receiver antennas in meters, respectively. The geoclimatic factor K for the average worst month from p L based upon empirical relationship of ITU-R Recommendation P [ITU, 2004b] is given as Table 1. [10] The p L means the percentage of time that the average refractivity gradient in the lowest 100 m of the atmosphere is less than 100 N in units/km and contour maps are provided in ITU-R Recommendation P [ITU, 2004c]. According to documentation, p L in Korea is 10 for the worst month, August [ITU, 2004c]. [11] To obtain a guarantee for transmission quality with a predefined percentage of time in unavailability for a given link, the required fade margin due to multipath fading is equal to fade depth exceeded for the percentage of time ( p w ) in the average worst month. So, equation (1) may be expressed by FM ¼ 10 log 10 Kd 3:6 f 0:89 1 þ e p 1:4 10 log 10 ðp w Þ ð2þ where FM = 10 log( p n p 0 ) and p w = P(p p 0 ) 100(%). [12] Second, let us examine the flat fade margin due to rainfall. The specific attenuation due to rain based upon ITU-R Recommendation P.838 [ITU, 2004d] is given by g R ¼ kr a ðdb=kmþ ð3þ where R is rainfall intensity in mm/hr, and k and a denote frequency dependent coefficients. [13] For any given path, the attenuation due to rain is calculated for an effective path length, which account for the distribution of the rain intensity rate. The attenuation A 0.01 due to rain exceeded for 0.01% of the worst month is A 0:01 ¼ g R d 1 þ d ð4þ d 0 in db, where d is an actual path length in km, and d 0 is given by d 0 ¼ 35e 0:015R 0:01 ; for R 0: ð5þ in mm/hr. However, 100 of d 0 is used in the case of R 0.01 : 100 mm/hr. The attenuation due to rain exceeded for other percentages of time p( ) of the worst month is A p ¼ A 0:01 0:12 p ð 0:546þ0:043 log 10 pþ ð6þ Table 1. Geoclimatic Factor K Figure 3. Concept of frequency selective fading caused by multipath. 3of10 Factor K = p L K = p L K = p L K = p L Description overland links for which the lower of the transmitting and receiving antennas is less than 700 m above mean sea level overland links for which the lower of the transmitting and receiving antennas is higher than 700 m above mean sea level links over medium-sized bodies of water, coastal areas beside such bodies of water, or regions of many lakes links over large bodies of water, or coastal areas beside such bodies of water

4 Figure 4. Wanted signal (AB and CD sites) and interference paths (AD and CB sites). Equation (6) is valid for radio links located in latitudes equal to or greater than 30 (north or south). 3. Frequency Coordination and Protection Ratio 3.1. Frequency Coordination [14] A typical case of a potential interference scenario for one direction of transmission with a single frequency is depicted in Figure 4. Assuming that link AB means an existing service and link CD is a proposed new service, the potential interference paths of AD and CB marked by dotted lines with the corresponding transmit and receive antenna discrimination angles (q and q 0 ) relative to the respective antenna boresight azimuth are also presented. Then, the received wanted or unwanted signal power can be expressed as P r ¼ P t þ G t þ G r L t L r L f ð7þ where P r is RF signal power at the input of the receiver (dbm), P t is RF signal power at the output of the transmitter (dbm), G t means the gain of the transmitting antenna in the azimuth of the receiver (dbi), G r means the gain of the receiving antenna in the azimuth of the transmitter (dbi), L t is feeder and branching losses associated with the transmitter (db), L r is feeder and branching losses associated with the receiver (db), and L f denotes the total transmission loss between the transmitter and receiver antennas (db). [15] For instance, let us consider link AB sites in Figure 4. The power ratio of wanted-to-unwanted signals at the input of a potential interfered receiver of B site can be expressed by C=I ¼ P wntd P unwntd ð8þ in db. In consequence, in order to make a successful frequency coordination, equation (8) is compared with the relevant protection ratio (PR), defined by a minimum relative power ratio of wanted and unwanted (interference) signals at the input of the interfered receiver. Therefore the power ratio of wanted-to-unwanted signal, C/I, should be greater than the protection ratio, which is given by C=I PR ð9þ 3.2. Protection Ratio and Net Filter Discrimination [16] In general, the system sensitivity T s is defined as the signal strength that a properly modulated signal must have to provide the desired output in a receiver system, and a maximum allowable interference power with respect to noise power in the interfered receiver system may comprise a single or multiple interference sources, which is expressed by I t. Then both parameters are given by [Suh, 2005] T s ¼ 10 logðktbþþnf þ S=NðBER 10 y Þ ð10þ I t ¼ 10 logðktbþþnf ðn=iþ MIA ð11þ where k is Boltzmann s constant ( J/K), T is Kelvin temperature (290 K), B is the receiver bandwidth in Hz, and NF is noise figure in db. S/N is the power ratio of signal to noise in db depending upon the modulation and coding schemes at a given bit error ratio (BER) of 10 y, and therefore ITU-R F.1101 is referenced for S/N of M-ary quadrature amplitude modulation (QAM) at BER N/I is the power ratio of noise to interference such as 10, 6 or 3 db which brings about 0.5, 1 or 2 db degradation in signal level due to interference, respectively. MIA is multiple interference 4of10

5 Figure 5. Pictorial concept for the protection ratio. allowance related with multiple interference sources, which usually permits 4.0 db. [17] In order to formulate the protection ratio including fade margin associated with availability and the net filter discrimination related with a function of rejecting adjacent channel interference. Considering the result of subtracting equation (11) from equation (10) as well as fade margin and the net filter discrimination, the protection ratio is given as follow [Suh, 2005] PR ¼ S=NðBER 10 y ÞþN=I þ MIA þ FM NFD ð12þ where FM is the fade margin of dispersive or flat fade, and NFD is the net filter discrimination depending upon transmitter spectrum mask and overall receiver filter characteristics. In consequence, to understand parameters related with equation (12), the pictorial concept of the protection ratio is characterized in Figure 5. [18] It is general practice in coexistence studies between transmitters and receivers of different symbol rate and modulation formats to use the concept of the net filter discrimination. The definition of the net filter discrimination is given by [ETSI, 2005] NFD ¼ 10 log P c ð13þ P a ¼ P c ¼ Z 1 0 Z 1 0 P a GðfÞjHðfÞj 2 df ð14þ Gðf Df ÞjHðfÞj 2 df ð15þ where P c is the total power received after cochannel RF, IF, and base band filtering, and P a is the total power received after offset RF, IF, and base band filtering. The functions of G( f ) and H( f ) are transmitter spectrum mask and overall receiver filter responses, respectively, and Df denotes the frequency separation between a desired signal and an interference signal. So, it can be readily expected that NFD yields 0 db for the cochannel interference with Df = 0. [19] Equation (13) is made on the basis of two the assumptions. First, adjacent channel cross polarization discrimination (XPD), if any, is not taken into consideration. Second, a single sideband interfering channel only is taken into account and for double side like-modulated interferences, the net filter discrimination with 3 db lower should be considered. Therefore, as pointed out in ITU-R Recommendation F.746 [ITU, 2003], equation (13) is produced purely by transmitter spectrum and by the overall receiver filtering, and it does not comprise any other decoupling such as antenna discrimination or the actual interfering power level. To compute the net filter discrimination numerically, a discrete form of equation (13) may be written by "!!# NFD ¼ 10 log X i¼n 1 i¼0 10 T ci þr ci 10 = X i¼n 1 i¼0 10 T oi þr ci 10 ð16þ where n denotes number of samples, jh( f )j 2 = R ci (db) is receiver mask sampled at a defined step frequency in cochannel, G( f )=T ci (db) means transmission mask 5of10

6 Figure 6. Algorithm for the protection ratio calculation. sampled at a defined step frequency in cochannel, and G( f Df )=T oi (db) is transmission mask sampled at a defined step frequency in offset. [20] As a consequence, the calculation for the protection ratio mentioned above can be summarized as seen in Figure 6, and in order to make proper frequency coordination it also describes that the resultant protection ratio should be larger than C/I obtained at the input of interfered receiver. 4. Simulated Results and Discussion [21] Prior to calculating the protection ratio, let us consider parameters involved in equation (12) for the frequency band of 6.2 GHz. So, N/I = 6 db, MIA = 4.0 db, Figure 7. Cochannel protection ratio. 6of10

7 Figure 8. Transmitter spectrum mask with MHz channel bandwidth. p L =10,p w = 0.01%, e p = 0 are chosen, and S/N is referred from ITU-R F.1101, for instance, which gives S/N = 23.8 db for 64-QAM at BER 10 6 [ITU, 2004e]. The value of N/I of 6.0 db equals to 1.0 db degradation in received signal due to interference. In addition, to find out the net filter discrimination for a given system, the overall receiver selectivity, given by the resultant response of RF, IF, and base band filtering, is required. However, it is assumed to have the same spectrum mask as the case of transmitter one for convenience here in the absence of related data. [22] Figure 7 shows the protection ratio for cochannel interference as a function of distance and M-ary QAM, and it illustrates the protection ratio of 74.9 db for 64-QAM and 60 km. The physical meaning of this value is that C/I of the interfered receiver in the existing network should have more than 74.9 db to assure coexistence for a new radio relay link design. [23] Next, let us examine the net filter discrimination in advance of calculating the adjacent channel protection ratio. According to the channel allocation of ITU-R Recommendation F [ITU, 2001], MHz per channel is assigned for high-capacity transmission in 6.2 GHz band. The transmitter spectrum mask used for calculating the net filter discrimination is the response of curve a in Figure 8 [ETSI, 2002] and the receiver spectrum mask noted by jh(f)j 2 = R ci (db) is the square of the overall receiver filter response, which is also taken as the same transmitter spectrum mask with a view to showing calculation procedure for the sake of convenience. The calculated net filter discrimination as a function of frequency offset Df is illustrated in Figure 9 and gives 3.6 and 26.5 db at offset frequency 15 and 30 MHz, respectively. Although the integral range for computing the net filter discrimination is from 0 to 1, it is assumed that the integration is actually performed from 30 MHz to +30 MHz because the cumulative power beyond that bandwidth is relatively negligible. Since the net filter discrimination at Df = 30 MHz in Figure 9 is 26.5 db, the adjacent channel protection ratio at 30 MHz can be easily obtained as 48.4 db, subtracting the net filter discrimination of 26.5 db from the cochannel protection ratio of 74.9 db. So, it is possible to summarize the resultant protection ratios corresponding to cochannel and adjacent channel as seen in Table 2. Therefore if one applies the same procedure to the remaining M-ary QAM or other modulation schemes, a variety of the protection ratio are easily produced by adding or subtracting S/N difference with respect to 64-QAM. [24] Also, in order to see the effect of protection ratios caused by rain as well as multipath, Figure 10 illustrates cochannel ratios at 11 GHz under the same conditions of Figure 7, where rain intensities of 40, 60, and 80 mm/hr are taken to be 0.01 percentage of time at the worst month. The protection ratio due to multipath fading gives 66.2 db for 30 km and 64-QAM, which is larger than that due to any rain intensity. If one considers the protection ratio due to 80 mm/hr, it leads to about 58.0 db reduced by 8.2 db from the protection ratio corresponding to multipath fading. It is known that multipath fading in the atmosphere and flat fading due to intense rain does not occur simultaneously. So, the above two results imply that a link designer should make much account of the protection ratio caused by multipath fading because the fade margin of rain is still less than that of multipath fading, especially for frequency band below 11 GHz. However, if frequency increases up to a millimeter wave, rain attenuation is rapidly increasing and the effect of multipath fading is negligibly small in 7of10

8 Figure 9. Net filter discrimination with offset frequency. comparison to rain fading because of short-distance transmission. [25] In addition, to show the required fade margin due to multipath or rainfall for assuring availability in percentage of time, Figure 11 depicts a sort of correction factor of the protection ratio with 4 curves corresponding to one multipath and 3 rain intensities at 11 GHz. As multipath fade margin is dominant, 32.4 db at 30 km based upon equation (2) is referenced as 0 db, and then the others are relatively plotted. If one wants to find the protection ratio for a given rain intensity, the corresponding value can be easily derived by using Figure 11 because parameters in equation (12) are constant except for fade margin here. So, it is worth noting that, for instance, even though the use of correction factor related with fade margin is illustrated, one may apply the same concept to various parameters such as modulation schemes, frequency, geoclimatic factor, N/I, percentage of time in unavailability, rain intensity etc. Consequently, this means that an easy and systematic extension for the protection ratio calculation could be realized by adopting correction factors presented here. [26] Finally, to examine the effect of flat fading only due to rainfall in the millimeter wave band, Figure 12 shows the cochannel protection ratio for 38 GHz radio system with a 4-PSK (phase shift keying) or on-off keying modulation, where rain intensities of 40, 60, and 80 mm/hr are taken to be 0.01 percentage of time at the worst month, resulting in availability of 99.99%. That frequency band is greatly used for short-distance radio links with low capacity for mobile base stations as well as high-speed Internet bridge. Since heavy rain attenuation in the millimeter wave limits transmitting distance even up to a few km, one may make little account of the fade margin due to multipath fading. The parameters involved in Figure 12 are the same as Figure 7 only except for S/N of 13.5 db at BER 10 6 without coding and MIA = 0 db. It is shown that fade margins for 40, 60, and 80 mm/h are about 27.7, 39.0, and 48.4 db at 3 km, respectively, and its corresponding protection ratios are 47.2, 58.5, and 67.9 db. As a consequence, it is interesting to note that only fade margin caused by rain attenuation should be considered for the protection ratio calculation in the millimeter wave. 5. Conclusions [27] In this paper, the efficient and generalized algorithm of the protection ratio calculation was suggested for the initial planning of frequency coordination in fixed radio relay networks, and some interesting results related with fade margin, net filter discrimination, and protection ratio were illustrated in view of direct applications of network design for frequency coordination up to the millimeter wave band. In addition, net filter discrimination, depending upon transmitter spectrum mask and overall receiver filter characteristics, was examined to Table 2. Adjacent and Cochannel Protection Ratios a Frequency Offset, MHz Protection Ratio, db 0 (cochannel) (first adjacent channel) 48.4 a Other parameters are f = 6.2 GHz, 64-QAM, d = 60 km, and p L = 10. 8of10

9 Figure 10. Cochannel protection ratio at 11 GHz. see the effect of the adjacent channel protection ratio caused by adjacent channel interferences. [28] According to computed results with regard to 6.2 GHz, 64-QAM, and 60 km at BER 10 6, the fade margin and the cochannel protection ratio were 41.1 and 74.9 db, respectively. For the net filter discrimination with respect to 30 MHz channel bandwidth, it provided 26.5 db at the first adjacent channel of 30 MHz, and this resulted in the adjacent channel protection ratio of 48.4 db. Moreover by introducing the protection ratio correction factor, it was shown that the protection ratio calculation can be systematically expanded for various parameters such as modulation schemes, frequency, geoclimatic factor, N/I, percentage of time in unavailability, rain intensity, etc. In addition, protection ratios for 11 and 38 GHz radio relay systems were calculated and reviewed to show the effect of fade margin caused by rainfall as the frequency increases. The suggested method provides an easy and systematic method to compute the protection ratio and can be applied to Figure 11. Protection ratio correction factor at 11 GHz. 9of10

10 Figure 12. Cochannel protection ratio at 38 GHz. frequency coordination in fixed wireless networks up to the millimeter wave band. References Australian Communications Authority (2003), Microwave fixed service frequency coordination (RALI, FX3), Aust. Commun. and Media Auth., Belconnen, ACT, Australia. European Telecommunications Standards Institute (ETSI) (2002), Fixed radio systems; point-to-point equipment; high capacity digital radio systems carrying SDH signals (up to 2 STM-1) in frequency bands with about 30 MHz channel spacing and using co-polar arrangements or co-channel dual polarized operation, ETSI EN , Sophia Antipolis, France. European Telecommunications Standards Institute (ETSI) (2005), Derivation of receiver interference parameters useful for planning fixed service point-to-point systems operating different equipment classes and/or capacities, ETSI TR , Sophia Antipolis, France. Freeman, R. L. (1997), Radio System Design for Telecommunications, John Wiley, Hoboken, N. J. International Telecommunication Union (ITU) (2001), Recommendation ITU-R F.383-5, Radio-frequency channel arrangements for high capacity radio-relay systems operating in the lower 6 GHz band, F Series, Geneva, Switzerland. International Telecommunication Union (ITU) (2003), Recommendation ITU-R F.746, Radio-frequency arrangements for fixed service systems, F Series, Geneva, Switzerland. International Telecommunication Union (ITU) (2004a), Recommendation ITU-R F.758-3, Considerations in the development of criteria for sharing between the terrestrial fixed service and other services, F Series, Geneva, Switzerland. International Telecommunication Union (ITU) (2004b), Recommendation ITU-R P , Propagation data and prediction methods required for the design terrestrial lineof-sight-systems, P Series, Geneva, Switzerland. International Telecommunication Union (ITU) (2004c), Recommendation ITU-R P.453-9, The radio refractive index: Its formula and refractive data, P Series, Geneva, Switzerland. International Telecommunication Union (ITU) (2004d), Recommendation ITU-R P.838, Specific attenuation model for use in prediction method, P Series, Geneva, Switzerland. International Telecommunication Union (ITU) (2004e), Recommendation ITU-R F.1101, Characteristics of digital fixed wireless systems below about 17 GHz, F Series, Geneva, Switzerland. National Spectrum Managers Association (1992), Report and tutorial: Carrier-to-interference objectives, Rep. WG , Vancounver, Wash., Jan. Radiocommunications Agency (2003), Technical frequency assignment criteria, report, Off. of Commun., London, June. Rummler, W. D. (1979), A new selective fading model: Application to propagation data, Bell Syst. Tech. J., 58, Suh, K. W. (2004), A study on the criteria of co-channel and adjacent channel interferences allowable in the M/W relay networks, report, Electron. and Telecommun. Res. Inst., Daejeon, South Korea. Suh, K. W. (2005), A study on calculation of NFD and protection ratio of fixed radio relay system for analyzing adjacent channel interference (in Korean), Korea Electromagn. Eng. Soc., 16(11), Townsend, A. A. R. (1988), Digital Line-of-Sight Radio Links: A Handbook, Prentice-Hall, Upper Saddle River, N. J. K. W. Suh, Electronics Engineering, Kangnam University, Yong-In, Kyeonggi-Do , South Korea. (kwsuh@ kangnam.ac.kr) 10 of 10