Development of Power Quality Diagnosis System for Power Quality Improvement

U-Yop Chung, Dong-Jun Won, Joong-Moon Kim, Seon-Ju Ahn, Student Member, IEEE, Seung-I1 Moon, Jang-Cheol Seo, and Jong-Woong Choe, Member, IEEE

Abstract- This paper represents the structure and functions of total power quality data management system including power quality monitoring system (PQMS), GUI system, and PQ diagnosis system. PQMS characterizes RMS value, harmonics, and frequency, and detects power quality (PQ)events. GUI system manages and stores power quality data received from a number of PQMSs and displays the power quality trends and events. PQ diagnosis system gives total solution of power quality problems such as statistical analysis, event prediction, power quality indexing, event identification, remedy suggestion, etc. Therefore, usen can improve their power quality level according to advices provided by PQ diagnosis system. This paper explains the functions of PQ diagnosis system in detail. Moreover, a prototype of PQ diagnosis system as well as PQMS and GUI system is constructed and tested.
Index Terms-power
sag,

quality, diagnosis, monitoring, voltage harmonics, event identification,power quality index

I. INTRODUCTION

significant issue for both power suppliers and customers. There have been three important changes in relation to power quality. First of all, the characteristics of load have become so complex that the voltage and current of the power line connected with these loads are easy to be distorted. Lately, for example, non-linear loads with power electronic interface that generate large harmonic current have been greatly increased in power system [ 11. Next, the end-user equipments have become more sensitive to power quality than before. Telecommunication devices, adj ustable-speed drives, process-control equipments and computers are notorious for their sensitivity to power quality. Those loads may suffer failure, misoperation, or hardware damage even when voltage and current deviate from its normal state in small quantities [2]. This means the fact that definition of power quality should be changed. In the past, it is sufficient that utility should provide electric power without
This work was suppared in pan by the LG Industrial Systems Co., Ltd.. Il-Yop Chung. Dong-Jun Won, Joang-Moan Kim. Son-Ju Ahn and Seung-Il Maon are with the Schwl of Electrical Engineering and Computer Science, S e w 1 Nationvl University, Seoul 151-742, Korea (e-mil:
iry"pepo~~rluh.a",,.~~.~) Jang-Cheol Se0 and long-Woang Choe are with the Elec~otechnologyR&D Center, LG Industrial System Co.. Ltd.. Sewl. Korea (e-mik [email protected])

I..

N the recent years, power quality (PQ) has become an

interruption. However, in recent years, customers require high-level power quality, so power quality encompasses various categories like sag, swell, interruption, harmonics, transient, flicker, and so on. IEEE standard 1159 defines power quality events in detail. Finally, deregulation of the electric power business brings the radical and abrupt changes in the whole power system including utility, facility, and customers. In the traditional electric power system, the price and service t e r m supplied by monopoly utility are set uniformly. However, in the deregulated power system, monopoly utility will be partitioned to diverse companies such as generation companies, network operators, power traders, and retail energy service companies, and they will supply unbundled power quality required by customers. Then, the level of power quality will he determined by the contract between supplier and customer. Each participant will have the responsibility not to degrade power quality level [3]. Therefore, it is important to calculate power quality level exactly and identify the source of faults that corrupt power quality. To solve above-mentioned various power quality problems, power quality should be measured accurately according to the exact definition of each power quality category, and then evaluated and diagnosed in versatile ways. Power quality monitoring system (PQMS), that characterizes power quality events and variations, has experienced rapid progress using high-tech detection functions. PQMS has many variations in their structure, but the recent trend is the permanent type that is installed at one site permanently to assess the power quality for 24 hours a day without a break. Moreover, in many cases, to manage local power system that has more than one PQMS efficiently, the analyzed data of each PQMS are sent to the graphic user interface (GUI) computer of the system supervisor using network connection [4]. This paper presents the power quality diagnosis system that makes power quality data stored in GUI system more informative. This system gives total solution such as event detection, power quality trend analysis, statistical analysis, event prediction, event cause identification, event location, remedy suggestion, and so forth. The structure of this system is explained and its main functions are also presented. Moreover, prototype of total power quality data management system that includes PQMS, GUI system, and PQ diagnosis system is developed and tested.

0-7803-7989-6/03/$17.00 02003 IEEE

1256

Authorized licensed use limited to: Saintgits College of Engineering. Downloaded on August 4, 2009 at 23:14 from IEEE Xplore. Restrictions apply.

11. SYSIEM CONFICURA-ITON

A. Overall System Configuration

The overall power quality data management system consists of three major components like PQMS, GUI system, and PQ diagnosis system as illustrated in Fig. 1 . PQMS evaluates power quality trends and detects power quality events after it analyzes the sampled voltage and current waveforms obtained from power line. The power quality data made by PQMS are transferred to CUI system through network line. The most recommended network is fast Ethernet using TCPnP though other network like serial communication can he used. The reason is that power quality data is very large in quantity. GUI system can receive the data from several PQMSs and store the data to suitable database. In addition, GUI system shows the event data, trend data, and waveform on monitor screen and alerts event happening in various ways like pager, cellular phone, or even alarm hell and lamp. The data in the database are used as the input of PQ diagnosis system. PQ diagnosis system is offered as an optional function of CUI system. Therefore, when this system is installed, PQ diagnosis menus are added and activated to main menus of GUI software. This system makes a diagnosis of power quality in many ways such as predicting probability of event, identifying event cause and location, proposing counter measures, and so forth.
POWW

and trend analysis. Final one is data transmit part that usually uses fast Ethernet, for example, 10 or 100 Mhps with TCPAP. Fig. 2 shows representative structure of PQMS with decentralized processing architecture. This structure integrates all of ahove-mentioned three parts into one PQMS. Most companies produce this kind of PQMS system because it is easy to integrate multiple functions such as A/D conversion, data processing, and communication into one hoard or equipment according to development of processing board technology.

Metering + Analyzing

Fig. 2. FQ monitorins system with decentralized pracessing a~cYtecfure

PQ Diagnosis

system

I; j
~

Fig. 3 shows PQMS with concentrated processing architecture that is advantageous to multiple point monitoring. The main idea is to use concise and cheap power quality meters (PQ meter) whose function is only to convert voltage and current to digital signal, and to concentrate data processing function to one power quality analyzer (PQA) that can accomplish high performance. Therefore, large amount of sampled raw data should be transferred from PQ meters to PQA. This is the disadvantage of this structure hecause communication burden is relatively heavy, hut current communication technology is improving fast enough to settle this problem. On the other hand, this structure is economical and competitive when many PQMSs are installed on a large scale because the price of each PQ meter is cheaper than PQMS with decentralized processing architecture [51.

Fig. 1. Interconnections of the overall power quality &la management system

The detailed functions of each individual component will

Metenng

+
Analynng

Fig 3 FQ mnitonng system with concentrated provssing ardutecture

he explained below. The software of PQMS has the process as illustrated in Fig.

B. Power Qualio Monitoring System

The hardware of PQMS can he divided into three parts. First one is A/D conversion part that converts voltage and current signal into digitized samples. Lately, for precise measurement, the sampling rate is more than 128 samples per cycle, that is, 7,680 Hz and the resolution is more than 16 hit. Second one is power quality analysis part that analyzes sampled data for the purpose of power quality event detection

4 and the main processing algorithms are characterizing and

PQ analysis [6]. First, characterizing is to calculate RMS values of voltage and current, harmonic components and frequency using various signal processing algorithms. For example, RMS values are measured in every half cycle, and harmonic components represented by the magnitude and phase angle of each individual harmonics are calculated using Fast Fourier Transform (FIT), especially 128-point 2-radix FFT [71. Furthermore, there are many algorithms about measuring frequency, for example, Discrete Fourier

1251

Authorized licensed use limited to: Saintgits College of Engineering. Downloaded on August 4, 2009 at 23:14 from IEEE Xplore. Restrictions apply.

Transform (DIT) [SI, Prony's estimation method [9], and Smart Discrete Fourier Transform @DIT) [IO]. The developed prototype system uses modified SDFT theory because this method gives accurate solution when frequency deviates from nominal value, its calculation speed is faster than DIT, and it is immune to harmonics.

down-sampled to 64 samples per cycle in order to relieve communication burden. This data are generated when voltage event happens or user of GUI system wants to see voltage and current waveforms. Harmonic raw data store each harmonic component from OLhorder to 50Ihorder harmonics.

C. GUISystem
-Event Detect

Fig. 4. Ihepracesr of PQ monitoring system

Second, PQ analysis is to calculate several indices and to classify PQ events according to IEEE standard 1159. Table I shows the output data of PQMS.
TABLE 1. STORE0 POWER QUALITY DATA A F l E R PQ MONITORING

Category Trend data

Event data

Raw data Harmonic data

Kind Vrms, Irms, power, energy, frequency, power factor, DF, THD, TEHD, TOHD, TDD, Kfactor, ilicker index, unbalance factor, etc. Sag, swell, interruption, undervoltage, overvoltage, unbalance, ilicker, transient, harmonics, over-frequency, under-frequency. event direction 64 samples per cycle 0-50'h harmonics

Trend data means the tendency of power quality variation for relatively long time period because it is generated twice in a second and averaged once in half hour. The items of trend D. Diagnosis System data are RMS value of voltage and current, power (unit: W, Var, and VA), energy (unit: Wh, Varh, and VAh), frequency, The process of PQ diagnosis system is illustrated in Fig. 5. power factor, total harmonic distortion (THD), total even As above-mentioned, the input of PQ diagnosis system is the harmonic distortion (TEHD), total odd harmonic distortion database that stores PQ data. PQ diagnosis system uses open (TOHD), distortion factor (DF), K-factor, frequency, database access (ODBC), so any database based on structured unbalance factor, and so forth. In addition, flicker indices query language (SQL) can be used, for example, Microsoft such as P,, and Pi, are calculated once every IO minute (PSJ SQL server, ORACLE, etc. and 2 hour (Pd. Event data are calculated according to IEEE definition. For Diagnosis Function PQ Inform. example, sag means l0-90% voltage reduction for 0.5 cycle -Event Tmnd Diagnosis -Graphs 1 minute. If voltage RMS decreases under sag condition, sag -Stochastic Diagnosis event data is generated. Each event data consists of event -Power Qualii Indexing -Rep& duration, event kind, start time, end time, mean value, and Fig. 5 . The pacess of FQ diagnosis system minor max value. In addition, event direction data presents the relative The main diagnosis functions are event trend diagnosis, location of the event source, that is, whether the source of stochastic diagnosis, power quality indexing, and event event is in the territory of the monitor or not, namely identification. The detailed functions will be explained in the upstream or downstream. Voltage event direction is decided next chapter. using various detection methods that utilize the current and The diagnosis data can he displayed through various power during or after the event. Harmonic event direction is graphs and reponed in documents. Especially, PQ diagnosis determined by the direction of harmonic power flow [ 111. system provides System Map that shows the configuration of Raw data are the voltage and current waveforms that are local power system and the interconnections of PQMSs. If PQ sampled 64 times per cycle. The voltage and current data of event happens, System Map notifies immediately the location PQ analysis algorithm in the PQ meter are 128 samples per of the PQMS that detects PQ event. Moreover, after event cycle. However, raw data transferred to GUI system are identification, the event source that deteriorates power quality

GUI system has three major functions: display function, PQMS management function, and DB management function. At first, GUI system displays various PQ trend graphs like voltage magnitude variation, simple PQ event data information, and the waveforms of voltage and current instantly. Secondly, because many PQMSs are connected to GUI system, GUI system should manage and control those PQMSs. GUI system sends standard time signal to PQMSs in order to synchronize the time of each PQMS to the time of GUI system. In addition, GUI system checks the status of PQMS periodically by communicating notification signal. Finally, GUI system manages data received from numbers of PQMSs. GUI system has five kinds of database such as point information database, trend database, event database, raw waveform database, and harmonic raw database. Point information database stores the information of the PQMSs like location, rated value, status and so on. The rest four database stores each kind data received from PQMS.

1258

Authorized licensed use limited to: Saintgits College of Engineering. Downloaded on August 4, 2009 at 23:14 from IEEE Xplore. Restrictions apply.

is indicated on System Map. Users can modify System Map easily because it is easy to add a new component to System Map. Furthermore, System Map has many components that represent the power system equipments such as transformer, distribution line, motor, inverter, general load and so on. In order to diagnoses more exactly, System Map needs some information about each power system component.

is used as an index that represents harmonic distortion of the site [12].

C. Power Quality Indexing

To evaluate PQ level, a number o f PQ indices have been developed. PQ indices can quantify complicated PQ problems and help customers understand PQ level simply. Therefore, these indices can be used as standard rules when customers contract electricity power with power companies. However, it is hard to catch all PQ phenomena exactly even in all exceptional cases, so pertinent PQ indices that tit the given PQ condition are needed. The representative conventional PQ index is SARFI, (System Average RMS variation Frequency Index) that represents the average number of specified RMS variation measurement events that occurred over the assessment period per customer served, where the specified disturbances are those with a magnitude less than x for sags or a magnitude greater than x for swells [131. Sag Score defined by Detroit Edison is more efficient to a contract. The Sag Score is obtained by subtracting average three-phase RMS voltage from 1.0 (PA.) 1141. However, this method does not consider the duration of the sag. The prototype PQ diagnosis system uses new voltage RMS index that is calculated by multiplying the voltage difference between contract RMS voltage and average three-phase RMS voltage by duration of the event. In addition, the prototype PQ diagnosis system provides also harmonic index and overall PQ event index. The harmonic index represents THD variation by subtracting average THD from contract THD, and the overall PQ event index can he calculated by weighing each event according to its priority. For instance, sag has 2 weights while interruption has 3 weights. These indices evaluate each PQ so simply that they are easy to be adjusted to the contract.

111. DIAGNOSIS FUNC~ONS

A. Event Trend Diagnosis

Event trend diagnosis function consists of hourly plot, weekly plot, monthly plot and time plot. Hourly plot, weekly plot, and monthly plot show the trend of events according to time, weekday, and month, respectively. For example, if the total number of events get higher during the daytime and early evening, and lower during the late evening and early morning, users can understand the fact that the diagnosed site has hourly trend through hourly plot. The above threc plots display the trend in the form of bar graph and indicate the peak time and value. With these plots, customers are assisted to identify the main cause of PQ events Additionally, customers can modify the patterns of the use of power to improve the PQ level of their site. In time plot, the number of events is displayed in the form of cumulative graph and the total number of events during the diagnosis period is presented. This plot gives customers clear information of the trend of events in time order. In addition, customers can find out the number of events occurred in a specific day.

B. Stochastic Diagnosis
Stochartic diagnosis function proffers standard charts, voltage event graph, and harmonic event graph. CBEMA, ITIC, and SEMI charts arc generally used as the standard of voltage supplied to computer based devices or semiconductor manufacturing equipments [2]. Voltage event graph analyzes sag and interruption that are most frequently occurred in the distribution system. Using the data of events, sag coordination chart that is proposed in IEEE Standard 493 and IEEE Standard 1346 is constructed. With this chart and voltage tolerance curve, the number o f trip per year of the equipment installed in the monitoring area can be estimated [2]. Therefore users can predict the extent of damage for voltage events. THD is usually measured to assess the harmonic distortion. In the harmonic event graph, the numbers of harmonic events are displayed in the form of bar graph according to the THD and the cumulative number is also displayed. CP95, which means the THD corresponding to 9 5 8 of cumulative number,

D. Event Identification
Event identification function has three categories such as event location identification, event cause identification, and solution suggestion as illustrated in Fig. 6.
SOlUtiOn

Suggestion

Fig. 6. The caago"es o f e v e 1 identiticalion

Event location identification that is to find the position of event source is very important to solve PQ problems because this function makes it possible to remove the faults rapidly and to find more suitable compensating method. To identify event location, PQ diagnosis system uses the event direction data generated by PQMS that are stored in database and system configuration data obtained from System Map. The more PQMSs are installed in the local system, the more exact

1259

Authorized licensed use limited to: Saintgits College of Engineering. Downloaded on August 4, 2009 at 23:14 from IEEE Xplore. Restrictions apply.

location can be obtained. Event source identification is also significant because this function gives the solution of improving PQ level. The prototype PQ diagnosis system identifies two kinds of events such as voltage RMS events and harmonic event. First, to identify the source of voltage RMS events, PQ diagnosis system calculates RMS continuously using sliding window from voltage waveform raw data, and then evaluates deviation of RMS to classify the faults. If, for example, the characteristics of fault correspond with the pattern only resulted from transformer fault, the kind of source is determined as transformer. Therefore, a lot of information about fault of equipment is needed for better determination. Event source identification algorithm is being studied especially in the recent years [151. Secondly, the source of harmonic event can he identified using harmonic raw data. For example, if the power level of the harmonics is high compared to rest harmonics and the components are 6 n A , the harmonic source may be six pulse switching devices [ 161. PQ diagnosis system also suggests the various solutions to improve PQ level like the sort and rating of various PQ compensating devices. This system helps users compare the PQ compensating devices and select the hest way they can use

event, so they can consider a counter plan to improve voltage quality

i i01 Melers . 3
PeakTlmE

Peak N u "

Period 200201 08-2002.1227

I
Fig. 8. The hourly plot far sag event

Fig. 9 shows the distribution of voltage RMS event on ITIC chart. The x and y axis of the charts represents the duration and the magnitude of voltage respectively. Events that are dotted outside the curve may he harmful to the device. Therefore, the number of these events is also displayed. In addition, other functions are programmed and tested.

Iv. TEST OF PROTOTYPE DEVELOPMENT

The prototype PQ data management system is constructed, and Fig. 7 shows the main screen of GUI system including PQ diagnosis system. The left window of main screen is the System Map editor and the right window shows the raw voltage waveform of sag event.

i i01 Meters ' 3

jj

Tmal h s m Numbrr: 11 ou d F " ' 13,

I
I/
'1

,
1

. .

:.......
..i

.......

~:
. i

......

j...~.. ~! ..........
;............

.....

........... ;...........
: :
~

: ............

:...

........

.................................

.............

I
I

...:.. . ...... .. .. . ... . . ..... .\.& ... ..i...-.. ......... u-r . *..

:\. , : I

o m

0 03

01

Period 2002 01 03 2002 1226

Fig 9 ITK Chan

V. CONCLUSION This paper represents the developed power quality diagnosis system that gives total solutions for power quality. Power quality diagnosis system has event trend diagnosis function, stochastic diagnosis function, power quality indexing function, and event identification function. Therefore, this system gives a lot of information to users to improve power quality. This paper explains the construction and functions of total power quality data management system including PQMS, GUI system, and diagnosis system. In

Fig. 8 shows the hourly plot for test sag data. Users can diagnose event trend for the any period that they like to diagnose. Then, users can understand the hourly trend of sag

1260

Authorized licensed use limited to: Saintgits College of Engineering. Downloaded on August 4, 2009 at 23:14 from IEEE Xplore. Restrictions apply.

addition, prototype PQ diagnosis system installed in total power quality data management system is tested and some functions are explained.

VI. REFERENCES
111 Badml H. Chowdhury, Power Quality, IEEE Potentials, V01.20. Issue: 2,AprRvlaypp. 5-11.2001 I21 Math H.J. Bollen. Under.xandinp Power Qunliry Problems: Volmge Saps and Interruptions, New York IEEE Press, 2000. 131 T.K. Abdel-Galil. E.F. El-Saadany. and M.M.A. Salam, Effect of New Deregulation Policy on Power Quality Monitoring and Mitigation Techniques. IEEE Transmission and Distribution Conference and Exporirion. V0l.l. pp. 554-560.2001. [41 Mark McGranaghan, Trends in Power Quality Monitoring, IEEE Power Engineering Review. Val.21. Issw: 10. Oct, pp.3-9 2001. I51 lmng-Moon k i m Dong-Jun Won, Il-Yop Chung. Sung-Wm Park. SeungII M m , Jang-Cheol Seo, and long-Wmng Chae. A New Concept on the System Architecture of the Power Quality Monitoring System hremorionol Conference on Eleclrical Engineering. pp.2243-2246,2002, I61 Dong-Iun Won, Il-Yop Chung. Jmng-Mwn KunSeung-II Moon, JangCheol Sea, and Jong-Woong Cha, Developmm of Power Quality Monitodng System with Cenml Processing Schem, IEEE PES Summer Meeting. pp. 915-919.2W2. I71 Alan V. Oppenheim and Ronald W. Schafer, Dircrere-Time Signal Processing. International Edition. Prentice-Hall International Inc. NI 1989. A New Measurement I81 A.G. Phadke. J.S.Thorp. and M.G. A&&, Technique for Tracking Voltage Phasars, Local System Frequency, and Rate of Change of Frequency. IEEE Tans. on Power Apparatus and Systems,Vol. PAS-102. No. 5, pp.1025-1038, May. 1983. I91 Tadeusz Lobos and Jacek R m n x , Real-Tim: Detemination of Pawer System Frequency, IEEE Tmnr. on lnstnrmentation and Meosuremenr, Vol. 46, No. 4, pp. 877-881, August 1997. I101 Jun.& Yan -e ~ L dChih-Wen~Liu A Prrcise Calculation of Power System Frequency. IEEE Trans. on Power Delivev, Vol. 16, No. 3, pp361-366. July. 2001, 1111 G.T. Heydt. Identification of H m n i c Sources by a state Esiition Technique, IEEE Trons. on Power D e l i v e y Vo1.4, No.1, January, pp.569-576, 1989. 1121 D. Daniel Sabin, Daniel L. Brmks, and Ashok Sundaram Indices far Assessing H m n i c Distortion f r o m Power Quality Measuremas: Definitions and Benchmark Data. IEEE Trons. on Power Delivery, Vol. 14. No.2, April, pp.489-496, 1999. 1131 D.L. Brooks. R.C. Dugan, Marek Waclawiak and Ashok Sundaram Indices for Assessing Utility Distribution System RMS Variation Performance, IEEE Trans. on Power Delivery, Vol. 13, No.1, pp. 254259, January 1998. I141 R.S. Thallam and G.T. Heydt. Power Acceptability and Voltage Sag Indices in the M Phase Sense, IEEE PES Summer Meeting, Vol. 2 pp. 905-910.2000. 1151 E m w i l Styvaklakis, Math H.J. Bollen, and k n c Y.H. Gu, Expen System for Classification and Analysis of Power System Events, IEEE Trans. on Power Delivery, Vo1.17. No.2, w.423-428. April 2002.
~~

1161 Asrat Teshom, H m n i c Source and Type Identification in B Radial Distribution System IEEE lndusrry Applirotion Sociery Annual Meeting. Vol. 2. pp1602-1609. 1991.

VII. BICGRAPHIES
11-Yop Chung was b m in KM-, an March 16, 1976. He received his B.S. and M.S. degrees in electrical engineering at Senul National University, Seoul, KM- in 1999 and 2001 respectively. He is cunenlly Ph. D candidate at Seoul National University. His s a c i a l field of interest includes power quality. custom p w e r device and dispersed generation.

Dong-Jun Won was bam in K M ~ on . January 1 , 197.5. He received the B.S. and M.S. degrees in electrical engineering at Seoul National University. Seoul. K M in ~ 1998 and 2000. respectively. Now he is U W.D. candidate at Seoul National University. His special field of interest includes power quality. FACTS and HVDC.
Jaong-Mwn Kim was hom in Korea, on November 30. 1971. He received the B.S. and M.S. degrees fromChnnhuk National University, Korea in 1996 and 1998. respectively. Now he is a Ph. D. candidate at Seoul National University. His special field of interest includes analysis and control of power system and power quality.

SeonJu Ahn was bom in Korea,an January 14, 1980. He received his B.S. degree in Electrical Engineering at Seoul National University, Seoul, Karea in 2002. He i s currently M S student at Seoul National University. His Special field of interest includes power quality and custom pwer device. Seung-Il Moon received the B.S. degree from S e a l National Univasity. Korea in 1985 and M.S. and Ph. D. degrees from The Ohio State University in 1989 and 1993. respectively. Currently, he is an Associate Professor of School of Electrical Engineering and Computer Science at Seoul National University. His special field of interest includes analysis. conVol and modelina of the mwer system FACTS and custom power.
Jang-Cheol Se0 received the B.S., M.S. and h . D . degrees from Seoul National University, Korea in 1993. 1995 and 2000, respectively. Now he is working at Elecuatechnolo~R&D Center. LG Industrial system Co.. Ltd. as research engineer. His research interests are analysis. modeling and control in FACTS.

Jong-Woong Choe received the B.S. degree from Pusan National University. Korea in 1981 and M.S.. Ph.D. degrees from ChungNam National University, Korea. He is a vice-president of Electrntechnolngy R&D Center, LG Industrial System Ca., Ltd. and U Board Member of Advanced Power Technology at LW, USA. His present research interests are the Power System Power Information N e t w ~ land ; Signal Processing.

1261

Authorized licensed use limited to: Saintgits College of Engineering. Downloaded on August 4, 2009 at 23:14 from IEEE Xplore. Restrictions apply.