Walsh Function Based Synthesis Method of PWM Pattern for Full-Bridge Inverter

Similar documents
Low Switching Frequency Active Harmonic Elimination in Multilevel Converters with Unequal DC Voltages

NATIONAL RADIO ASTRONOMY OBSERVATORY Green Bank, West Virginia SPECTRAL PROCESSOR MEMO NO. 25. MEMORANDUM February 13, 1985

antenna antenna (4.139)

RC Filters TEP Related Topics Principle Equipment

To: Professor Avitabile Date: February 4, 2003 From: Mechanical Student Subject: Experiment #1 Numerical Methods Using Excel

IEE Electronics Letters, vol 34, no 17, August 1998, pp ESTIMATING STARTING POINT OF CONDUCTION OF CMOS GATES

Parameter Free Iterative Decoding Metrics for Non-Coherent Orthogonal Modulation

Shunt Active Filters (SAF)

Figure.1. Basic model of an impedance source converter JCHPS Special Issue 12: August Page 13

Graph Method for Solving Switched Capacitors Circuits

A Comparison of Two Equivalent Real Formulations for Complex-Valued Linear Systems Part 2: Results

STUDY OF MATRIX CONVERTER BASED UNIFIED POWER FLOW CONTROLLER APPLIED PI-D CONTROLLER

Passive Filters. References: Barbow (pp ), Hayes & Horowitz (pp 32-60), Rizzoni (Chap. 6)

Harmonic Balance of Nonlinear RF Circuits

Control of Venturini Method Based Matrix Converter in Input Voltage Variations

The Performance Improvement of BASK System for Giga-Bit MODEM Using the Fuzzy System

PRACTICAL, COMPUTATION EFFICIENT HIGH-ORDER NEURAL NETWORK FOR ROTATION AND SHIFT INVARIANT PATTERN RECOGNITION. Evgeny Artyomov and Orly Yadid-Pecht

Digital Transmission

Adaptive Modulation for Multiple Antenna Channels

NMR Part II, Quadrature Detection

Section 5. Signal Conditioning and Data Analysis

Fuzzy Logic Controlled Shunt Active Power Filter for Three-phase Four-wire Systems with Balanced and Unbalanced Loads

Comparison of Reference Compensating Current Estimation Techniques for Shunt Active Filter

An Algorithm Forecasting Time Series Using Wavelet

Voltage Quality Enhancement and Fault Current Limiting with Z-Source based Series Active Filter

Uncertainty in measurements of power and energy on power networks

Calculation of the received voltage due to the radiation from multiple co-frequency sources

Rejection of PSK Interference in DS-SS/PSK System Using Adaptive Transversal Filter with Conditional Response Recalculation

Figure 1. DC-DC Boost Converter

A MODIFIED DIFFERENTIAL EVOLUTION ALGORITHM IN SPARSE LINEAR ANTENNA ARRAY SYNTHESIS

Figure 1. DC-DC Boost Converter

Dual Functional Z-Source Based Dynamic Voltage Restorer to Voltage Quality Improvement and Fault Current Limiting

Dynamic Optimization. Assignment 1. Sasanka Nagavalli January 29, 2013 Robotics Institute Carnegie Mellon University

DIMENSIONAL SYNTHESIS FOR WIDE-BAND BAND- PASS FILTERS WITH QUARTER-WAVELENGTH RES- ONATORS

PULSEWIDTH-modulated (PWM) voltage-source inverters

THE GENERATION OF 400 MW RF PULSES AT X-BAND USING RESONANT DELAY LINES *

Australian Journal of Basic and Applied Sciences

Optimal Placement of PMU and RTU by Hybrid Genetic Algorithm and Simulated Annealing for Multiarea Power System State Estimation

A Review of Multilevel Selective Harmonic Elimination PWM: Formulations, Solving Algorithms, Implementation and Applications

Application of Intelligent Voltage Control System to Korean Power Systems

Implementation of Fan6982 Single Phase Apfc with Analog Controller

Selective Harmonic Mitigation Technique for Cascaded H-Bridge Converters with Equal DC Link Voltages

A study of turbo codes for multilevel modulations in Gaussian and mobile channels

Understanding the Spike Algorithm

Impact of Interference Model on Capacity in CDMA Cellular Networks. Robert Akl, D.Sc. Asad Parvez University of North Texas

Compressive Direction Finding Based on Amplitude Comparison

29. Network Functions for Circuits Containing Op Amps

Research of Dispatching Method in Elevator Group Control System Based on Fuzzy Neural Network. Yufeng Dai a, Yun Du b

Power System State Estimation Using Phasor Measurement Units

Evaluate the Effective of Annular Aperture on the OTF for Fractal Optical Modulator

Target Response Adaptation for Correlation Filter Tracking

A new family of linear dispersion code for fast sphere decoding. Creative Commons: Attribution 3.0 Hong Kong License

Prevention of Sequential Message Loss in CAN Systems

The Application of Interpolation Algorithms in OFDM Channel Estimation

Network Reconfiguration in Distribution Systems Using a Modified TS Algorithm

Th P5 13 Elastic Envelope Inversion SUMMARY. J.R. Luo* (Xi'an Jiaotong University), R.S. Wu (UC Santa Cruz) & J.H. Gao (Xi'an Jiaotong University)

Analysis of Time Delays in Synchronous and. Asynchronous Control Loops. Bj rn Wittenmark, Ben Bastian, and Johan Nilsson

Current control with asymmetrical regular sampled pulse width modulator applied in parallel active filter

Design of Shunt Active Filter for Harmonic Compensation in a 3 Phase 3 Wire Distribution Network

University of Bristol - Explore Bristol Research. Peer reviewed version. Link to published version (if available): /ECCE.2016.

POLYTECHNIC UNIVERSITY Electrical Engineering Department. EE SOPHOMORE LABORATORY Experiment 1 Laboratory Energy Sources

INSTANTANEOUS TORQUE CONTROL OF MICROSTEPPING BIPOLAR PWM DRIVE OF TWO-PHASE STEPPING MOTOR

Inverse Halftoning Method Using Pattern Substitution Based Data Hiding Scheme

A Current Differential Line Protection Using a Synchronous Reference Frame Approach

Space Time Equalization-space time codes System Model for STCM

Harmonic Modeling of Inrush Current in Core Type Power Transformers using Hartley Transform

Research on Controller of Micro-hydro Power System Nan XIE 1,a, Dezhi QI 2,b,Weimin CHEN 2,c, Wei WANG 2,d

Test 2. ECON3161, Game Theory. Tuesday, November 6 th

NOVEL ITERATIVE TECHNIQUES FOR RADAR TARGET DISCRIMINATION

Mesh Equations. Introduction

A Novel Soft-Switching Two-Switch Flyback Converter with a Wide Operating Range and Regenerative Clamping

Control of Chaos in Positive Output Luo Converter by means of Time Delay Feedback

Digital Differential Protection of Power Transformer Using Matlab

Research on Peak-detection Algorithm for High-precision Demodulation System of Fiber Bragg Grating

Design and Implementation of DDFS Based on Quasi-linear Interpolation Algorithm

EE 508 Lecture 6. Degrees of Freedom The Approximation Problem

ANNUAL OF NAVIGATION 11/2006

FAST ELECTRON IRRADIATION EFFECTS ON MOS TRANSISTOR MICROSCOPIC PARAMETERS EXPERIMENTAL DATA AND THEORETICAL MODELS

MTBF PREDICTION REPORT

Controller Design Using Coefficient Diagram Methods for Matrix Converter Based Unified Power Flow Controllers

MASTER TIMING AND TOF MODULE-

Simulation of Distributed Power-Flow Controller (Dpfc)

High Speed, Low Power And Area Efficient Carry-Select Adder

PSO Selection of Surviving Nodes in QRM Detection for MIMO Systems

problems palette of David Rock and Mary K. Porter 6. A local musician comes to your school to give a performance

ECE315 / ECE515 Lecture 5 Date:

Optimal Allocation of Static VAr Compensator for Active Power Loss Reduction by Different Decision Variables

Latency Insertion Method (LIM) for IR Drop Analysis in Power Grid

Reconstruction of the roadway coverage parameters from radar probing measurements

AC-DC CONVERTER FIRING ERROR DETECTION

A High-Sensitivity Oversampling Digital Signal Detection Technique for CMOS Image Sensors Using Non-destructive Intermediate High-Speed Readout Mode

Available online at ScienceDirect. Procedia Computer Science 85 (2016 )

Generalized Incomplete Trojan-Type Designs with Unequal Cell Sizes

An Effective Approach for Distribution System Power Flow Solution

Network Theory. EC / EE / IN. for

Multi-transmitter aperture synthesis with Zernike based aberration correction

Model Reference Current Control of a Unipolar Induction Motor Drive

Error Probability of RS Code Over Wireless Channel

Design of RF and Microwave Filters

The Impact of Spectrum Sensing Frequency and Packet- Loading Scheme on Multimedia Transmission over Cognitive Radio Networks

Transcription:

Walsh Functon Based Synthess Method of PWM Pattern for Full-Brdge Inverter Sej Kondo and Krt Choesa Nagaoka Unversty of Technology 63-, Kamtomoka-cho, Nagaoka 9-, JAPAN Fax: +8-58-7-95, Phone: +8-58-7-957 E-mal: kondo@voscc.nagaokaut.ac.jp Abstract-Ths paper apples the Walsh functon to smplfy the on-off pattern synthess of a sngle phase full-brdge PWM nverter wth low harmoncs. Owng to the dgtal nature of the Walsh functon, the synthess process becomes straghtforward. It s the remarkable feature of the proposed method over a conventonal method usng Fourer seres. Several examples are shown n the paper, whch verfes the valdty of the proposed synthess method. Ⅰ. INTRODUCTION The Fourer seres expanson s conventonally used to derve optmal PWM pattern wth low harmoncs []. The trgonometrc functon used n the Fourer seres causes a problem. It s that an equaton ncludng trgonometrc functon requres a numercal teraton process to fnd ts soluton. The teraton loop shown n Fg. s necessary for searchng optmal PWM pattern wth low harmoncs and t takes long calculaton tme. FFT order, where n s the number of zero axs crossng for one cycle. There are many way to create wal(n,t). In ths paper, the Walsh functon wal(n,t) s ndrectly generated by the recurrence relaton of H( ): H(, t) = t otherwse. [ n/]+ p H ( n+ p, t) = (-) {H( n, ( t+ ))} + n+ p (-) { H( n, ( t ))}, n=,,,3,..., p=([n/]%)= or, where, [n/] s maxmum nteger that s smaller than n/, and (m%) s the remnder of m dvded by. Then the functon H( ) gves the Walsh functon wal( ) as: () () wal(n,t) = H(n+p,t). (3) FFT - - wal(,t) wal(,t)=sal(,t) wal(,t)=cal(,t) Fg. Iteraton loop Outlne for searchng optmal PWM pattern The Walsh functon has only two level sgnal of ± and s sutable for the applcaton to swtchng process []. As t wll be descrbed n secton Ⅲ, the equaton relatng the on-off swtchng tmng wth the Walsh spectrum results n a lner algebrac equaton whch can be solved drectly wthout any teraton process. Hence, the PWM pattern synthess usng the Walsh functon becomes straght-forward and ts computaton s smpler and faster than that usng trgonometrc functon. Ⅱ. FUNDAMENTALS OF WALSH FUNCTION A. Defnton of Walsh Functon The Walsh functon forms an ordered set of rectangular waveforms takng only two ampltude values, + and -, over one normalzed frequency perod [, ] as llustrated n Fg.. The Walsh functon, wal(n,t), has Walsh s orgnal - - - - - - wal(3,t)=sal(,t) wal(,t)=cal(,t) wal(5,t)=sal(3,t) wal(6,t)=cal(3,t) wal(7,t)=sal(,t) 3 Fg. Walsh Functon Smlar to the trgonometrc functon, the Walsh functon can be classfed nto odd-functon and even-functon: sal(m, t) = wal(m-, t) [odd-functon], cal(m, t) = wal(m, t) [even-functon]. () The sal() and cal() are correspondng to sne-walsh and cosne-walsh, respectvely. Unlke the sne and cosne functons n the Fourer seres, the sal() and cal() have rectangular waveforms. Ths knd of bnary nature of the Walsh functon makes the calculaton of the seres expanson smpler than that of the Fourer seres expanson. t

B. Walsh Seres Expanson An approprately contnuous functon x(t) can be expressed by the Walsh seres n the smlar way as the Fourer seres expanson: xt () = Awal(,) t + { Acal(,) t + Bsal (,)} t, (5) where, A = x t t ()d A = x()cal(,)d t t t B = x t t t () sal(,)d C. Comparson between Fourer and Walsh Expansons of PWM pattern In order to compare calculaton tme between Fourer and Walsh expansons, t s assumed that the f(t)s a PWM swtchng pattern of whch frst quarter cycle s shown n Fg. 3(b) and whole one cycle s shown n Fg. 3(a). The Fourer and Walsh seres-expansons can be derved as follows. (6) Walsh seres expanson: f() t = B sal(,) t 35,,,... B = f ()sal(,)dt t t B = f ()sal(,) t t dt+ f ()sal(,) t t dt for, B + 3 3 = ( ) ( ). (8) For the Fourer expanson, the computaton of C n (7) requres the calculaton of cos(), whch takes relatvely long tme. On the other hand, for the Walsh expanson, the computaton of B n (8) requres only plus/mnus operatons. Ths beneft of the Walsh functon can make the calculaton faster. By the computer smulaton result, t was found that the computaton tme rato of (8)/(7) was.9. Ⅲ. PROPOSED SYNTHESIS METHOD OF PWM PATTERN / 3/ /T / - (a). One cycle of Full-brdge PWM nverter 3 /T (b). Frst quarter cycle Fg.3 Example of PWM pattern Fourer seres expanson: f() t = C sn( ωt) 35,,,... C = f ()sn( t ω t ) dt C for, C = f ()sn( t ωt) dt + f ()sn( t ωt) dt { = (cos( ω ) cos( ω)) ω (cos( ω ) cos( ω )) 3 3 } (7) A. Outlne of Proposed Method Fg. llustrates the outlne of the PWM pattern synthess process proposed n ths chapter. The goal of the synthess s to make the PWM pattern close to the commanded output waveform whch s llustrated as a sne-wave n Fg.. In the followng, t s assumed that the commanded output waveform s a sne-wave to make the explanaton smple. At the frst step, the values of B, whch are the Walsh seres coeffcents of the commanded waveform, are calculated. Ths wll be descrbed n the secton-b of ths chapter. Secondary, the equaton whch relates the PWM on-off tmng to the Walsh coeffcents B s derved. Ths wll be gven n the secton-c. Here at ths pont, f B =B for =,3,5,...,, the PWM wave should concde wth the commanded wave. Ths s true n theoretcal sense. But ths knd of coeffcents matchng can not be acheved actually, because the swtchng frequency of the PWM nverter can not be nfnty. Therefore, we lmt the maxmum value of for the coeffcent matchng takng account of the swtchng speed of the power devce. At the last step, the swtchng tmng of the PWM pattern can be determned drectly from the Walsh seres coeffcents B of the commanded waveform, whch s based on a equaton derved n the secton-d of ths chapter.

C. Walsh seres expanson of PWM pattern Walsh seres expanson Walsh seres expanson B B B = B Fg. 3 shows an example of whole one cycle of Full-brdge PWM nverter waveform whch has 3-level. By usng the symmetrcal property of the PWM pattern, t s enough to calculate only the frst quarter cycle. The PWM pattern s settng to be f(t) whch s the functon of. [ ] = N [ K ] [ B ] M [ T ] Fg. Outlne of propose PWM pattern synthess method B. Walsh Seres Expanson of Commanded Waveform In order to make the synthess smple, t s assumed that the commanded waveform of the PWM nverter output has the property of quarter-wave symmetry. As an example, the case when the output command equals to Msn( π t), where M s modulaton rato, wll be explaned n what follows. The Walsh seres expanson s gven as: Msn( π t) = { MB sal (, t)}, (9) = B = sn( πt) sal(, t) dt = 3,,,,... () Snce the output command s assumed to be the sne-wave, n the rght hand sde of (9), the coeffcent B - remans but the coeffcents A and A and B appeared n (5) and (6) become zero. The calculated values of B - for M= are shown n Fg. 5. B- B Ampltude B Ampltude B B 3 B 5 B 7 B 9 B B 3 B 5 6.37963e- -.67569e- -5.8833e- -.79e- -.565e- 6.8687577e-3 -.7638e- -6.76e- B 7 B 9 B B 3 B 5 B 7 B 9 B 3 -.539788e-3.833e-3 -.73538e-3.568e-3-7.88793e-3.3397e-3 -.5966e- -.9599e-.6.5..3... 3 order of harmonc - Fg. 5. Calculated values of coeffcents B to B 3 (M=) f () t = where t /. t = [, ],[ 3, ],... otherwse The Walsh seres expanson of f(t) s gven as: () f() t = { B sal(,)} t, () B = f ( t ) sal (, t ) dt = 3,,,,... D. Proposed Synthess Method of PWM pattern (3) In order to derve the synthess method of the on-off swtchng functon of the PWM nverter, the frst quarter cycle of the output command s dvded nto N subntervals, where N should be chosen to be an nteger of power of two. The sample quarter cycle of the PWM swtchng functon f(t) for N= s shown n Fg.6. - - - - K K K 3 K K K K3 K K 3 K3 K 33 K 3 K K K 3 K f(t) 3 T T T 3 T Fg.6 PWM desgn (N=) sal(,t) sal(3,t) sal(5,t) sal(7,t) PWM To make the swtchng functon f(t) close to the output command Msn( π t), the B n () should be set equal to the MB n (9), that s, MB = B. () As shown n Fg.6, the value of sal functon s + or - for N subntervals, whch makes the calculaton smple. Lettng the value of the sal functon for each subnterval be K j :

K j = j sal(, ). (5) N Accordng to the property of sal(), the followng relatons hold: Kj = Kj, (6) N ( Kj ) = N. (7) 3,,,... From (3) and (), the followng equaton holds: M B = f () t sal(,) t dt N d N N = K f ( t) t + K f ( t) dt+... N N ( N ) N N d NN d. (8) N + K f () t t + K f () t t Substtutng () nto (8), the followng equaton can be obtaned: M B K T K T where, T = ( ) + ( ) +... + K( N ) ( TN N ) + KN( N TN ) = K( ) + K( ) +... + K( N ) ( N ) + KN( N) + KT+ K( T ) +... + K( N ) ( TN ) + KN( TN ) (9) N =,,3,...,N/. () Rewrtng (9) n matrx form, B M B3... B N K K... K K K... K =... K K... K K K... K K K... K... KN KN... K N N N N NN N N N NN +... T T... TN. () The smple form of the above equaton s gven as: [ ] [ ] [ ] [ ] [ ] M B = K + K T. () By the property of the Walsh functon, the [K] matrx s a complete orthogonal matrx, and the relatons (6) and (7) hold. Takng these propertes nto account, the nverse matrx of [K] can be smply calculated as: [ K] = N [ K]. (3) Solvng () wth respect to the [] and usng (3), the swtchng tmng vector [] can be calculated as: [ ] = [ ] [ ] [ ] N K B M T. () The proposed synthess method s summarzed as follows: step-. calculate B - matrx by (), or use data lsted n Fg. 5 f N less than 6. step-. calculate T - by (), step-3. calculate [K] by (5), step-. then, determne the [] by (). Ⅳ. SIMULATION RESULTS The harmoncs of the proposed method and the conventonal trangular wave modulaton method wll be compared n the condton when the swtchng frequences of both methods are kept the same. A. The Case for Pure Sne-Wave Output In the frst quarter of one cycle the number of swtchng angle of both methods s set to N. For N= the calculaton of the Walsh s calculaton method and the conventonal trangular wave modulaton method are shown n Fg.7 and Fg.8 respectvely. For N=6 the calculaton of the Walsh s calculaton method and the conventonal trangular wave modulaton method are shown n Fg.9 and Fg.. As shown n Fg.7(c) and Fg.9(c), the relaton between modulaton rato and swtchng angle of the Walsh s calculaton method are lnear. Comparng Fg.7(b) and Fg.9(b), the calculaton wth larger N results n less harmoncs shfted to hgher frequency range. Comparng Fg.7(b) and Fg.8(b), the Walsh s calculaton method can decrease the harmoncs (3 rd,5 th,7 th ) more than the conventonal trangular wave modulaton method. For the calculaton by N=6, the harmoncs of both method are almost the same as shown n Fg.9(b) and Fg.(b). B. The case for the thrd harmonc ncludng In ths case, the fundamental wave ncludng the thrd harmonc s a command waveform for a sample smulaton. The ampltude of the thrd harmonc s settng to /6 on the fundamental component, then the command waveform s sn(πt)+/6sn(6πt). Both calculaton methods are compared n the same condton when N=6. The Walsh s calculaton method s shown n Fg. and the conventonal trangular wave modulaton method n Fg.. The harmoncs of both calculatons method are shown n TableⅠ.

..5..5..5 (a). PWM waveform (M=.). Ampltude Modulaton rato M.8.6... 5 5 5 3 35 5 5..9.8.7.6.5. Fourer s harmonc order (b). Fourer spectrum of waveform (a)...5..5..5 Quarter cycle of swtchng angle () (c). Relaton between modulaton rato M and relaton Fg. 7 Walsh s calculaton waveform when N=..5..5..5 (a). PWM waveform (M=.). Ampltude Modulaton rato M.8.6... 5 5 5 3 35 5 5..9.8.7.6.5. Fourer s harmonc order (b). Fourer spectrum of waveform (a)...5..5..5 Quarter cycle of swtchng angle ( ) (c).relaton between modulaton rato M and relaton Fg. 9 Walsh s calculaton waveform when N=6..5..5..5 Ampltude Modulaton rato M..8.6.. (a). PWM waveform (M=.). 5 5 5 3 35 5 5..9.8.7.6.5. Fourer's harmonc order (b). Fourer spectrum of waveform (a)...5..5..5 Quarter cycle of swtchng angle ( ) (c). Relaton between modulaton rato M and relaton Fg. 8 Trangular s modulaton waveform when swtchng angle=...5..5..5 Ampltude Modulaton rato M..8.6.. (a). PWM waveform (M=.). 5 5 5 3 35 5 5..9.8.7.6.5. Fourer's harmonc order (b). Fourer spectrum of waveform (a)...5..5..5 Quarter cycle of swtchng angle () (c). Relaton between modulaton rato M and relaton. Fg. Trangular s modulaton waveform when swtchng angle=6

..5..5..5 (a). PWM waveform. Ampltude.8.6... 5 5 5 3 35 5 5 Fourer's harmonc order (b). Fourer spectrum of waveform (a). Fg. The Walsh s calculaton for 3 rd harmonc ncludng when N=6..5..5..5 (a). PWM waveform. Ampltude.8.6... 5 5 5 3 35 5 5 Fourer's harmonc order (b). Fourer spectrum of waveform (a). harmoncs n lower length of the center (5 th to 3 st ) are smaller but the ampltude of harmoncs n the hgher length of the center (33 rd to 3 rd ) are larger than the ampltude of harmoncs calculated by trangular wave modulaton method. Harmoncs (th) 3 5 7 9 3 5 7 9 3 5 7 9 3 33 35 37 39 3 5 7 9 TABLE Ⅰ Fourer spectrum Walsh s calculaton.996.69.3....3.3.....6.9.9..8..5.35..3... Trangular s modulaton..67..........9.33.3.....3.33.9....3 Fg. Trangular waveform modulaton for 3 rd harmonc ncludng when N=6 From TableⅠ, by the calculaton of the trangular wave modulaton method we can verfy that the ampltude of fundamental and thrd harmonc matches to the command waveform. The ampltude of harmoncs n swtchng length are symmetry, when the 3 st and 33 rd harmoncs are suppose to be a center of swtchng. By the calculaton of the propose method, we found the ampltude of fundamental and thrd harmonc does not match to the command waveform. That s because we lmt the number N of Walsh seres coeffcents (B ) to desgned the PWM pattern. For matchng the ampltude of fundamental and/or the thrd harmonc, we should adjust nfntely large number of Walsh seres coeffcents. Also, the ampltude of harmoncs n swtchng length are not symmetry as the ampltude of harmoncs calculated by the trangular wave modulaton method. The ampltude of Ⅴ. CONCLUSION Ths paper proposes the algorthm to desgn full-brdge PWM pattern wth low harmoncs by usng the Walsh functon. The proposed algorthm process becomes straghtforward, n opposton to the algorthm process of the conventonal method usng Fourer seres, whch requres an teraton process for searchng the PWM pattern. The Walsh functon, whch has only two level sgnals of + and -, s sutable to desgn an algorthm process smpler and faster than that usng trgonometrc functon. REFERENCES [] I. Takahash, H. Mochkawa: A New Control of PWM Inverter Waveform for Mnmum Loss Operaton of an Inducton Motor Drve, IEEE Trans. on IA, Vol. IA-, No., p. 58 (985) [] T.J. Lang and R.G. Hoft Walsh Functon Method of Harmonc Elmnaton, pp.87-853, APEC conf., March 993 proceedngs.