MODULAR DECOMPOSITION OF THE NOR-TSUM MULTIPLE-VALUED PLA

Similar documents
Revision: August 19, E Main Suite D Pullman, WA (509) Voice and Fax

Revisiting the Socrates Example

Bootstrap Method > # Purpose: understand how bootstrap method works > obs=c(11.96, 5.03, 67.40, 16.07, 31.50, 7.73, 11.10, 22.38) > n=length(obs) >

ENSC Discrete Time Systems. Project Outline. Semester

ENG2410 Digital Design Sequential Circuits: Part B

5 th grade Common Core Standards

Sections 15.1 to 15.12, 16.1 and 16.2 of the textbook (Robbins-Miller) cover the materials required for this topic.

A proposition is a statement that can be either true (T) or false (F), (but not both).

MODULE FOUR. This module addresses functions. SC Academic Elementary Algebra Standards:

ENG2410 Digital Design Arithmetic Circuits

ECEN 4872/5827 Lecture Notes

Medium Scale Integrated (MSI) devices [Sections 2.9 and 2.10]

Differentiation Applications 1: Related Rates

Homology groups of disks with holes

Dead-beat controller design

ECE 2100 Circuit Analysis

NUMBERS, MATHEMATICS AND EQUATIONS

Chapter Summary. Mathematical Induction Strong Induction Recursive Definitions Structural Induction Recursive Algorithms

Admissibility Conditions and Asymptotic Behavior of Strongly Regular Graphs

Department of Electrical Engineering, University of Waterloo. Introduction

The standards are taught in the following sequence.

NOTE ON APPELL POLYNOMIALS

Verification of Quality Parameters of a Solar Panel and Modification in Formulae of its Series Resistance

CHAPTER 2 Algebraic Expressions and Fundamental Operations

13. PO TREATMENT OF DEPT (DISTORTIONLESS ENHANCEMENT POLARIZATION TRANSFER)

Cambridge Assessment International Education Cambridge Ordinary Level. Published

B. Definition of an exponential

Determining the Accuracy of Modal Parameter Estimation Methods

MODULE ONE. This module addresses the foundational concepts and skills that support all of the Elementary Algebra academic standards.

SUPPLEMENTARY MATERIAL GaGa: a simple and flexible hierarchical model for microarray data analysis

Chapter 3 Kinematics in Two Dimensions; Vectors

Preparation work for A2 Mathematics [2017]

CHAPTER 3 INEQUALITIES. Copyright -The Institute of Chartered Accountants of India

Physics 2B Chapter 23 Notes - Faraday s Law & Inductors Spring 2018

Chapter 9 Vector Differential Calculus, Grad, Div, Curl

A Matrix Representation of Panel Data

Theoretical study of third virial coefficient with Kihara potential

Lecture 13: Electrochemical Equilibria

Section I5: Feedback in Operational Amplifiers

Enhancing Performance of MLP/RBF Neural Classifiers via an Multivariate Data Distribution Scheme

initially lcated away frm the data set never win the cmpetitin, resulting in a nnptimal nal cdebk, [2] [3] [4] and [5]. Khnen's Self Organizing Featur

7 TH GRADE MATH STANDARDS

Supplementary Course Notes Adding and Subtracting AC Voltages and Currents

ENG2410 Digital Design Sequential Circuits: Part A

REPRESENTATIONS OF sp(2n; C ) SVATOPLUK KR YSL. Abstract. In this paper we have shown how a tensor product of an innite dimensional

Thermodynamics and Equilibrium

Computational modeling techniques

A new Type of Fuzzy Functions in Fuzzy Topological Spaces

CS 477/677 Analysis of Algorithms Fall 2007 Dr. George Bebis Course Project Due Date: 11/29/2007

Keysight Technologies Understanding the Kramers-Kronig Relation Using A Pictorial Proof

1 The limitations of Hartree Fock approximation

Preparation work for A2 Mathematics [2018]

Building to Transformations on Coordinate Axis Grade 5: Geometry Graph points on the coordinate plane to solve real-world and mathematical problems.

MODULE 1. e x + c. [You can t separate a demominator, but you can divide a single denominator into each numerator term] a + b a(a + b)+1 = a + b

Chapter 3: Cluster Analysis

Modeling the Nonlinear Rheological Behavior of Materials with a Hyper-Exponential Type Function

1996 Engineering Systems Design and Analysis Conference, Montpellier, France, July 1-4, 1996, Vol. 7, pp

[COLLEGE ALGEBRA EXAM I REVIEW TOPICS] ( u s e t h i s t o m a k e s u r e y o u a r e r e a d y )

Supplementary Course Notes Adding and Subtracting AC Voltages and Currents

Technical Bulletin. Generation Interconnection Procedures. Revisions to Cluster 4, Phase 1 Study Methodology

2.161 Signal Processing: Continuous and Discrete Fall 2008

Perturbation approach applied to the asymptotic study of random operators.

Lecture 6: Phase Space and Damped Oscillations

Least Squares Optimal Filtering with Multirate Observations

Current/voltage-mode third order quadrature oscillator employing two multiple outputs CCIIs and grounded capacitors

This is the most commonly used defuzzification technique. In this method, the overlapping area is counted twice.

Inflow Control on Expressway Considering Traffic Equilibria

FINITE BOOLEAN ALGEBRA. 1. Deconstructing Boolean algebras with atoms. Let B = <B,,,,,0,1> be a Boolean algebra and c B.

Distributions, spatial statistics and a Bayesian perspective

A Few Basic Facts About Isothermal Mass Transfer in a Binary Mixture

ChE 471: LECTURE 4 Fall 2003

On Topological Structures and. Fuzzy Sets

YEAR 6 (PART A) Textbook 6A schema

ALE 21. Gibbs Free Energy. At what temperature does the spontaneity of a reaction change?

An Introduction to Matrix Algebra

COMP 551 Applied Machine Learning Lecture 9: Support Vector Machines (cont d)

Section 6-2: Simplex Method: Maximization with Problem Constraints of the Form ~

, which yields. where z1. and z2

AMERICAN PETROLEUM INSTITUTE API RP 581 RISK BASED INSPECTION BASE RESOURCE DOCUMENT BALLOT COVER PAGE

the results to larger systems due to prop'erties of the projection algorithm. First, the number of hidden nodes must

On Boussinesq's problem

Lecture 17: Free Energy of Multi-phase Solutions at Equilibrium

ECE 2100 Circuit Analysis

Pattern Recognition 2014 Support Vector Machines

Name: Period: Date: ATOMIC STRUCTURE NOTES ADVANCED CHEMISTRY

Last Updated: Oct 14, 2017

Lead/Lag Compensator Frequency Domain Properties and Design Methods

THE FINITENESS OF THE MAPPING CLASS GROUP FOR ATOROIDAL 3-MANIFOLDS WITH GENUINE LAMINATIONS

Unit 2 Expressions, Equations, and Inequalities Math 7

Copyright Paul Tobin 63

Surface and Contact Stress

Level Control in Horizontal Tank by Fuzzy-PID Cascade Controller

Lyapunov Stability Stability of Equilibrium Points

CHAPTER 4 DIAGNOSTICS FOR INFLUENTIAL OBSERVATIONS

Designing Polymorphic Circuits with Periodical Weight Adjustment

Floating Point Method for Solving Transportation. Problems with Additional Constraints

Tree Structured Classifier

Sequential Allocation with Minimal Switching

This section is primarily focused on tools to aid us in finding roots/zeros/ -intercepts of polynomials. Essentially, our focus turns to solving.

Transcription:

MODUAR DECOMPOSITION OF THE NOR-TSUM MUTIPE-AUED PA T. KAGANOA, N. IPNITSKAYA, G. HOOWINSKI k Belarusian State University f Infrmatics and Radielectrnics, abratry f Image Prcessing and Pattern Recgnitin. Brvky St., 6, 6, Minsk, Republic f Belarus, Phne: (37517) 49-19-81, Fax: (37517) 495-16, E-mail: jack@expert.belpak.minsk.by r pttsina@risq.belcaf.minsk.by k Institute f Cmputer Science & Infrmatin Systems, Technical University, ul. Zlnierska 49, Szczecin, Pland, Fax. (+4891) 48-76439, E-mail: hlwins@dedal.man.szczecin.pl ABSTRACT. A methd fr designing PA-based cmbinatinal circuits by mdular decmpsitin is presented. Main subjects are 1) Specific prperties f TSUM peratr, ) MIN-TSUM and NOR-TSUM expansins with respect t the bund set, X 1 f variables, 3) Realizatin f functins by multiple-valued PA-based cmbinatinal circuits, 4) Cmparisn with ther methds. Experimental investigatins shw that the size f suggested cmbinatinal circuit is the same as the size f multiple-valued PA implementing a multiple-valued lgic functin with large number f variables. KEY WORDS. Multiple-valued lgic, mdular decmpsitins, lgic design, multiple-valued prgrammable lgic array (M PA) 1. Intrductin Multiple-valued lgic is ne f the active area f research. It is knwn that multiple-valued lgic allws intercnnectins t be reduced int a chip, as well as a reductin in the numbers f inputs and utputs. Different M PA structures t implement the multiple-valued lgic (M) functins have been prpsed 1,, 3, 4]. These M PA's realize the lgic functins using the MIN (r-valued minimum), MAX (r-valued maximum), MODSUM (mdular sum), linear SUM and TSUM (truncated sum) arrays. The number f inputs, utputs and prduct terms f suggested M PA's have physical restrictins. These structures d nt permits the M functins with the large number f variables t be implemented because f the number f inputs is smaller than the number f input variables. The slutin f this prblem is t use the functinal decmpsitin technique. The decmpsitin f a lgic functin can lead t a systematic design which incrprates readily designed simpler functins. Ideally, a systematic design wuld psses such desirable prperties as regularity and testability. One apprach t decmpsitin f multiple-valued lgic functins (M functins) is t extend disjunctive decmpsitin fr the binary case 5]. The main drawback f this apprach is fllws. Since nly a very small number f functins will have a specific decmpsitin prperty 5], if the functin des nt have the desired prperty, the analysis des nt yield a design. In ther hand, the specific prperty f M functins is that an arbitrary lgic functin can be given by different set f multiple-valued peratrs. These determines the basic multiple-valued gates frm which a cmplex digital circuit implemented given M functin is synthesized. In this paper a NOR-TSUM M PA, prpsed by Pelay and etc., is assumed t be the cmpnent used t implement the functin. A decmpsitin apprach fr the mdular design f cmbinatinal multiple-valued circuits based n the NOR-TSUM PA is described. The primary bjective f this apprach is t develp prcedures which are applicable t the design f PA-based cmbinatinal circuits and t extent the class f decmpsable functins. The rest f the paper is structured as fllws. Sectin prvides definitins and ntatins, the distinctive features f TSUM peratr are investigated in Sectin 3, Sectin 4 is devted t MIN-TSUM and NOR-TSUM decmpsitins f the M functin; the M PA - based technique is explained in Sectin 5, experimental results are disscussed in Sectin 6, the Sectin 7 draws a cnclusin.. Definitin and ntatins et us examine sme r-valued peratrs by which we can represent any r-valued lgic functin. et x and y be tw r-valued variables. Then, (a) (b) x Λ y=x y=min(x, y); x y=tsum(x+y, );, if x = s,

(c) x s =, therwise, s {,1,...,}. A prduct f the literals x σ 1 1 x σ... x σ n n, is reffered t as a prduct term (als called term r prduct fr shrt). In a prduct term, all the peratins are AND peratins. A prduct term that includes literals fr all variables x 1, x,..., x n is called a full term. An r-valued minimum f term and r-valued cnstant, f(σ) x 1 σ 1 x σ... x n σ n is dented as an r-valued minimum term (als called r-valued term). An r-valued term that cntains literals fr all variables x 1, x,..., x n is called a full r-valued term. A truncated sum f r-valued terms is called a truncated sum-f-minimums (TSOM) expressin: ( 1) f(x) = l f(σ) x σ 1 1 x σ... x σ n n where the symbl l stands fr TSUM peratr, f(σ) is the value f functin evaluated with X=σ, σ=(σ 1 σ...σ n ) r, f(σ) E r ={, 1,..., }, x i E r, σ i E r. In ther hand, this expressin may be written as fllws 1]: n ( ) f(x) = l f(σ)λ( x i σ i ). i=1 where the symbl stands fr binary NOR peratr. Expressin ( ) states that any r-valued functin may be expanded as a truncated sum f binary NOR terms weighted by the r-valued cnstants f(σ). Definitin 1. et X is the set f r-valued variables. {X 1, X } is a partitin f X when X 1 X = and X 1 X =X. T represent r-valued functins with minimal expressins, the fllwing ntatins are adpted: E r : the set f cnstants is an r-valued Rsser algebra, E r ={, 1,..., } X : the set f n r-valued variables {x 1,x,...,x n }, where n is the number f elements f X. X 1 : the bund set f variables; X : the free set f variables; d(x): the number f elements f the set X. σ : (σ 1 σ σ n ), where σ i E r, i=1,,..., n. τ : (τ 1 τ τ n1 ) σ, where τ i E r, i=1,,...,n 1, d(x 1 )=n 1. η : (η 1 η η n ) σ, where η i E r, i=1,,...,n, d(x )=n. Definitin. et f(x) be an r-valued functin and {X 1, X } is a partitin f X. Then the prjectin f f(x) ver X 1 =τ, f(τ, X ) is the value f f(x) evaluated with X 1 =τ. Definitin 3. An r-valued functin f(x) is said t have a generalized decmpsitin with a bund set X 1 and a free set X if there exist r-valued functins h 1,h,...,h k and g such that ( 3) f(x)=g(h 1 (X 1 ), h (X 1 ),..., h k (X 1 ), X ) where {X 1, X } is a partitin f X. n 1 n X 1 H k h 1 h k n G g(h 1 (X 1 ),..., h k (X 1 ), X ) X H k f(τ, X ) f(τ, X ) n 1 G g(h 1 (X 1 ),..., h k (X 1 ), X ) X X 1 (a) Fig. 1 Circuit diagram f f(x)=g(h 1 (X 1 ), h (X 1 ),..., h k (X 1 ), X ). (a) h i is cded as r-valued variables, (b) h i =f(τ, X ) is examined as the prjectins f f(x) ver X 1 =τ. There are tw ways t realize generalized decmpsitin f M functins. The first way is applied in classical decmpsitin methds. Here, r-valued functins h i are cded as r-valued variables f g (Fig. 1(a)). By secnd way, r-valued functins h 1 (X 1 ), h (X 1 ),..., h k (X 1 ) can be examined as the prjectin f f(x) ver X 1 =τ, f(τ,x ) (Fig. 1(b)). In this paper the circuit diagram f f(x)=g(h 1 (X 1 ), h (X 1 ),..., h k (X 1 ), X ), shwn in Fig. 1 (a), will be cnsider mre detail. (b)

3. The specific prperties f TSUM peratr In this sectin we derive the specific prperties f TSUM peratr. Based n these features the MIN- TSUM and NOR-TSUM expansins with respect t a grup f variables (Therem 1 and Therem ) are btained which will be used in the next sectin fr designing the PA-based cmbinatinal circuits. et x, y and z be r-valued lgic variables. The fllwing peratins hld fr multiple-valued algebra: 1. Assciative laws: y) x= (y z). Cmmutative laws: y = y 3. Identities: 3A. () x= 3B. = 3C. 1... = 3D. 1 Λ 1 Λ... () Λ =x 4. Distributive laws: 4A. Λy) z=( z)(y z) 4B. y)λz (Λz) (yλz) 4C. y)λz i =(Λz i ) (yλz i ). All specific prperties f TSUM peratr are prved in Appendix 1. 4. Decmpsitin f NOR-TSUM and MIN-TSUM expansins In this sectin we will describe the new representatins f NOR-TSUM and MIN-TSUM expansins ver the bund set, X 1 f variables. The Q σ (X) term is a binary functin defined by the Blean prduct f literals and taking values in the {, } set: Q σ (X) = x σ 1 1 x σ... x σ n n Therem A. et f(x) be an r-valued lgic functin and {X 1, X } be a partitin f X. et n 1 =d(x 1 ) and n =d(x ). Then the MIN-TSUM expansin f f(x) ver X 1 is given by r n1-1 ( 4) f(x) = l Q τ (X 1 ) l f(τ,η)λq η (X ) τ= η= where f(σ) =f(τ,η) is the value f f(x) evaluated with X 1 =τ and X =η, i.e. simple the value f f(x) fr σ, σ=(σ 1 σ...σ n ) r be the input cmbinatin f X, σ={τ, η}, Q σ (X) = Q τ (X 1 )ΛQ η (X ). (Prf) Transfrm expressin ( 1) using cnjunctin and truncated sum prperties. By definitin f Q σ (X) term, expressin ( 1) can be represented as: f(x) = l f(σ)λq σ (X) Because {X 1, X } is a partitin f X, the functin Q σ (X) can be expressed as Q σ (X)=Q τ (X 1 )ΛQ η (X ), where σ={τ, η} is the input cmbinatin f X and τ and η are the input cmbinatins f X 1 and X, respectively. Therefre, f(x) = l f(σ) Λ Q τ (X 1 ) Λ Q η (X ) Bth cnjunctin and truncated sum are cmmutative f(x) = l Q τ (X 1 )Λ f(σ) Λ Q η (X ) ( 9), then Nte, that Q τ (X 1 ) and Q η (X ) are the binary functins and f(σ) takes the value f, 1,..., (). Because σ={τ, η} and bth cnjunctin and truncated sum are submitted t all distributive laws ( 14, ( 16), s r n1-1 f(x) = l Q τ (X 1 ) l f(τ,η)λq η (X ) τ= η=

Hence, the therem is prved. Q.E.D. Nte that the prjectin f f(x) ver X 1 fr MIN-TSUM expansin is determined as fllws: f(τ, X ) = l f(τ,η)λq η (X ) η= Example A. Cnsider a 3-valued 3 variable lgic functin f(x 1,x,x 3 ) (r=3, n=3), shwn in Table 1. Than the MIN-TSUM expansin with respect t x 1, x variables fr this functin is given as x 1 x 3 x f(x)= x 1 x ( x 3 x 1 3 1 x 3 ) x 1 1 x (1 x 3 1 x 1 3 1 x 3 ) x 1 x ( x 3 x 1 3 x 3 ) x 1 x 1 ( x 3 1 x 1 3 x 3 ) x 1 1 x 1 ( x 3 x 1 3 x 3 ) x 1 x 1 (1 x 3 x 1 3 1 x 3 ) x 1 x (1 x 3 x 1 3 1 x 3 ) x 1 1 x (1 x 3 x 1 3 x 3 ) x 1 x ( x 3 x 1 3 x 3 ). (End f example) Therem B. et f(x) be an r-valued lgic functin and {X 1, X } be a partitin f X. et n 1 =d(x 1 ) and n =d(x ). Then the expansin f f(x) ver X 1 fr NOR-TSUM expressin is given by r n1-1 n 1 n ( 5) f(x) = l ( x j τ j ) l f(τ,η)λ( x i η i ) τ= j=1 η= i=1 where the symbl stands fr binary NOR peratr. (Prf) By therem a any r-valued functin can be represented as fllws: r n1-1 ( 6) f(x) = l Q τ (X 1 ) l f(τ,η)λq η (X ) τ= η= Each term in previus expressin may be cnverted as fllws 1] ( 7) Q τ (X 1 ) = x 1 τ 1 Λx τ Λ... Λx n1 τ n1 = x 1 τ 1 x τ... x n1 τ n1 = x i τ i i=1 Taking int accunt the expressin ( 7) we can make the substitutin in ( 6). Hence we have a therem. Q.E.D. Nte that the prjectin f f(x) ver X 1 fr NOR-TSUM expansin is determined as fllws: n ZebpZ 1 Decmpsitin table fr f(x 1, x, x 3 ) 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 f(τ,x ) = l f(τ,η)λ( x i η i ) η= i=1 Example B. The NOR-TSUM expansin with respect t x 1, x variables fr functin described in example a is ffered as f(x)= x 1 x Λ( Λx 3 Λx 3 1 Λx 3 ) x 11 x Λ(1 Λx 3 Λx 3 1 Λ x 3 ) x 1 x Λ(1Λ x 3 Λ x 3 1 1Λ x 3 ) x 1 x 1 Λ(1 Λx 3 Λ x 3 1 1 Λx 3 ) x 11 x 1 Λ( Λx 3 Λ x 3 1 Λ x 3 ) x 1 x 1 Λ(Λ x 3 1Λ x 3 1 Λ x 3 ) x 1 x Λ( Λx 3 Λx 3 1 Λ x 3 ) x 11 x Λ(1Λ x 3 1Λ x 3 1 1 Λx 3 ) x 1 x Λ( Λx 3 Λx 3 1 1 Λx 3 ). (End f example) 5. Design methd fr M PA-based cmbinatinal circuits In this sectin we present a design methd fr M PA based cmbinatinal circuits using decmpsitin f NOR-TSUM and MIN-TSUM expansins. This nvel methd incrprates the cncept f systematically reexpressing the given M functin. In ur apprach, we d nt attempt t rewrite the functin in terms f functins which are then used as new variables in the lgic equatin. Rather, we identify the minimum number f prjectins f f(x) ver X 1 and reexpressed the given functin by btained partitin f X. Each f fucntin prjectin is then utilized as r- valued cnstant in the lgic equatin. The functin is reespressed recursively until it is represented entirely by the functin prjectins. In this case the partitin f X cmpses p subsets f variables {X 1, X,..., X p }, where p is the number f functin reexpressins. In this paper we cnsider nly partitin f X cntaining tw subsets. Therefre, functin f(x) is reexpressed nly nce (Therem 1 and Therem ). n 1

The r-valued cmbinatinal circuit cmpses f tw level. On the first level the prjectins f f(x) ver X 1 are implemented by M PA. et us cnsider the M PA prpsed in 1], and called NOR-TSUM PA. In this case the number f inputs crrespnds t the number f variables in the free set, X. The number f utputs is determined by the number f different prjectins in the NOR-TSUM expansin. On the secnd level f cmbinatinal circuit the functin prjectins ver X 1 are identified as r-valued cnstants. The number f inputs in the secnd NOR-TSUM PA is determined by the number f variables in bund set X 1. Fig. (a) illustrates the structure f tw-level PA-based cmbinatinal circuit. Nte that instead f M PA we can use any kind f universal lgic mdule such as T-gate (multiplexer), prgrammable lgic array (PA), lk-up table type FPGA (field prgrammable gate array) and etc.. Example 3. ZebpZ shws a 3-valued 5 variable functin, f(x 1, x, x 3, x 4, x 5 ). A 3 variable M PA is assumed t be the cmpnent used t implement the functin. Since the M PA uses nt mre than 3 variables, nly 3 variables are used t partitin functin f(x). N attempt is made t try 4 variables in an attempt t decmpse the functin. Selecting x 1 and x as the variables f the free set, X, the functin f(x) is rewritten as a 3 variable functin and its functinal values becme the prjectins f f(x) ver X 1. Then, there are 5 different prjectins f f(x) ver X 1 =τ, f(τ, X ), that is the number f prduct terms in the secnd M PA. The r- valued cmbinatinal circuit implementing f(x) functin is shwn in Fig. (b). ZebpZ Decmpsitin table f f(x 1, x, x 3, x 4, x 5 ) x x 1 1 1 11 1 1 x 5 x 4 x 3 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 a b b b b a c a b d d d d e d a b b b b a c a b 3 «3 3 3 15 5 3 «3 3 3 15 «15 5 3 3 3 4 4 4 5 5 5 15 768 Iτ ; 768 Iτ ; 768 I; 768 I; E Fig. The structure f PA-based cmbinatinal circuit (a) cmmn case, (b) fr functin f(x 1, x, x 3, x 4, x 5 ). E ZebpZ 3 PA by methd 1] Cmbinatinal circuit by the prpsed methd Functin The first PA The secnd PA In Out W * S * In Out W 1 S 1 In Out W S S 1 +S X+Y

S 4 1 13 153 1 8 136 1 1 17 153 S 1 4 1 35 459 4 15 3 1 16 7 57 S 4 1 11 187 3 4 76 1 1 17 46 X*Y M 4 1 7 13 4 3 6 1 9 153 13 M 1 4 1 156 3 16 64 48 1 16 7 3 M 4 1 138 549 15 74 94 1 15 55 549 M 3 4 1 57 99 1 9 754 1 14 38 99 3 digit shift register R 4 1 5 18 3 3 6 1 4 68 18 R 1 4 1 5 51 3 34 1 1 17 51 R 4 1 5 51 3 34 1 1 17 51 R 3 4 1 5 18 3 3 6 1 4 68 51 6. Experimental results ZebpZ 3 cmpares the size f synthesized circuits with the size f M PA implemented the initial functin. This table shws that the present methd prduces the cmpetitive results with ther methds. Because the prpsed decmpsitin apprach permits nly ne r-valued functin t be decmpsed the functins frm the standardized benchmarks are separately implemented. Here such benchmarks as -figure quaternary adder, -figure quaternary multiplier, 3 digit shift register are used fr experimental investigatins. 7. Cnclusin A new technique fr realizing PA-based cmbinatinal circuits has been prpsed. This technique yields multiple-valued circuit f reduced size. This has been illustrated by cnsidering the realizatin f functins frm standardized benchmarks. Within suggested apprach the nvel MIN-TSUM and NOR- TSUM expansins with respect t X 1 are derived. The specific prperties f TSUM peratr are investigated t. It is shwn that truncated sum f r-valued variables des nt submit ne f the distributive laws. It needs t take int accunt the cnsideratin at examining the MIN-TSUM and NOR-TSUM expansins. Future wrk is carried ut in frame f generalizatin the prpsed apprach fr system f M functins. It leads t investigate the perfrmance f new apprach n ther standardized benchmark prblem. 8. Acknwledgment The authrs sincerely thank Dr.. Shmerk frn Technical University (Szczecin, Pland) fr his helpful cmments and suggestins n this paper. Appendix 1 1. Assciative laws: ( 8) y) x= (y z) (Prf) 1 ) 3=MIN (x 1 +x, ) x 3 =MIN (x 1 +x +x 3, +x 3, ) = MIN(x 1 +x +x 3, ), (a) because r- 1+z fr all values f z. In ther hand, 1 ( 3)=1 MIN(x +x 3, )=MIN (x 1 +x +x 3, +x 1, ) = MIN(x 1 +x +x 3, ), (b) because +x fr all values f. Since expressins (a) and (b) are equivalent, then TSUM peratr is assciative. Q.E.D.. Cmmutative laws: ( 9) y = y (Prf) This expressin is prved using the definitin f TSUM peratr. 3. Identities: ( 1) () x= (Prf) () x=min(+x, ) =, because +x fr all values f. Q.E.D. ( 11) = (Prf) x=min(+x, ) =MIN(x, ) =. Q.E.D. ( 1) 1... = (Prf) This expressin describes all input cmbinatin f x, therefre by definitin f literal functin Eq. ( 1) takes value (). Q.E.D. ( 13) 1 Λ 1 Λ... () Λ =x, where the symbl Λ stands fr r-valued minimum.

(Prf) x= Λ 1 Λ1 Λ... () Λ =1 Λ 1 Λ... () Λ, because Λ =. Q.E.D. 4. Distributive laws. ( 14) Λy) z=( z)(y z), where x, y, z {, 1,..., } (Prf) a) x Λ y z = MIN (MIN (x, y) + z, ) = MIN (x+z, y+z, ) b) ( z)(y z)=min (x+z,) Λ MIN (y+z,)=min(min(x+z,),min (y+z,))=min (x+z, y+z, r- 1). Since expressins (a) and (b) are equivalent, then TSUM peratr submit t the distributive law. Q.E.D. ( 15) y)λz (Λz) (yλz). (Prf) a) (x y) Λ z = MIN (x+y, ) Λ z = MIN( MIN(x+y, ), z))= MIN (x+y, z, ). b) (Λz) (yλz)=min (x, z) MIN(y, z) =MIN(MIN(x, z)+min(y, z), ) = MIN(x+y, z+x, z, y+z, ) Because, the expressins (a) and (b) are nt equivalent, TSUM peratr des nt submit t the distributive laws. Q.E.D. ( 16) y)λz i =(Λz i ) (yλz i ). (Prf) a) (x y) Λ z i = MIN (x+y, ) Λ z i = MIN(MIN(x+y, ), z i )) = MIN (x+y, z i, ). b) (Λz i ) (yλz i )= MIN(x, z i ) MIN(y, z i ) =MIN(MIN(x, z i )+MIN(y, z i ), ) = MIN(x+y, z i +x, z i, y+z i, ) Because z i {,}, then x _keb z=, x+z i =MAX(x, y) = x+, therwise. But +, Eq. (b) takes int accunt variable z i, therefre x+z i and z i d nt examine. The expressin y+z i is analyzed by analgy with previus expressin. (c) MIN (x+y, z i +x, z i, y+z i, )=MIN(x+y, z, ). Since expressins (a) and (c) are equivalent, then TSUM peratr submits t the distributive law. Q.E.D. References 1 Pelay F.J. and thers, CMOS Current - Mde Multiple-alued PA s. IEEE Trans. Circut and Syst., 1991, l. 38, pp. 434-441.. Asari K., C. Eswaran, An Optimisatin Technique fr the Design f Multiple-alued PA s. IEEE Trans. Cmput., vl.43, N. 1, 1994, pp.118-1. 3. Abd-El-Barr M. and H. Chy, Use f Simulated Annealing t Reduce -bit Decder PAs. Int. J. Electrnics, 1993, l. 74, NO. 3, pp.441-45. 4. Deng X., T.Hanyu and M.Kameyama, Design And Evaluatin Of A Current-Mde Multiple-alued PA Based On A Resnant Tunnelling transistr Mdel. IEE Prc.- Circuits Devices Syst., l. 141, N.6, 1994, pp.445-45. 5. Abugharieh S.B., S.C. ee, A Fast Algrithm Fr The Disjunctive Decmpsitin Of M-alued Functins. Part 1. The Decmpsitin Algrithm. IEEE Trans. n Cmputers. 1995, pp. 118-15.