6.1 signal flow graph representation of linear constant- coefficient difference equations 6.2 basic...

Post on 22-Dec-2015

238 views 0 download

Tags:

transcript

6.1 signal flow graph representation of linear constant- coefficient difference equations6.2 basic structures for IIR system

6.2.1 direct forms6.2.2 cascade forms6.2.3 parallel forms6.2.4.lattice structure

6.3 basic structures for FIR system 6.3.1 direct forms6.3.2 cascade forms6.3.3 structures for linear-phase FIR system6.2.4.lattice structure

CHAPTER 6 structures for discrete-time system

6.1 signal flow graph representation of linear constant-coefficient difference equations

Figure 6.8

Figure 6.9

node : source 、 sink 、 network

branch : constant 、 z-1 、 1 、 -1

The value at each node in a graph is the sum of the outputs of all the branches entering the node.

w2[n]

w1[n] w3[n]

- 1 w4[n]

6

w5[n] 3

2

z-1

y[n] x[n]

z-1

z-1

EXAMPLE

)()(2)()(

)(3)(

)()(

)()(

)()(6)(

)()()(

5311

45

31

4

13

451

2

21

zWzWzWzzY

zWzW

zWzzW

zWzW

zWzWzzW

zWzXzW

21

1

181

)42()()(

zz

zzXzY

21

1

181

)42(

)(

)()(

zz

z

zX

zYzH

]1[4][2]2[18]1[][ nxnxnynyny

y[n]

w2[n]

w1[n] w3[n]

- 1 z-1

w4[n]

6 z-1

w5[n] 3

2

z-1

x[n]

transposed form

transpose : reverse the directions of all branches reverse the roles of the input and output

The relationship between the input and output does not change.

6.2 basic structures for IIR system 6.2.1 direct forms

][][][],[][

][][][

10

01

knyanwnyknxbnw

knxbknyany

N

kk

M

kk

M

kk

N

kk

1.direct I

),()()(),()()(

)()(

1

1

1

21

21

1

0

1

0

zHzWzYzHzXzW

zHzH

za

zb

za

zb

zHN

k

kk

kM

k

kN

k

kk

kM

k

k

Figure 6.14

strongpoint : simple ;shortcoming : more delay ; be sensitive to word length ; be inconvenient to adjust zeros and poles

][][][],[][10

knyanwnyknxbnwN

k

k

M

k

k

double x[3],y[2];while(!eof(in_file)){ for(k=3;k>0;k++) //M=3

x[k]=x[k-1];x[0]=getc(in_file)-128;for(k=2;k>0;k++) /N=2 y[k]=y[k-1];for(k=0,y[0]=0;k<=3;k++)

y[0]+=b[k]*x[k];for(k=1;k<=2;k++)

y[0]+=a[k]*y[k]putc(out_file,y[0]+128);

}

realization of direct form in c program

][][][][][3

0

2

1

knxkbknykanykk

M

k

k

N

k

k

M

k

kkN

k

kk

N

k

kk

kM

k

k

knwbny

knwanxnw

zHzWzY

zHzXzW

zbzH

za

zHwhere

zHzHzHzH

za

zb

zH

0

1

1

2

0

1

1

2

1221

1

0

][][

][][][

)()()(

)()()(

,

1

1:

)()()()(

1

2.direct II ( canonic direct form )

b0 w[n]

z-1

z-1

z-1

y[n]

a1

a2

aN

z-1

z-1

z-1

aN-1

bM

bM-1

b2

b1

x[n]

M

k

k

N

k

k knwbnyknwanxnw01

][][][][][ ,

Figure 6.15

strongpoint : delay is reduced half

2/]1),([

12

21

1

22

110

2/]1),([

12

21

1

22

11

0

2/]1),([

1

1

0

1

1

1

1

NMMAX

k kk

kkk

NMMAX

k kk

kkNMMAX

k

kN

k

kk

kM

k

k

zz

zbzbb

zz

zzbzH

za

zb

zH

6.2.2 cascade forms

reason of being nonuniform : pairing manner of real zeros and poles ; order of cascade connection; pairing manner of zeros and poles 。

Figure 6.18

strongpoint : lower sensitivity to coefficient quantization than that of direct form; search the least-error ones because of the effects of limited word length;

pairing manner of real zeros and poles ; order of cascade connection;

be convenient to adjust zeros and poles; time division multiplexing using a second order loop.

shortcoming : not as fast as parallel form

NM

k

N

k kk

kokkkN

k

kk

kM

kk

zz

zzC

za

zb

zH0

]2

1[

12

21

1

11

1

0

11

6.2.3 parallel forms

strongpoint : lower sensitivity to coefficient quantization than that of direct form; less error because of the effects of limited word length;

be convenient to adjust poles; fast hardware realization.shortcoming : can not adjust zeros; can not be used in the filters with high precision requirement of zero location, such as notch filter and narrowband bandstop filter.

α 2[(N+1)/2]

z-1

α 1[(N+1)/2] γ 1[(N+1)/2]

z-1

γ 0[(N+1)/2]

α 21

z-1

γ 11 α 11

z-1

γ 01

C0

y[n] x[n]

Figure 6.20

6.2.4. supplement: lattice structure

321

3

1

8

5

24

131

1)(

zzzzH

y[n]

x[n]

-z-1 -z-1 -z-1

-1/4

1/4

-1/2

1/2

-1/3

1/3

EXAMPLE

strongpoint : be not sensitive to limited word length

knxkhnyzkhzHM

k

kM

k

00

)(,

6.3 basic structures for FIR system

6.3.1 direct forms

Figure 6.31

transversal filter structure

while(!eof(in_file)){ for(k=3;k>0;k++) //M=3

x[k]=x[k-1];x[0]=getc(in_file)-128;for(k=0,y[0]=0;k<=3;k++)

y[0]+=h[k]*x[k];putc(out_file,y[0]+128);

}

realization of direct form in c program][][][

3

0

knxkhnyk

]2

1[

1

22

110

]2

1[

1

22

11

0

'''1]0[

M

kkkk

M

kkk

kM

k

zbzbbzbzbhzkhzH

Figure 6.33

6.3.2 cascade forms

strongpoint : be convenient to adjust zeros; time division multiplexing using a second order loop.

12/

0

12/

0

12/

0

12/

0 12/

0

]2/[]2/[])[][]([

][][]2/[]2/[][][

][][]2/[]2/[][][

][][][

.1

M

k

M

k

M

k

M

k

M

Mk

M

k

MnxMhkMnxknxkh

kMnxkMhMnxMhknxkh

knxkhMnxMhknxkh

knxkhny

evenisM

6.3.3 structures for linear-phase FIR system

Figure 6.34

2/)1(

0

2/)1(

0

2/)1(

0

2/)1(

0 2/)1(

0

])[][]([

][][][][

][][][][

][][][

:.2

M

k

M

k

M

k

M

k

M

Mk

M

k

kMnxknxkh

kMnxkMhknxkh

knxkhknxkh

knxkhny

oddisM

Figure 6.35

strongpoint : multiplication operation is reduced half

6.3.4 supplement: lattice structure

321

3

1

8

5

24

131)( zzzzH

1/4

1/4

y[n] x[n]

-z-1 -z-1 -z-1

1/3

1/3

1/2

1/2

EXAMPLE

6.4 effects of limited word length

different infinite-precision realization structures : the same result, different operation quantity, speed, storage space;different finite-precision realization structures : different results, different error of frequency response, different difficulties to adjust frequency response 。

Reasons of error :1. coefficient quantity of filter’s : frequency response alters , even is instable; the more dense zeros and poles are, the more sensitive to effects of limited word length 2. round in operations

High-order IIR should try to avoid using direct form;FIR ( generally, zeros distribute uniformly ) use linear-phase direct form widely.

summary

6.1 signal flow graph representation 6.2 basic structures for IIR system

6.2.1 direct forms6.2.2 cascade forms6.2.3 parallel forms

6.3 basic structures for FIR system 6.3.1 direct forms6.3.2 cascade forms6.3.3 structures for linear-phase FIR syste

m

consideration :complexity of realization;

operation quantity and storage quantity ;

sensitivity to effects of limited word length.

requirement : transformation between system function and flow graph representation; strongpoints and shortcomings of different structures.

exercises

6.226.256.29