LN10 Imlementation Structures v3

Embed Size (px)

Citation preview

  • 7/25/2019 LN10 Imlementation Structures v3

    1/31

    ilolu 1

    IMPLEMENTATION STRUCTURES FOR DISCRETE-TIME SYSTEMS

    FINITE PRECISION NUMERICAL EFFECTS-NUMBER REPRESENTATIONS

    QUANTIZATION IN IMPLEMENTING SYSTEMS

    REALIZABLE POLE LOCATIONS

    DIRECT FORM-I, DIRECT FORM-II

    CASCADE FORM

    PARALLEL FORMS

    TRANSPOSED FORMS

    FIR STRUCTURES

    GENERALIZED LINEAR PHASE FIR STRUCTURES

    DETERMINATION OF THE SYSTEM FUNCTION FROM A FLOW GRAPH

  • 7/25/2019 LN10 Imlementation Structures v3

    2/31

    ilolu 2

    Although two structures may be equivalent with regard to their input-output

    characteristics for infinite precision representations of coefficients and

    variables, they may have vastly different behavior when the numerical precision

    is limited.

    Oppenheim, Schafer, 3rded., p. 403

  • 7/25/2019 LN10 Imlementation Structures v3

    3/31

    ilolu 3

    FINITE PRECISION NUMERICAL EFFECTS

    NUMBER REPRESENTATIONS

    A real number in twos complement form (infinite precision)

    ( + 2= )

    : arbitrary scale factor

    0 0 1 0

  • 7/25/2019 LN10 Imlementation Structures v3

    4/31

    ilolu 4

    Quantized form ( +1bits, finite precision )

    ( + 2

    = )

    Quantization step size,

    2

  • 7/25/2019 LN10 Imlementation Structures v3

    5/31

    ilolu 5

    THE ROLE OFIn A/D conversion

    [ , ] volts 1 1 binary numbersEx: A 14 bit A/D converter is specified to have a dynamic range of 5volts.Assuming uniform quantization what are the values of 14 binary bits when its

    input is 3.111 Volt?

    Solution:

    5 1 3 2 5 23.111

    5097.1

    5 0 9 7 2 + 2 + 2 + 2 + 2 + 2 + 2 + 2 0

    1

    0

    MATLAB code to check

    x = 5*(2^12+2^9+2^8+2^7+2^6+2^5+2^3+2^0)*2^-13d = 5*2^-13(3.111-x)/d

    Result

    x = 3.110961914062500

    d = 6.103515625000000e-04

    ans = 0.062400000000343

  • 7/25/2019 LN10 Imlementation Structures v3

    6/31

    ilolu 6

    In fixed-point arithmetic, it is common to assume that each binary number has a

    scale factor of

    2

    For example

    2 .

    In floating-point arithmetic,

  • 7/25/2019 LN10 Imlementation Structures v3

    7/31

    ilolu 7

    QUANTIZATION IN IMPLEMENTING SYSTEMS

  • 7/25/2019 LN10 Imlementation Structures v3

    8/31

    ilolu 8

    REALIZABLE POLE LOCATIONS

  • 7/25/2019 LN10 Imlementation Structures v3

    9/31

    ilolu 9

    {0,18 ,

    28 ,

    38 ,

    48 ,

    58 ,

    68 ,

    78}

    0 , 18 , 12 , 38 , 12 , 58 , 32 , 78 cos {0, 18 , 28 , 38 , 48 , 58 , 68 , 78}

  • 7/25/2019 LN10 Imlementation Structures v3

    10/31

    ilolu 10

  • 7/25/2019 LN10 Imlementation Structures v3

    11/31

    ilolu 11

  • 7/25/2019 LN10 Imlementation Structures v3

    12/31

    ilolu 12

  • 7/25/2019 LN10 Imlementation Structures v3

    13/31

    ilolu 13

  • 7/25/2019 LN10 Imlementation Structures v3

    14/31

    ilolu 14

    DIRECT FORM-I, DIRECT FORM-II

    N

    k

    k

    k

    M

    k

    k

    k

    N

    N

    M

    M

    za

    zb

    zazaza

    zbzbzbbzH

    0

    0

    2

    2

    1

    1

    2

    2

    1

    10

    ...1

    ...

    with 10 a ,

  • 7/25/2019 LN10 Imlementation Structures v3

    15/31

    ilolu 15

    Direct Form - I

    0b

    1b

    2b

    1Nb

    Nb

    1a

    2a

    1

    Na

    Na

    1z

    1

    z

    1z

    1z

    1

    z

    1z

    nx ny nv

    Direct Form - II

    2a

    1

    Na

    Na

    nx ny0b

    1b

    2

    b

    1Nb

    Nb

    1z

    1z

    1z

    nw

    1 a

  • 7/25/2019 LN10 Imlementation Structures v3

    16/31

    ilolu 16

    Ex:

    21

    21

    125.075.01

    21

    zz

    zzzH

    Direct Form - I

    Direct Form - II

    2 75.0

    125.0

    1z

    1

    z1

    z

    1z

    nx ny

    75.0

    125.0

    nx ny

    2

    1z

    1z

    1a

  • 7/25/2019 LN10 Imlementation Structures v3

    17/31

    ilolu 17

    CASCADE FORM

    SN

    k kk

    kkk

    N

    k

    kk

    N

    k

    k

    M

    k

    kk

    M

    k

    k

    N

    N

    MM

    zaza

    zbzbb

    zdzdzc

    zgzgzf

    A

    zazaza

    zbzbzbbzH

    1

    2

    2

    1

    1

    2

    2

    1

    10

    1

    11

    1

    1

    1

    11

    1

    1

    2

    2

    1

    1

    2

    2

    1

    10

    1

    111

    111

    ...1

    ...

    21

    21

  • 7/25/2019 LN10 Imlementation Structures v3

    18/31

    ilolu 18

    Ex: Cascade form of a 6thorder system.

    2ndorder subsystems have Direct Form-II realizations.

    11a

    21a

    nx ny01b

    11b

    21b

    1z

    1

    z

    nw1

    12a

    22a

    02b

    12b

    22b

    1z

    1

    z

    nw2

    13a

    23a

    03b

    13b

    23b

    1z

    1

    z

    nw3

    ny1 ny2 ny3

  • 7/25/2019 LN10 Imlementation Structures v3

    19/31

    ilolu 19

    Ex: Cascade form of a 2ndorder system.

    11

    11

    21

    21

    25.015.01

    11

    125.075.01

    21

    zz

    zz

    zz

    zzzH

    nx ny

    5.0

    1z

    1z

    1z

    1z

    25.0

  • 7/25/2019 LN10 Imlementation Structures v3

    20/31

    ilolu 20

    PARALLEL FORMS

    21

    111

    1

    1

    1

    0

    1

    11

    1

    1

    N

    kkk

    kk

    N

    k k

    k

    N

    k

    k

    zdzd

    zeB

    zc

    AzCzH

    P

    21 2NNNNM

    P

  • 7/25/2019 LN10 Imlementation Structures v3

    21/31

    ilolu 21

    11a

    21a

    nx ny

    01e

    11e

    1z

    1z

    nw1

    ny1

    12a

    22a

    02e

    12e

    1z

    1z

    nw2

    ny2

    13a

    23a

    03e

    13e

    1

    z

    1z

    nw3

    ny3

    0C

  • 7/25/2019 LN10 Imlementation Structures v3

    22/31

    ilolu 22

    Ex:

    21

    1

    21

    21

    125.075.01

    878

    125.075.01

    21

    zz

    z

    zz

    zz

    zH

    nx ny

    75.0

    125.0

    7

    8

    1z

    1z

    8

  • 7/25/2019 LN10 Imlementation Structures v3

    23/31

    ilolu 23

    Ex:

    11

    21

    21

    25.01

    25

    5.01

    188

    125.075.01

    21

    zz

    zz

    zzzH

    nx

    ny

    5.0

    18

    1z

    8

    25.0

    25

    1z

  • 7/25/2019 LN10 Imlementation Structures v3

    24/31

    ilolu 24

    TRANSPOSED FORMS

    For a single input, single output (SISO) linear flow graph: Reverse all branch

    directions, interchange the input and output node assignments, keep

    transmittences the same, then the system function remains unchanged

  • 7/25/2019 LN10 Imlementation Structures v3

    25/31

    ilolu 25

    Ex:

    21

    21

    125.075.01

    21

    zz

    zzzH

    Direct Form II

    Transposed Direct Form II

    Transposed Direct Form II, redrawn.

    75.0

    125.0

    nx ny

    2

    1z

    1z

    nx ny

    75.0

    125.0

    2

    1z

    1z

    nx ny

    75.0

    125.0

    2

    1z

    1z

  • 7/25/2019 LN10 Imlementation Structures v3

    26/31

    ilolu 26

    FIR STRUCTURES

    M

    k

    knxkhny0

    Direct Form

    Transposed Direct Form

    nx

    ny

    0h

    1z

    1z

    1z

    1z

    1h 2h 1Mh Mh

    nx

    ny

    Mh

    1z

    1z

    1z

    1z

    1Mh 2Mh 1h 0h

  • 7/25/2019 LN10 Imlementation Structures v3

    27/31

    ilolu 27

    GENERALIZED LINEAR PHASE FIR STRUCTURES

    Odd Length Filters (Type-I and Type-III)

    : even (filter order)

    Note that 02

    Mh for Type-III filters!

    -1 multiplications in parentheses are for Type-III (odd symmetry) filters!

    1z

    0h 1h 2h 12

    Mh

    2

    Mh

    nx

    ny

    1z

    1z

    1z

    1z

    1z

    1z

    1z

    1 1 1 1

  • 7/25/2019 LN10 Imlementation Structures v3

    28/31

    ilolu 28

    EVEN LENGTH FILTERS (TYPE-II AND TYPE-IV)

    : odd (filter order)

    -1 multiplications in parentheses are for Type-IV (odd symmetry) filters!

    1z

    0h 1h 2h 2

    1Mh

    nx

    ny

    1z

    1z

    1z

    1

    z1

    z1

    z

    1 1 1 1

  • 7/25/2019 LN10 Imlementation Structures v3

    29/31

    ilolu 29

    DETERMINATION OF THE SYSTEM FUNCTION FROM A FLOW GRAPH

    nxnwnw 41

    zXzWzW 41 (a)

    zWzW12

    (b)

    zXzWzW 23 (c)

    zWzzW3

    1

    4

    (d)

    zWzWzY 42 (e)

    a b zXzWzW 42 (f)

    cd zXzWzzW 21

    4 (g)

    f,g

    zX

    z

    zzW

    1

    1

    2 1

    1

    (h)

    f,g

    zXz

    zzW

    1

    1

    4

    1

    1

    (i)

    h,ie

    zXzz

    zXz

    zzzY

    1

    1

    1

    11

    1

    1

    11

    1z

    1 nw1

    nw3

    nw2

    nw4

    nx ny

  • 7/25/2019 LN10 Imlementation Structures v3

    30/31

    ilolu 30

    Ex: a) Given the following flow graph of an LTI filter, determine its transfer

    function H(z).

    b)Plot the Direct Form II structure for the filter H1(z)=(1-2z-1)H(z), where H(z) is

    the filter in part-a.

    x[n] y[n]

    z-1z-1

    -1

    -0.5

    -0.5

  • 7/25/2019 LN10 Imlementation Structures v3

    31/31

    Ex: Consider the following system function with real valued coefficients

    22

    11

    2112

    1)(

    zbzb

    zzbbzH

    a)Find and plot the direct form II structure for H(z). Determine the number

    of multiplications, additions and delay terms.

    b)

    Find and plot the signal flow graph of a new filter structure such that thereare two multiplications only. You can have more delay terms than those in

    part a. (multiplication by 1 or -1 does not count).

    a) num. of multiplications=4

    num. of additions=4

    num. of delay terms = 2

    b)

    x[n]

    y[n]z-1

    z-1

    -b1

    -b2

    b1

    b2

    x[n]y[n]

    z-1

    z-1

    -b1

    -1

    -1

    b2

    z-2