L2_f05

download L2_f05

of 26

Transcript of L2_f05

  • 8/10/2019 L2_f05

    1/26

    EECS 247 Lecture 2: Filters 2005 H.K. Pag e 1

    EE247 - Lecture 2

    Filters

    Material covered today:

    Nomenclature

    Filter specifications

    Quality factor

    Frequency characteristics

    Group delay

    Filter types Butterworth

    Chebyshev I

    Chebyshev II

    Elliptic

    Bessel

    Group delay comparison example

    EECS 247 Lecture 2: Filters 2005 H.K. Pag e 2

    Filters

    Filters Provide frequency selectivity and/or phase shaping

    nV

    FilteroutV

    ( )jH

    ( )jH

  • 8/10/2019 L2_f05

    2/26

  • 8/10/2019 L2_f05

    3/26

    EECS 247 Lecture 2: Filters 2005 H.K. Pag e 5

    0

    x 10Frequency (Hz)

    Lowpass Filter Frequency Characteristics

    ( )jH

    ( )jH

    ( )0H

    Passband Ripple (Rpass)

    TransitionBand

    cPassband

    Passband

    Gain

    s topfStopband

    Frequency

    StopbandRejection

    f

    ( )jH

    3dBf

    dB3

    EECS 247 Lecture 2: Filters 2005 H.K. Pag e 6

    Quality Factor (Q)

    The term Quality Factor (Q) has different definitions:

    Component quality factor (inductor & capacitor

    Q)

    Pole quality factor

    Bandpass filter quality factor

    Next 3 slides clarifies each

  • 8/10/2019 L2_f05

    4/26

    EECS 247 Lecture 2: Filters 2005 H.K. Pag e 7

    Component Quality Factor (Q)

    For any component with a transfer function:

    Quality factor is defined as:

    ( )( ) ( )

    ( )( )

    Energy StoredAverage Powe r Diss ipation

    1H jR jX

    XQ

    R

    =+

    =

    EECS 247 Lecture 2: Filters 2005 H.K. Pag e 8

    Inductor & Capacitor Quality Factor

    Inductor Q :

    Capacitor Q :

    Rs LL L1 LY Q

    Rs j L Rs

    = =+

    Rp

    C

    C C1Z Q CRp1 jRp

    C

    = =+

  • 8/10/2019 L2_f05

    5/26

    EECS 247 Lecture 2: Filters 2005 H.K. Pag e 9

    Pole Quality Factor

    x

    j

    P

    xPo lex

    Q2

    =

    s-Plane

    EECS 247 Lecture 2: Filters 2005 H.K. Page 10

    Bandpass Filter Quality Factor (Q)

    -20

    -15

    -10

    -5

    0

    0.1 1 10f1 fcenter f2

    0

    -3dB

    f = f2 - f1

    ( )H jf

    Frequency

    Magnitude(dB)

    Q= fcenter/f

  • 8/10/2019 L2_f05

    6/26

    EECS 247 Lecture 2: Filters 2005 H.K. Page 11

    Consider a continuous time filter with s-domain transfer function G(s):

    Let us apply a signal to the filter input composed of sum of two

    sinewaves at slightly different frequencies (

  • 8/10/2019 L2_f05

    7/26

    EECS 247 Lecture 2: Filters 2005 H.K. Page 13

    What is Group Delay?

    Signal Magnitude and Phase Impairment

    { ]}[vOUT(t) = A1 G(j) sin t +()

    +

    { ]}[+ A2 G[j(+)]sin (+) t +d()

    d()

    +()

    -( )

    If the second term in the phase of the 2nd sin wave is non-zero, then the

    filters output at frequency + is time-shifted differently than thefilters output at frequency Phase distortion

    If the second term is zero, then the filters output at frequency +

    and the output at frequency are each delayed in time by -()/ PD -()/ is called the phase delay and has units of time

    EECS 247 Lecture 2: Filters 2005 H.K. Page 14

    Phase distortion is avoided only if:

    Clearly, if ()=k, k a constant, no phase distortion

    This type of filter phase response is called linear phase

    Phase shift varies linearly with frequency

    GR -d()/d is called the group delay and also has unitsof time. For a linear phase filter GR PD =k

    GR= PD implies linear phase

    Note: Filters with()=k+c are also called linear phase filters, buttheyre not free of phase distortion

    What is Group Delay?Signal Magnitude and Phase Impairment

    d()d

    ()- = 0

  • 8/10/2019 L2_f05

    8/26

    EECS 247 Lecture 2: Filters 2005 H.K. Page 15

    What is Group Delay?

    Signal Magnitude and Phase Impairment

    IfGR= PD No phase distortion

    [ )](vOUT(t) = A1 G(j) sin t -GR +

    [+ A2 G[j(+)] sin (+) )]( t -GR

    If alsoG( j)=G[ j(+)] for all input frequencies withinthe signal-band, vOUT is a scaled, time-shifted replica of the

    input, with no signal magnitude distortion :

    In most cases neither of these conditions are realizable exactly

    EECS 247 Lecture 2: Filters 2005 H.K. Page 16

    Phase delay is defined as:

    PD -()/ [ time]

    Group delay is defined as :

    GR -d()/d [time]

    If ()=k, k a constant, no phase distortion

    For a linear phase filter GR PD =k

    SummaryGroup Delay

  • 8/10/2019 L2_f05

    9/26

    EECS 247 Lecture 2: Filters 2005 H.K. Page 17

    Filter Types

    Lowpass Butterworth Filter

    Maximally flat amplitude

    within the filter passband

    Moderate phase distortion

    -60

    -40

    -20

    0

    Magnitude(dB)

    1 2

    -400

    -200

    Normalized Frequency

    Phase(degrees)

    5

    3

    1

    N

    ormalizedGroupDelay

    0

    0

    Example: 5th Order Butterworth filter

    N

    0

    d H( j )0

    d

    =

    =

    EECS 247 Lecture 2: Filters 2005 H.K. Page 18

    Lowpass Butterworth Filter

    All poles

    Poles located on the unit

    circle with equal angless-plane

    j

    Example: 5th Order Butterworth Filter

  • 8/10/2019 L2_f05

    10/26

    EECS 247 Lecture 2: Filters 2005 H.K. Page 19

    Filter Types

    Chebyshev I Lowpass Filter

    Chebyshev I filter

    Ripple in the passband

    Sharper transition band

    compared to Butterworth

    Poorer group delay

    As more ripple is allowed in

    the passband:

    Sharper transition band

    Poorer phase response

    1 2

    -40

    -20

    0

    Normalized Frequency

    Magnitude(dB)

    -400

    -200

    0

    Phase(degrees)

    0

    Example: 5th Order Chebyshev filter

    35

    0 No

    rmalizedGroupDelay

    EECS 247 Lecture 2: Filters 2005 H.K. Page 20

    Chebyshev I Lowpass Filter Characteristics

    All poles

    Poles located on an ellipse

    inside the unit circle

    Allowing more ripple in the

    passband:

    _Narrower transition band

    _Sharper cut-off

    _Higher pole Q

    _Poorer phase response

    Example: 5th Order Chebyshev I Filter

    s-planej

    Chebyshev I LPF 3dB passband ripple

    Chebyshev I LPF 0.1dB passband ripple

  • 8/10/2019 L2_f05

    11/26

    EECS 247 Lecture 2: Filters 2005 H.K. Page 21

    Bode Diagram

    Frequency [Hz]

    Phase(deg)

    Magnitude(dB)

    0 0.5 1 1.5 2

    -360

    -270

    -180

    -90

    0

    -60

    -40

    -20

    0

    Filter Types

    Chebyshev II Lowpass

    Chebyshev II filter

    Ripple in stopband

    Sharper transition

    band compared to

    Butterworth

    Passband phase

    more linear

    compared to

    Chebyshev I

    Example: 5th Order Chebyshev II filter

    EECS 247 Lecture 2: Filters 2005 H.K. Page 22

    Filter Types

    Chebyshev II Lowpass

    Example:

    5th Order

    Chebyshev II Filter

    s-plane

    j

    Both poles & zeros

    No. of poles n

    No. of zeros n-1

    Poles located both inside

    & outside of the unit circle

    Zeros located onj axis

    Ripple in the stopband

    only

    poles

    zeros

  • 8/10/2019 L2_f05

    12/26

    EECS 247 Lecture 2: Filters 2005 H.K. Page 23

    Filter Types

    Elliptic Lowpass Filter

    Elliptic fi lter

    Ripple in passband

    Ripple in the stopband

    Sharper transition band

    compared to Butterworth &

    both Chebyshevs

    Poorest phase response

    Magnitude(dB)

    Example: 5th Order Elliptic filter

    -60

    1 2Normalized Frequency

    0-400

    -200

    0

    Phase(degrees)

    -40

    -20

    0

    EECS 247 Lecture 2: Filters 2005 H.K. Page 24

    Filter Types

    Elliptic Lowpass Filter

    Example: 5th Order Elliptic Filter

    s-plane

    j

    Pole

    Zero

    Both poles & zeros

    No. of poles n

    No. of zeros n-1

    Zeros located onj axis

    Sharp cut -off

    _Narrower transition

    band

    _Pole Q higher

    compared to the

    previous filters

  • 8/10/2019 L2_f05

    13/26

    EECS 247 Lecture 2: Filters 2005 H.K. Page 25

    Filter Types

    Bessel Lowpass Filter

    s-planej

    Bessel

    All poles

    Maximally flat group

    delay

    Poor amplitude

    attenuation

    Poles outside unit circle

    (s-plane)

    Relatively low Q poles

    Example: 5th Order Bessel filter

    Pole

    EECS 247 Lecture 2: Filters 2005 H.K. Page 26

    Magnitude Response of a Bessel Filter as aFunction of Filter Order (n)

    Normalized Frequency

    Magnitude[dB]

    0.1 100-100

    -90

    -80

    -70

    -60

    -50

    -40

    -30

    -20

    -10

    0

    n=1

    2

    3

    4

    5

    76

    Filte

    rOrder

    Increased

    1 10

  • 8/10/2019 L2_f05

    14/26

    EECS 247 Lecture 2: Filters 2005 H.K. Page 27

    Filter Types

    Comparison of Various Type LPF Magnitude Response

    -60

    -40

    -20

    0

    Normalized Frequency

    Magnitude(dB)

    1 20

    BesselButterworth

    Chebyshev I

    Chebyshev II

    Elliptic

    All 5th order filters with same corner freq.

    Magnitude(dB)

    EECS 247 Lecture 2: Filters 2005 H.K. Page 28

    Filter Types

    Comparison of Various LPF Singularities

    s-plane

    j

    Poles Bessel

    Poles Butterworth

    Poles Elliptic

    Zeros Elliptic

    Poles Chebyshev I 0.1dB

  • 8/10/2019 L2_f05

    15/26

    EECS 247 Lecture 2: Filters 2005 H.K. Page 29

    Comparison of Various LPF Groupdelay

    Bessel

    Butterworth

    Chebyshev I

    0.5dB Passband Ripple

    Ref: A. Zverev,Handbook of filter synthesis, Wiley, 1967.

    1

    12

    1

    28

    1

    1

    10

    5

    4

    EECS 247 Lecture 2: Filters 2005 H.K. Page 30

    Group Delay Comparison

    Example

    Lowpass filter with 100kHz corner frequency

    Chebyshev I versus Bessel

    Both filters 4th order- same -3dBpoint

    Passband ripple of 1dBallowed for Chebyshev I

  • 8/10/2019 L2_f05

    16/26

    EECS 247 Lecture 2: Filters 2005 H.K. Page 31

    Magnitude Response

    Bode Magnitude Diagram

    Frequency [Hz]

    Magnitude(dB)

    104

    105

    106

    -70

    -60

    -50

    -40

    -30

    -20

    -10

    0

    4th Order Chebychev 14th Order Bessel

    EECS 247 Lecture 2: Filters 2005 H.K. Page 32

    Phase Response

    0 0.5 1 1.5 2

    x 105

    -350

    -300

    -250

    -200

    -150

    -100

    -50

    0

    Frequency [Hz]

    Phase[degrees]

    4th Order Chebychev 14th Order Bessel

  • 8/10/2019 L2_f05

    17/26

    EECS 247 Lecture 2: Filters 2005 H.K. Page 33

    Group Delay

    104

    105

    106

    0

    2

    4

    6

    8

    10

    12

    14

    Frequency [Hz]

    GroupDelay[s

    ]

    4th Ord. Chebychev 14th Ord. Bessel

    EECS 247 Lecture 2: Filters 2005 H.K. Page 34

    Normalized Group Delay

    104

    105

    106

    0

    0.5

    1

    1.5

    2

    2.5

    3

    Frequency [Hz]

    GroupDelay[normalized]

    4th Ord. Chebychev 14th Ord. Bessel

  • 8/10/2019 L2_f05

    18/26

    EECS 247 Lecture 2: Filters 2005 H.K. Page 35

    Step Response

    Time (sec)

    Amplitude

    0 0.5 1 1.5 2

    x 10-5

    0

    0.2

    0.4

    0.6

    0.8

    1

    1.2

    1.44th Order Chebychev 14th Order Bessel

    EECS 247 Lecture 2: Filters 2005 H.K. Page 36

    Intersymbol Interference (ISI)ISI Broadening of pulses resulting in interference between successive transmitted

    pulses

    Example: Simple RC filter

  • 8/10/2019 L2_f05

    19/26

    EECS 247 Lecture 2: Filters 2005 H.K. Page 37

    Pulse Broadening

    Bessel versus Chebyshev

    1.1 1.2 1. 3 1 .4 1. 5 1 .6 1.7 1. 8 1. 9 2

    x 10-4

    -1.5

    -1

    -0.5

    0

    0.5

    1

    1.5

    1.1 1.2 1. 3 1 .4 1. 5 1 .6 1. 7 1 .8 1.9 2

    x 10-4

    -1.5

    -1

    -0.5

    0

    0.5

    1

    1.5

    8th order Bessel 4th order Chebyshev I

    Chebyshev has more pulse broadening compared to Bessel More ISI

    InputOutput

    EECS 247 Lecture 2: Filters 2005 H.K. Page 38

    0 0.2 0.4 0.6 0.8 1 1.2 1.4

    x 10-4

    -1.5

    -1

    -0.5

    0

    0.5

    1

    1.5

    0 0.2 0.4 0.6 0.8 1 1.2 1.4

    x 10-4

    -1.5

    -1

    -0.5

    0

    0.5

    1

    1.5

    0 0.2 0.4 0.6 0.8 1 1.2 1.4

    x 10-4

    -1.5

    -1

    -0.5

    0

    0.5

    1

    1.5

    1111011111001010000100010111101110001001

    1111011111001010000100010111101110001001 1111011111001010000100010111101110001001

    Response to Random Data

    Chebyshev versus Bessel

    4th order Bessel 4th order Chebyshev I

    Input Signal:

    Symbol rate 1/130kHz

  • 8/10/2019 L2_f05

    20/26

    EECS 247 Lecture 2: Filters 2005 H.K. Page 39

    Measure of Signal Degradation

    Eye Diagram

    Eye diagram is a useful graphical illustration for signal degradation

    Consists of many overlaid traces of a signal using an oscilloscope

    where the symbol timing serves as the scope trigger

    It is a visual summary of all possible intersymbol interference

    waveforms The vertical opening immunity to noise

    Horizontal opening timing jitter

    EECS 247 Lecture 2: Filters 2005 H.K. Page 40

    Measure of Signal Degradation

    Eye Diagram

    Random data with symbol rates: 1/50kHz

    1/100kHz

    1/130kHz

    Frequency [Hz]

    Magnitude(dB)

    2 4 6 8 10 12 14

    x 104

    -10

    -9

    -8

    -7

    -6

    -5

    -4

    -3

    -2

    -1

    0

    BesselChebychev 1

    2 4 6 8 10 12 14

    x 104

    0

    0.5

    1

    1.5

    2

    2.5

    3

    Frequency [Hz]

    GroupDelay[normalized]

    4th Ord. Chebychev 14th Ord. Bessel

  • 8/10/2019 L2_f05

    21/26

    EECS 247 Lecture 2: Filters 2005 H.K. Page 41

    1 1.1 1.2 1.3 1.4 1.5 1.6 1.7

    x 10-5

    -0.8

    -0.6

    -0.4

    -0.2

    0

    0.2

    0.4

    0.6

    0.8

    Time

    4th Order Chebychev

    0.8 0.9 1 1.1 1.2 1.3 1.4 1.5

    x 10-5

    -0.8

    -0.6

    -0.4

    -0.2

    0

    0.2

    0.4

    0.6

    0.8

    Time

    4th Order Bessel

    0.6 0.8 1 1.2 1.4 1.6

    x 10-5

    -1

    -0.8

    -0.6

    -0.4

    -0.2

    0

    0.2

    0.4

    0.6

    0.8

    1

    Time

    Input Signal

    Eye Diagram

    Chebyshev versus Bessel

    4th order Chebyshev I4th order Bessel

    Input Signal

    Random data

    Symbol rate:1/ 130kHz

    EECS 247 Lecture 2: Filters 2005 H.K. Page 42

    Eye Diagrams

    2.5 3 3.5 4 4.5 5

    x 10

    -5

    -1

    -0.8

    -0.6

    -0.4

    -0.2

    0

    0.2

    0.4

    0.6

    0.8

    1

    Time

    4th Order Chebychev

    2.5 3 3.5 4 4.5 5

    x 10

    -5

    -1

    -0.8

    -0.6

    -0.4

    -0.2

    0

    0.2

    0.4

    0.6

    0.8

    1

    Time

    4th Order Bessel

    100%

    Eye openin g

    70%

    Eye openin g

    Random data maximum power symbol rate 1/50kHz

  • 8/10/2019 L2_f05

    22/26

    EECS 247 Lecture 2: Filters 2005 H.K. Page 43

    Eye Diagrams

    Filter with constant group delay More open eye Lower BER (bit-error-rate)

    Random data maximum symbol rate 1/100kHz

    1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2

    x 10-5

    -1

    -0.8

    -0.6

    -0.4

    -0.2

    0

    0.2

    0.4

    0.6

    0.8

    1

    Time

    4th Order Bessel

    1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2

    x 10-5

    -1

    -0.8

    -0.6

    -0.4

    -0.2

    0

    0.2

    0.4

    0.6

    0.8

    1

    Time

    4th Order Chebychev

    80%

    Eye opening

    40%

    Eye openin g

    EECS 247 Lecture 2: Filters 2005 H.K. Page 44

    Summary

    Filter Types

    Filters with high signal attenuation per pole _poor phase response

    For a given signal attenuation requirement ofpreserving constant groupdelay Higher orderfilter In the case of passive filters _ higher component count

    Case of integrated active filters _ higher chip area &power dissipation

    In cases where filter is followed by ADC and DSP Possible to digitally correct for phase non-linearities

    incurred by the analog circuitry by using phase equalizers

  • 8/10/2019 L2_f05

    23/26

    EECS 247 Lecture 2: Filters 2005 H.K. Page 45

    RLC Filters

    Bandpass filter:

    Singularities: Complex conjugate poles + zeros and zero & infinity

    o

    so RC

    2 2in oQ

    o

    oo

    VV s s

    1 LC

    RQ RCL

    =+ +

    =

    = =

    oVR

    CLnV

    EECS 247 Lecture 2: Filters 2005 H.K. Page 46

    RLC Filters

    Design a bandpass filter with:

    Center frequency of 1kHz

    Q of 20

    Assume that the inductor has series R resulting in an

    inductor Q of 40

    What is the effect of finite inductor Q on the overall Q?

    oVR

    CLnV

  • 8/10/2019 L2_f05

    24/26

    EECS 247 Lecture 2: Filters 2005 H.K. Page 47

    RLC Filters

    Effect of Finite Component Q

    ideal f i l t ind.f i l t

    1 1 1Q QQ

    = +

    Q=20 (ideal L)

    Q=13.3 (Qind.=40)

    eComponent Q must be much higher compared

    to desired filter Q

    EECS 247 Lecture 2: Filters 2005 H.K. Page 48

    RLC Filters

    Question:

    Can RLC filters be integrated on-chip?

    oVR

    CLnV

  • 8/10/2019 L2_f05

    25/26

    EECS 247 Lecture 2: Filters 2005 H.K. Page 49

    Monolithic Inductors

    Feasible Quality Factor & Value

    vRef: Radio Frequency Filters, Lawrence Larson; Mead workshop presentation 1999

    c Feasible monolithic inductor in CMOS tech.

  • 8/10/2019 L2_f05

    26/26

    EECS 247 Lecture 2: Filters 2005 H.K. Page 51

    Monolithic Filters

    Desirable to integrate filters with critical frequencies