Scribd Logo
Upload

Welcome to Scribd!

  • 161 views

    All Pass Comb Filters

    Uploaded by

    pho144
    The document discusses different types of comb filters, including: 1) Feedforward comb filters, which have a difference equation of y(n) = b0x(n) + bMx(n-M) and can create peaks and nulls in the frequency response. 2) Feedback comb filters, which can be implemented using direct form II structures and have a transfer function of H(z) = b0/(1 + aMz-M). 3) Nested allpass filters can be designed by replacing delays in stable allpass filters with other stable allpass filters, preserving the allpass property and stability. 4) Feedback delay networks are multi-input multi-output extensions of feedback

    Copyright:

    Attribution Non-Commercial (BY-NC)
    Download as PDF, TXT or read online from Scribd

    All Pass Comb Filters

    Uploaded by

    pho144
    161 views13 pages
    The document discusses different types of comb filters, including: 1) Feedforward comb filters, which have a difference equation of y(n) = b0x(n) + bMx(n-M) and can create peaks and nulls in the frequency response. 2) Feedback comb filters, which can be implemented using direct form II structures and have a transfer function of H(z) = b0/(1 + aMz-M). 3) Nested allpass filters can be designed by replacing delays in stable allpass filters with other stable allpass filters, preserving the allpass property and stability. 4) Feedback delay networks are multi-input multi-output extensions of feedback

    Copyright

    © Attribution Non-Commercial (BY-NC)

    Available Formats

    PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    The document discusses different types of comb filters, including: 1) Feedforward comb filters, which have a difference equation of y(n) = b0x(n) + bMx(n-M) and can create peaks and nulls in the frequency response. 2) Feedback comb filters, which can be implemented using direct form II structures and have a transfer function of H(z) = b0/(1 + aMz-M). 3) Nested allpass filters can be designed by replacing delays in stable allpass filters with other stable allpass filters, preserving the allpass property and stability. 4) Feedback delay networks are multi-input multi-output extensions of feedback

    Copyright:

    Attribution Non-Commercial (BY-NC)
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    161 views13 pages

    All Pass Comb Filters

    Uploaded by

    pho144
    The document discusses different types of comb filters, including: 1) Feedforward comb filters, which have a difference equation of y(n) = b0x(n) + bMx(n-M) and can create peaks and nulls in the frequency response. 2) Feedback comb filters, which can be implemented using direct form II structures and have a transfer function of H(z) = b0/(1 + aMz-M). 3) Nested allpass filters can be designed by replacing delays in stable allpass filters with other stable allpass filters, preserving the allpass property and stability. 4) Feedback delay networks are multi-input multi-output extensions of feedback

    Copyright:

    Attribution Non-Commercial (BY-NC)
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    You are on page 1of 13

    Comb Filters

    Feedforward Comb Filter


    b0 bM x(n) z M y(n)

    b0 = Feedforward coecient bM = Delay output coecient M = Delay-line length in samples Dierence Equation y(n) = b0 x(n) + bM x(n M ) Transfer Function H(z) = b0 + bM z M Frequency Response H(ejT ) = b0 + bM ejM T
    31

    Gain Range for Feedforward Comb Filter

    b0 bM x(n) z M y(n)

    y(n) = b0 x(n) + bM x(n M ) For a sinewave input, with b0, bM > 0: Gain is maximum (b0 + bM ) when a whole number of periods ts in M samples: 2 k T = k , k = 0, 1, 2, . . . M (the DFT basis frequencies for length M DFTs) Gain is minimum (|b0 bM |) when an odd number of half-periods ts in M samples: k T = (2k + 1) , k = 0, 1, 2, . . . M

    32

    Feed-Forward Comb-Filter Amplitude Response


    a) 2

    Magnitude (Linear)

    1.5

    0.5

    0.1

    0.2

    0.3 0.4 0.5 0.6 0.7 Normalized Frequency (cycles per sample))

    0.8

    0.9

    b)

    10 5 Magnitude (dB) 0 5 10 15 20 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 Normalized Frequency (cycles per sample)) 0.8

    g=0.1 g=0.5 g=0.9


    0.9 1

    Linear (top) and decibel (bottom) amplitude scales H(z) = 1 + gz M M =5 g = 0.1, 0.5, 0.9 G() = H(ejT ) = 1 + gejM T 2 cos(M T /2) when g = 1 In angers, these nulls slowly move with time
    33

    Feedback Comb Filter


    b0 y(n) x(n) z M

    aM

    aM = Feedback coecient (need |aM | < 1 for stability) M = Delay-line length in samples Direct-Form-II Dierence Equation (see gure): v(n) = x(n) aM v(n M ) y(n) = b0 v(n) Direct-Form-I Dierence Equation (commute gain b0 to the input): y(n) = b0 x(n) aM y(n M ) Transfer Function b0 H(z) = 1 + aM z M Frequency Response H(e
    jT

    b0 )= 1 + aM ejM T
    34

    Simplied Feedback Comb Filter Special case: b0 = 1, aM = g y(n) = x(n) + g y(n M ) 1 H(z) = 1 g z M Impulse response is a series of echoes, exponentially decaying and uniformly spaced in time: 1 H(z) = = 1 + g z M + g 2 z 2M + 1 g z M (n) + g (n M ) + g 2 (n 2M ) + = [1, 0, . . . , 0, g, 0, . . . , 0, g 2, 0, . . . ]
    M 1 M 1

    Models a plane wave between parallel walls Models wave propagation on a guitar string g = round-trip gain coecient: two wall-to-wall traversals (two wall reections) two string traversals (two endpoint reections)

    35

    Simplied Feedback Comb Filter, Contd


    y(n) x(n) z M

    y(n) = x(n) + g y(n M ) H(z) = 1 1 g z M

    For a sinewave input and 0 < g < 1: Gain is maximum [1/(1 g)] when a whole number of periods ts in M samples: 2 k T = k , k = 0, 1, 2, . . . M These are again the DFTM basis frequencies Gain is minimum [1/(1 + g)] when an odd number of half-periods ts in M samples: k T = (2k + 1) , k = 0, 1, 2, . . . M
    36

    Feed-Back Comb-Filter Amplitude Response

    a)

    12 10 Magnitude (Linear) 8 6 4 2 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 Normalized Frequency (cycles per sample)) 0.8 0.9 1

    b)

    20 15 Magnitude (dB) 10 5 0 5 10 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 Normalized Frequency (cycles per sample)) 0.8 0.9 1

    g=0.1 g=0.5 g=0.9

    Linear (top) and decibel (bottom) amplitude scales H(z) = M = 5,


    1 1gz M

    g = 0.1, 0.5, 0.9


    1gejM T 1
    g=1

    G() = H(ejT ) =

    2 sin( M T ) 2

    37

    Inverted-Feed-Back Comb-Filter Amplitude Response

    a)

    12 10 Magnitude (Linear) 8 6 4 2 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 Normalized Frequency (cycles per sample)) 0.8 0.9 1

    b)

    20 15 Magnitude (dB) 10 5 0 5 10 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 Normalized Frequency (cycles per sample)) 0.8 0.9 1

    g=0.1 g=0.5 g=0.9

    Linear (top) and decibel (bottom) amplitude scales H(z) = M = 5,


    1 1gz M

    g = 0.1, 0.5, 0.9


    1gejM T
    38

    G() = H(ejT ) =

    g = 1

    2 cos( M T ) 2

    Schroeder Allpass Filters


    b0

    x(n)

    z M

    y(n)

    aM

    Used extensively in articial reverberation Transfer function: b0 + z M H(z) = 1 + aM z M To obtain an allpass lter, set b0 = aM Proof: H(e
    jT

    a + ejM T a + ejM T = jM T ) = jM T 1 + ae e +a a + ejM T = =1 jM T a+e

    39

    First-Order Allpass Filter Transfer function: k1 + z 1 H1(z) = S1(z) = 1 + k1z 1

    (a)
    k1 x(n) z 1 y(n)

    k1

    (b)
    x(n) k1 k1 y(n) z 1

    (a) Direct form II lter structure (b) Two-multiply lattice-lter structure


    40

    Nested Allpass Filter Design Any delay-element or delay-line inside a stable allpass-lter can be replaced by any stable allpass-lter to obtain a new stable allpass lter: z 1 Ha(z) z 1 (The pure delay on the right-hand-side guarantees no delay-free loops are introduced, so that the original structure can be used) Proof: 1. Allpass Property: Note that the above substitution is a conformal map taking the unit circle of the z plane to itself. Therefore, unity gain for |z| = 1 is preserved under the mapping. 2. Stability: Expand the transfer function in series form: S [Ha(z)z 1]1
    2 = s0+s1Ha(z)z 1+s2Ha (z)z 2+

    where sn = original impulse response. In this form, it is clear that stability is preserved if Ha(z) is stable.

    41

    Nested Allpass Filters H2(z) = S1 [z S2(z)]


    (a)
    x(n) z 1 k2

    k1 + z 1S2(z) = 1 + k1z 1S2(z)

    k1 y(n)

    z 1 S2 (z) k2

    k1 S2 (z)

    (b)
    x(n) k1 k1 k2 k2 z 1 1

    H2 (z) =

    1 1 + a1 z 1 + a2 z 2

    H0 (z) =

    a2 + a1 z + z 1 + a1 z 1 + a2 z 2
    0

    y(n) 2

    z 1

    2 ya (n)

    Ha (z) =

    b0 + b1 z 1 + b2 z 2 1 + a1 z 1 + a2 z 2

    (a) Nested direct-form-II structures (b) Two-multiply lattice-lter structure (equivalent)

    42

    Feedback Delay Network (FDN)


    x1 (n) x2 (n) x3 (n)

    u1 (n) u2 (n) u3 (n)

    z M1 z M2 z M3 q11 q12 q13 q21 q22 q23 q31 q32 q33

    y1 (n) y2 (n) y3 (n)

    g1 g2 g3

    Order 3 MIMO FDN Vectorized Feedback Comb Filter Closely related to state-space representations of LTI systems (vectorized one-pole lter) Transfer function, stability analysis, etc., essentially identical to corresponding state-space methods

    43

    You might also like