Scribd
Upload
107 views

Fir Cascade Lattice Structure

This document discusses FIR cascaded lattice filters. It describes how to determine the transfer functions H(z) and G(z) given the lattice coefficients (lattice → TF). It also describes how to determine the lattice coefficients given desired transfer functions H(z) and G(z) (TF → lattice). An example is provided to illustrate the TF → lattice process. Special cases like symmetric lattices are also mentioned.

Uploaded by

Aryan Verma
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
107 views

Fir Cascade Lattice Structure

This document discusses FIR cascaded lattice filters. It describes how to determine the transfer functions H(z) and G(z) given the lattice coefficients (lattice → TF). It also describes how to determine the lattice coefficients given desired transfer functions H(z) and G(z) (TF → lattice). An example is provided to illustrate the TF → lattice process. Special cases like symmetric lattices are also mentioned.

Uploaded by

Aryan Verma
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 9

DSP: FIR Cascaded Lattice Filters

Digital Signal Processing


FIR Cascaded Lattice Filters

D. Richard Brown III

D. Richard Brown III 1/8


DSP: FIR Cascaded Lattice Filters

FIR Cascaded Lattice Structures

x[n] y[n]
δ0
δ1 δ2 δ3

γ1 γ2 γ3
γ0
z −1 z −1 z −1 w[n]

This approach can realize two FIR transfer functions:

Y (z) W (z)
H(z) = and G(z) = .
X(z) X(z)

Given a lattice structure block diagram, how do we determine H(z) and


G(z) (lattice → TF)?
Given a desired FIR H(z) and G(z) (or h[n] and g[n]), how do we
determine the lattice coefficients (TF → lattice)?
D. Richard Brown III 2/8
DSP: FIR Cascaded Lattice Filters

FIR Cascaded Lattice Structures (Lattice → TF)

Hℓ−1 (z) Hℓ (z)

δℓ

γℓ

Gℓ−1 (z) z −1 Gℓ (z)

Analysis of single lattice segment:

Hℓ (z) = Hℓ−1 (z) + z −1 δℓ Gℓ−1 (z)


Gℓ (z) = γℓ Hℓ−1 (z) + z −1 Gℓ−1 (z)

Since H0 (z) = δ0 X(z) and G0 (z) = γ0 X(z), this recursion can be applied
for ℓ = 1, 2, . . . , N to determine the transfer functions H(z) and G(z)
from the lattice coefficients.
D. Richard Brown III 3/8
DSP: FIR Cascaded Lattice Filters

Lattice → TF Example
x[n] y[n]
δ0
δ1 δ2 δ3

γ1 γ2 γ3
γ0
z −1 z −1 z −1 w[n]

We can write
H0 (z) = δ0
G0 (z) = γ0
H1 (z) = H0 (z) + δ1 z −1 G0 (z) = δ0 + δ1 z −1 γ0
G1 (z) = γ1 H0 (z) + z −1 G0 (z) = γ1 δ0 + z −1 γ0
H2 (z) = H1 (z) + δ2 z −1 G1 (z) = δ0 + δ1 z −1 γ0 + δ2 z −1 (γ1 δ0 + z −1 γ0 )
G2 (z) = γ2 H1 (z) + z −1 G1 (z) = γ2 (δ0 + δ1 z −1 γ0 ) + z −1 (γ1 δ0 + z −1 γ0 )
and so on.
D. Richard Brown III 4/8
DSP: FIR Cascaded Lattice Filters

FIR Cascaded Lattice Structures (TF → Lattice)


To get a recursion for computing lattice coefficients given H(z) and G(z), we can
redo our analysis, solving instead for Hℓ−1 (z) and Gℓ−1 (z), to get
Hℓ−1 (z) = Kℓ [Hℓ (z) − δℓ Gℓ (z)]
Gℓ−1 (z) = Kℓ z [Gℓ (z) − γℓ Hℓ (z)]
1
where Kℓ = 1−δℓ γℓ .

D. Richard Brown III 5/8


DSP: FIR Cascaded Lattice Filters

FIR Cascaded Lattice Structures (TF → Lattice)


To get a recursion for computing lattice coefficients given H(z) and G(z), we can
redo our analysis, solving instead for Hℓ−1 (z) and Gℓ−1 (z), to get
Hℓ−1 (z) = Kℓ [Hℓ (z) − δℓ Gℓ (z)]
Gℓ−1 (z) = Kℓ z [Gℓ (z) − γℓ Hℓ (z)]
1
where Kℓ = 1−δℓ γℓ .

Suppose, as an example, ℓ = 2 with Hℓ (z) = h0 + h1 z −1 + h2 z −2 and


Gℓ (z) = g0 + g1 z −1 + g2 z −2 . Then
 
H1 (z) = K2 h0 + h1 z −1 + h2 z −2 − δ2 (g0 + g1 z −1 + g2 z −2 )
 
G1 (z) = K2 z g0 + g1 z −1 + g2 z −2 − γ2 (h0 + h1 z −1 + h2 z −2 )
Note that H1 (z) and G1 (z) can only be order 1 transfer functions. Hence we
need the z −2 and z +1 terms to vanish, which happens when
h2
z −2 term in H1 (z) vanishes ⇔ h2 − δ2 g2 = 0 ⇔ δ2 =
g2
+1 g0
z term in G1 (z) vanishes ⇔ g0 − γ2 h0 = 0 ⇔ γ2 =
h0
D. Richard Brown III 5/8
DSP: FIR Cascaded Lattice Filters

FIR Cascaded Lattice Structures (TF → Lattice)


Outline of procedure:
1. Set L equal to the filter order (impulse response length minus 1).
2. Set HL (z) = H(z) and GL (z) = G(z).
3. Set ℓ = L.
4. Solve for δℓ and γℓ .
5. Compute Kℓ = 1−δ1ℓ γℓ .
6. Compute Hℓ−1 (z) and Gℓ−1 (z). Both transfer functions should be
order ℓ − 1. Note coefficients are affected by Kℓ (you can’t just say
Hℓ−1 (z) is equal to the first ℓ − 1 coefficients of Hℓ (z) or that
Gℓ−1 (z) is equal to the first ℓ − 1 coefficients of Gℓ (z)).
7. Decrement ℓ.
8. If ℓ ≥ 1 go to step 4.
9. The final gains at the input are simply δ0 = H0 (z) and γ0 = G0 (z)
(these are both zero-order transfer functions).
D. Richard Brown III 6/8
DSP: FIR Cascaded Lattice Filters

TF → Lattice Example
Suppose H(z) = 1 + 2z −1 + 3z −2 and G(z) = 4 + 5z −1 + 6z −2 .
1. We set ℓ = L = 2 and solve δ2 = 36 = 21 and γ2 = 14 = 4.
1
2. We compute K2 = 1−2
= −1.
3. We compute
 
1 1
H1 (z) = −1 · 1 + 2z −1 + 3z −2 − (4 + 5z −1 + 6z −2 ) = 1 + z −1
2 2
and
 
G1 (z) = −1z 4 + 5z −1 + 6z −2 − 4(1 + 2z −1 + 3z −2 ) = 3 + 6z −1
1
1 3
4. Now set ℓ = 1 and solve δ1 = 2
6
= 12
and γ1 = 1
= 3.
1 4
5. Compute K1 = 3
1− 12
= 3
.
6. Compute  
4 1 1
H0 (z) = · 1 + z −1 − (3 + 6z −1 ) = 1 = δ0
3 2 12
and  
4 1
G0 (z) = z 3 + 6z −1 − 3(1 + z −1 ) = 6 = γ0
3 2
D. Richard Brown III 7/8
DSP: FIR Cascaded Lattice Filters

Special Case: Symmetric Lattice

Often, we only need a single transfer function and can use the symmetric
lattice
x[n] y[n]
1
−k1 −k2 −k3

−k1 −k2 −k4


1
z −1 z −1 z −1

The procedures for TF → lattice and lattice → TF are similar to what


we’ve covered. Also see Matlab functions tf2latc and latc2tf.

D. Richard Brown III 8/8

You might also like