You are on page 1of 10

EE 802-Advanced Digital Signal Processing

Dr. Amir A. Khan Office : A-218, SEECS 9085-2162; amir.ali@seecs.edu.pk

1
Lecture 22 EE-802 ADSP SEECS-NUST

Lecture Outline
Lattice Structures for FIR Filters
(Chapter 9 Proakis)

2
Lecture 22 EE-802 ADSP SEECS-NUST

From Direct Form to Lattice


Lattice structures are extensively used in DSP especially for the adaptive filter description Consider an FIR filter Impulse Response?

Direct form

Same as filter coeffs

3
Lecture 22 EE-802 ADSP SEECS-NUST

Lattice Structure for FIR Filters


One step forward predictor First order FIR
Reflection coefficient Take o/p on upper branch

2nd order FIR

Discard lower branch

4
Lecture 22 EE-802 ADSP SEECS-NUST

Lattice Structure for FIR Filters


2nd order FIR

Filter coefficients of A2(z)

Reflection coefficients of lattice filter can be linked to the direct form Filter coefficients
Lecture 22 EE-802 ADSP

SEECS-NUST

m-stage Lattice Structure


Relation between mth order direct form FIR and mth order lattice filter

Output of M-1 stage FIR filter

6
Lecture 22 EE-802 ADSP SEECS-NUST

Filter Coefficients for Upper and Lower Branches

Coefficients {1 K1} = {1, a1(1)} Coefficients {K1 1} = {a1(1), 1}

Filter coefficients of lower branch Filter coefficients of upper branch


Lecture 22 EE-802 ADSP

Filter coefficients of upper and lower branches are in reversed order 7


SEECS-NUST

Forward/Backward Prediction Inter-relationship of Upper/Lower Branches

Filter coefficients of Bm(z) Linear Prediction Context

Forward Prediction

x(n) x(n) Prediction Error

Backward Prediction x(n m) x(n m) Prediction Error

Bm(z) is reciprocal filter of Am(z)


8 Their product is an all pass system
Lecture 22 EE-802 ADSP SEECS-NUST

Lattice Structure FIR-Matrix Form

Divide both side by X(z)

Matrix form of a lattice structure


9
Lecture 22 EE-802 ADSP SEECS-NUST

Conversion of Lattice Coefficients to Direct Form Coefficients and Vice Versa

10
Lecture 22 EE-802 ADSP SEECS-NUST

You might also like