DSP: FIR Filters with Generalized Linear Phase
Digital Signal Processing
FIR Filters with Generalized Linear Phase
D. Richard Brown III
D. Richard Brown III 1/7
DSP: FIR Filters with Generalized Linear Phase
Generalized Linear Phase
Definition
A system has linear phase if its phase response θ(ω) = ∠H(ejω ) = −cω
for all ω and any constant c.
In general, a linear phase system has frequency response
H(ejω ) = |H(ejω )|e−jωc
and delays all frequencies by the same amount of time.
Definition
A system has generalized linear phase if its frequency response can be
written as H(ejω ) = A(ejω )e−jωc+jb where c and b are constants and
A(ejω ) is a real (not necessarily positive) function.
GLP systems are sometimes called “affine phase” systems and have
constant group delay except at discontinuities in the phase response.
D. Richard Brown III 2/7
DSP: FIR Filters with Generalized Linear Phase
Type I Causal FIR Generalized Linear-Phase Systems
Characteristics:
◮ Odd number of coefficients.
◮ Symmetric
Example (filter order M = 2): h[n] = {1, 2, 1}.
Frequency response
M
X
H(ejω ) = h[n]e−jωn
n=0
= h[0](e−jω0 + e−jωM ) + h[1](e−jω1 + e−jω(M −1) ) + · · · + h[M/2]e−jωM/2
= e−jωM/2 h[0](ejωM/2 + e−jωM/2 ) + · · · + h[M/2]
= e−jωM/2 (h[0] · 2 cos(ωM/2) + h[1] · 2 cos(ω(M/2 − 1)) · · · + h[M/2])
M/2
X
= e−jωM/2 a1 [k] cos(ωk)
k=0
where (
h[M/2] k=0
a1 [k] =
2h[M/2 − k] otherwise.
D. Richard Brown III 3/7
DSP: FIR Filters with Generalized Linear Phase
Type II Causal FIR Generalized Linear-Phase Systems
Characteristics:
◮ Even number of coefficients.
◮ Symmetric
Example (filter order M = 3): h[n] = {1, 2, 2, 1}.
Frequency response
M
X
H(ejω ) = h[n]e−jωn
n=0
= h[0](e−jω0 + e−jωM ) + · · · + h[(M − 1)/2](e−jω(M −1)/2 + e−jω(M +1)/2) )
= e−jωM/2 h[0](ejωM/2 + e−jωM/2 ) + · · · + h[(M − 1)/2](ejω/2 + e−jω/2 )
= e−jωM/2 (h[0] · 2 cos(ωM/2) + · · · + h[(M − 1)/2] · 2 cos(ω/2))
(M −1)/2
X 1
= e−jωM/2 a2 [k] cos ω k +
2
k=0
where
a2 [k] = 2h[(M − 1)/2 − k].
D. Richard Brown III 4/7
DSP: FIR Filters with Generalized Linear Phase
Type III Causal FIR Generalized Linear-Phase Systems
Characteristics:
◮ Odd number of coefficients.
◮ Anti-symmetric
Example (filter order M = 2): h[n] = {1, 0, −1} (h[M/2] must be zero).
Frequency response
M
X
H(ejω ) = h[n]e−jωn
n=0
= h[0](e−jω0 − e−jωM ) + · · · + h[1](e−jω1 − e−jω(M −1) ) + · · · + h[M/2]e−jωM/2
= e−jωM/2 h[0](ejωM/2 − e−jωM/2 ) + h[1](ejω(M/2−1) − e−jω(M/2−1) ) + . . .
= e−jωM/2 (h[0] · 2j sin(ωM/2) + h[1] · 2j sin(ω(M/2 − 1)) + . . . )
M/2−1
X
= je−jωM/2 a3 [k] sin(ω(k + 1))
k=0
where
a3 [k] = 2h[M/2 − k − 1].
D. Richard Brown III 5/7
DSP: FIR Filters with Generalized Linear Phase
Type IV Causal FIR Generalized Linear-Phase Systems
Characteristics:
◮ Even number of coefficients.
◮ Anti-symmetric
Example (filter order M = 3): h[n] = {1, −2, 2, −1}.
Frequency response
M
X
H(ejω ) = h[n]e−jωn
n=0
= h[0](e−jω0 − e−jωM ) + · · · + h[(M − 1)/2](e−jω(M −1)/2 − e−jω(M +1)/2) )
= e−jωM/2 h[0](ejωM/2 − e−jωM/2 ) + · · · + h[(M − 1)/2](ejω/2 − e−jω/2 )
= e−jωM/2 (h[0] · 2j sin(ωM/2) + · · · + h[(M − 1)/2] · 2j sin(ω/2))
(M −1)/2
X 1
= je−jωM/2 a4 [k] sin ω k +
2
k=0
where
a4 [k] = 2h[(M − 1)/2 − k].
D. Richard Brown III 6/7
DSP: FIR Filters with Generalized Linear Phase
Summary
Type Impulse response symmetry Impulse response length
I symmetric M + 1 odd (order is even)
II symmetric M + 1 even (order is odd)
III antisymmetric M + 1 odd (order is even)
IV antisymmetric M + 1 even (order is odd)
Type LPF HPF BPF BSF Comment
I Y Y Y Y Most versatile.
II Y N Y N Zero at z = −1.
III N N Y N Zeros at z = ±1.
IV N Y Y N Zero at z = 1.
All have the same constant group delay, τg (ω) = M 2 , which is an integer
for Type I and Type III but is not an integer for Type II and Type IV.
D. Richard Brown III 7/7
