Linear Phase Filters

Delay functions
 The frequency response is defined as
𝑗𝜔 𝑗𝜔 𝑗∠𝐻 𝑒 𝑗𝜔
𝐻 𝑒 = 𝐻 𝑒 𝑒 = 𝐻 𝑒 𝑗𝜔 𝑒 𝑗𝜃 𝜔

Amplitude response Phase response

𝜃 𝜔 = radians
𝜔 = radians/sec
 Inorder to examine the linear and non-linear phase characteristics two delay
functions are defined
−𝜃 𝜔
Phase delay, 𝜏𝑝 =
𝜔
Negative of ratio of phase and frequency of a filter
− ⅆ𝜃 𝜔
Group delay, 𝜏𝑔 =
ⅆ𝜔
Negative slope of the phase response
 Linear phase filters
 Filters for which 𝝉𝒑 and 𝝉𝒈 are independent of frequency are referred to
as linear phase filters
 A filter is said to have a linear phase response if its phase satisfies one of the
following conditions
𝜃 𝜔 = −𝛼𝜔 … … … … … . 𝐸𝑞(1)
𝜃 𝜔 = −𝛼𝜔+𝛽 … … … … . 𝐸𝑞(2)
 If a filter satisfies the condition given in 𝐸𝑞(1) it will have both constant
group delay and constant phase delay
 For condition in 𝐸𝑞(1) to be satisfied
− the impulse response of the filter must have positive symmetry with centre of
𝑵−𝟏
symmetry at 𝒏 =
𝟐
𝑁−1
− ℎ 𝑛 = ℎ(𝑁 − 1 − 𝑛) ; where 𝑛=0,1,2…… for 𝑁 odd
2
𝑁
and 𝑛=0,1,2…… − 1 for 𝑁 even
2
− the phase response is a function of filter length
𝑁−1
i.e., 𝛼 =
2
𝑁 = odd and symmetric impulse response
𝑁=9

𝑁−1
 centre of symmetry at 𝑛 =
2
9−1
𝑛= =4
2

 ℎ 𝑛 = ℎ(𝑁 − 1 − 𝑛) ;
𝑁−1
where 𝑛=0,1,2…… for 𝑵 odd
2
𝒏 = 0,1,2,3,4
𝒉 𝟎 = ℎ 9 − 1 − 0 = 𝒉(𝟖)
𝒉 𝟏 = ℎ 9 − 1 − 1 = 𝒉(𝟕)
𝒉 𝟐 = ℎ 9 − 1 − 2 = 𝒉(𝟔)
𝒉 𝟑 = ℎ 9 − 1 − 3 = 𝒉(𝟓)
𝒉 𝟒 = ℎ 9 − 1 − 4 = 𝒉(𝟒)
𝑁 = even and symmetric impulse response
𝑁=8

𝑁−1
 centre of symmetry at 𝑛 =
2
8−1
𝑛= = 3.5
2

 ℎ 𝑛 = ℎ(𝑁 − 1 − 𝑛) ;
𝑁
where 𝑛=0,1,2…… − 1 for 𝑵 even
2
𝒏 = 0,1,2,3
𝒉 𝟎 = ℎ 8 − 1 − 0 = 𝒉(𝟕)
𝒉 𝟏 = ℎ 8 − 1 − 1 = 𝒉(𝟔)
𝒉 𝟐 = ℎ 8 − 1 − 2 = 𝒉(𝟓)
𝒉 𝟑 = ℎ 8 − 1 − 3 = 𝒉(𝟒)
 If a filter satisfies the condition given in 𝐸𝑞(2) it will have a constant
group delay only
 For condition in 𝐸𝑞(2) to be satisfied
− the impulse response of the filter must have negative symmetry (anti-
𝑵−𝟏
symmetry) with centre of symmetry at 𝒏 =
𝟐
𝑁−1
ℎ 𝑛 = −ℎ(𝑁 − 1 − 𝑛) ; where 𝑛=0,1,2…… for 𝑁 odd
2
𝑁
and 𝑛=0,1,2…… − 1 for 𝑁 even
2

− the phase response is a function of filter length

𝑁−1 𝜋
i.e., 𝛼 = and 𝛽 =
2 2
𝑁 = odd and antisymmetric impulse response
𝑁=9
𝑁−1
 centre of symmetry at 𝑛 =
2
9−1
𝑛= =4
2

 ℎ 𝑛 = −ℎ(𝑁 − 1 − 𝑛) ;
𝑁−1
where 𝑛=0,1,2…… for 𝑵 odd
2
𝒏 = 0,1,2,3,4
𝒉 𝟎 = −ℎ 9 − 1 − 0 = −𝒉(𝟖)
𝒉 𝟏 = −ℎ 9 − 1 − 1 = −𝒉(𝟕)
𝒉 𝟐 = −ℎ 9 − 1 − 2 = −𝒉(𝟔)
𝒉 𝟑 = −ℎ 9 − 1 − 3 = −𝒉(𝟓)
𝑁−1
𝒉 𝟒 = −ℎ 9 − 1 − 4 = −𝒉 𝟒 = 𝟎 becoz 𝑛=0 ℎ 𝑛 =0

𝑁−1
Note that 𝑛=0 ℎ 𝑛 = 0 for filters with antisymetric impulse response
𝑁 = even and antisymmetric impulse response
𝑁=8

𝑁−1
 centre of symmetry at 𝑛 =
2
8−1
𝑛= = 3.5
2

 ℎ 𝑛 = −ℎ(𝑁 − 1 − 𝑛) ;
𝑁
where 𝑛=0,1,2…… − 1 for 𝑵 even
2
𝒏 = 0,1,2,3
𝒉 𝟎 = −ℎ 8 − 1 − 0 = −𝒉(𝟕)
𝒉 𝟏 = −ℎ 8 − 1 − 1 = −𝒉(𝟔)
𝒉 𝟐 = −ℎ 8 − 1 − 2 = −𝒉(𝟓)
𝒉 𝟑 = −ℎ 8 − 1 − 3 = −𝒉(𝟒)

𝑁−1
Note that 𝑛=0 ℎ 𝑛 = 0 for filters with antisymetric impulse response
Linear phase and symmetric impulse response
 For FIR filters with linear phase
𝜃 𝜔 = −𝛼𝜔; −𝜋 ≤ 𝜔 ≤ 𝜋 … … … … … … . 𝐸𝑞(3)
 From the definition of phase delay and group delay
−𝜃 𝜔
𝜏𝑝 = =𝛼
𝜔
− ⅆ𝜃 𝜔
𝜏𝑔 = =𝛼
ⅆ𝜔
𝑁−1
 𝐻 𝑒 𝑗𝜔 = 𝑛=0 ℎ(𝑛)𝑒 −𝑗𝜔𝑛 = ±𝐻(𝑒 𝑗𝜔 )𝑒 𝑗𝜃 𝜔 = ±𝐻(𝑒 𝑗𝜔 )𝑒 −𝑗𝛼𝜔
𝑁−1

ℎ(𝑛) 𝑐𝑜𝑠𝜔𝑛 − 𝑗𝑠𝑖𝑛𝜔𝑛 = ±𝐻(𝑒 𝑗𝜔 ) 𝑐𝑜𝑠𝛼𝜔 − 𝑗𝑠𝑖𝑛𝛼𝜔

𝑛=0

 Equation real and imaginary parts

𝑁−1
±𝐻(𝑒 𝑗𝜔 ) 𝑐𝑜𝑠𝛼𝜔 = 𝑛=0 ℎ(𝑛)𝑐𝑜𝑠𝜔𝑛 … … … … 𝐸𝑞(4)
𝑁−1
±𝐻(𝑒 𝑗𝜔 ) 𝑠𝑖𝑛𝛼𝜔 = 𝑛=0 ℎ 𝑛 𝑠𝑖𝑛𝜔𝑛 … … … … 𝐸𝑞(5)
 𝑁−1
𝐸𝑞(5) 𝑠𝑖𝑛𝛼𝜔 𝑛=0 ℎ 𝑛 𝑠𝑖𝑛𝜔𝑛
→ = 𝑁−1
𝐸𝑞(4) 𝑐𝑜𝑠𝛼𝜔 𝑛=0 ℎ(𝑛)𝑐𝑜𝑠𝜔𝑛

 Cross multiplying →
𝑁−1 𝑁−1
sin 𝛼𝜔 𝑛=0 ℎ 𝑛 𝑐𝑜𝑠𝜔𝑛 − 𝑐𝑜𝑠𝛼𝜔 𝑛=0 ℎ 𝑛 𝑠𝑖𝑛𝜔𝑛 = 0
𝑁−1

ℎ 𝑛 sin 𝛼𝜔. 𝑐𝑜𝑠𝜔𝑛 − 𝑐𝑜𝑠𝛼𝜔. 𝑠𝑖𝑛𝜔𝑛 = 0

𝑛=0
𝑁−1

ℎ 𝑛 sin 𝛼 − 𝑛 𝜔 = 0 … … … … … … … 𝐸𝑞(6)
𝑛=0

 One solution of 𝐸𝑞. (6) exist when

𝑁−1
𝛼= and
2

ℎ 𝑛 = ℎ(𝑁 − 1 − 𝑛) for 0 ≤ 𝑛 ≤ 𝑁 − 1
Proof
𝑁−1

ℎ 𝑛 sin 𝛼 − 𝑛 𝜔
𝑛=0
𝑁−1
𝑁−1
= ℎ 𝑛 sin −𝑛 𝜔
2
𝑛=0
𝑁−1
𝑁−1−𝑛−𝑛
= ℎ 𝑛 sin 𝜔
2
𝑛=0
ℎ 𝑛 = ℎ(𝑁 − 1 − 𝑛) → 𝑁 − 1 − 𝑛 = 𝑛
𝑁−1
𝑛−𝑛
= ℎ 𝑛 sin 𝜔
2
𝑛=0
𝑁−1

= ℎ 𝑛 sin 0
𝑛=0
=0
Linear phase and antisymmetric impulse response
 For FIR filters with linear phase
𝜃 𝜔 = −𝛼𝜔 + 𝛽; −𝜋 ≤ 𝜔 ≤ 𝜋 … … … … … … . 𝐸𝑞(7)
 From the definition of phase delay and group delay
−𝜃 𝜔 𝛽
𝜏𝑝 = =𝛼−
𝜔 𝜔
− ⅆ𝜃 𝜔
𝜏𝑔 = =𝛼
ⅆ𝜔
𝑁−1
 𝐻 𝑒 𝑗𝜔 = 𝑛=0 ℎ(𝑛)𝑒 −𝑗𝜔𝑛
= ±𝐻(𝑒 𝑗𝜔 )𝑒 𝑗𝜃 𝜔
= ±𝐻(𝑒 𝑗𝜔 )𝑒 𝑗(−𝛼𝜔+𝛽)
𝑁−1

ℎ(𝑛) 𝑐𝑜𝑠𝜔𝑛 − 𝑗𝑠𝑖𝑛𝜔𝑛 = ±𝐻(𝑒 𝑗𝜔 ) cos 𝛽 − 𝛼𝜔 + 𝑗𝑠𝑖𝑛(𝛽 − 𝛼𝜔)

𝑛=0

 Equation real and imaginary parts

𝑁−1
±𝐻(𝑒 𝑗𝜔 )cos(𝛽 − 𝛼𝜔) = 𝑛=0 ℎ(𝑛)𝑐𝑜𝑠𝜔𝑛 … … … … 𝐸𝑞(8)
𝑁−1
±𝐻 𝑒 𝑗𝜔 𝑠𝑖𝑛 𝛽 − 𝛼𝜔 = − 𝑛=0 ℎ 𝑛 𝑠𝑖𝑛𝜔𝑛 … … … … 𝐸𝑞(9)
 𝑁−1
𝐸𝑞(9) 𝑠𝑖𝑛 𝛽 − 𝛼𝜔 − 𝑛=0 ℎ 𝑛 𝑠𝑖𝑛𝜔𝑛
→ = 𝑁−1
𝐸𝑞(8) cos(𝛽 − 𝛼𝜔) 𝑛=0 ℎ(𝑛)𝑐𝑜𝑠𝜔𝑛

 Cross multiplying →
𝑁−1 𝑁−1

𝑠𝑖𝑛 𝛽 − 𝛼𝜔 ℎ(𝑛)𝑐𝑜𝑠𝜔𝑛 + cos 𝛽 − 𝛼𝜔 ℎ 𝑛 𝑠𝑖𝑛𝜔𝑛 = 0

𝑛=0 𝑛=0
𝑁−1 𝑁−1

ℎ(𝑛)𝑠𝑖𝑛 𝛽 − 𝛼𝜔 𝑐𝑜𝑠𝜔𝑛 + ℎ 𝑛 𝑠𝑖𝑛𝜔𝑛cos(𝛽 − 𝛼𝜔) = 0

𝑛=0 𝑛=0
𝑁−1

ℎ(𝑛)𝑠𝑖𝑛 𝛽 − (𝛼 − 𝑛)𝜔 = 0
𝑛=0
𝑁−1
𝜋
𝛽= → ℎ 𝑛 𝑐𝑜𝑠 (𝛼 − 𝑛)𝜔 = 0 … … . . 𝐸𝑞(10)
2
𝑛=0
 𝐸𝑞(10) is satisified when
𝑁−1
𝛼= and
2

ℎ 𝑛 = −ℎ(𝑁 − 1 − 𝑛) for 0 ≤ 𝑛 ≤ 𝑁 − 1
Proof
ℎ 𝑛 = −ℎ 𝑁 − 1 − 𝑛 → 𝑛 = 𝑁 − 1 − 𝑛
𝑁−1
𝑁−1
𝐹𝑟𝑜𝑚 𝐸𝑞(10) ℎ 𝑛 𝑐𝑜𝑠 ( − 𝑛)𝜔
2
𝑛=0
𝑁−1 𝑁−1
𝑁−1−𝑛−𝑛 𝑛−𝑛
= ℎ 𝑛 𝑐𝑜𝑠 𝜔= ℎ 𝑛 𝑐𝑜𝑠 𝜔
2 2
𝑛=0 𝑛=0

𝑁−1 𝑁−1

ℎ 𝑛 cos 0 = ℎ 𝑛 = 0 𝑓𝑜𝑟 𝑎𝑛𝑡𝑖𝑠𝑦𝑚𝑚𝑒𝑡𝑟𝑖𝑐 𝑓𝑖𝑙𝑡𝑒𝑟𝑠

𝑛=0 𝑛=0
Frequency response of Linear phase Filters
Case i: Symmetric impulse response, N odd
𝑁=9

ℎ 𝑛 is splitted with respect to the center of symmetry 𝑛 = 4

𝑁−3
𝑛 = 0 𝑡𝑜 3 → 𝑛 = 0 𝑡𝑜
2
𝑁+1 ← 𝑓𝑖𝑛ⅆ𝑖𝑛𝑔 𝑙𝑖𝑚𝑖𝑡𝑠 𝑜𝑓 𝑠𝑢𝑚𝑚𝑎𝑡𝑖𝑜𝑛
𝑛 = 5 𝑡𝑜 8 → 𝑛 = 𝑡𝑜 𝑁 − 1
2
𝑁−1
 𝐻 𝑒 𝑗𝜔 = 𝑛=0 ℎ(𝑛)𝑒 −𝑗𝜔𝑛

𝑁−3
2 𝑁−1
𝑁−1 𝑁−1
−𝑗𝜔 2
= ℎ(𝑛)𝑒 −𝑗𝜔𝑛 + ℎ 𝑒 + ℎ(𝑛)𝑒 −𝑗𝜔𝑛
2
𝑛=0 𝑁+1
𝑛= 2
 Let 𝑚 = 𝑁 − 1 − 𝑛 → 𝑛 = 𝑁 − 1 − 𝑚
𝑁+1 𝑁+1 𝑁−3
𝑛= →𝑚 =𝑁−1− =
2 2 2
𝑛 =𝑁−1→𝑚 =𝑁−1− 𝑁−1 =0
𝑁−3 𝑁−3
2 2
𝑗𝜔 −𝑗𝜔𝑛
𝑁 − 1 −𝑗𝜔 𝑁−1
𝐻 𝑒 = ℎ(𝑛)𝑒 +ℎ 𝑒 2 + ℎ(𝑁 − 1 − 𝑚) 𝑒 −𝑗𝜔(𝑁−1−𝑚)
2
𝑛=0 𝑚=0

 Replace dummy variable with original variable n

𝑁−3 𝑁−3
2 2
𝑁 − 1 −𝑗𝜔 𝑁−1
𝐻 𝑒 𝑗𝜔 = ℎ(𝑛)𝑒 −𝑗𝜔𝑛 +ℎ 𝑒 2 + ℎ(𝑁 − 1 − 𝑛) 𝑒 −𝑗𝜔(𝑁−1−𝑛)
2
𝑛=0 𝑛=0

 For symmetric impulse response ℎ 𝑛 = ℎ(𝑁 − 1 − 𝑛)

𝑁−3 𝑁−3
2 2
𝑁 − 1 −𝑗𝜔 𝑁−1
𝐻 𝑒 𝑗𝜔 = ℎ(𝑛)𝑒 −𝑗𝜔𝑛 +ℎ 𝑒 2 + ℎ(𝑛) 𝑒 −𝑗𝜔(𝑁−1−𝑛)
2
𝑛=0 𝑛=0
𝐻 𝑒 𝑗𝜔 =
𝑁−3 𝑁−3
2 2
−𝑗𝜔
𝑁−1
𝑗𝜔
𝑁−1 𝑁−1 𝑁−1
=𝑒 2 ℎ(𝑛)𝑒 2 −𝑛 + ℎ(𝑛)𝑒
−𝑗𝜔 2 −𝑛 +ℎ
2
𝑛=0 𝑛=0
𝑁−3
2
−𝑗𝜔
𝑁−1 𝑁−1 𝑁−1
=𝑒 2 2ℎ 𝑛 cos 𝜔 −𝑛 +ℎ ……Eq(11)
2 2
𝑛=0

Note that the phase is Amplitude response

in the form of 𝑒 −𝑗𝛼𝜔 𝐴 𝜔 = 𝐻(𝑒 𝑗𝜔 )
Case ii: Symmetric impulse response, N even
𝑁=8

ℎ 𝑛 is splitted with respect to the center of symmetry 𝑛 =3.5

𝑁−2
𝑛 = 0 𝑡𝑜 3 → 𝑛 = 0 𝑡𝑜
2
𝑁 ← 𝑓𝑖𝑛ⅆ𝑖𝑛𝑔 𝑙𝑖𝑚𝑖𝑡𝑠 𝑜𝑓 𝑠𝑢𝑚𝑚𝑎𝑡𝑖𝑜𝑛
𝑛 = 4 𝑡𝑜 7 → 𝑛 = 𝑡𝑜 𝑁 − 1
2
𝑁−1
 𝐻 𝑒 𝑗𝜔 = 𝑛=0 ℎ(𝑛)𝑒
−𝑗𝜔𝑛

𝑁−2
2 𝑁−1

= ℎ(𝑛)𝑒 −𝑗𝜔𝑛 + ℎ(𝑛)𝑒 −𝑗𝜔𝑛

𝑛=0 𝑁
𝑛= 2
 Let 𝑚 = 𝑁 − 1 − 𝑛 → 𝑛 = 𝑁 − 1 − 𝑚
𝑁 𝑁 𝑁 𝑁−2
𝑛 = →𝑚 =𝑁−1− = −1=
2 2 2 2
𝑛 =𝑁−1→𝑚 =𝑁−1− 𝑁−1 =0
𝑁−2 𝑁−2
2 2
𝐻 𝑒 𝑗𝜔 = ℎ(𝑛)𝑒 −𝑗𝜔𝑛 + ℎ(𝑁 − 1 − 𝑚) 𝑒 −𝑗𝜔(𝑁−1−𝑚)
𝑛=0 𝑚=0

 Replace dummy variable with original variable n

𝑁−2 𝑁−2
2 2
𝐻 𝑒 𝑗𝜔 = ℎ(𝑛)𝑒 −𝑗𝜔𝑛 + ℎ(𝑁 − 1 − 𝑛) 𝑒 −𝑗𝜔(𝑁−1−𝑛)
𝑛=0 𝑛=0

 For symmetric impulse response ℎ 𝑛 = ℎ(𝑁 − 1 − 𝑛)

𝑁−2 𝑁−2
𝐻 𝑒 𝑗𝜔 = 2
𝑛=0 ℎ(𝑛)𝑒 −𝑗𝜔𝑛 + 2
𝑛=0 ℎ(𝑛) 𝑒 −𝑗𝜔(𝑁−1−𝑛)
 𝑁−2 𝑁−2
𝑁−1 2 2
𝑁−1 𝑁−1
−𝑗𝜔 𝑗𝜔
𝐻 𝑒 𝑗𝜔 = 𝑒 2 ℎ(𝑛)𝑒 2 −𝑛 + ℎ(𝑛)𝑒
−𝑗𝜔 2 −𝑛
𝑛=0 𝑛=0
𝑁−2
2
𝑁−1
−𝑗𝜔 2 𝑁−1
= 𝑒 2ℎ(𝑛) cos 𝜔 −𝑛 …………Eq(13)
2
𝑛=0
Note that the phase is
in the form of 𝑒 −𝑗𝛼𝜔
Amplitude response
𝐴 𝜔 = 𝐻(𝑒 𝑗𝜔 )

 Amplitude response is
Antisymmetric with 𝜔 = 𝜋
Non zero at 𝜔 = 0
Zero at 𝜔 = 𝜋
Case iii: Antisymmetric impulse response, N odd
𝑁=9

ℎ 𝑛 is splitted with respect to the center of symmetry 𝑛 = 4

𝑁−3
𝑛 = 0 𝑡𝑜 3 → 𝑛 = 0 𝑡𝑜
2
𝑁+1 ← 𝑓𝑖𝑛ⅆ𝑖𝑛𝑔 𝑙𝑖𝑚𝑖𝑡𝑠 𝑜𝑓 𝑠𝑢𝑚𝑚𝑎𝑡𝑖𝑜𝑛
𝑛 = 5 𝑡𝑜 8 → 𝑛 = 𝑡𝑜 𝑁 − 1
2
𝑁−1
 𝐻 𝑒 𝑗𝜔 = 𝑛=0 ℎ(𝑛)𝑒
−𝑗𝜔𝑛

𝑁−3
2 𝑁−1

= ℎ(𝑛)𝑒 −𝑗𝜔𝑛 + ℎ(𝑛)𝑒 −𝑗𝜔𝑛

𝑛=0 𝑁+1
𝑛= 2

 Let 𝑚 = 𝑁 − 1 − 𝑛 → 𝑛 = 𝑁 − 1 − 𝑚
𝑁+1 𝑁+1 𝑁−3
𝑛= →𝑚 =𝑁−1− =
2 2 2
𝑛 =𝑁−1→𝑚 =𝑁−1− 𝑁−1 =0
 𝑁−3 𝑁−3
2 2
𝐻 𝑒 𝑗𝜔 = ℎ(𝑛)𝑒 −𝑗𝜔𝑛 + ℎ(𝑁 − 1 − 𝑚) 𝑒 −𝑗𝜔(𝑁−1−𝑚)
𝑛=0 𝑚=0

 Replace dummy variable with original variable n

𝑁−3 𝑁−3
2 2
𝐻 𝑒 𝑗𝜔 = ℎ(𝑛)𝑒 −𝑗𝜔𝑛 + ℎ(𝑁 − 1 − 𝑛) 𝑒 −𝑗𝜔(𝑁−1−𝑛)
𝑛=0 𝑛=0

 For antisymmetric impulse response ℎ 𝑛 = −ℎ 𝑁 − 1 − 𝑛

𝑁−3 𝑁−3
2 2
𝐻 𝑒 𝑗𝜔 = ℎ(𝑛)𝑒 −𝑗𝜔𝑛 − ℎ(𝑛) 𝑒 −𝑗𝜔(𝑁−1−𝑛)
𝑛=0 𝑛=0
𝑁−3 𝑁−3
𝑁−1 2 2
𝑁−1 𝑁−1
−𝑗𝜔 𝑗𝜔
=𝑒 2 ℎ(𝑛)𝑒 2 −𝑛 − ℎ(𝑛)𝑒
−𝑗𝜔 2 −𝑛
𝑛=0 𝑛=0
 𝑁−3
2
𝑗𝜔 −𝑗𝜔
𝑁−1 𝑁−1
𝐻 𝑒 =𝑒 2 𝑗 2ℎ(𝑛) sin 𝜔 −𝑛 ……………Eq(14)
2
𝑛=0

Note the imaginary part

𝜋
𝑗2 𝜋 𝜋
𝑗=𝑒 = 𝑐𝑜𝑠 + 𝑗𝑠𝑖𝑛 = 0 + 𝑗1 →
2 2
𝑁−1 𝜋 𝑁−3
−𝑗𝜔 𝑗 𝑁−1
𝐻 𝑒 𝑗𝜔 =𝑒 2 𝑒 2 2
𝑛=0 2ℎ(𝑛) sin 𝜔 −𝑛
2
𝑁−3
2
𝜋 𝑁−1
𝑗 2− 2 𝜔 𝑁−1
= 𝑒 2ℎ(𝑛) sin 𝜔 −𝑛
2
𝑛=0

Note that the phase is

Amplitude response
in the form of 𝑒 𝑗 𝛽−𝛼𝜔
𝐴 𝜔 = 𝐻(𝑒 𝑗𝜔 )
 Amplitude response is
Antisymmetric with 𝜔 = 𝜋
Zero at 𝜔 = 0 and 𝜔 = 𝜋
Case iv: Antisymmetric impulse response, N even
𝑁 =8

ℎ 𝑛 is splitted with respect to the center of symmetry 𝑛 =3.5

𝑁−2
𝑛 = 0 𝑡𝑜 3 → 𝑛 = 0 𝑡𝑜
2
𝑁 ← 𝑓𝑖𝑛ⅆ𝑖𝑛𝑔 𝑙𝑖𝑚𝑖𝑡𝑠 𝑜𝑓 𝑠𝑢𝑚𝑚𝑎𝑡𝑖𝑜𝑛
𝑛 = 4 𝑡𝑜 7 → 𝑛 = 𝑡𝑜 𝑁 − 1
2
𝑁−1
 𝐻 𝑒 𝑗𝜔 = 𝑛=0 ℎ(𝑛)𝑒
−𝑗𝜔𝑛
𝑁−2
2 𝑁−1

= ℎ(𝑛)𝑒 −𝑗𝜔𝑛 + ℎ(𝑛)𝑒 −𝑗𝜔𝑛

𝑛=0 𝑁
𝑛= 2
 Let 𝑚 = 𝑁 − 1 − 𝑛 → 𝑛 = 𝑁 − 1 − 𝑚
𝑁 𝑁 𝑁 𝑁−2
𝑛 = →𝑚 =𝑁−1− = −1=
2 2 2 2
𝑛 =𝑁−1→𝑚 =𝑁−1− 𝑁−1 =0
 𝑁−2 𝑁−2
2 2
𝐻 𝑒 𝑗𝜔 = ℎ(𝑛)𝑒 −𝑗𝜔𝑛 + ℎ(𝑁 − 1 − 𝑚) 𝑒 −𝑗𝜔(𝑁−1−𝑚)
𝑛=0 𝑚=0
 Replace dummy variable with original variable n
𝑁−2 𝑁−2
2 2
𝐻 𝑒 𝑗𝜔 = ℎ(𝑛)𝑒 −𝑗𝜔𝑛 + ℎ(𝑁 − 1 − 𝑛) 𝑒 −𝑗𝜔(𝑁−1−𝑛)
𝑛=0 𝑛=0
 For antisymmetric impulse response ℎ 𝑛 = −ℎ 𝑁 − 1 − 𝑛
𝑁−2 𝑁−2
2 2
𝐻 𝑒 𝑗𝜔 = ℎ(𝑛)𝑒 −𝑗𝜔𝑛 − ℎ(𝑛) 𝑒 −𝑗𝜔(𝑁−1−𝑛)
𝑛=0 𝑛=0
𝑁−2 𝑁−2
2 2
𝑁−1 𝑁−1 𝑁−1
−𝑗𝜔 𝑗𝜔
=𝑒 2 ℎ(𝑛)𝑒 2 −𝑛 − ℎ(𝑛)𝑒
−𝑗𝜔 2 −𝑛
𝑛=0 𝑛=0
𝑁−2
2
−𝑗𝜔
𝑁−1 𝑁−1
=𝑒 2 𝑗 2ℎ(𝑛) sin 𝜔 −𝑛 ……………Eq(15)
2
𝑛=0
Note the imaginary part
 𝜋
𝑗2 𝜋 𝜋
𝑗=𝑒 = 𝑐𝑜𝑠 + 𝑗𝑠𝑖𝑛 = 0 + 𝑗1 →
2 2
𝑁−2
2
−𝑗𝜔
𝑁−1
𝑗
𝜋 𝑁−1
𝐻 𝑒 𝑗𝜔 =𝑒 2 𝑒 2 2ℎ(𝑛) sin 𝜔 −𝑛
2
𝑛=0
𝑁−2
2
𝜋 𝑁−1
𝑗 2− 2 𝜔 𝑁−1
= 𝑒 2ℎ(𝑛) sin 𝜔 −𝑛
2
𝑛=0
Note that the phase is
in the form of 𝑒 𝑗 𝛽−𝛼𝜔 Amplitude response
𝐴 𝜔 = 𝐻(𝑒 𝑗𝜔 )

 Amplitude response is
Symmetric with 𝜔 = 𝜋
Zero at 𝜔 = 0
Non zero at 𝜔 = 𝜋
Frequency response of digital filters

Non zero at 𝜔 = 0 Zero at 𝜔 = 0

Zero at 𝜔 = 𝜋 Zero at 𝜔 = 𝜋

Zero at 𝜔 = 0 Non zero at 𝜔 = 0

Non zero at 𝜔 = 𝜋 Non zero at 𝜔 = 𝜋
Properties of Linear phase Filters
 Case i [Amplitude of frequency response : Non zero at 𝜔 = 0 and 𝜔 = 𝜋]
linear phase filters are versatile and can be used to design almost all types of
filters
 Case ii [Amplitude of frequency response : Non zero at 𝜔 = 0 and Zero at 𝜔 =
𝜋] linear phase filter are suitable for lowpass filters but cannot be used to
design a highpass filter
 Case iii [Amplitude of frequency response : Zero at 𝜔 = 0 and 𝜔 = 𝜋] linear
phase filter are suitable for designing bandpass filters but cannot be used to
design a bandstop filter
• Case iv [Amplitude of frequency response : Zero at 𝜔 = 0 Non zero at 𝜔 = 𝜋]
linear phase filter are suitable for designing highpass filters but cannot be used
to design a low pass filter

Note : For more understanding compare the values of frequency response at 𝜔 =

0 and 𝜔 = π
Note : The frequency response of case iii and case iv filters are imaginary and are
suitable for use as
− Differentiators 𝐻 𝑗𝜔 = 𝑗𝜔
− Hilbert transformers 𝐻 𝑗𝜔 = −𝑗 𝑠𝑔𝑛 𝜔
−1 𝑓𝑜𝑟 𝑥 < 0
𝑤ℎ𝑒𝑟𝑒 𝑠𝑖𝑔𝑛𝑢𝑚 𝑓𝑢𝑛𝑐𝑡𝑖𝑜𝑛 𝑠𝑔𝑛 𝑥 = 0 𝑓𝑜𝑟 𝑥 = 0
1 𝑓𝑜𝑟 𝑥 > 0
Thank You