Convolution and

Fourier Transform
 Continuous Time Signals
• Recall the property of delta function:
∞

න 𝑥(𝜏)𝛿 𝜏 𝑑𝜏 = 𝑥(0)
−∞
• In general,
∞

𝑥 𝑡 = න 𝑥(𝑡 − 𝜏)𝛿 𝜏 𝑑𝜏
−∞
• Change of dummy variable,
∞

𝑥 𝑡 = න 𝑥(𝜏)𝛿 𝑡 − 𝜏 𝑑𝜏
−∞
• Notice that the last integral can be viewed as a limiting sum of weighted
and shifted unit impulses.
 Convolution
• We define the convolution of 2 functions 𝑥1(𝑡)
and 𝑥2(𝑡) as
∞
𝑥1 ∗ 𝑥2 𝑡 = න 𝑥1 𝜏 𝑥2 𝑡 − 𝜏 𝑑𝜏
−∞
• Notice that
∞ ∞
𝑥1 ∗ 𝑥2 𝑡 = න 𝑥1 𝜏 𝑥2 𝑡 − 𝜏 𝑑𝜏 = න 𝑥2 𝜏 𝑥1 𝑡 − 𝜏 𝑑𝜏 = 𝑥2 ∗ 𝑥1 (𝑡)
−∞ −∞

• Notice that 𝑥(𝑡) can be viewed as a

convolution of the delta function with itself.
 Example
• 𝑥(𝑡) ∗ 𝑥(𝑡)
• 𝑥(𝑡)
• 𝑥(−𝑡)
 CT LTI System
• Consider a continuous time linear time-invariant system 𝒯 . .
• Denote its response to the unit impulse 𝛿(𝑡) as 𝒯 𝛿(𝑡) = ℎ(𝑡).
• We want to find the response to a general input 𝑥(𝑡).
∞
• View 𝑥(𝑡) as ‫׬‬−∞ 𝑥(𝜏)𝛿 𝑡 − 𝜏 𝑑𝜏, which is a limiting linear combination of
impulses.
• By linearity,
∞ ∞

𝒯 𝑥(𝑡) = 𝒯 න 𝑥(𝜏)𝛿 𝑡 − 𝜏 𝑑𝜏] = න 𝑥 𝜏 𝒯 𝛿 𝑡 − 𝜏 𝑑𝜏

−∞ −∞
• By time-invariance,
∞ ∞

𝒯 𝑥(𝑡) = න 𝑥 𝜏 𝒯 𝛿 𝑡 − 𝜏 𝑑𝜏 = න 𝑥 𝜏 ℎ 𝑡 − 𝜏 𝑑𝜏 = 𝑥 ∗ ℎ(𝑡)
−∞ −∞
• Hence, the output is the convolution of the input 𝑥(𝑡) and the impulse response
ℎ(𝑡).
 More on Convolution
• Commutative property
• 𝑥∗𝑦 =𝑦∗𝑥
-> No problem about which one goes first in your calculations.
• Distributive
• 𝑥∗ 𝑦+𝑧 =𝑥∗𝑦+𝑥∗𝑧
-> Parallel implementation of LTI systems
• Associative
• 𝑥 ∗ 𝑦 ∗ 𝑧 = (𝑥 ∗ 𝑦) ∗ 𝑧
-> just write 𝑥 ∗ 𝑦 ∗ 𝑧
• Proofs by writing out the detailed expressions
with appropriate changes of variables.
Continuous Time Fourier Transform
• Definition:
∞
𝑋(𝑓) = න 𝑥(𝑡)𝑒 −𝑗2𝜋𝑓𝑡 𝑑𝑡
−∞
∞
𝑥(𝑡) = න 𝑋(𝑓)𝑒 𝑗2𝜋𝑓𝑡 𝑑𝑓
−∞
• Frequency contents of a signal
 Examples
• 𝑥 𝑡 = 𝑒 −𝑎𝑡 𝑢 𝑡 where 𝑎 > 0
∞
−𝑎𝑡 −𝑗2𝜋𝑓𝑡
1
𝑋 𝑓 =න 𝑒 𝑒 𝑑𝑡 =
0 𝑎 + 𝑗2𝜋𝑓
• 𝑥 𝑡 =𝛿 𝑡
∞
𝑋 𝑓 = න 𝛿(𝑡)𝑒 −𝑗2𝜋𝑓𝑡 𝑑𝑡 = 1
−∞

1
1, 𝑡 <
2
• 𝑥 𝑡 =൞ 1
0, 𝑡 ≥
2
1/2
sin 𝜋𝑓
𝑋 𝑓 =න 𝑒 −𝑗2𝜋𝑓𝑡 𝑑𝑡 =
−1/2 𝜋𝑓
 Examples
• 𝑥 𝑡 = 𝑒 𝑎𝑡 𝑢 −𝑡 where 𝑎 > 0
0
1
𝑋 𝑓 = න 𝑒 𝑎𝑡 𝑒 −𝑗2𝜋𝑓𝑡 𝑑𝑡 =
−∞ 𝑎 − 𝑗2𝜋𝑓
• http://en.wikipedia.org/wiki/Wikipedia:Featured_picture_candidates/Fourier_transform_time_and_frequency_domains
 Properties
• Linearity
𝐹
𝑎𝑥 𝑡 + 𝑏𝑦 𝑡 ՞ 𝑎𝑋 𝑓 + 𝑏𝑌(𝑓)
• Duality
𝐹
𝑋(𝑡) ՞ 𝑥(−𝑓)
𝐹
– Since 𝛿(𝑡) ՞ 1,
𝐹
1 ՞ 𝛿 −𝑓 = 𝛿(𝑓)

• Time shifting
𝐹
𝑥(𝑡 − 𝑡0 ) ՞ 𝑒 −𝑗2𝜋𝑓𝑡0 𝑋(𝑓)
• Frequency shifting
𝐹
𝑗2𝜋𝑓 𝑡
𝑒 0 𝑥(𝑡) ՞ 𝑋(𝑓 − 𝑓0 )
 Properties
• Conjugation
𝐹
𝑥 ∗ (𝑡) ՞ 𝑋 ∗ (−𝑓)
– For real 𝑥 𝑡 , 𝑋 −𝑓 = 𝑋 ∗ (𝑓).
• Time / frequency scaling
1
𝐹 𝑓
𝑥 𝑎𝑡 ՞ 𝑋( )
|𝑎| 𝑎
– Put 𝑎 = −1 for time reversal.
• Differentiation in time
𝑑 𝐹
𝑥(𝑡) ՞ 𝑗2𝜋𝑓𝑋(𝑓)
𝑑𝑡
• Differentiation in frequency
𝐹 𝑗 𝑑
𝑡𝑥(𝑡) ՞ 𝑋(𝑓)
2𝜋 𝑑𝑓
 Properties
• Convolution
𝐹
𝑥(𝑡) ∗ ℎ 𝑡 ՞ 𝑋 𝑓 𝐻(𝑓)

• Multiplication
𝐹
𝑥 𝑡 𝑦(𝑡) ՞ 𝑋(𝑓) ∗ 𝑌(𝑓)

• Parseval’s relation
∞ ∞
න 𝑥1 (𝑡)𝑥2∗ (𝑡) 𝑑𝑡 = න 𝑋1 (𝑓)𝑋2∗ (𝑓) 𝑑𝑓
−∞ −∞
 Properties
• Integration in time
𝑒 −𝑎𝑡 , 𝑡>0
– Consider 𝑦 𝑡 = ൞ 0, 𝑡 = 0 where 𝑎 > 0.
−𝑒 𝑎𝑡 , 𝑡<0
𝐹 −𝑗4𝜋𝑓
𝑦(𝑡) ՞ 2
𝑎 + 4𝜋 2 𝑓 2
– Take limit as 𝑎 → 0,
1, 𝑡>0 𝐹 1
𝑠𝑔𝑛 𝑡 = ቐ 0, 𝑡 = 0՞
−1, 𝑡 < 0 𝑗𝜋𝑓
– Shifting up and scaling,
1 1𝐹 1 1
𝑢 𝑡 = 𝑠𝑔𝑛 𝑡 + ՞ + 𝛿(𝑓)
2 2 𝑗2𝜋𝑓 2
– Integration as convolution with 𝑢(𝑡),
𝑡 𝐹 1 1
න 𝑥(𝑠) 𝑑𝑠 = 𝑥 𝑡 ∗ 𝑢 𝑡 ՞ 𝑋 𝑓 + 𝑋(0)𝛿(𝑓)
−∞ 𝑗2𝜋𝑓 2
 Examples
• Complex exponentials
𝐹
𝑒 𝑗2𝜋𝑓0𝑡 ՞ 𝛿 𝑓 − 𝑓0
• Sinusoids
𝑒 𝑗2𝜋𝑓0 𝑡 +𝑒 −𝑗2𝜋𝑓0 𝑡 𝐹 1 1
cos 2𝜋𝑓0 𝑡 = ՞ 𝛿 𝑓 − 𝑓0 + 𝛿 𝑓 + 𝑓0
2 2 2
𝑒 𝑗2𝜋𝑓0 𝑡 −𝑒 −𝑗2𝜋𝑓0 𝑡 𝐹 1 1
sin 2𝜋𝑓0 𝑡 = ՞ 𝛿 𝑓 − 𝑓0 − 𝛿 𝑓 + 𝑓0
2𝑗 2𝑗 2𝑗
• Modulation
𝐹 1 1
𝑥(𝑡) cos 2𝜋𝑓0 𝑡 ՞ 𝑋(𝑓 − 𝑓0 )+ 𝑋(𝑓 + 𝑓0 )
2 2
 Examples
• Triangular pulse
1
1, 𝑡 <
2 𝐹 sin 𝜋𝑓
– Recall 𝑥 𝑡 = ൞ 1 ՞𝑋 𝑗𝜔 =
𝜋𝑓
0, 𝑡 ≥
2

𝐹 sin 𝜋𝑓 2
– Consider 𝑥 ∗ 𝑥 𝑡 ՞
𝜋𝑓
The Fourier transform of a triangular pulse is a sinc-square
function.
 Fourier Transform of Periodic
Functions
• For a periodic signal 𝑥(𝑡),
– expressing it as a Fourier series:
∞

𝑥 𝑡 = ෍ 𝑎𝑘 𝑒 𝑗2𝜋𝑘𝑓0𝑡
𝑘=−∞
– taking CTFT on both sides:
∞

𝑋(𝑓) = ෍ 𝑎𝑘 𝛿 𝑓 − 𝑘𝑓0
𝑘=−∞
 Example
• Impulse train
∞

𝑥 𝑡 = ෍ 𝛿(𝑡 − 𝑛𝑇)
𝑛=−∞
• Fourier series
∞
1 2𝜋𝑘
𝑗 𝑇 𝑡
𝑥 𝑡 = ෍ 𝑒
𝑇
𝑘=−∞
• Fourier transform
∞
1 𝑘
𝑋 𝑓 = ෍ 𝛿(𝑓 − )
𝑇 𝑇
𝑘=−∞