Fourier Transform Cheat Sheet
Fourier Transform Cheat Sheet
Convolution:
e2πit = cos 2πt + i sin 2πt
Z ∞
e2πit
+e −2πit
e 2πit −2πit
−e (f ∗ g)(x) = f(x − y)g(y) dy
cos 2πt = , sin 2πt = −∞
2 2i
Polar form: – f ∗g = g∗f
p
z = x + iy z = reiθ , r= x2 + y2 , θ = tan−1 (y/x) – (f ∗ g) ∗ h = (f ∗ g) ∗ h
– f ∗ (g + h) = f ∗ g + f ∗ h
Symmetric sum of complex exponentials (special
case of geometric series): Smoothing: If f (or g) is p-times continuously dif-
ferentiable, p ≥ 0, then so is f ∗ g and
X
N
sin(2N + 1)πt
e2πint = dk dk
sin πt (f ∗ g) = ( k f) ∗ g
n=−N k
dx dx
Fourier series If f(t) is periodic with period T its
Fourier series is Convolution Theorem:
∞
X F(f ∗ g) = (Ff)(Fg)
f(t) = cne2πint/T
n=−∞ F(fg) = Ff ∗ Fg
1
Autocorrelation: Let g(x) be a function satisfying Linearity: F{αf(x) + βg(x)} = αF (s) + βG(s)
R∞ 2
|g(x)| dx < ∞ (finite energy) then 1
−∞ Stretch: F{g(ax)} = |a|
G( as )
Z ∞
(g ? g)(x) = g(y)g(y − x) dy Shift: F{g(x − a)} = e−i2πas G(s)
−∞
1 −i2πsb/a
= g(x) ∗ g(−x) Shift & stretch: F{g(ax − b)} = |a| e G( as )
Cross correlation: Let g(x) and h(x) be functions with Rayleigh (Parseval):
finite energy. Then Z ∞ Z ∞
2
Z ∞ |g(x)| dx = |G(s)|2 ds
−∞ −∞
(g ? h)(x) = g(y)h(y + x) dy Z ∞ Z ∞
−∞ f(x)g(x) dx = F (s)G(s) ds
Z ∞
−∞ −∞
= g(y − x)h(y) dy
−∞
Modulation:
= (h ? g)(−x)
F{g(x) cos(2πs0 x)} = 12 [G(s − s0 ) + G(s + s0 )]
Rectangle and triangle functions
( ( Autocorrelation: F{g ? g} = |G(s)|2
1, |x| < 12 1 − |x|, |x| ≤ 1
Π(x) = Λ(x) =
0, |x| ≥ 21
0, |x| ≥ 1 Cross Correlation: F{g ? f} = G(s)F (s)
sin πs Derivative:
FΠ(s) = sinc s = , FΛ(s) = sinc2 s
πs – F{g0 (x)} = 2πisG(s)
Scaled rectangle function – F{g(n) (x)} = (2πis)n G(s)
( i n (n)
1, |x| < p – F{xng(x)} = ( 2π ) G (s)
Πp (x) = Π(x/p) = 2 ,
p
0, |x| ≥ 2 Moments: Z ∞
f(x)dx = F (0)
FΠp (s) = p sinc ps −∞
Z ∞
Scaled triangle function i 0
xf(x)dx = F (0)
( −∞ 2π
1 − |x/p|, |x| ≤ p Z ∞
Λp (x) = Λ(x/p) = , i n (n)
0, |x| ≥ p xnf(x)dx = ( ) F (0)
−∞ 2π
2
FΛp (s) = p sinc ps Miscellaneous:
Gaussian Z x
2 2 G(s)
F(e−πt ) = e−πs F g(ξ)dξ = 12 G(0)δ(s) +
−∞ i2πs
Gaussian with variance σ2
The Delta Function: δ(x)
1 2 2 2
σ 2 s2
g(t) = √ e−x /2σ Fg(s) = e−2π 1
σ 2π Scaling: δ(ax) = |a|
δ(x)
One-sided exponential decay R∞
Sifting: −∞
δ(x − a)f(x)dx = f(a)
(
0, t < 0, 1 R∞
f(t) = Ff(s) = −∞
δ(x)f(x + a)dx = f(a)
e−at , t ≥ 0. a + 2πis
Convolution: δ(x) ∗ f(x) = f(x)
Two-sided exponential decay
δ(x − a) ∗ f(x) = f(x − a)
2a
F(e−a|t| ) =
a2 + 4π2s2 δ(x − a) ∗ δ(x − b) = δ(x − (a + b))
Fourier Transform Theorems Product: h(x)δ(x) = h(0)δ(x)
2
Fourier Transform: Fδ = 1 The complex Fourier series representation:
∞
X
F(δ(x − a)) = e−2πisa n
p(x) = αne2πi L x
n=−∞
Derivatives:
R∞ where
– −∞
δ (n)(x)f(x)dx = (−1)n f (n) (0)
– δ 0(x) ∗ f(x) = f 0 (x) 1 n
αn = F( )
L L
– xδ(x) = 0
0 Z L/2
– xδ (x) = −δ(x) 1 n
= p(x)e−2πi L xdx
L −L/2
Fourier transform of cosine and sine
1 Linear Systems Let L be a linear system, w(t) = Lv(t),
F cos 2πat =
(δ(s − a) + δ(s + a)) with impulse response h(t, τ ) = Lδ(t − τ ).
2
1 Superposition integral:
F sin 2πat = (δ(s − a) − δ(s + a))
2i Z ∞
Unit step and sgn w(t) = v(τ )h(t, τ )dτ
( −∞
0, t ≤ 0 1 1
H(t) = FH(s) = δ(s) + A system is time-invariant if:
1, t > 0 2 πis
( w(t − τ ) = L[v(t − τ )]
−1, t<0 1
sgn t = Fsgn (s) =
1, t>0 πis In this case L(δ(t − τ ) = h(t − τ ) and L acts by
convolution:
The Shah Function: III(x) Z ∞
∞
X ∞
X w(t) = Lv(t) = v(τ )h(t − τ )dτ
III(x) = δ(x − n) , IIIp (x) = δ(x − np) −∞
3
The vector complex exponentials are orthogonal: Inverse Hilbert Transform H−1 f = −Hf
( 1
0, k 6≡ ` mod N Impulse response: − πx
ωk · ω` =
N, k ≡ ` mod N Transfer function: i sgn (s)
Causal functions: g(x) is causal if g(x) = 0 for x < 0.
The DFT of order N accepts an N -tuple as input
A casual signal Fourier Transform G(s) = R(s) +
and returns an N -tuple as output. Write an N -tuple
iI(s), where I(s) = H{R(s)}.
as f = (f[0], f[1], . . ., f[N − 1]).
Analytic signals: The analytic signal representation
N
X −1
of a real-valued function v(t) is given by:
Ff= f[k]ω−k
k=0 Z(t) = F −1 {2H(s)V (s)}
Inverse DFT: = v(t) − iHv(t)
1 X
N −1
Narrow Band Signals: g(t) = A(t) cos[2πf0 t + φ(t)]
F −1 f = f[k]ωk
N
k=0 Analytic approx: z(t) ≈ A(t)ei[2πf0 t+φ(t)]
Envelope: | A(t) |=| z(t) |
Periodicity of inputs and outputs: If F = F f then
both f and F are periodic of period N . Phase: arg[z(t)] = 2πf0 t + φ(t)
1 d
Instantaneous freq: fi = f0 + 2π dt
φ(t)
Convolution
Higher Dimensional Fourier Transform In n-
X
N −1
dimensions:
(f ∗ g)[n] = f[k] g[n − k] Z
k=0
Ff(ξ) = e−2πix·ξ f(x) dx
Rn
Discrete δ:
( Inverse Fourier Transform:
1, m≡k mod N Z
−1
δ k [m] = F f(x) = e2πix·ξ f(ξ) dξ
0, m 6≡ k mod N Rn
In 2-dimensions (in coordinates):
DFT of the discrete δ F δ k = ω−k Z ∞ Z ∞
DFT of vector complex exponential F ωk = N δ k Ff(ξ1 , ξ2) = e−2πi(x1 ξ1 +x2 ξ2 ) f(x1 , x2) dx1dx2
−∞ −∞
Reversed signal: f − [m] = f[−m] F f − = (F f)− The Hankel Transform (zero order):
DFT Theorems Z ∞
F (ρ) = 2π f(r)J0 (2πrρ)rdr
Linearity: F {αf + βg) = αF f + βF g 0
4
Shift: Let (τb f)(x) = f(x − b). Then
F(τb f)(ξ) = e−2πiξ·b Ff(ξ)