Skip to content

Agenda

DT Periodicity DT Fourier (DTFS) - DFT - FFT

DT Periodicity

we say x:Z->R or C is p-periodic if Ep element of {1, 2, 3, ...} s.t. x[n+p] = x[n] for all n element of Z

The smallest T> 0 for which the above is true is called the fundamental period.

w_0 = 2pi/p fundamental freq rad/sample p:samples

Ex: x[n] = 1 An element of Z

p = 1 sample fundamental period w_0 = 2pi/p = 2pi rad/sample

draw the signal x[n] in ascii below

Ex: x[n] = cos(pi/5 n)

For dt, you first figure out the period, and then the frequency. for CT its not restricted, because the freq is right in the exp.

to find the period,

x[n+p] = cos(pi/5 (n + p)) = cos(pi/5 n + pi/5 p) = cos(pi/5 n) = x[n]

cos(pi/5 n + pi/5 p) = cos(pi/5 n)

so pi/5 p = 2pik where k elem of z

so p = 2 * 5 * k = 10k

to find the fundamental period, make k smallest int to make p an integer, we let k be 1

so fundamental period is 10 so w_0 = 2pi/10 = pi/5

one more

Ex: x[n] = cos(3pi/5 n)

x[n + p] = cos(3pi/5 n + 3pi/5 p) = cos(3pi/5 n) = x[n]

so 3pi/5 p = 2pik

so p = 10k/3 so k = 3 to make p an int -> p becomes 10 samples

then w_0 = 2pi/10 = pi/5 rad/sample

Ex: x[n] = e^in

x[n + p] = e^i(n+p) = e^in * e^ip

want this to be e^in

so e^ip = 1

so p = 2pik

no integer value for k exists that can make p an integer

therfore x is not periodic

lets look at a complex exponential

lets say e^i(pi/4)n

draw on the unit circle in ascii: - x[0] - x[1] - x[2] - ... - x[7]

Ex: e^iw_0n -> e^iw_0(n+p) = e^iw_0n * e^iw_0p

x[n+p] = x[p] => e^iw_0p = 1 => w_0p = 2pik

w_0 = 2pik/p

w_0 must be a rational multiple of pi

Fourier Analysis

  1. Decomposition of a signal into a linear combination of its constituent frequencies.
    • We have already seen how we can split a sine or a cosine into a linear combination of its constituents
  2. We decompose a signal into a linear combination of an orthogonal basis
    • We aren't decomposing it into any random OB, but one where each basis signal is related directly to (represents) a constituent frequency
    • It's a change of basis, to put it more simply.

those are the two viewpoints.

Fourier Analysis requires 4 types of transforms

DT signals CT signals Periodic signals Aperiodic signals.

make a punnett Square in ascii and add the elements

DT-P = DTFS CT-P = CTFS DT-AP = DTFT CT-AP = CTFT

Ex: 2 periodic DT signals where x[n+2] + x[n] for all n element of Z

p = 2 w_0 = 2pi/2 = pi

imagine it oscillates between x[0] = 4, x[1] = 2, x[2] = 4, x[-1] = 2, etc. Draw in ascii

on the interval [0, 1] if i were to ask you "can you find a 2d cartesian vector representation of the signal in that period." would you be able to make it?

x = [4 2]

x = 4[1 0] + 2[0 1]

can you draw each of these signals in the period of interest in ascii?

for phi_0[n]: x[0] = 1, x[1] = 0, x[-1] = 0, x[2] = 1, ...

for phi_1[n]: x[0] = 0, x[1] = 1, x[-1] = 1, x[2] = 0, ...

x = 4phi_0 + 2phi_1

In Cartesian space, in R^2, we have this vector x=[4, 2], and what we have done so far is decompose it into a linear combination of phi_0 = [1 0] and phi_1 = [0, 1]

Now we are going to turn it into a problem. We ask to do something else.

lets represent the signal with x = [4 2]: explain the vector in terms of the basis {[1 1] and [1 -1]} this is an orthogonal basis

[4 2] = [a b] [[1 1] [1 -1]]

we are looking for the coefficient in the eqn x = X_oPsi_0 + X_1Psi_1 where psi_0 ortho psi_1

=

to determine X_0 project x onto psi_0

X_0 = proj_psi_0 X_0 = / X_1 = /

4_0 =