inverse dft. frequency to time domain sometimes calculations are easier in the frequency domain then...

8
Inverse DFT

Upload: myron-clark

Post on 14-Jan-2016

214 views

Category:

Documents


2 download

TRANSCRIPT

Page 1: Inverse DFT. Frequency to time domain Sometimes calculations are easier in the frequency domain then later convert the results back to the time domain

Inverse DFT

Page 2: Inverse DFT. Frequency to time domain Sometimes calculations are easier in the frequency domain then later convert the results back to the time domain

Frequency to time domain

• Sometimes calculations are easier in the frequency domain then later convert the results back to the time domain

• Convert Time -> Frequency with DFT

• Convert Frequency -> Time with the Inverse Discrete Fourier Transform

Page 3: Inverse DFT. Frequency to time domain Sometimes calculations are easier in the frequency domain then later convert the results back to the time domain

• From Last week, the DFT is:

21

0

( ) ( )j hnNN

n

F h x n e

• The IDFT is:

21

0

1( ) ( )

j hnNN

h

x n F h eN

Where x is effectively a row matrix of size Nh is the required harmonicN is number of Fourier coefficientsF(h) is the complex DFT value

Page 4: Inverse DFT. Frequency to time domain Sometimes calculations are easier in the frequency domain then later convert the results back to the time domain

• To speed up the manual analysis, remember:

0

2

3

2

2

cos( ) sin( )

cos(0) sin(0) 1

cos( ) sin( )2 2

cos( ) sin( ) 1

3 3cos( ) sin( )

2 2

cos(2 ) sin(2 ) 1

j

j

j

j

j

j

e j

e j

e j j

e j

e j j

e j

• Relate this to the argand diagram…

Page 5: Inverse DFT. Frequency to time domain Sometimes calculations are easier in the frequency domain then later convert the results back to the time domain

• Similarly

0

2

3

2

2

cos( ) sin( )

cos(0) sin(0) 1

cos( ) sin( )2 2

cos( ) sin( ) 1

3 3cos( ) sin( )

2 2

cos(2 ) sin(2 ) 1

j

j

j

j

j

j

e j

e j

e j j

e j

e j j

e j

• So the vector rotates clockwise

Page 6: Inverse DFT. Frequency to time domain Sometimes calculations are easier in the frequency domain then later convert the results back to the time domain

Example• Consider the 4 DFT values generated from last

week’s example: {2,1+j,0,1-j}

21

0

2 01

0

2 1 2 0 2 1 2 334 4 4

0

3 3

2 2 2 2

1( ) (0) ( ) for 4

1 2 (1 ) 0 (1 )(0) (0) 1

4 4

1 1(1) (1) 2 (1 ) 0 (1 )

4 4

1 22 0

4

j hnNN

h

j hNN

h

j j j j

N

h

j j j j

x nT x X k e NN

j jx F e

x F e e j e j e

j j je je e je

00

4

Show that the other 2 terms for k=2 and k=3 are 0 and 1

j j j

Page 7: Inverse DFT. Frequency to time domain Sometimes calculations are easier in the frequency domain then later convert the results back to the time domain

DFT processing cost

• DFT processing cost is expensive– Each term is a product of a complex number– Each term is added so for an 8 point DFT

need 8 multiplies and 7 adds (N and N-1)– There are 8 harmonic components to be

evaluated (h=0 to 7)– So an 8 point DFT requires 8x8 complex

multiplications and 8x7 complex additions– An N point transform needs N2 Complex

multiplications and N(N-1) complex adds

Page 8: Inverse DFT. Frequency to time domain Sometimes calculations are easier in the frequency domain then later convert the results back to the time domain

Fast Fourier Transform

• Processing cost for DFT is:

2 2log Complex multiplications log Complex additions2

NN N N

2 Complex multiplications ( 1) Complex additionsN N N

• Processing cost for FFT is:

• 1024 point:

DFT: 1048576x and 1047552+

FFT: 5120x and 10240+