EECE 301 Signals & Systems Prof.

Mark Fowler
Note Set #15
C-T Signals: Fourier Transform Properties Reading Assignment: Section 3.6 of Kamen and Heck

1/39

Course Flow Diagram

The arrows here show conceptual flow between ideas. Note the parallel structure between the pink blocks (C-T Freq. Analysis) and the blue blocks (D-T Freq. Analysis). New Signal Models
Ch. 1 Intro
C-T Signal Model Functions on Real Line System Properties LTI Causal Etc D-T Signal Model Functions on Integers

Ch. 3: CT Fourier Signal Models

Fourier Series Periodic Signals Fourier Transform (CTFT) Non-Periodic Signals

Ch. 5: CT Fourier System Models

Frequency Response Based on Fourier Transform

Ch. 6 & 8: Laplace Models for CT Signals & Systems

Transfer Function

New System Model

New System Model

Ch. 2 Diff Eqs

C-T System Model Differential Equations D-T Signal Model Difference Equations Zero-State Response Zero-Input Response Characteristic Eq.

Ch. 2 Convolution
C-T System Model Convolution Integral D-T System Model Convolution Sum

New System Model

New Signal Model Powerful Analysis Tool

Ch. 4: DT Fourier Signal Models

DTFT (for Hand Analysis) DFT & FFT (for Computer Analysis)

Ch. 5: DT Fourier System Models

Freq. Response for DT Based on DTFT

Ch. 7: Z Trans. Models for DT Signals & Systems

Transfer Function

New System Model

New System Model2/39

Fourier Transform Properties

Note: There are a few of these we wont cover. see Table on Website or the inside front cover of the book for them. I prefer that you use the tables on the website they are better than the books

As we have seen, finding the FT can be tedious (it can even be difficult) Butthere are certain properties that can often make things easier. Also, these properties can sometimes be the key to understanding how the FT can be used in a given application.

So even though these results may at first seem like just boring math they are important tools that let signal processing engineers understand how to build things like cell phones, radars, mp3 processing, etc.
3/39

1. Linearity (Supremely Important) If x (t ) X ( ) &

Gets used virtually all the time!!

y (t ) Y ( )

then [ax (t ) + by (t )] [aX ( ) + bY ( )] Another way to write this property:

F{ax (t ) + by (t )} = aF{x (t )}+ bF{y (t )}

To see why: F {ax (t ) + by (t )} =

[ax(t ) + by (t )]e jt dt
jt

Use Defn of FT

By standard Property of Integral of sum of functions

= a x(t )e

dt + b y (t )e jt dt
= Y ( )

= X ( )
By Defn of FT

4/39

Example Application of Linearity of FT: Suppose we need to find the FT of the following signal 2 x(t )

1
t

2 1

1 2

Finding this using straight-forward application of the definition of FT is not difficult but it is tedious:

F{x (t )} = e
2

jt

dt + 2 e
1

jt

dt + e jt dt
1

So we look for short-cuts: One way is to recognize that each of these integrals is basically the same Another way is to break x(t) down into a sum of signals on our table!!!

5/39

Break a complicated signal down into simple signals before finding FT:
p4 (t )

1
t

2 1

2
p2 (t )

Add to get

2 x(t ) 1 2
t

Mathematically we write: x(t ) = p4 (t ) + p2 (t )

X ( ) = P4 ( ) + P2 ( )

From FT Table we have a known result for the FT of a pulse, so

2 + 2 sinc X ( ) = 4 sinc
6/39

2. Time Shift (Really Important!)

Used often to understand practical issues that arise in audio, communications, radar, etc.

If

x (t ) X ( ) then

x (t c ) X ( )e jc

Note: If c > 0 then x(t c) is a delay of x(t) So what does this mean??
jc = X ( ) First it does nothing to the magnitude of the FT: X ( )e

That means that a shift doesnt change how much we need of each of the sinusoids we build with
jc } = X ( ) + e jc Second it does change the phase of the FT: { X ( )e

= X ( ) + c
Line of slope c Phase shift increases linearly as the frequency increases This gets added to original phase

Shift of Time Signal

Linear Phase Shift of Frequency Components

7/39

Example Application of Time Shift Property: Room acoustics. Practical Questions: Why do some rooms sound bad? Why can you fix this by using a graphic equalizer to boost some frequencies and cut others? Very simple case of a single reflection:
Reflecting Surface speaker ear

x(t c)
x(t )

c > 0 Delayed signal 0 < < 1 Attenuated signal

So You hear:

y (t ) = x(t ) + x(t c) instead of just x(t)

Use linearity and time shift to get the FT at your ear:

Y ( ) = F{x (t ) + x (t c )} = F{x (t )}+ F{x (t c )} = X ( ) + X ( )e jc

Y ( ) = X ( ) 1 + e jc

This is the FT of what you hear It gives an equation that shows how the reflection affects what you hear!!!!

8/39

The big picture!

Y ( ) = X ( ) 1 + e jc
H ( )

H ( ) changes shape of X ( )

The room changes how much of each frequency you hear

Lets look closer at |H()| to see what it does

Using Eulers formula gives Rectangular Form

H ( ) = 1 + e jc = 1 + cos(c ) j sin( c )

= (1 + cos(c )) 2 + 2 sin 2 (c ) = 1 + 2 cos(c ) + 2 cos2 (c ) + 2 sin 2 (c )

mag =

(Re )2 + (Im )2
Expand 1st squared term

= 2
Use Trig ID

| H ( ) |= (1 + 2 ) + 2 cos(c )
9/39

The big picture revisited:

Y ( ) = X ( ) (1 + ) + 2 cos(c )
2

Effect of the room what does it look like as a function of frequency?? The cosine term makes it wiggle up and down and the value of c controls how fast it wiggles up and down Spacing = 1/c Hz
Dip-to-Dip Peak-to-Peak The next 3 slides explore these effects

c controls spacing between dips/peaks

controls depth/height of dips/peaks

What is a typical value for delay c??? Speed of sound in air 340 m/s Typical difference in distance 0.167m

c=

0.167m = 0.5msec 340m/s

Spacing = 2 kHz
10/39

Attenuation: = 0.2

Delay: c = 0.5 ms (Spacing = 1/0.5e-3 = 2 kHz)

FT magnitude at the speaker (a made-up spectrum but kind of like audio)

|H()| the effect of the room

FT magnitude at your ear room gives slight boosts and cuts at closely spaced locations

Longer delay causes closer spacing so more dips/peaks over audio range!

11/39

Attenuation: = 0.8

Delay: c = 0.5 ms (Spacing = 1/0.5e-3 = 2 kHz)

FT magnitude at the speaker

|H()| the effect of the room

FT magnitude at your ear room gives large boosts and cuts at closely spaced locations

Stronger reflection causes bigger boosts/cuts!!

12/39

Attenuation: = 0.2

Delay: c = 0.1 ms (Spacing = 1/0.1e-3 = 10 kHz)

FT magnitude at the speaker

|H()| the effect of the room

FT magnitude at your ear room gives small boosts and cuts at widely spaced locations

Shorter delay causes wider spacing so fewer dips/peaks over audio range!

13/39

Matlab Code to create the previous plots

14/39

3. Time Scaling (Important) Q: If x(t ) X ( ) , then x( at ) ??? for A: x(at )

a0

1 X a a

If the time signal is Time Scaled by a

Then The FT is Freq. Scaled by 1/a

An interesting duality!!!

15/39

To explore this FT propertyfirst, what does x(at) look like?

x(t )
Original Signal

t 1
x(2t )

3 4

7 |a| > 1 squishes horizontally t

Time-Scaled w/ a = 2

3.5 1
1 x t 2

Time-Scaled w/ a = 1/2

|a| < 1 stretches horizontally t 1 2 3 4 5 6 7 8 9 10 11 12 13 14

|a| > 1 makes it wiggle faster need more high frequencies |a| < 1 makes it wiggle slower need less high frequencies
16/39

When |a| > 1 |1/a| < 1

Time Signal is Squished

x(at )

1 X a a
FT is Stretched Horizontally and Reduced Vertically

x(t )

Original Signal & Its FT

A

t
Squished Signal & Its FT
x(2t )
.5A

X ( )

1 X 2 2

17/39

When |a| < 1 |1/a| > 1

Time Signal is Stretched

x(at )

1 X a a
FT is Squished Horizontally and Increased Vertically

x(t )

Original Signal & Its FT

A

t
1 x t 2

X ( )

Stretched Signal & Its FT

2A

2X (2 )

Rough Rule of Thumb we can extract from this property: Duration Bandwidth Duration Bandwidth Very Short Signals tend to take up Wide Bandwidth
18/39

4. Time Reversal (Special case of time scaling: a = 1)

x(t ) X ( )
double conjugate Note: X ( ) =

x(t )e

j ( ) t

dt = x(t )e + jt dt

= No Change

= x(t )e + jt dt

Conjugate changes to j

= x(t) if x(t) is real

= x (t )e jt dt = X ( )

So if x(t) is real, then we get the special case:

Recall: conjugation doesnt change abs. value but negates the angle

X ( ) = X ( )

x ( t ) X ( )

X ( ) = X ( )
19/39

5. Multiply signal by tn

d n X ( ) t x (t ) ( j ) d n
n n

n = positive integer

This property is mostly useful for finding the FT of typical signals. Example Find X() for this x(t) -1
x(t )

1 -1
t

t
1

1 -1 Notice that:

t
1
p2 (t )

x(t ) = tp2 (t )

-1 1

t
1
20/39

-1

So we can use this property as follows:

From entry on FT From entry #8 in Table 3.2 with = = 2. Table with 2.

d d X ( ) = j P2 ( ) = j 2 sinc d d cos( ) sin( ) = j 2 2

Now how do you get the derivative of the sinc???

Use the definition of sinc and then use the rule for the derivative of a quotient you learned in Calc I:

d f ( x) g ( x) = dx

g ( x)

df ( x ) dg ( x ) f ( x) dx dx g 2 ( x)
21/39

6. Modulation Property Super important!!!

Essential for understanding practical issues that arise in communications, radar, etc.

There are two forms of the modulation property 1. Complex Exponential Modulation simpler mathematics, doesnt directly describe real-world cases 2. Real Sinusoid Modulation mathematics a bit more complicated, directly describes real-world cases

Eulers formula connects the two so you often can use the Complex Exponential form to analyze real-world cases

22/39

Complex Exponential Modulation Property

x(t )e j0t X ( 0 )
Multiply signal by a complex sinusoid Shift the FT in frequency

x (t )

X ( )

F{x (t )e jot } = X ( o )
A

o
23/39

Real Sinusoid Modulation Based on Euler, Linearity property, & the Complex Exp. Modulation Property

1 F{x (t ) cos(0t )} = F x (t )e j0t + x (t )e j0t 2 1 = F x (t )e j0t + F x (t )e j0t 2 1 = [X ( o ) + X ( + o )] 2

The Result:

Eulers Formula

Linearity of FT

[ {[

]} {[

]}]
Comp. Exp. Mod.

x (t ) cos(0t )

1 [X ( + 0 ) + X ( 0 )] 2
Shift Down Shift Up

Related Result:

j [X ( + 0 ) X ( 0 )] x (t ) sin(0t ) 2
Exercise:

x (t ) cos(0t + 0 ) ??

24/39

Visualizing the Result

x (t ) cos(0t )

1 [X ( 0 ) + X ( + 0 )] 2
Shift up Shift down

X ( )

Shift Up

Shift Down
F{x (t ) cos(0t )}

Interesting This tells us how to move a signals spectrum up to higher frequencies without changing the shape of the spectrum!!! What is that good for??? Well only high frequencies will radiate from an antenna and propagate as electromagnetic waves and then induce a signal in a receiving antenna. 25/39

Application of Modulation Property to Radio Communication FT theory tells us what we need to do to make a simple radio system then electronics can be built to perform the operations that the FT theory calls for:
Sound and microphone

Transmitter
Modulator

antenna

x (t )

amp

multiply
cos(0t )

amp

oscillator FT of Message Signal

X ( )

F{x (t ) cos(0t )}

Choose f0 > 10 kHz to enable efficient radiation (with 0 = 2f0 ) AM Radio: around 1 MHz FM Radio: around 100 MHz
26/39

Cell Phones: around 900 MHz, around 1.8 GHz, around 1.9 GHz etc.

The next several slides show how these ideas are used to make a receiver: Receiver Amp & Filter
De-Modulator Speaker

multiply oscillator

Signals from Other Transmitters

cos(0t )

Amp & Filter

Signals from Other Transmitters

0
The Filter removes the Other signals (Well learn about filters later)

27/39

Receiver

De-Modulator

Speaker

multiply oscillator

cos(0t )

Amp & Filter

By the Real-Sinusoid Modulation Property the De-Modulator shifts up & down: Shifted Up

2 0

2 0 2 0

0 0

0
Shifted Down

2 0

0
Add gives double

2 0 0

2 0

28/39

Receiver Amp & Filter

Extra Stuff we dont want

De-Modulator

Speaker

multiply oscillator

cos(0t )

Amp & Filter

2 0 0
The Filter removes the Extra Stuff

2 0

2 0 0

2 0

Speaker is driven by desired message signal!!!

29/39

So what have we seen in this example: Using the Modulation property of the FT we saw 1. Key Operation at Transmitter is up-shifting the message spectrum: a) FT Modulation Property tells the theory then we can build b) modulator = oscillator and a multiplier circuit 2. Key Operation at Transmitter is down-shifting the received spectrum a) FT Modulation Property tells the theory then we can build b) de-modulator = oscillator and a multiplier circuit c) But the FT modulation property theory also shows that we need filters to get rid of extra spectrum stuff i. So one thing we still need to figure out is how to deal with these filters ii. Filters are a specific system and we still have a lot to learn about Systems iii. That is the subject of much of the rest of this course!!!
30/39

7. Convolution Property (The Most Important FT Property!!!) The ramifications of this property are the subject of the entire Ch. 5 and continues into all the other chapters!!! It is this property that makes us study the FT!!

Mathematically we state this property like this:

x (t ) h (t )
Another way of stating this is:

X ( ) H ( )

F{x (t ) h(t )} = X ( ) H ( )
31/39

Now what does this mean and why is it so important??!! Recall that convolution is used to described what comes out of an LTI system:
x(t ) y (t ) = x(t ) h(t )

h(t) Now we can take the FT of the input and the output to see how we can view the system behavior in the frequency domain:
x(t ) y (t ) = x(t ) h(t )

h(t) FT
X ( )

FT
Y ( ) = X ( ) H ( )

Use the Conv. Property!! Systems H() changes the shape of the inputs X() via multiplication to create outputs Y()

It is easier to think about and analyze the operation of a system using this frequency domain view because visualizing multiplication is easier than visualizing convolution
32/39

Lets revisit our Room Acoustics example:

speaker ear

x (t )
amp

y (t ) = x (t ) * hroom (t )

Recall:

Y ( ) = X ( ) 1 + e jc
Hroom()
Plot of |Hroom()|

What we hear is not right!!!

33/39

So, we fix it by putting in an equalizer (a system that fixes things) Equalizer

x (t )

heq (t )

amp

x2 (t ) = x (t ) * heq (t )
X 2 ( ) = H eq ( ) X ( )
(by convolution property)

y (t ) = x2 (t ) * hroom (t ) = [x (t ) * heq (t )]* hroom (t )

Y ( ) = H room ( ) X 2 ( ) = H room ( ) H eq ( ) X ( )
(by convolution property, applied twice!)

Then: Y ( ) = X ( ) H eq ( ) 1 + e jc
Recall: Peaks and dips

Want this whole thing to be = 1 so Y ( ) = X ( )

34/39

Equalizers |Heq()| should peak at frequencies where the rooms |Hroom()| dips and vice versa

Room

Equalizer

Room & Equalizer

35/39

8. Multiplication of Signals

1 1 x(t ) y (t ) X ( ) Y ( ) = 2 2

X ( )Y ( )d

This is the dual of the convolution property!!!

Convolution in the Time-Domain

gives

Multiplication in the Frequency-Domain

Multiplication in the Time-Domain

gives

Convolution in the Frequency-Domain

36/39

9. Parsevals Theorem (Recall Parsevals Theorem for FS!)

1 x(t ) dt = 2
2

X ( ) d
2

Energy computed in time domain

Energy computed in frequency domain

x(t ) dt
= energy at time t

= energy at freq.

d X ( ) 2
2

Generalized Parsevals Theorem:

1 x(t ) y(t )dt = 2

X ( )Y ( )d

37/39

10. Duality:

X ( ) = x(t )e jt dt

x (t )
1 x(t ) = 2

X ( )

X ( )e jt d

Both FT & IFT are pretty much the same machine:

c f ( )e j d

So if there is a time-to-frequency property we would expect a virtually similar frequency-to-time property Illustration: Delay Property:

x(t c) X ( )e jc

Modulation Property: Other Dual Properties:

x(t )e j0t

X ( 0 )

(Multiply by tn) vs. (Diff. in time domain) (Convolution) vs. (Mult. of signals)
38/39

Also, this duality structure gives FT pairs that show duality. Suppose we have a FT table that a FT Pair A we can get the dual Pair B using the general Duality Property: 1. Take the FT side of (known) Pair A and replace by t and move it to the time-domain side of the table of the (unknown) Pair B. 2. Take the time-domain side of the (known) Pair A and replace t by , multiply by 2, and then move it to the FT side of the table of the (unknown) Pair B. Here is an example We found the FT pair for the pulse signal:

Pair A

p (t )
Step 1

sinc 2
Step 2

Here we have used the fact that p(-) = p()

Pair B

t sinc 2

2 p ( )
39/39