Frequency Response

Lecture #12
Chapter 10

BME 310 Biomedical Computing - 299

J.Schesser
 Ideal Filters
• We want to study H(jω) functions which
provide frequency selectivity such as:
– Low Pass
– High Pass
– Band Pass
• However, we will look a ideal filtering, that is,
filter which have ideal performance but are
very difficult to construct.

BME 310 Biomedical Computing - 300

J.Schesser
 A simple Filter – Ideal Delay
• Ideal Delay Filter => y(t) = x(t – td): the output is same as the input
except shifted in time by an amount td seconds.
• The impulse response is just h(t) = δ(t – td)
• The frequency response is then
 
H ( j )   h(t )e  jt dt    (t  t d )e  jt dt Alternatively, if x(t )  e j (t  ) , then
 
y (t )  x(t  t d )  e j[ (t td )  ]  e  jtd e j (t  )
 e  jtd
But y (t )  H ( j )e j (t  )
And therefore,
H ( j )  e  jtd

• The Frequency Response of an Ideal Delay filter has a constant

magnitude with a phase that is linear with frequency
• Therefore, it does not affect the magnitude of the input. It only effects the
phase by an amount of -ωtd
BME 310 Biomedical Computing - 301
J.Schesser
 Example
A signal of the form x(t )  10e j 4 e j 200t is input to an ideal
delay filter with delay of 0.001 sec.
The frequency response is : H ( j )  e  j 0.001
H ( j 200 )  e  j 200 ( 0.001)
Then the output signal becomes :
y (t )  H ( j ) x(t )  H ( j 200 )10e j 4 e j 200t
 e  j 200 ( 0.001)10e j 4 e j 200t  e  j 0.2 10e j 4 e j 200t

j ( 200t   0.2 )
 10e 4
 10e j ( 200t  0.05 )

j [ 200 ( t  0.001)  ]
Or rewritten as : y (t )  10e 4

BME 310 Biomedical Computing - 302

J.Schesser
 Ideal Low Pass Filter
• This filter only passes frequencies below a value ωco and
attenuates all frequencies above ωco.
• We call ωco the cutoff frequency.
• Therefore, the frequency response of a low pass filter is:

1   co
H lp ( j ) 
0   co
ωco ωco

BME 310 Biomedical Computing - 303

J.Schesser
 Ideal High Pass Filter
• This filter only passes frequencies above a value ωco and
attenuates all frequencies below ωco.
• We call ωco the cutoff frequency.
• Therefore, the frequency response of a high pass filter is:

1   co
H hp ( j ) 
0   co
ωco ωco

BME 310 Biomedical Computing - 304

J.Schesser
 Ideal Band Pass Filter
• This filter only passes frequencies above a value ωco1 and
below a value ωco2 and attenuates all other frequencies
outside this range.
• We call ωco1 the lower (or low) cutoff frequency and
ωco2 the upper (or high) cutoff frequency.
• Therefore, the frequency response of a bandpass filter is:

0   co1
H bp ( j )  1 co1    co 2
0   co 2
-ωco2 ωco1 ωco1 ωco2

BME 310 Biomedical Computing - 305

J.Schesser
 Application of Ideal Filters
• We will apply a band pass filter to a periodic
square wave filter out its fundamental
frequency.
• Let our input signal have a period of
To = 500µs or fo=2kHz => ωo=2π(2000) rad/sec
and its form over one period is:
2 0 ≤ t < To / 2 x(t)
2

x(t) = t

0 To / 2 ≤ t < To -2To -To –To/2 0 To/2 To 2To

BME 310 Biomedical Computing - 306

J.Schesser
 Application of Ideal Filters
• Since x(t) is a period, let’s calculate the Fourier
series for to decompose the input into its
frequency components.
 1 1 2 2  j /2
x(t )  a e
k  
k
jko t
ak 
jk
[1 e  jk ] 
jk
[1  (1)k ]  
jk k
e for odd values of k

To 2 To 2 0 for even values of k

1 1
 x(t )e  2e
 jk o t  jko t
ak  dt  dt 1 o
T 2
1
To 2
To To 2
To 0 a0  
To To 2
x(t)dt 
To  0
2 dt  1
2 To 2
2  jk o t To 2 1  jk  jk 0
 e |0  [e To 2
e To
]
To ( jko )  jk 1

1
 [1  e  jk ]
jk
Recall that e  jk  cos k  j sin k  1 for even values of k
 1 for odd values of k 0
 jk
 (1) k -4 -3 -2 -1 0 1 2 3 4
or e
k

BME 310 Biomedical Computing - 307

J.Schesser
 Application of Ideal Filters
• Now let’s apply an ideal band pass filter with low
frequency cutoff of 1,250 Hz and high frequency cutoff
of 2,750 Hz which has a bandwidth of 1500 Hz and is
centered around 2000 Hz which is the fundamental
frequency of this square wave.

0
-12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12

Hz

BME 310 Biomedical Computing - 308

J.Schesser
 Application of Ideal Filters
• If the filter is LTI, then the output signal is also
periodic with same fundamental frequency.
Therefore, y(t) can be written as a Fourier Series.

y(t)  be
k
k
jk ot

y(t)  b1e j 2  (2000)t  b1e  j 2  (2000)t

By superposition of these complex exponential signals 2  j /2 j 2  (2000)t 2
 e e  e  j /2 e  j 2  (2000)t )
at k o  (1)  (1)
2
 ( e j(2  (2000)t /2)  e  j(2  (2000)t /2) )
bk  H ( jk o )ak 
2
But since H ( j ) is only defined for 1250    2750,  ( e j(2  (2000)t /2)  e j e  j(2  (2000)t /2) )

then only the terms for which k  1 will be left 
2
( e j(2  (2000)t /2)  e  j(2  (2000)t /2) )

upon this multiplication 4
 cos(2 (2000)t   / 2)
2  j /2 
bk  H ( jk o )ak  e for k  1
k
BME 310 Biomedical Computing - 309
J.Schesser
 Time Domain or Frequency
Domain
• We have seen that a LTI can be represented by its
impulse response in the time domain and by its
frequency response in the frequency domain.
• In general when working with sinusoids (or complex
exponentials) either single or summed signals, it is
easier to work in the Frequency Domain.
• If the signal consists of impulses, step functions, or
other non-sinusoidal signals (e.g., signals which are
progressive integrations of the impulse function),
convolution of the impulse response (Time Domain)
is usually easiest.
BME 310 Biomedical Computing - 310
J.Schesser
 An Example
• An LTI system has an impulse response of
h(t) = δ (t) - 200πe-200πtu(t)
• The following signal is applied:
x(t)=10+20δ(t - 0.1)+40cos(200π t+0.3π) for all t
• The input has 3 parts: a constant, an impulse
and a cosine wave. We will take each part
separately and use the easiest method to find
the solution.
BME 310 Biomedical Computing - 311
J.Schesser
 An Example
• Let’s first find the frequency response of the system
from the impulse response:

H ( j )   [ (t )  200e  200t u (t )]e  jt dt

 
   (t )e  j t
dt  200  e  200t u (t )e  jt dt
 

200
 1  200  e ( 200  j )t dt  1  e ( 200  j )t |0
0
200  j
200 j
 1 
200  j 200  j

BME 310 Biomedical Computing - 312

J.Schesser
 An Example
• Now let’s take the first (constant, ω = 0) part and the third (cosine) part and
evaluate the solution using the frequency response:
The first part of the input :10
j
H ( j ) 
200  j
j0
10  H ( j 0)10  10  0
200  j 0
The third part of the input : 40cos(200t  0.3 )
j
H ( j ) 
200  j
40cos(200t  0.3 )  40 H ( j 200 ) cos[200t  0.3  H ( j 200 )]

1
j 200 j 2  1 
H ( j 200 )   
200  j 200 1  j  2 4
2
4
1 40
40cos(200t  0.3 )  40 cos[200t  0.3  .25 ]  cos[200t  0.55 ]
2 2
BME 310 Biomedical Computing - 313
J.Schesser
 An Example
• Now for the second part of the input (the impulse
function), we will apply the impulse response:
The second part of the input : 20 (t  0.1)
20 (t  0.1)  20h(t  0.1)  20[ (t  0.1)  200e  200 (t 0.1)u (t  0.1)]
20 (t  0.1)  20 (t  0.1)  4000e  200 (t 0.1)u (t  0.1)

• The Complete solution by superposition is:

y (t )  0
 20 (t  0.1)  4000 e 200 ( t 0.1)u (t  0.1)
40
 cos(200 t  0.55 )
2
BME 310 Biomedical Computing - 314
J.Schesser
 Homework
• Exercises:
– 10.4-10.7
• Problems:
– 10.5, 10.6,
– 10.7 Use Matlab to plot x(t); show your code

BME 310 Biomedical Computing - 315

J.Schesser