You are on page 1of 5

EE453: Lecture 03: Low-pass signal filtering Fall 2014

3 Low-pass signal filtering


The Shannon ideal low-pass filter cannot be realized in practice since it is not causal (h(t) 6= 0 t <
0 ), here we shall examine practical filters:

3.1 Frequency response of LTI systems


Given an LTI system with transfer function H(s), H(jω) is called the frequency response (FR
of the system:
Y (jω) = H(jω)X(jω)
|Y (jω)| = |H(jω)||X(jω)| Magnitude relationship
∠Y (jω) = ∠H(jω) + ∠X(jω) phase relationship
The FR determine How the input signal is transformed on the basis of its frequency contents. For
each frequency in the input:

1. Its magnitude is multiplied by the magnitude of system’s FR at that frequency.

2. it’s phase shift is increased/decreased (by addition of subtraction) by the phase of system’s
FR at that frequency.

note 1 for an LTI systems, its output will never contain a frequency that is not available in the
input (magnitude relationship), e,i LTI systems do not generate new frequencies.

All-pass filter and linear phase Consider the LTI system with the following frequency re-
sponse:
H(jω) = e−jωt0
Its magnitude and phase shift are:

|H(jω)| = 1 ∠H(jω) = −ωt0

All the input’s frequencies pass with not amplitude alteration, However their phase linearly decrease
(lag) proportionally to the frequency.
If the input is x(t)
Y (jω) = X(jω)e−jωt0 ⇒ y(t) = x(t − t0 )

Note 2: A linear phase system leads to an output that is just a time delayed input

General band-pass systems For band-pass systems, The phase characteristics is only mean-
ingful within the pass-band, since outside the pass-band, the magnitude is zero.

Note 3: A linear phase system (filter) is one with a linear phase response in the pass-band

Note 4: A nonlinear phase in filters is not genrally desirable as it leads to (phase) distortion

7
EE453: Lecture 03: Low-pass signal filtering Fall 2014

3.1.1 Wave and frequency response


In general an input signal has a continuum of frequencies.
A sine (cosine) signal on the other hand is a simple signal that contains only one(positive)
frequency. Thus, it is very suitable to show the effect (and meanning) of the FR of an LTI system
on the input signal.

A
x(t) = A cos(ω0 t) = (ejω0 (t + e−jω0 t
2
Z Z
A
y(t) = h(τ )x(t − τ ) dτ = h(τ )(ejω0 (t−τ ) + e−jω0 (t−τ ) )dτ
2
 Z Z 
A jω0 t −jω0 τ −jω0 t jω0 τ
y(t) = e h(τ )e dτ + e h(τ )e dτ
2
A  jω0 t  A h i
y(t) = e H(jω0 ) + e−jω0 t H(−jω0 ) = |H(jω0 )| ejω0 t ejφ + e−jω0 t e−jφ
2 2
Where φ = ∠H(jω0 )
y(t) = A|H(jω0 )| cos(ω0 t + ∠H(jω0 ))
• Since the input contains only one (positive) frequency ω0 , the output must contain only one
frequency , and its ω0 , and thus is also a cosine.
• the output’s magnitude will be the input magnitude multiplied by the magnitude of H(jω)
at ω0 .
• its phase shift will and the input phase shift increase by the phase shift of of H(jω) at ω0 .

3.1.2 Frequency response identification


Using the system output to sine wave formula, one can devise a practical method to determine an
LTI system’s frequency response at discrete number of frequencies:
x(t) = A cos(ω0 t) ⇒ y(t) = A|H(jω0 )| cos(ω0 t + ∠H(jω0 ))
Let the output be in general B cos(ωt + φ).
Given a set of discrete frequencies S = {ω0 , ω1 , ω2 , . . . , ωn }
• Repeat for i = 0 → n
1. Adjust input frequency ω = ωi , (x(t) = A cos(ωi t))
2. Measure the output signal’s magnitude B and phase shift φ
3. Compute and record the frequency response: |H(jωi )| = B/A and ∠H(jωi ) = φ

3.2 Low-pass filter


3.2.1 Control Applications
Low-pass filters are very useful, some of the main applications are:
• Can be used to reconstruct a signal from its samples.
• can be used filter out noise (usually of a high frequency nature) present the the sensor’s
output.
• Implement an Anti-aliasing filter: it is used to remove the part of the signal spectrum
that is above the Nyquist frequency ωs /2 before sampling, Thus insuring that no aliasing
(folding) distortion will occur.
A low-pass filter can be implemented as an LTI system with IR h(t) and FR H(jω).

8
EE453: Lecture 03: Low-pass signal filtering Fall 2014

3.2.2 Low-pass filters characteristics:


The characteristics and performance of a the filter are then defined primarily by frequency response,
and secondly by its time response (step response):

Amplitude spectrum characteristics√ Aside from the cut-off frequency, which is the frequency
at which the amplitude drops to 2/2 (or −3db). The amplitude spectrum characteristics expresses
how close the filter’s amplitude is to the ideal filter. A practical filter will go through three phases:
1. pass-band: is it represented by the frequency range [0, ωp ], inside which the amplitude is
within ±Gp of 1.
2. Stop-band: it is represented by the frequency ωst after which the amplitude is always less
than a specified gain Gst .
3. Transition-band: is it represented by the intermediate frequency range [ωp , ωst ] between the
pass-band and stop-band . Ideally this range should as small as possible
The next figure illustrates these parameters. The shaded area is forbidden.

|H(jω)|

1 + Gp
1
1 − Gp

(2)
2

Gst

ω
0 ωpωc ωst
pass-band transition-band stop-band

Phase spectrum characteristics The phase characteristics measures how close to linear is the
phase whithin the band-pass:
d∠H(jω)
Φ(ω) = 0 < ω < ωc

For a linear phase filter ∠H(jω) = −ωt0
Φ(ω) = −t0 Constant

Time response measure the the amount of ringing in the step response of the filter.
The ringing is the amount of oscillation in the step response before settling. Therefore is usually
measured as the settling time which is also proportional to the overshoot.

9
EE453: Lecture 03: Low-pass signal filtering Fall 2014

3.2.3 Typical Low-pass filters


There are mainly five families of filters,each of which has particular advantages and drawbacks:
1. Butterworth filters family
2. Chebyshev I filters family
3. Chebyshev II (inverse chebyshev) filters family
4. Elliptic (or cauer) filters family
5. bessel (or linear phase)
The amplitude spectrum of the filters (2nd order) are compared for a the cutoff frequency ωc =
1rad/s on a logarithmic frequency scale, in the next figure

The step response is shown next

10
EE453: Lecture 03: Low-pass signal filtering Fall 2014

The following table compared the different filters classes:

Ripple in pass-band Ripple in stop-band cutoff sharpness phase linearity ringing


Butterworth No (max flatness) No . - -
Chebyshev I yes no + --- ---
Chebyshev II No yes + -- --
Elliptic yes yes ++ ---- ----
Bessel No No - Linear + No +

3.3 First-order low-pass filter


The simplest kind of filters is the first-order , its TF is:
1
H(s) =
sτ + 1
1 −t
h(t) = e τ us (t)
τ
t
(1 − e− τ )us (t)
The step response exhibits no ringing for any value of τ .
its frequency response is:
1
H(jω) =
jωτ + 1
1
|H(jω)| = p ∠H(jω) = − tan−1 (ωτ )
(ωτ )2 + 1
• for ω >> ωc , |H(jω)| = 0 and ∠H(jω) = −π/2

• for ω << ωc , |H(jω)| = 1 and ∠H(jω) = 0



• for ω = ωc , |H(jω)| = 2/2 and ∠H(jω) = −π/4

The plot of the amplitude and phase response with τ = 1, ωc = 1rad/s is shown using a
logarithmic frequency scale

11

You might also like