You are on page 1of 22

Signals and

Electronic
System Design
© Linear

Laplace transform for


Passive Filter Design
Mario Mata
mario.mata@gcu.ac.uk
M204
Contents
• Filters
• Design example: RC low-pass filter
• Load effects
• Design example: LR low-pass filter
• Design example: RC high-pass filter
• Second order transfer functions. Cascading
• Design example: RLC low-pass filter
• Transfer function normalization. Resonance. Q factor
• Transfer function normalization. Time response
• Other 2nd order topologies

M Mata 2 Signals and electronic systems design


Summary
Complex impedances: V (s) Basic circuits theory:
Z
I (s) - Series/parallel
Kirchoff’s laws
• Resistor vt   Ri t 
-
- Thévenin’s theorem
i(t) - Superposition
+ v(t) - V s 
ZR  R
I s 
Circuit’s transfer function:
Output ( s ) Num( s )
dv t  1 H (s)  
• Capacitor  i (t ) Input ( s ) Den( s )
dt C
i(t) - It’s a model in Laplace domain
+ v(t) - V s  1 - Den(s)=0  circuit poles
ZC   𝒔 = −𝜶 ± 𝒋𝝎𝒅
I s  sC

Time domain: Frequency domain:


vt   L
di (t )
• Inductor dt
- Static gain H(0)
- Final value v(∞)= VinH(0)
- Dynamic gain |H(jw)|
- Phase arg(H(jw))
i(t) - Time constant t|1/a| - Cut-off frequencies (wc=a)
- Settling time tS=3t - Bode diagram
+ v(t) - V s  - Oscillation period TOSC=2p/wd
ZL   sL
I s  - Peak time tP=3/wd 𝜶
- Overshooting 𝑶𝑺 = 𝒆−𝝅𝝎 d
M Mata 3 Signals and electronic systems design
Filters
• Filters are circuits showing a certain behaviour in
frequency domain: they block (attenuate) non-desired
frequencies but let the interesting ones through. Uses:
– Removing noise from signal (sensors, music, …)
– Selecting (tuning) bands in communications
– Removing undesired harmonics in electric power systems
• Depending on the cut-off frequencies fC of the poles/zeroes
of the filter’s tf, filters can behave as:
– Low-pass: allows frequencies below fC, blocks above fC
– High-pass: blocks frequencies below fC, allows above fC
– Band-pass: allows frequencies fC1 < f < fC2, blocks otherwise
– Band-reject: blocks frequencies fC1 < f < fC2, allows otherwise
• Attenuation amount depends on the dynamic gain slope
(dB/dec), and distance from each input frequency to fC
Filter ORDER n ≡ dynamic gain slope is 20n dB/dec
M Mata 4 Signals and electronic systems design
RC Low Pass Filter
• Basic low-pass filter:
Vi s 
 Vi s 
1
i(t) I (s) 
ZR  R Z R  ZC R  1 sC
1
ZC 

VO s   Z C I ( s )  Vi s   Vi s 
sC 1 sC 1
R  1 sC 1  sRC

• Transfer function:
1
Vo s  Pole: s  
H s  
1 sC 1
 
Vi s  R  1 sC 1  sRC RC

H s  
1
– Time constant: t  RC
1t s
1 1
– Cut-off frequency = |pole|: wC  
t RC
M Mata 5 Signals and electronic systems design
RC Low Pass Filter
• Basic low-pass filter: attenuates frequencies above wC
V s 
H s   o 
1 1 1
  w 
1

1
  
1
Hz 
Vi s  1  sRC 1  ts 1  s
rad / s f
RC t 2pRC
C C

wC
1
• Frequency response (Bode plot): 𝐻 𝑗𝜔 =
1 + 𝑗𝜔/𝜔𝐶
1
Gain: 𝐻 𝑗𝜔 = 2
Delay (phase): 𝜙 𝜔 = arg 𝐻 𝑗𝜔 = atan 𝜔/𝜔𝐶
1 + 𝜔/𝜔𝐶
Low frequencies gain:
|H(j0)| = 1 = 0db 0

-10
Gain (dB)

-20dB/dec
-20

-30
High frequencies gain 0
Low frequencies delay: -40
H(j0) = 1  0 deg 0
Phase (deg)

-45deg/dec
-45
High frequencies delay:
90 deg
-90 -2 Frequency (w/wC)
-1 0 1 2
10 10 10 10 10
w=wC
M Mata 6 Signals and electronic systems design
RC Low Pass Filter
• Notice that it draws current from the input signal:
i(t)
I ( s )  Vi s 
1
R  1 sC

Vi s 
• The Input impedance is calculated as: Z IN   R  1 sC
I (s)
– It depends on frequency. Substituting s = jw:
 , w  0
𝑍𝐼𝑁 = 𝑅 + 1/𝑗𝜔𝐶 Z  
 R, w  
IN

• Depending on the application, we’ll be interested on having


a large or a small input impedance:
– For signal processing, we need a large ZIN so only a small current
is drawn from the signal source. This circuit suits this application
(as input impedance is at least R)
– For power line filtering, we need a small ZIN to minimize energy
loss in R (P = i2R) when the current is delivered to any load at v0.
Not suitable!
M Mata 7 Signals and electronic systems design
Design example: RC Low Pass Filter
Design an RC Low-Pass signal filter with fC= 100kHz
1 1 1
wC  2pf C   RC    1.59 x10 6
RC 2pf C 200000p

• Solution has 2 degrees of freedom (variables) R and C, but only one


restriction ( equation to satisfy) RC = 1.59x10-6) 
we may arbitrarily choose one value (R or C) and get the other! I.e.:
R  1W  C  1.59 x10 6 F  1.59 F
• Is this a good solution for signal filtering?
• NO! Input impedance at high frequencies is tiny (~1W):
Z IN  R  1 jwC w Z IN  R  1W Let the input voltage amplitude be Vi = 3V.

If the signal is a high frequency one, the filter
could draw up to i ≈ 3V / 1W = 3A!
• Better choice: use a larger R (input current is smaller):
R  1kW  C  1.59 x10 9 F  1.59nF
R  10kW  C  1.59 x10 10 F  0.159nF
M Mata 8 Signals and electronic systems design
Load effects
• When any load is connected to the filter (i.e. the next
circuit), a load impedance ZLOAD appears in parallel with C:
i(t) +

𝑍𝐶 ||𝑍𝐿𝑂𝐴𝐷
v0 ZLOAD 𝑉0 𝑠 = 𝑉𝑖 𝑠
𝑍𝑅 + 𝑍𝐶 ||𝑍𝐿𝑂𝐴𝐷
-

• The tf changes, affecting the gain and cut-off frequency of


the overall filter + load system. For a resistive load RLOAD:
𝑅𝐿𝑂𝐴𝐷 /𝑠𝐶
𝑅𝐿𝑂𝐴𝐷 + 1/𝑠𝐶 1
𝐻 𝑠 = =
𝑅𝐿𝑂𝐴𝐷 /𝑠𝐶 𝑅𝐶𝑠 + 1 + 𝑅/𝑅𝐿𝑂𝐴𝐷
𝑅+
𝑅𝐿𝑂𝐴𝐷 + 1/𝑠𝐶
1 1 + 𝑅/𝑅𝐿𝑂𝐴𝐷 1 + 𝑅/𝑅𝐿𝑂𝐴𝐷
𝐻 0 = 𝑃𝑜𝑙𝑒: 𝑠 = − → 𝜔𝐶 =
1 + 𝑅/𝑅𝐿𝑂𝐴𝐷 𝑅𝐶 𝑅𝐶

• This must be taken into account at design time to accommodate for the
expected RLOAD, OR the next circuit should be designed to present a
large input impedance (RLOAD>>R)
M Mata 9 Signals and electronic systems design
LR Low Pass Filter
Vi s  Vi s 
 Vi s 
1
i(t) I (s)  Z IN   R  Ls
ZL  ZR Ls  R I (s)

VO s   Z R I ( s )  Vi s   Vi s 
R 1
R  Ls 1 s L
R
• It provides the same low-pass filtering that the RC:
V s 
H s   o 
1
Vi s  1  s L
H s  
1
w 
1 R
 rad / s  𝑓𝐶 =
𝑅
𝐻𝑧
 t t 2𝜋𝐿
C
R 1 s L
t L R
𝑅, 𝜔 → 0
𝑍𝐼𝑁 = 𝑅 + 𝑗𝜔𝐿 =
• Input impedance is now different: ∞, 𝜔 → ∞
– Good for power transmission, low frequency (50-60Hz) sees R
– R can actually be the supplied load, in that case there’re no losses
in the filter itself!
• But coils are bulkier and more costly than capacitors, as
large ones require a ferromagnetic core and lots of wire
(which will also cause a non-zero series resistance in the coil)
M Mata 10 Signals and electronic systems design
RC High Pass Filter
• Basic high-pass filter:
Vi s 
 Vi s 
1
i(t) I (s) 
Z R  ZC R  1 sC

VO s   Z R I ( s )  Vi s   Vi s 
R RCs
R  1 sC 1  RCs

V s  H s  
st
H s   o 
• Transfer function: sRC
Vi s  1  sRC 1  st
– Time constant: t  RC

Vi s 
• Input impedance: Z IN   R  1 sC
I (s)

M Mata 11 Signals and electronic systems design


RC High Pass Filter
• It’s frequency response corresponds to a high-pass filter
(frequencies below the cut-off frequency are attenuated):
s
V s  wC
H s   o  rad / s fC  1 Hz
sRC 1 1
 wC  
Vi s  1  sRC 1  s RC t 2pRC
wC
0 High frequencies gain:
|H(j∞)| = 1 = 0db
-10
Gain (dB)

-20
+20dB/dec
-30

-40

-50 High frequencies delay:


90
H(j∞) = 1  0 deg
Phase (deg)

45
-45deg/dec

0 Frequency w/wC
-2 -1 0 1 2
10 10 10 10 10
w=wC
M Mata 12 Signals and electronic systems design
Cascading 1st order filters
• If we connect a low-pass filter in series (cascade) with a
high-pass filter, we get a wide band-pass filter:
H s  
Vi 1 s Vo s
s  1 s 100  1 s  1s 100  1
0

-10
Gain (dB)

-20
+20dB/dec -20dB/dec
-30

-40

-50
90
Phase (deg)

45

-45

-90 Frequency w (rad/sec)


-2 -1 0 1 2 3 4
10 10 10 10 10 10 10

• Cascading several low-pass (or high pass) stages


increases the overall order (strenght) of the filter
M Mata 13 Signals and electronic systems design
2nd Order RLC Low Pass Filter
Vi s 
i(t) I (s) 
Z L  Z R || Z C

VO ( s )  I ( s )Z R || Z C 

𝑍𝑅 ||𝑍𝐶
𝑉0 𝑠 = 𝑉𝑖 𝑠
𝑍𝐿 + 𝑍𝑅 ||𝑍𝐶

𝑅
R. 1 R 1 + 𝑠𝑅𝐶 = 𝑅
Z R || Z C  sC  𝑉0 𝑠 = 𝑉𝑖 𝑠
R 1 1  sRC 𝐿𝑠 +
𝑅 𝑅𝐿𝐶𝑠 2 + 𝐿𝑠 + 𝑅
sC 1 + 𝑠𝑅𝐶

1
𝑉0 𝑠 𝑅 𝐿𝐶
𝐻 𝑠 = = =
𝑉𝑖 𝑠 𝑅𝐿𝐶𝑠 2 + 𝐿𝑠 + 𝑅 𝑠 2 + 1 𝑠 + 1
𝑅𝐶 𝐿𝐶

Vi s  RLCs 2  Ls  R  R, w  0
Z IN  
Input impedance: Z IN  
I (s) 1  sRC , w  
(adequate for power transmission)
M Mata 14 Signals and electronic systems design
2nd Order tf normalization - frequency
RLC filter: Normalized 2nd order transfer function:
w02
H s  
1 LC
H s   Q ≡ Quality factor
 2 1 1  w0 w0 ≡ Natural frequency
 s  s   s2  s  w02
 RC LC  Q

Comparing both circuit and normalized transfer functions:

(undamped) natural frequency Q-factor


1
w0  Q
R
 w0 RC  R
C
LC w0 L L

This factorization is useful to design high-order filters


In filter design Q is usually called Quality factor

M Mata 15 Signals and electronic systems design


2nd Order RLC Low Pass Filter
Second order  2 poles Q ≡ quality factor
• Real poles: no oscillations. Q<1/2
• Complex poles: oscillations. Q>1/2. Q > 0.707  resonance
20
Q=2
Q=4
0 Q=1

-20 Q=1/3

-40
-40dB/dec
-60

-80
0

-45 Q=1

-90 Q=1/3
Q=2
-90deg/dec
-135 Q=4
Frequency
(w/wC)
-180
-2 -1 0 1 2
10 10 10 10 10
M Mata 16 Signals and electronic systems design
2nd Order tf normalization – time domain
2nd order tf can written under an alternative normalization:
𝝎𝟐𝟎 𝝎𝟐𝟎 w0 ≡ undamped natural frequency
𝑯 𝒔 = 𝝎𝟎 = 𝟐 𝟐 Q ≡ quality factor
𝟐
𝒔 + 𝒔 + 𝝎𝟎 𝒔 + 𝟐𝜻𝝎𝟎 𝒔 + 𝝎𝟎
𝟐
𝑸 𝜻=
𝟏
≡ Damping factor
𝟐𝑸
Poles: Poles:
s
w0
2Q
 1  1  4Q 2
 𝑠 = 𝜁𝜔0 −1 ± 1 −
1
𝜁2

Poles are real or complex depending on Q (or z):


Factorization:
2 real
a1  a 2
Overdamped z  1 Q  12 poles H (s) 
s  -a1 , - a 2 s  a1 s  a 2 
Double
Critically a2
z  1 Q  12 real pole H (s) 
damped s  -a , - a s  a 2
Complex
Underdamped z  1 Q  2 a2 2
1 poles H (s) 
s  -a  j s  a 2   2
M Mata 17 Signals and electronic systems design
2nd Order step time-response XIm(s)

9 Re(s)
In(t)=u(t) H (s)  ;z  0 Undamped
Out(t) s 3
2 2

10
H (s)  ; z  1/ 3
s  12  32
Underdamped
X Im(s)

Re(s)

t/a

1 Im(s)

H (s) 
0.75
; z  1.15
Im(s)
H (s)  ;z 1
s  1.5s  0.5 X X
Re(s) s  12 X
Re(s)

Overdamped Critically damped


M Mata 18 Signals and electronic systems design
Step response depending on damping
w0 2 z = 1 (Q = 1/2) gets minimum rise time
H (s)  2
s  2zw 0 s  w0
2
without overshooting (critically damped)
1.8
z1/8
1.6
z1/4
1.4

1.2 z1/2
Output amplitude

z1
0.8

0.6

0.4
z2
0.2

0
0 5 10 15 20 25 30 35 40 45
Normalized time (t * w0)
M Mata 19 Signals and electronic systems design
2nd-order LPF frequency response and poles

r r

wr =

M Mata 20 Signals and electronic systems design


Second Order RLC Low Pass Filter Example
• Design a RLC low pass signal filter with natural frequency
w0 = 103 rad/s, no overshooting and minimum rise-time
• Possible solution:
1 1 6
w0   10  LC  2  10
3

LC w0
1
z 1 Q   QR C 1
2 L 2
• 2 restrictions, 3 variables  arbitrarily choose one of them
– Try to get feasible component values and a large input impedance
 R, w  0 » Choose a large R. I.e. R=1kW 
Z IN 
, w   C 1 L
 C 
L 2000 2000 2
6 L2 L 2
LC  10  2
 10 6  L  2 H C 2
 2
 5 x10  7  0.5F
2000 2000 2000

R and C are feasible, but L is quite large! Use software to iterate calculations
(i.e. a spreadsheet)

M Mata 21 Signals and electronic systems design


Other 2nd order filter topologies

High-pass filter: Using the circuit load as R:

Band-pass filter:

As exercises, get the tf


for each of them
Source: Wikipedia

M Mata 22 Signals and electronic systems design

You might also like