This action might not be possible to undo. Are you sure you want to continue?
II.
PASSIVE FILTERS
Frequencyselective or ﬁlter circuits pass to the output only those input signals that are in a
desired range of frequencies (called pass band). The amplitude of signals outside this range
of frequencies (called stop band) is reduced (ideally reduced to zero). Typically in these
circuits, the input and output currents are kept to a small value and as such, the current
transfer function is not an important parameter. The main parameter is the voltage transfer
function in the frequency domain, H
v
(jω) = V
o
/V
i
. As H
v
(jω) is complex number, it has
both a magnitude and a phase, ﬁlters in general introduce a phase diﬀerence between input
and output signals.
To minimize the number of subscripts, hereafter, we will drop subscript v of H
v
. Furthermore,
we concentrate on the the ”openloop” transfer functions, H
vo
, and denote this simply by
H(jω). The impact of loading is sperately discussed.
Pass
Band Band
Stop
 H(j )  ω
ω ω
c
ω  H(j ) 
ω ω
c
Κ
0.7Κ
2.1 LowPass Filters
An ideal lowpass ﬁlter’s transfer function is shown. The
frequency between the pass andstop bands is called the
cutoﬀ frequency (ω
c
). All of the signals with frequen
cies below ω
c
are transmitted and all other signals are
stopped.
In practical ﬁlters, pass and stop bands are not clearly
deﬁned, H(jω) varies continuously from its maximum
toward zero. The cutoﬀ frequency is, therefore, deﬁned
as the frequency at which H(jω) is reduced to 1/
√
2 =
0.7 of its maximum value. This corresponds to signal
power being reduced by 1/2 as P ∝ V
2
.
o

+
i

+
V
V
L
R
Lowpass RL ﬁlters
A series RL circuit as shown acts as a lowpass ﬁlter. For
no load resistance (“openloop” transfer function), V
o
can
be found from the voltage divider formula:
V
o
=
R
R +jωL
V
i
→ H(jω) =
V
o
V
i
=
R
R +jωL
=
1
1 +j(ωL/R)
We note
H(jω) =
1
_
1 + (ωL/R)
2
ECE65 Lecture Notes (F. Najmabadi), Spring 2006 21
It is clear that H(jω) is maximum when denominator is smallest, i.e., ω →0 and H(jω)
decreases as ω is increased. Therefore, this circuit allows “lowfrequency” signals to pass
through while “blocking” highfrequency signals (i.e., reduces the amplitude of the voltage
of the highfrequency signals). The reference to deﬁne the “low” and “high”frequencies is
the cutoﬀ frequency: “low”frequencies mean frequencies much lower than ω
c
.
To ﬁnd the cutoﬀ frequency, we note that the H(jω)
Max
= 1 occurs at ω = 0 (alterna
tively ﬁnd d H(jω) /dω and set it equal to zero to ﬁnd ω = 0 which maximizes H(jω)).
Therefore,
H(jω)
max
= 1
H(jω)
ω=ωc
=
1
√
2
H(jω)
max
=
1
√
2
1
_
1 + (ω
c
L/R)
2
=
1
√
2
−→ 1 +
_
ω
c
L
R
_
2
= 2 →
ω
c
L
R
= 1
Therefore,
ω
c
=
R
L
and H(jω) =
1
1 +jω/ω
c
Input Impedance: Using the deﬁnition of the input impedance, we have:
Z
i
=
V
i
I
i
= jωL +R
The value of the input impedance depends on the frequency ω. For good voltage coupling,
we need to ensure that the input impedance of this ﬁlter is much larger than the output
impedance of the previous stage. Since we do not know the frequency of the input signal,
we need to ensure that good voltage coupling criteria is satisﬁed for all frequencies (or all
possible values of Z
i
). As such, the minimum value of Z
i
is an important number. Z
i
is
minimum when the impedance of the inductor is zero (ω →0).
Z
i

min
= R
o
Z
L
R
Output Impedance: The output impdenace can be
found by “killing” the source and ﬁnding the equivalent
impdenace between output terminals:
Z
o
= jωL R
ECE65 Lecture Notes (F. Najmabadi), Spring 2006 22
where the source resistance is ignored. Again, the value of the output impedance depends on
the frequency ω. For good voltage coupling, we need to ensure that the output impedance
of this ﬁlter is much smaller than the input impedance of the next stage for all frequencies,
the maximum value of Z
o
is an important number. Z
o
is maximum when the impedance of
the inductor is inﬁnity (ω →∞).
Z
o

max
= R
Bode Plots and Decibel
The voltage transfer function of a twoport network (and/or the ratio of output to input
powers) is usually expressed in Bel:
Number of Bels = log
10
_
P
o
P
i
_
or Number of Bels = 2 log
10
¸
¸
¸
¸
V
o
V
i
¸
¸
¸
¸
because P ∝ V
2
. Bel is a large unit and decibel (dB) is usually used:
Number of decibels = 20 log
10
¸
¸
¸
¸
V
o
V
i
¸
¸
¸
¸
or
¸
¸
¸
¸
V
o
V
i
¸
¸
¸
¸
dB
= 20 log
10
¸
¸
¸
¸
V
o
V
i
¸
¸
¸
¸
There are several reasons why decibel notation is used:
1) Historically, the analog systems were developed ﬁrst for audio equipment. Human ear
“hears” the sound in a logarithmic fashion. A sound which appears to be twice as loud
actually has 10 times power, etc. Decibel translates the output signal to what ear hears.
2) If several twoport network are placed in a cascade (output of one is attached to the input
of the next), the overall transfer function, H, is equal to the product of all transfer functions:
H(jω) = H
1
(jω) ×H
2
(jω) ×...
20 log
10
H(jω) = 20 log
10
H
1
(jω) + 20 log
10
H
2
(jω) +...
H(jω)
dB
= H
1
(jω)
dB
+H
2
(jω)
dB
+...
making it easier to ﬁnd the overall response of the system.
3) Plot of H(jω)
dB
versus frequency has special properties that make analysis simpler. For
example, the plot asymptotes to straight lines at low and high frequencies as is shown below.
ECE65 Lecture Notes (F. Najmabadi), Spring 2006 23
Also, using dB deﬁnition, we see that, there is a 3 dB diﬀerence between maximum gain and
gain at the cutoﬀ frequency:
20 log H(jω
c
) −20 log H(jω)
max
= 20 log
_
H(jω
c
)
H(jω)
max
_
= 20 log
_
1
√
2
_
≈ −3 dB
Bode plots are plots of H(jω)
dB
(magnitude) and
H(jω) (phase) versus frequency in a
semilog format (i.e., ω axis is a log axis). Bode plots of ﬁrstorder lowpass RL ﬁlters are
shown below (W denotes ω
c
).
H(jω)
dB
H(jω)
At high frequencies, ω/ω
c
1,
H(jω) ≈
1
ω/ω
c
→ H(jω)
dB
= 20 log
_
1
ω/ω
c
_
= 20 log(ω
c
) −20 log(ω)
which is a straight line with a slope of 20 dB/decade in the Bode plot. It means that if ω
is increased by a factor of 10 (a decade), H(jω)
dB
changes by 20 dB.
At low frequencies, ω/ω
c
1, H(jω) ≈ 1 which is also a straight line in the Bode plot.
The intersection of these two “asymptotic” values is at 1 = 1/(ω/ω
c
) or ω = ω
c
. Because of
this, the cutoﬀ frequency is also called the “corner” frequency.
The behavior of the phase of H(jω) can be found by examining
H(jω) = −tan
−1
(ω/ω
c
). At
low frequencies, ω/ω
c
1,
H(jω) ≈ 0 and at high frequencies, ω/ω
c
1,
H(jω) ≈ −90
◦
.
At cutoﬀ frequency,
H(jω) ≈ −45
◦
.
ECE65 Lecture Notes (F. Najmabadi), Spring 2006 24
General ﬁrstorder lowpass ﬁlters
As we discussed before, transfer functions characterize a twoport network. As such, it is
useful to group twoport networks into families based on their voltage transfer functions.
To facilitate this grouping, the convention is to simplify the voltage transfer function to a
form such that the “Real” part of the denominator of H(jω) is unity (i.e., the denominator
should be 1 + j · · · or 1 −j · · · ). As we will see later in this section, this grouping will also
help reduce the math that we do in analyzing various circuits.
The lowpass RL ﬁlter discussed before is part of the family of ﬁrstorder lowpass ﬁlters
(ﬁrst order means that ω appears in the denominator with an exponent of 1 or −1. In
general, the voltage transfer function of a ﬁrstorder lowpass ﬁlter is in the form:
H(jω) =
K
1 +jω/ω
c
The maximum value of H(jω) = K is called the ﬁlter gain. Note that the exponent of ω
in the denominator is +1 so that H(jω) decreases with frequency (thus,a lowpass ﬁlter):
H(jω) =
K
_
1 + (ω/ω
c
)
2
H(jω) = −
K
K
tan
−1
_
ω
ω
c
_
For RL ﬁlter, K = 1, and ω
c
= R/L. Note that K can be negative, and in that case, the
“minus” sign adds 180
◦
phase shift to the transfer function as is denoted by K/K factor
above.

o
i
+
+

V
R
C
V
Lowpass RC ﬁlters
A series RC circuit as shown also acts as a lowpass ﬁlter.
For no load resistance (“openloop” transfer function), V
o
can be found from the voltage divider formula:
V
o
=
1/(jωC)
R + 1/(jωC)
V
i
=
1
1 +j(ωRC)
V
i
H(jω) =
1
1 +jωRC
We see that the voltage transfer function of this circuit is similar to transfer function of a
general ﬁrstorder lowpass ﬁlter. So, this is a lowpass ﬁlter with K = 1 and ω
c
= 1/RC.
(Note: we identiﬁed the circuit and found the cutoﬀ frequency without doing any math!).
ECE65 Lecture Notes (F. Najmabadi), Spring 2006 25
We could, of course, do the math following the procedure in analyzing the lowpass RL ﬁlter
to get the same answer. (Exercise: Show this.).
Following the same procedure as for RL ﬁlters, we ﬁnd input and output Impedances
Z
i
= R +
1
jωC
and Z
i

min
= R
Z
o
= R
1
jωC
and Z
o

max
= R

o

+
i
L
+
V
C
R
V
R
Terminated RL and RC lowpass ﬁlters
Now let us examin the eﬀect of a load on the perfor
mance of our RL and RC ﬁlters. For this example,
a resistive load is considered but the analysis can be
easily extended to an impedance load. For example,
consider the terminated RC ﬁlter shown:
From the circuit,
H(jω) =
V
o
V
i
=
1/(jωC) R
L
R + [1/(jωC) R
L
]
=
R
/R
1 +j(ωR
C)
with R
= R R
L
This is similar to the transfer function for unterminated RC ﬁlter but with resistance R
being replaced by R
. Therefore,
ω
c
=
1
R
C
=
1
(R R
L
)C
and H(jω) =
R
/R
1 +jω/ω
c
We see that the impact of the load is to reduce the ﬁlter gain (K = R
/R < 1) and to shift
the cutoﬀ frequency to a higher frequency as R
= R R
L
< R.
Input Impedance: Z
i
= R +
1
jωC
R
L
Z
i

min
= R
Output Impedance: Z
o
= R
1
jωC
Z
o

max
= R
We could have arrived at the same results using the the relationship between openloop,
H
o
(jω), and terminated, H(jω), transfer functions of a twoport network:
H(jω) =
Z
L
Z
L
+Z
o
H
o
(jω) =
R
L
R
L
+R
1
jωC
×
1
1 +jωRC
ECE65 Lecture Notes (F. Najmabadi), Spring 2006 26
(Exercise: show this.) Also, note that the output impdenace of the terminated circuit is
exactly the same as the openloop version.
Furthermore, it can be seen that as long as R
L
Z
o
or R
L
Z
o

max
= R (our condition
for good voltage coupling), R
≈ R and the terminated RC ﬁlter will look exactly like an
unterminated ﬁlter – The ﬁlter gain is one, the shift in cutoﬀ frequency disappears, and
input and output resistances become the same as before.
Terminated RL lowpass ﬁlters
The parameters of the terminated RL ﬁlters can be found similarly:
Voltage Transfer Function: H(jω) =
V
o
V
i
=
1
1 +jω/ω
c
, ω
c
= (R R
L
)/L.
Input Impedance: Z
i
= jωL +R R
L
, Z
i

min
= R R
L
Output Impedance: Z
o
= (jωL) R, Z
o

max
= R
Here, the impact of load is to shift the cutoﬀ frequency to a lower value. Filter gain is not
aﬀected. Again for R
L
Z
o
or R
L
Z
o

max
= R (our condition for good voltage coupling),
the shift in cutoﬀ frequency disappears and the ﬁlter will look exactly like an unterminated
ﬁlter.
Exercise: Derive above equations for the transfer function and input and output impde
nacess.
ECE65 Lecture Notes (F. Najmabadi), Spring 2006 27
2.2 Firstorder high pass ﬁlters
In general, the voltage transfer function of a ﬁrstorder highpass ﬁlter is in the form:
H(jω) =
K
1 −jω
c
/ω
It is a ﬁrstorder ﬁlter because ω appears in the denominator with an exponent of −1. It
is a highpass ﬁlter because H = 0 for ω = 0 and H is constant for highfreqnecies.
Paramter ω
c
is the cutoﬀ freqnecy of the ﬁlter (Exercise: prove that H(jω
c
) is 1/
√
2 = 0.7
of H(jω)
Max
.)
The maximum value of H(jω) = K is called the ﬁlter gain.
H(jω) =
K
_
1 + (ω
c
/ω)
2
H(jω) = +
K
K
tan
−1
_
ω
c
ω
_
Bode Plots of ﬁrstorder highpass ﬁlters (K = 1) are shown below. The asymptotic behavior
of this class of ﬁlters is:
At low frequencies, ω/ω
c
1, H(jω) ∝ ω (a +20dB/decade line) and
H(jω) = 90
◦
At high frequencies, ω/ω
c
1, H(jω) ∝ 1 (a line with a slope of 0) and
H(jω) = 0
◦
H(jω)
H(jω)
ECE65 Lecture Notes (F. Najmabadi), Spring 2006 28
o
+
i

+

V
V
C
R
Highpass RC ﬁlters
A series RC circuit as shown acts as a highpass ﬁlter.
The openloop voltage transfer function of this ﬁlter is:
H(jω) =
V
o
V
i
=
R
R + 1/(jωC)
=
1
1 −j(1/ωRC)
Therefore, this is a ﬁrstorder highpass ﬁlter with K = 1 and ω
C
= 1/RC. Input and output
impdenaces of this ﬁlter can be found similar to the procedure used for lowpass ﬁlters:
Input Impedance: Z
i
= R +
1
jωC
and Z
i

min
= R
Output Impedance: Z
o
= R
1
jωC
and Z
o

max
= R
 
i
+
o
+
V
L
R
V
Highpass RL ﬁlters
A series RL circuit as shown also acts as a highpass ﬁlter.
Again, we ﬁnd the openloop tranfunction to be:
ω
c
=
R
L
H(jω) =
1
1 −jω
c
/ω
Input Impedance: Z
i
= R +jωL and Z
i

min
= R
Output Impedance: Z
o
= R jωL and Z
o

max
= R
Exercise: Compute the voltage transfer function and input and output impdenaces of
terminated RC and RL ﬁlters.
ECE65 Lecture Notes (F. Najmabadi), Spring 2006 29
2.3 Bandpass ﬁlters
A band pass ﬁlter allows signals with a range of frequencies (pass band) to pass through and
attenuates signals with frequencies outside this range.
Band
 H(j )  ω
Pass
u
ω ω ω
l
ω
l
: Lower cutoﬀ frequency;
ω
u
: Upper cutoﬀ frequency;
ω
0
≡
√
ω
l
ω
u
: Center frequency;
B ≡ ω
u
−ω
l
: Band width;
Q ≡
ω
0
B
: Quality factor.
As with practical low and highpass ﬁlters, upper and lower cutoﬀ frequencies of practical
band pass ﬁlter are deﬁned as the frequencies at which the magnitude of the voltage transfer
function is reduced by 1/
√
2 (or 3 dB) from its maximum value.
Secondorder bandpass ﬁlters:
Secondorder band pass ﬁlters include two storage elements (two capacitors, two inductors,
or one of each). The transfer function for a secondorder bandpass ﬁlter can be written as
H(jω) =
K
1 +jQ
_
ω
ω
0
−
ω
0
ω
_
H(jω) =
K
¸
1 +Q
2
_
ω
ω
0
−
ω
0
ω
_
2
H(jω) = −
K
K
tan
−1
_
Q
_
ω
ω
0
−
ω
0
ω
__
The maximum value of H(jω) = K is called the ﬁlter gain. The lower and upper cutoﬀ
frequencies can be calculated by noting that H(jω)
max
= K, setting H(jω
c
) = K/
√
2 and
solving for ω
c
. This procedure will give two roots: ω
l
and ω
u
.
H(jω
c
) =
1
√
2
H(jω)
max
=
K
√
2
=
K
¸
1 +Q
2
_
ω
c
ω
0
−
ω
0
ω
c
_
2
Q
2
_
ω
c
ω
0
−
ω
0
ω
c
_
2
= 1 → Q
_
ω
c
ω
0
−
ω
0
ω
c
_
= ±1
ω
2
c
−ω
2
0
±
ω
c
ω
0
Q
= 0
ECE65 Lecture Notes (F. Najmabadi), Spring 2006 30
The above equation is really two quadratic equations (one with + sign in front of fraction
and one with a − sign). Solving these equation we will get 4 roots (two roots per equation).
Two of these four roots will be negative which are not physical as ω
c
> 0. The other two
roots are the lower and upper cutoﬀ frequencies (ω
l
and ω
u
, respectively):
ω
l
= ω
0
¸
1 +
1
4Q
2
−
ω
0
2Q
ω
u
= ω
0
¸
1 +
1
4Q
2
+
ω
0
2Q
Bode plots of a secondorder ﬁlter is shown below. Note that as Q increases, the bandwidth
of the ﬁlter become smaller and the H(jω) becomes more picked around ω
0
.
H(jω)
db
H(jω)
Asymptotic behavior:
At low frequencies, ω/ω
0
1, H(jω) ∝ ω (a +20dB/decade line), and
H(jω) →90
◦
At high frequencies, ω/ω
0
1, H(jω) ∝ 1/ω (a 20dB/decade line), and
H(jω) →−90
◦
At ω = ω
0
, H(jω) = K (purely real) H(jω) = K (maximum ﬁlter gain), and
H(jω) = 0
◦
.
There are two ways to solve secondorder ﬁlter circuits. 1) One can try to write H(jω) in
the general form of a secondorder ﬁlters and ﬁnd Q and ω
0
. Then, use the formulas above
to ﬁnd the lower and upper cutoﬀ frequencies. 2) Alternatively, one can directly ﬁnd the
upper and lower cutoﬀ frequencies and use ω
0
≡
√
ω
l
ω
u
to ﬁnd the center frequency and
B ≡ ω
u
− ω
l
to ﬁnd the bandwidth, and Q ≡= ω
0
/B to ﬁnd the quality factor. The two
examples below show the two methods. Note that one can always ﬁnd ω
0
and k rapidaly as
H(jω
0
) is purely real and H(jω
0
) = k
ECE65 Lecture Notes (F. Najmabadi), Spring 2006 31
o

+

+
i
C
V
R V
L
Series RLC Bandpass ﬁlters
Using voltage divider formula, we have
H(jω) =
V
o
V
i
=
R
R +jωL + 1/(jωC)
H(jω) =
R
R +j
_
ωL −
1
ωC
_
There are two approaches to ﬁnd ﬁlter parameters, K, ω
0
, ω
u
, and ω
l
.
Method 1: We transform the transfer function in a form similar to general form of the
transfer function for second order bandpass ﬁlters:
H(jω) =
K
1 +jQ
_
ω
ω
0
−
ω
0
ω
_
Note that the denominator of the general form is in the form 1 +j . . . Therefore, we divide
top and bottom of transfer function of series RLC bandpass ﬁlters by R:
H(jω) =
1
1 +j
_
ωL
R
−
1
ωRC
_
Comparing the above with the general form of the transfer function, we ﬁnd K = 1. To ﬁnd
Q and ω
0
, we note that the imaginary part of the denominator has two terms, one positive
and one negative (or one that scales as ω and the other that scales as 1/ω) similar to the
general form of transfer function of 2ndorder bandpass ﬁlters (which includes Qω/ω
0
and
−Qω
0
/ω). Equating these similar terms we get:
Qω
ω
0
=
ωL
R
→
Q
ω
0
=
L
R
Qω
0
ω
=
1
ωRC
→ Qω
0
=
1
RC
We can solve these two equations to ﬁnd:
ω
0
=
1
√
LC
Q =
ω
0
R/L
=
¸
L
R
2
C
ECE65 Lecture Notes (F. Najmabadi), Spring 2006 32
The lower and upper cutoﬀ frequencies can now be found from the formulas on page 31.
Method 2: In this method, we directly calculate the ﬁlter parameters similar to the proce
dure followed for general form of transfer function in page 30. Some simpliﬁcations can be
made by noting: 1) At ω = ω
0
, H(jω) is purely real and 2) K = H(jω = jω
0
).
Starting with the transfer function for the series RLC ﬁlter:
H(jω) =
R
R +j
_
ωL −
1
ωC
_
We note that the transfer function is real if coeﬃcient of j in the denominator is exactly
zero (note that this happens for ω = ω
0
), i.e.,
ω
0
L −
1
ω
0
C
= 0 −→ ω
0
=
1
√
LC
Also
K = H(jω = jω
0
) =
R
R
= 1
The cutoﬀ frequencies can then be found by setting:
H(jω
c
) =
K
√
2
=
1
√
2
1 +
_
ω
c
L
R
−
1
ω
c
RC
_
2
= 2
which can be solved to ﬁnd ω
u
and ω
l
.
Input and Output Impedance of bandpass RLC ﬁlters
Z
i
= jωL +
1
jωC
+R = j
_
ωL −
1
ωC
_
+R
Z
i

min
= R occurs at ω = ω
0
Z
o
=
_
jωL +
1
jωC
_
R → Z
o

max
= R
ECE65 Lecture Notes (F. Najmabadi), Spring 2006 33
WideBand BandPass Filters
Bandpass ﬁlters can be constructed by putting a highpass and a lowpass ﬁlter back to
back as shown below. The highpass ﬁlter sets the lower cutoﬀ frequency and the lowpass
ﬁlter sets the upper cutoﬀ frequency of such a bandpass ﬁlter.
ω  H (j ) 
2
ω
1
 H (j )  X
ω =
l
ω =
ω
u
ω
ω  H (j ) 
ω
ω  H (j ) 
ω
1
ω ω ω
2
c2 c1
c2 c1
o
−
+
i
−
+
−
+
1
High−Pass Low−Pass
2 1
1
2
V
V
C
V
R C
R
An example of such a bandpass ﬁlter is
two RC lowpass and highpass ﬁlters put
back to back. These ﬁlters are widely
used (when appropriate, see below) instead
of an RLC ﬁlter as inductors are usually
bulky and take too much space on a cir
cuit board.
In order to have good voltage coupling in the above circuit, the input impedance of the
highpass ﬁlter (actually Z
i

min
= R
1
) should be much larger than the output impedance of
the lowpass ﬁlter (actually Z
o

max
= R
2
), or we should have R
1
R
2
. In that case we can
use unterminated transfer functions:
H(jω) = H
1
(jω) ×H
2
(jω) =
1
1 +jω/ω
c2
×
1
1 −jω
c1
/ω
ω
c1
= 1/(R
1
C
1
) ω
c2
= 1/(R
2
C
2
)
H(jω) =
1
(1 +jω/ω
c2
)(1 −jω
c1
/ω)
=
1
(1 +ω
c1
/ω
c2
) +j(ω/ω
c2
−ω
c1
/ω)
Again, we can ﬁnd the ﬁlter parameters by either of two methods above. Transforming the
transfer function to a form similar to the general form (left for students) gives:
K =
1
1 +ω
c1
/ω
c2
Q =
_
ω
c1
/ω
c2
1 +ω
c1
/ω
c2
ω
0
=
√
ω
c1
ω
c2
ECE65 Lecture Notes (F. Najmabadi), Spring 2006 34
One should note that the Bode plots of previous page are “asymptotic” plots. The real
H(jω) diﬀers from these asymptotic plots, for example, H(jω) is 3 dB lower at the cut
oﬀ frequency. A comparison of “asymptotic” Bode plots for ﬁrstorder highpass ﬁlters are
given in page 28. It can be seen that H
1
(jω) achieves its maximum value (1 in this case)
only when ω/ω
c1
< 1/3. Similarly for the low pass ﬁlter, H
2
(jω) achieves its maximum
value (1 in this case) only when ω/ω
c2
> 3. In the bandpass ﬁlter above, if ω
c2
ω
c1
(i.e.,
ω
c2
≥ 10ω
c1
), the center frequency of the ﬁlter will be at least a factor of three away from
both cutoﬀ frequencies and H(jω) = H
1
 ×H
2
 achieves its maximum value of 1. If ω
c2
is not ω
c1
(i.e., ω
c2
< 10ω
c1
), H
1
and H
2
will not reach their maximum of 1 and the ﬁlter
H(jω)
max
= H
1
 ×H
2
 will be less than one. This can be seen by examining the equation
of K above which is always less than 1 and approaches 1 when ω
c2
ω
c1
.
More importantly, we can never make a “narrow” band ﬁlter by putting two ﬁrstorder high
pass and lowpass ﬁlters back to back. When ω
c2
is not ω
c1
, H(jω)
max
becomes smaller
than 1. Since the cutoﬀ frequencies are located 3 dB below the maximum values, the cutoﬀ
frequencies will not be ω
c1
and ω
c2
(those frequencies are 3 dB lower than H(jω)
max
= 1).
The lower cutoﬀ frequency moves to a value lower than ω
c1
and the upper cutoﬀ frequency
moves to a value higher than ω
c2
. This can be seen by examining the quality factor of this
ﬁlter at the limit of ω
c2
= ω
c1
Q =
_
ω
c1
/ω
c2
1 +ω
c1
/ω
c2
=
1
1 + 1
= 0.5
while our asymptotic description of previous page indicated that when ω
c2
= ω
c1
, bandwidth
becomes vanishingly small and Q should become very large.
Because these ﬁlters work only when ω
c2
ω
c1
, they are called “wideband” ﬁlters. For
these wideband ﬁlters (ω
c1
ω
c2
), we ﬁnd from above:
K = 1 Q =
_
ω
c1
/ω
c2
ω
0
=
√
ω
c2
ω
c1
H(jω) =
1
1 +j(ω/ω
c2
−ω
c1
/ω)
We then substitute for Q and ω
0
in the expressions for cutoﬀ frequencies (page 31) to get:
ω
u
= ω
0
¸
1 +
1
4Q
2
+
ω
0
2Q
=
ω
0
2Q
_
_
1 + 4Q
2
+ 1
_
ω
l
= ω
0
¸
1 +
1
4Q
2
+
ω
0
2Q
=
ω
0
2Q
_
_
1 + 4Q
2
−1
_
ECE65 Lecture Notes (F. Najmabadi), Spring 2006 35
Ignoring 4Q
2
term compared to 1 (because Q is small),we get:
ω
u
=
ω
0
Q
=
√
ω
c2
ω
c1
_
ω
c1
/ω
c2
= ω
c2
For ω
l
, if we ignore 4Q
2
term compared to 1, we will ﬁnd ω
l
= 0. We should, therefore,
expand the square root by Taylor series expansion to get the ﬁrst order term:
ω
u
≈
ω
0
2Q
_
1 +
1
2
4Q
2
−1
_
=
ω
0
2Q
×2Q
2
= ω
0
Q = ω
c2
What are WideBand and NarrowBand Filters? Typically, a wideband ﬁlter is
deﬁned as a ﬁlter with ω
c2
ω
c1
(or ω
c2
≥ 10ω
c1
). In this case, Q ≤ 0.35 (prove this!). A
narrowband ﬁlter is usually deﬁned as a ﬁlter with B ω
0
(or B ≤ 0.1ω
0
). In this case,
Q ≥ 10.
Example: Design a bandpass ﬁlter with cutoﬀ frequencies of 160 Hz and 8 kHz. The load
for this circuit is 1 MΩ.
As this is wideband, bandpass ﬁlter (ω
u
/ω
l
= f
u
/f
l
= 50 1), we use two low and
highpass RC ﬁlter stages similar to circuit above. The prototype of the circuit is shown
below:
o
−
+
i
−
+
−
+
1
High−Pass Low−Pass
2 1
1 2
V
V
C
V
R C
R
The highpass ﬁlter sets the lower cutoﬀ
frequency, and the 1 MΩ load sets the out
put impedance of this stage. Thus:
Z
o

max
= R
1
1 MΩ →R
1
≤ 100 kΩ
ω
c
(Highpass) = ω
l
=
1
R
1
C
1
= 2π ×160 →R
1
C
1
= 1 ×10
−3
kΩ
One should choose R
1
as close as possible to 100 kΩ (to make the C
1
small) and R
1
C
1
=
1×10
−3
using commercial values of resistors and capacitors. A good set here are R
1
= 100 kΩ
and C
1
= 10 nF.
The lowpass ﬁlter sets the upper cutoﬀ frequency. The load for this component is the input
resistance of the highpass ﬁlter, Z
i

min
= R
1
= 100 kΩ. Thus:
Z
o

max
= R
2
100kΩ →R
2
≤ 10 kΩ
ω
c
(Lowpass) = ω
u
=
1
R
2
C
2
= 2π ×8 ×10
3
→R
2
C
2
= 2 ×10
−5
ECE65 Lecture Notes (F. Najmabadi), Spring 2006 36
As before, one should choose R
2
as close as possible to 10 kΩ and R
2
C
2
= 2 × 10
−5
using
commercial values of resistors and capacitors. A good set here are R
2
= 10 kΩ and C
2
= 2 nF.
In principle, we can switch the position of lowpass and highpass ﬁlter stages in a wide
band, bandpass ﬁlter. However, the lowpass ﬁlter is usually placed before the highpass
ﬁlter because the value of capacitors in such an arrangement will be smaller. (Try redesigning
the above circuit with lowpass and highpass ﬁlter stages switched to see that one capacitor
become much smaller and one much larger.)
Exercise: Design an RLC ﬁlter with the speciﬁcations in the previous example. (Hint: Do
not set R = 100 kΩ as this would make the value of the inductor very large.)
2.4 Exercise Problems
Problem 1. Design a RLC bandpass ﬁlter with a lower cutoﬀ frequency of 1 kHz and a
bandwidth of 3 kHz. What is the center frequency and Q of this ﬁlter?
Problem 2. We have an ampliﬁer that ampliﬁes a 1 kHz signal from a detector. The load
for this ampliﬁer can be modeled as a 50 kΩ resistor. The ampliﬁer output has a large
amount of 60 Hz noise. We need to reduce the amplitude of noise by a factor of 10. Design
a ﬁrstorder passive ﬁlter which can be placed between the ampliﬁer and the load and does
the job. Would this ﬁlter aﬀect the 1 kHz signal that we are interested in? If so, by how
much?
Problem 3. The tuner for an FM radio requires a bandpass ﬁlter with a central frequency
of 100 MHz (frequency of a FM station) and a bandwidth of 2 MHz. a) Design such a ﬁlter.
b) What are its cutoﬀ frequencies?
Problem 4. A telephone line carries both voice band (04 kHz) and data band (25 kHz to
1 MHz). Design a ﬁlter that lets the voice band through and rejects the data band. The
ﬁlter must meet the following speciﬁcations: a) For the voice band, the change in transfer
function should be at most 1 dB; and b) The transfer function should be as small as possible
at 25 kHz, the low end of the data band.
ECE65 Lecture Notes (F. Najmabadi), Spring 2006 37
2.5 Solution to Exercise Problems
Problem 1. Design a RLC bandpass ﬁlter with a lower cutoﬀ frequency of 1 kHz and a
bandwidth of 3 kHz. What is the center frequency and Q of this ﬁlter?
o

+

+
i
C
V
R V
L
The circuit prototype is:
For a 2nd order bandpass ﬁlter:
B(Hz) = f
u
−f
l
f
u
= 1 + 3 = 4 kHz
B(rad/s) = 2πB(Hz) = 1.88 ×10
4
ω
u
= 2πf
u
= 2.51 ×10
4
ω
l
= 2πf
l
= 6.28 ×10
3
ω
0
=
√
ω
u
ω
l
= 1.26 ×10
4
B(rad/s) =
ω
0
Q
→ Q =
1.26 ×10
4
1.88 ×10
4
= 0.67
For the series RLC circuit:
ω
0
=
1
√
LC
→ LC =
1
ω
2
0
C =
1
Lω
2
0
=
1
10 ×10
−3
×(1.26 ×10
4
)
2
= 0.63 µF
Q =
ω
0
R/L
→
R
L
=
ω
0
Q
= B(rad/s)
R = LB = 10 ×10
−3
×1.88 ×10
4
= 188 Ω
Therefore, using commercial values, the design values are L = 10 mH, R = 180 Ω, and
C = 0.68 µF.
ECE65 Lecture Notes (F. Najmabadi), Spring 2006 38
Problem 2. We have an ampliﬁer that ampliﬁes a 1 kHz signal from a detector. The load
for this ampliﬁer can be modeled as a 50 kΩ resistor. The ampliﬁer output has a large
amount of 60 Hz noise. We need to reduce the amplitude of noise by a factor of 10. Design
a ﬁrstorder passive ﬁlter which can be placed between the ampliﬁer and the load and does
the job. Would this ﬁlter aﬀect the 1 kHz signal that we are interested in? If so, by how
much?
We want to have 1 kHz signals to go through but reduce 60 Hz signals, so we need a highpass
ﬁlter. The prototype of the circuit is shown below. For this circuit:
Inverting
Amp.
i
L
o

+

+
R
V
C
R
V
H(jω) =
V
o
V
i
=
1
1 −jω
c
/ω
ω
c
=
1
RC
Z
i

min
= R
Z
o

max
= R
As the output impedance of the inverting ampliﬁer circuit is “zero”, we do not need to worry
about the input impedance of our ﬁlter. The output impedance of the ﬁlter is restricted by
Z
o

max
= R 50 kΩ → R ≤ 5 kΩ
This ﬁlter should reduce the amplitude of 60 Hz (ω
60
= 2π × 60 = 120π rad/s) signal by a
factor of 10, i.e.,
H(jω = jω
60
) =
¸
¸
¸
¸
V
o
V
i
¸
¸
¸
¸
60 Hz
=
1
_
1 + (ω
c
/ω
60
)
2
= 0.1
1 + (ω
c
/ω
60
)
2
= 100 →
1
RC
= ω
c
≈ 10ω
60
= 3751 rad/s → RC = 2.67 ×10
−4
Reasonable choices are R = 3.9 kΩ (to keep it below 5 kΩ) and C = 68 nF (f
c
≈ 600 Hz).
The impact on 1 kHz signal (ω
1000
= 2000π rad/s) can be found from:
H(jω = jω
1000
) =
1
_
1 + (ω
c
/ω
1000
)
2
=
1
_
1 + (3751/6283)
2
= 0.86
So the amplitude of 1 kHz signal is reduced by 14% (or by 1.3 dB).
ECE65 Lecture Notes (F. Najmabadi), Spring 2006 39
Problem 3. The tuner for an FM radio requires a bandpass ﬁlter with a central frequency
of 100 MHz (frequency of a FM station) and a bandwidth of 2 MHz. a) Design such a ﬁlter.
b) What are its cutoﬀ frequencies?
o
−
+
i
−
+
V
V R
L
C
Because this is not a wideband ﬁlter, the
simplest ﬁlter will be an RLC ﬁlter as is
shown. For this ﬁlter:
ω
0
=
1
√
LC
= 2π100 ×10
6
Q =
ω
0
B
=
¸
L
R
2
C
=
2π100 ×10
6
2π ×2 ×10
6
= 50
Using a L = 1 µH inductor:
1
LC
= 4π
2
10
16
→
1
C
= 4π
2
10
16
×10
−
6 → C = 2.5 ×10
−12
F
Choose: C = 2.2 pF
L
R
2
C
= 2, 500 → R
2
=
L
2, 500C
=
10
−6
2, 500 ×2.2 ×10
−12
= 182 → R = 13.5 Ω
Choose: R = 13 Ω (L = 1 µH and C = 2.2 pF).
To ﬁnd the cutoﬀ freqnecies, we not:
B = f
u
−f
l
= 2 MHz
f
0
=
_
f
u
f
L
= 100 MHz
Solution of the above two equations in two unknowns will give f
l
≈ 99 MHz and f
u
≈
101 MHz.
ECE65 Lecture Notes (F. Najmabadi), Spring 2006 40
Problem 4. A telephone line carries both voice band (04 kHz) and data band (25 kHz to
1 MHz). Design a ﬁlter that lets the voice band through and rejects the data band. The
ﬁlter must meet the following speciﬁcations: a) For the voice band, the change in transfer
function should be at most 1 dB; and b) The transfer function should be as small as possible
at 25 kHz, the low end of the data band.
We need a lowpass ﬁlter as it should allow lowfrequency signals (voice band) to go through
while eliminating highfrequency signals (data band). The prototype of an RC lowpass ﬁlter
is shown and its transfer function is:
H(jω) =
1
1 +jω/ω
c
=
1
1 +jf/f
c
i
o
+
−
+
−
ω  H(j ) 
f (kHz)
4 25
1
2
0
−1dB
V
V
R
C
The cutoﬀ frequency of the ﬁlter is not given and
it should be found from the speciﬁcations. First,
we need the change in transfer function to be at
most 1 dB for the frequency range of 04 kHz.
The transfer function of ﬁlters that satisfy this
constraint is the curve labeled “1” in the ﬁgure
and any transfer function located to the right of
this curve (such as transfer function labeled “2”).
Second, the transfer function should be as small as possible at 25 kHz. This requires that
we choose the cutoﬀ frequency as small as possible. Therefore, the transfer function of our
ﬁlter should be curve labeled “1” as it has the smallest possible value at 25 kHz:
20 log (H(jf = 4 kHz)) = −1 dB → H(jf = 4 kHz) = 0.891
Using the expression for H(jω), we have:
H(jf = 4 kHz) =
1
_
1 + (f/f
c
)
2
= 0.891
f/f
c
= 0.509 → f
c
=
f
0.509
= 7.85 kHz
f
c
=
1
2πRC
= 7.85 ×10
3
→ RC = 2.0 ×10
−5
Choosing C = 1 nF, we have R = 2.0 ×10
4
. The commercial values then are C = 1 nF and
R = 20 kΩ.
ECE65 Lecture Notes (F. Najmabadi), Spring 2006 41
It is clear that H(jω) is maximum when denominator is smallest, i.e., ω → 0 and H(jω) decreases as ω is increased. Therefore, this circuit allows “lowfrequency” signals to pass through while “blocking” highfrequency signals (i.e., reduces the amplitude of the voltage of the highfrequency signals). The reference to deﬁne the “low” and “high”frequencies is the cutoﬀ frequency: “low”frequencies mean frequencies much lower than ωc . To ﬁnd the cutoﬀ frequency, we note that the H(jω)M ax = 1 occurs at ω = 0 (alternatively ﬁnd d H(jω) /dω and set it equal to zero to ﬁnd ω = 0 which maximizes H(jω)). Therefore, H(jω)max = 1
1 1 H(jω)ω=ωc = √ H(jω)max = √ 2 2 1 =√ 2 1 + (ωc L/R)2 Therefore, ωc = R L and H(jω) = 1 1 + jω/ωc 1 −→ ωc L 1+ R
2
=2
→
ωc L =1 R
Input Impedance: Using the deﬁnition of the input impedance, we have: Zi = Vi = jωL + R Ii
The value of the input impedance depends on the frequency ω. For good voltage coupling, we need to ensure that the input impedance of this ﬁlter is much larger than the output impedance of the previous stage. Since we do not know the frequency of the input signal, we need to ensure that good voltage coupling criteria is satisﬁed for all frequencies (or all possible values of Zi ). As such, the minimum value of Zi is an important number. Zi is minimum when the impedance of the inductor is zero (ω → 0). Zi min = R
L
Output Impedance: The output impdenace can be found by “killing” the source and ﬁnding the equivalent impdenace between output terminals: Zo = jωL R
R
Zo
ECE65 Lecture Notes (F. Najmabadi), Spring 2006
22
A sound which appears to be twice as loud actually has 10 times power. 20 log10 H(jω) = 20 log10 H1 (jω) + 20 log10 H2 (jω) + . 3) Plot of H(jω)dB versus frequency has special properties that make analysis simpler. the plot asymptotes to straight lines at low and high frequencies as is shown below. Zo max = R Bode Plots and Decibel The voltage transfer function of a twoport network (and/or the ratio of output to input powers) is usually expressed in Bel: Number of Bels = log10 Po Pi or Number of Bels = 2 log10 Vo Vi because P ∝ V 2 . Najmabadi).where the source resistance is ignored.. making it easier to ﬁnd the overall response of the system. 2) If several twoport network are placed in a cascade (output of one is attached to the input of the next). the maximum value of Zo is an important number. we need to ensure that the output impedance of this ﬁlter is much smaller than the input impedance of the next stage for all frequencies. Decibel translates the output signal to what ear hears.... For good voltage coupling. Zo is maximum when the impedance of the inductor is inﬁnity (ω → ∞). Spring 2006 23 . ECE65 Lecture Notes (F.. is equal to the product of all transfer functions: H(jω) = H1 (jω) × H2 (jω) × . the overall transfer function. Bel is a large unit and decibel (dB) is usually used: Number of decibels = 20 log10 Vo Vi or Vo Vi = 20 log10 dB Vo Vi There are several reasons why decibel notation is used: 1) Historically. Again. Human ear “hears” the sound in a logarithmic fashion. H.. For example. etc. the analog systems were developed ﬁrst for audio equipment. H(jω)dB = H1 (jω)dB + H2 (jω)dB + . the value of the output impedance depends on the frequency ω.
The behavior of the phase of H(jω) can be found by examining H(jω) = − tan−1 (ω/ωc). H(jω) ≈ −90◦ . Because of this. ω/ωc 1. there is a 3 dB diﬀerence between maximum gain and gain at the cutoﬀ frequency: 20 log H(jωc ) − 20 log H(jω)max = 20 log 1 H(jωc ) = 20 log √ H(jω)max 2 ≈ −3 dB Bode plots are plots of H(jω)dB (magnitude) and H(jω) (phase) versus frequency in a semilog format (i. H(jω)dB changes by 20 dB. we see that. At low frequencies. It means that if ω is increased by a factor of 10 (a decade). The intersection of these two “asymptotic” values is at 1 = 1/(ω/ωc) or ω = ωc .. H(jω) ≈ 0 and at high frequencies. using dB deﬁnition. H(jω)dB = 20 log 1 = 20 log(ωc ) − 20 log(ω) ω/ωc which is a straight line with a slope of 20 dB/decade in the Bode plot. H(jω) ≈ −45◦ . Bode plots of ﬁrstorder lowpass RL ﬁlters are shown below (W denotes ωc ). ω/ωc H(jω) ≈ 1 ω/ωc → 1. the cutoﬀ frequency is also called the “corner” frequency.Also. ω/ωc 1.e. At low frequencies. ω/ωc 1. H(jω)dB H(jω) At high frequencies. H(jω) ≈ 1 which is also a straight line in the Bode plot. At cutoﬀ frequency. Spring 2006 24 . ECE65 Lecture Notes (F. Najmabadi). ω axis is a log axis).
Note that K can be negative.a lowpass ﬁlter): H(jω) = H(jω) = − 1 + (ω/ωc)2 K ω tan−1 K ωc K For RL ﬁlter. transfer functions characterize a twoport network. this grouping will also help reduce the math that we do in analyzing various circuits. (Note: we identiﬁed the circuit and found the cutoﬀ frequency without doing any math!).General ﬁrstorder lowpass ﬁlters As we discussed before. it is useful to group twoport networks into families based on their voltage transfer functions. and ωc = R/L. For no load resistance (“openloop” transfer function). To facilitate this grouping. Vo can be found from the voltage divider formula: Vo = 1/(jωC) 1 Vi = Vi R + 1/(jωC) 1 + j(ωRC) 1 H(jω) = 1 + jωRC + V i + C Vo  We see that the voltage transfer function of this circuit is similar to transfer function of a general ﬁrstorder lowpass ﬁlter. the convention is to simplify the voltage transfer function to a form such that the “Real” part of the denominator of H(jω) is unity (i. As we will see later in this section. this is a lowpass ﬁlter with K = 1 and ωc = 1/RC. K = 1. and in that case. ECE65 Lecture Notes (F. Najmabadi). the voltage transfer function of a ﬁrstorder lowpass ﬁlter is in the form: H(jω) = K 1 + jω/ωc The maximum value of H(jω) = K is called the ﬁlter gain. the denominator should be 1 + j · · · or 1 − j · · · ). As such. Note that the exponent of ω in the denominator is +1 so that H(jω) decreases with frequency (thus. The lowpass RL ﬁlter discussed before is part of the family of ﬁrstorder lowpass ﬁlters (ﬁrst order means that ω appears in the denominator with an exponent of 1 or −1. So. Spring 2006 25 . In general.e.. the “minus” sign adds 180◦ phase shift to the transfer function as is denoted by K/K factor above. R Lowpass RC ﬁlters A series RC circuit as shown also acts as a lowpass ﬁlter.
(Exercise: Show this. Ho (jω). consider the terminated RC ﬁlter shown: From the circuit. For this example. a resistive load is considered but the analysis can be easily extended to an impedance load.). For example. and terminated.We could. transfer functions of a twoport network: H(jω) = ZL Ho (jω) = ZL + Z o RL RL + R 1 jωC × 1 1 + jωRC 26 ECE65 Lecture Notes (F. Najmabadi). H(jω). we ﬁnd input and output Impedances Zi = R + Zo = R 1 jωC 1 jωC and and Zi min = R Zo max = R R + V i Terminated RL and RC lowpass ﬁlters Now let us examin the eﬀect of a load on the performance of our RL and RC ﬁlters. Spring 2006 . Input Impedance: Zi = R + Output Impedance: Zo = R 1 jωC 1 jωC RL Zi min = R Zo max = R We could have arrived at the same results using the the relationship between openloop. of course. H(jω) = Vo 1/(jωC) RL R /R = = Vi R + [1/(jωC) RL ] 1 + j(ωR C) + C Vo RL with R =R RL This is similar to the transfer function for unterminated RC ﬁlter but with resistance R being replaced by R . Therefore. ωc = 1 1 = RC (R RL )C and H(jω) = R /R 1 + jω/ωc We see that the impact of the load is to reduce the ﬁlter gain (K = R /R < 1) and to shift the cutoﬀ frequency to a higher frequency as R = R RL < R. do the math following the procedure in analyzing the lowpass RL ﬁlter to get the same answer. Following the same procedure as for RL ﬁlters.
ECE65 Lecture Notes (F. Vi 1 + jω/ωc Zi min = R Zo max = R ωc = (R RL RL )/L. Again for RL Zo or RL Zo max = R (our condition for good voltage coupling). Terminated RL lowpass ﬁlters The parameters of the terminated RL ﬁlters can be found similarly: Voltage Transfer Function: H(jω) = Input Impedance: Zi = jωL + R Output Impedance: Zo = (jωL) Vo 1 = . Spring 2006 27 . RL .) Also. note that the output impdenace of the terminated circuit is exactly the same as the openloop version. R. and input and output resistances become the same as before. Furthermore.(Exercise: show this. Exercise: Derive above equations for the transfer function and input and output impdenacess. the impact of load is to shift the cutoﬀ frequency to a lower value. Here. Filter gain is not aﬀected. R ≈ R and the terminated RC ﬁlter will look exactly like an unterminated ﬁlter – The ﬁlter gain is one. the shift in cutoﬀ frequency disappears and the ﬁlter will look exactly like an unterminated ﬁlter. Najmabadi). it can be seen that as long as RL Zo or RL Zo max = R (our condition for good voltage coupling). the shift in cutoﬀ frequency disappears.
Spring 2006 28 . ω/ωc 1. H(jω) ∝ ω (a +20dB/decade line) and H(jω) = 90◦ At high frequencies. H(jω) = 1 + (ωc /ω)2 K H(jω) = + K ωc tan−1 K ω Bode Plots of ﬁrstorder highpass ﬁlters (K = 1) are shown below.2 Firstorder high pass ﬁlters In general. Najmabadi). the voltage transfer function of a ﬁrstorder highpass ﬁlter is in the form: H(jω) = K 1 − jωc /ω It is a ﬁrstorder ﬁlter because ω appears in the denominator with an exponent of −1. It is a highpass ﬁlter because H = 0 for ω = 0 and H is constant for highfreqnecies.7 of H(jω)M ax.2.) The maximum value of H(jω) = K is called the ﬁlter gain. H(jω) ∝ 1 (a line with a slope of 0) and H(jω) = 0◦ H(jω) H(jω) ECE65 Lecture Notes (F. √ Paramter ωc is the cutoﬀ freqnecy of the ﬁlter (Exercise: prove that H(jωc ) is 1/ 2 = 0. ω/ωc 1. The asymptotic behavior of this class of ﬁlters is: At low frequencies.
Najmabadi). The openloop voltage transfer function of this ﬁlter is: Vo R 1 H(jω) = = = Vi R + 1/(jωC) 1 − j(1/ωRC) + Vi  C + R Vo  Therefore.Highpass RC ﬁlters A series RC circuit as shown acts as a highpass ﬁlter. Spring 2006 29 . ECE65 Lecture Notes (F. Again. Input and output impdenaces of this ﬁlter can be found similar to the procedure used for lowpass ﬁlters: Input Impedance: Zi = R + 1 jωC 1 jωC and and Zi min = R Zo max = R R + V i Output Impedance: Zo = R Highpass RL ﬁlters A series RL circuit as shown also acts as a highpass ﬁlter. we ﬁnd the openloop tranfunction to be: ωc = R L H(jω) = 1 1 − jωc /ω and and Zi min = R Zo max = R + L Vo  Input Impedance: Zi = R + jωL jωL Output Impedance: Zo = R Exercise: Compute the voltage transfer function and input and output impdenaces of terminated RC and RL ﬁlters. this is a ﬁrstorder highpass ﬁlter with K = 1 and ωC = 1/RC.
Secondorder bandpass ﬁlters: Secondorder band pass ﬁlters include two storage elements (two capacitors. Najmabadi).3 Bandpass ﬁlters A band pass ﬁlter allows signals with a range of frequencies (pass band) to pass through and attenuates signals with frequencies outside this range. This procedure will give two roots: ωl and ωu . two inductors.and highpass ﬁlters.2. or one of each). Spring 2006 30 . Upper cutoﬀ frequency. upper and lower cutoﬀ frequencies of practical band pass ﬁlter are deﬁned as the frequencies at which the magnitude of the voltage transfer √ function is reduced by 1/ 2 (or 3 dB) from its maximum value.  H(j ω)  Pass Band ωl ωu ω B ≡ ω u − ωl : ω0 Q≡ : B As with practical low. 1 K H(jωc) = √ H(jω)max = √ = 2 2 ωc ω0 2 =1 − ω0 ωc ωc ω0 2 2 =0 ωc − ω 0 ± Q K ωc ω0 1 + Q2 − ω0 ωc ωc ω0 − ω0 ωc = ±1 2 Q2 → Q ECE65 Lecture Notes (F. ωl : ωu : ω0 ≡ √ ω l ωu : Lower cutoﬀ frequency. The lower and upper cutoﬀ √ frequencies can be calculated by noting that H(jω)max = K. Quality factor. Band width. Center frequency. The transfer function for a secondorder bandpass ﬁlter can be written as K ω ω0 − 1 + jQ ω0 ω K H(jω) = ω0 ω 1 + Q2 − ω0 ω H(jω) = 2 H(jω) = − K ω ω0 tan−1 Q − K ω0 ω The maximum value of H(jω) = K is called the ﬁlter gain. setting H(jωc ) = K/ 2 and solving for ωc .
ω/ω0 1. ω/ω0 At high frequencies. respectively): ωl = ω 0 1 + ω0 1 − 2 4Q 2Q ωu = ω 0 1 + ω0 1 + 2 4Q 2Q Bode plots of a secondorder ﬁlter is shown below. Note that as Q increases. H(jω)db H(jω) Asymptotic behavior: At low frequencies. H(jω) ∝ 1/ω (a 20dB/decade line). There are two ways to solve secondorder ﬁlter circuits. Solving these equation we will get 4 roots (two roots per equation). The other two roots are the lower and upper cutoﬀ frequencies (ωl and ωu . and Q ≡= ω0 /B to ﬁnd the quality factor.The above equation is really two quadratic equations (one with + sign in front of fraction and one with a − sign). the bandwidth of the ﬁlter become smaller and the H(jω) becomes more picked around ω0 . The two examples below show the two methods. Then. and H(jω) → 90◦ 1. use the formulas above to ﬁnd the lower and upper cutoﬀ frequencies. 1) One can try to write H(jω) in the general form of a secondorder ﬁlters and ﬁnd Q and ω0 . H(jω) = K (purely real) H(jω) = K (maximum ﬁlter gain). and H(jω) → −90◦ At ω = ω0 . Two of these four roots will be negative which are not physical as ωc > 0. one can directly ﬁnd the √ upper and lower cutoﬀ frequencies and use ω0 ≡ ωl ωu to ﬁnd the center frequency and B ≡ ωu − ωl to ﬁnd the bandwidth. Spring 2006 31 . H(jω) ∝ ω (a +20dB/decade line). 2) Alternatively. and H(jω) = 0◦ . Note that one can always ﬁnd ω0 and k rapidaly as H(jω0 ) is purely real and H(jω0 ) = k ECE65 Lecture Notes (F. Najmabadi).
Spring 2006 . ωu . we divide top and bottom of transfer function of series RLC bandpass ﬁlters by R: H(jω) = 1 ωL 1 1+j − R ωRC Comparing the above with the general form of the transfer function. To ﬁnd Q and ω0 . K. Therefore. Method 1: We transform the transfer function in a form similar to general form of the transfer function for second order bandpass ﬁlters: H(jω) = K ω ω0 1 + jQ − ω0 ω Note that the denominator of the general form is in the form 1 + j . ω0 . and ωl . we have H(jω) = H(jω) = R Vo = Vi R + jωL + 1/(jωC) R R + j ωL − 1 ωC + Vi  There are two approaches to ﬁnd ﬁlter parameters. Equating these similar terms we get: Qω ωL Q L = → = ω0 R ω0 R Qω0 1 1 = → Qω0 = ω ωRC RC We can solve these two equations to ﬁnd: ω0 = √ 1 LC Q= ω0 = R/L L R2 C 32 ECE65 Lecture Notes (F. Najmabadi). . . we ﬁnd K = 1.L C + R Vo  Series RLC Bandpass ﬁlters Using voltage divider formula. we note that the imaginary part of the denominator has two terms. one positive and one negative (or one that scales as ω and the other that scales as 1/ω) similar to the general form of transfer function of 2ndorder bandpass ﬁlters (which includes Qω/ω0 and −Qω0 /ω).
Starting with the transfer function for the series RLC ﬁlter: H(jω) = R R + j ωL − 1 ωC We note that the transfer function is real if coeﬃcient of j in the denominator is exactly zero (note that this happens for ω = ω0 ). i. Spring 2006 33 . Method 2: In this method. H(jω) is purely real and 2) K = H(jω = jω0 ). Najmabadi). ω0 L − Also K = H(jω = jω0 ) = R =1 R 1 =0 ω0 C −→ ω0 = √ 1 LC The cutoﬀ frequencies can then be found by setting: K 1 H(jωc) = √ = √ 2 2 1+ ωc L 1 − R ωc RC 2 =2 which can be solved to ﬁnd ωu and ωl .The lower and upper cutoﬀ frequencies can now be found from the formulas on page 31. Input and Output Impedance of bandpass RLC ﬁlters Zi = jωL + Zi min = R Zo = jωL + 1 1 + R = j ωL − +R jωC ωC occurs at ω = ω0 1 jωC R → Zo max = R ECE65 Lecture Notes (F. Some simpliﬁcations can be made by noting: 1) At ω = ω0 .e.. we directly calculate the ﬁlter parameters similar to the procedure followed for general form of transfer function in page 30.
These ﬁlters are widely used (when appropriate. Spring 2006 . we can ﬁnd the ﬁlter parameters by either of two methods above. Transforming the transfer function to a form similar to the general form (left for students) gives: 1 1 + ωc1 /ωc2 ωc1 /ωc2 1 + ωc1 /ωc2 √ K= Q= ω0 = ωc1 ωc2 34 ECE65 Lecture Notes (F. or we should have R1 R2 . Najmabadi).WideBand BandPass Filters Bandpass ﬁlters can be constructed by putting a highpass and a lowpass ﬁlter back to back as shown below. The highpass ﬁlter sets the lower cutoﬀ frequency and the lowpass ﬁlter sets the upper cutoﬀ frequency of such a bandpass ﬁlter. the input impedance of the highpass ﬁlter (actually Zi min = R1 ) should be much larger than the output impedance of the lowpass ﬁlter (actually Zo max = R2 ). + C2 V1 − Low−Pass High−Pass R1 + Vo − In order to have good voltage coupling in the above circuit. In that case we can use unterminated transfer functions: H(jω) = H1 (jω) × H2 (jω) = ωc1 = 1/(R1 C1 ) 1 1 × 1 + jω/ωc2 1 − jωc1 /ω ωc2 = 1/(R2 C2 ) H(jω) = 1 1 = (1 + jω/ωc2)(1 − jωc1 /ω) (1 + ωc1 /ωc2 ) + j(ω/ωc2 − ωc1 /ω) Again. see below) instead of an RLC ﬁlter as inductors are usually bulky and take too much space on a circuit board.  H (j ω)  2  H (j ω)  1  H (j ω)  X  H (j ω)  1 2 ω c2 ω ω c1 ω R2 + V − i ω l = ωc2 ω u = ωc1 C1 ω An example of such a bandpass ﬁlter is two RC lowpass and highpass ﬁlters put back to back.
H2 (jω) achieves its maximum value (1 in this case) only when ω/ωc2 > 3. Spring 2006 . For these wideband ﬁlters (ωc1 ωc2 ).. if ωc2 ωc1 (i. The lower cutoﬀ frequency moves to a value lower than ωc1 and the upper cutoﬀ frequency moves to a value higher than ωc2 . If ωc2 is not ωc1 (i.e. the cutoﬀ frequencies will not be ωc1 and ωc2 (those frequencies are 3 dB lower than H(jω)max = 1). Similarly for the low pass ﬁlter. When ωc2 is not ωc1 . bandwidth becomes vanishingly small and Q should become very large. H(jω)max becomes smaller than 1. This can be seen by examining the quality factor of this ﬁlter at the limit of ωc2 = ωc1 ωc1 /ωc2 1 + ωc1 /ωc2 1 = 0. for example. H(jω) is 3 dB lower at the cutoﬀ frequency. the center frequency of the ﬁlter will be at least a factor of three away from both cutoﬀ frequencies and H(jω) = H1  × H2  achieves its maximum value of 1.5 1+1 Q= = while our asymptotic description of previous page indicated that when ωc2 = ωc1 .One should note that the Bode plots of previous page are “asymptotic” plots. It can be seen that H1 (jω) achieves its maximum value (1 in this case) only when ω/ωc1 < 1/3. they are called “wideband” ﬁlters. More importantly. A comparison of “asymptotic” Bode plots for ﬁrstorder highpass ﬁlters are given in page 28. Since the cutoﬀ frequencies are located 3 dB below the maximum values.. The real H(jω) diﬀers from these asymptotic plots. we ﬁnd from above: K=1 H(jω) = Q= ωc1 /ωc2 ω0 = √ ωc2 ωc1 1 1 + j(ω/ωc2 − ωc1 /ω) We then substitute for Q and ω0 in the expressions for cutoﬀ frequencies (page 31) to get: ωu = ω 0 1 + ωl = ω 0 1 + ω0 1 ω0 = + 2 4Q 2Q 2Q 1 ω0 ω0 + = 2 4Q 2Q 2Q 1 + 4Q2 + 1 1 + 4Q2 − 1 35 ECE65 Lecture Notes (F. This can be seen by examining the equation of K above which is always less than 1 and approaches 1 when ωc2 ωc1 .e. Because these ﬁlters work only when ωc2 ωc1 . In the bandpass ﬁlter above. H1 and H2 will not reach their maximum of 1 and the ﬁlter H(jω)max = H1  × H2  will be less than one. we can never make a “narrow” band ﬁlter by putting two ﬁrstorder highpass and lowpass ﬁlters back to back. ωc2 < 10ωc1 ). ωc2 ≥ 10ωc1 ). Najmabadi).
therefore. Spring 2006 36 . we will ﬁnd ωl = 0. and the 1 MΩ load sets the output impedance of this stage. We should. The prototype of the circuit is shown below: R2 C1 + C2 V1 − Low−Pass High−Pass R1 + Vo − + V − i The highpass ﬁlter sets the lower cutoﬀ frequency. In this case.we get: √ ωc2 ωc1 ω0 ωu = = ωc2 = Q ωc1 /ωc2 For ωl . Q ≥ 10. The load for this component is the input resistance of the highpass ﬁlter. Najmabadi). The lowpass ﬁlter sets the upper cutoﬀ frequency. expand the square root by Taylor series expansion to get the ﬁrst order term: ωu ≈ ω0 1 ω0 × 2Q2 = ω0 Q = ωc2 1 + 4Q2 − 1 = 2Q 2 2Q What are WideBand and NarrowBand Filters? Typically. A narrowband ﬁlter is usually deﬁned as a ﬁlter with B ω0 (or B ≤ 0. As this is wideband.1ω0 ). if we ignore 4Q2 term compared to 1. Example: Design a bandpass ﬁlter with cutoﬀ frequencies of 160 Hz and 8 kHz. Thus: Zo max = R1 1 MΩ → R1 ≤ 100 kΩ ωc (Highpass) = ωl = 1 = 2π × 160 R1 C1 → R1 C1 = 1 × 10−3 kΩ One should choose R1 as close as possible to 100 kΩ (to make the C1 small) and R1 C1 = 1×10−3 using commercial values of resistors and capacitors. we use two low. In this case. Thus: Zo max = R2 100kΩ 1 = 2π × 8 × 103 R2 C2 → R2 ≤ 10 kΩ → R2 C2 = 2 × 10−5 ωc (Lowpass) = ωu = ECE65 Lecture Notes (F. bandpass ﬁlter (ωu /ωl = fu /fl = 50 1).35 (prove this!). a wideband ﬁlter is deﬁned as a ﬁlter with ωc2 ωc1 (or ωc2 ≥ 10ωc1 ). Zi min = R1 = 100 kΩ. The load for this circuit is 1 MΩ. A good set here are R1 = 100 kΩ and C1 = 10 nF.and highpass RC ﬁlter stages similar to circuit above.Ignoring 4Q2 term compared to 1 (because Q is small). Q ≤ 0.
the low end of the data band. we can switch the position of lowpass and highpass ﬁlter stages in a wideband. ECE65 Lecture Notes (F. The tuner for an FM radio requires a bandpass ﬁlter with a central frequency of 100 MHz (frequency of a FM station) and a bandwidth of 2 MHz. The ampliﬁer output has a large amount of 60 Hz noise. Najmabadi).) 2. bandpass ﬁlter. (Try redesigning the above circuit with lowpass and highpass ﬁlter stages switched to see that one capacitor become much smaller and one much larger. the lowpass ﬁlter is usually placed before the highpass ﬁlter because the value of capacitors in such an arrangement will be smaller. We need to reduce the amplitude of noise by a factor of 10. by how much? Problem 3. b) What are its cutoﬀ frequencies? Problem 4. The ﬁlter must meet the following speciﬁcations: a) For the voice band. Would this ﬁlter aﬀect the 1 kHz signal that we are interested in? If so. Design a ﬁrstorder passive ﬁlter which can be placed between the ampliﬁer and the load and does the job.As before.) Exercise: Design an RLC ﬁlter with the speciﬁcations in the previous example. What is the center frequency and Q of this ﬁlter? Problem 2. Design a RLC bandpass ﬁlter with a lower cutoﬀ frequency of 1 kHz and a bandwidth of 3 kHz. Spring 2006 37 . A good set here are R2 = 10 kΩ and C2 = 2 nF.4 Exercise Problems Problem 1. and b) The transfer function should be as small as possible at 25 kHz. In principle. However. (Hint: Do not set R = 100 kΩ as this would make the value of the inductor very large. A telephone line carries both voice band (04 kHz) and data band (25 kHz to 1 MHz). the change in transfer function should be at most 1 dB. a) Design such a ﬁlter. one should choose R2 as close as possible to 10 kΩ and R2 C2 = 2 × 10−5 using commercial values of resistors and capacitors. The load for this ampliﬁer can be modeled as a 50 kΩ resistor. Design a ﬁlter that lets the voice band through and rejects the data band. We have an ampliﬁer that ampliﬁes a 1 kHz signal from a detector.
and C = 0. What is the center frequency and Q of this ﬁlter? L C + R Vo  The circuit prototype is: For a 2nd order bandpass ﬁlter: B(Hz) = fu − fl ωu = 2πfu = 2. Najmabadi).2. using commercial values.26 × 104 = 0.88 × 104 For the series RLC circuit: ω0 = √ 1 1 → LC = 2 ω0 LC 1 1 C= = = 0.28 × 103 1.5 Solution to Exercise Problems Problem 1.26 × 104 )2 Q= ω0 R/L → R ω0 = = B(rad/s) L Q R = LB = 10 × 10−3 × 1.26 × 104 B(rad/s) = ω0 Q → Q= fu = 1 + 3 = 4 kHz 4 + Vi  B(rad/s) = 2πB(Hz) = 1.88 × 10 ωl = 2πfl = 6. the design values are L = 10 mH. R = 180 Ω.67 1. Spring 2006 38 . ECE65 Lecture Notes (F.63 µF 2 Lω0 10 × 10−3 × (1. Design a RLC bandpass ﬁlter with a lower cutoﬀ frequency of 1 kHz and a bandwidth of 3 kHz.68 µF.51 × 104 √ ω0 = ωu ωl = 1.88 × 104 = 188 Ω Therefore.
Problem 2. The prototype of the circuit is shown below. Vi R + Vo R L This ﬁlter should reduce the amplitude of 60 Hz (ω60 = 2π × 60 = 120π rad/s) signal by a factor of 10.9 kΩ (to keep it below 5 kΩ) and C = 68 nF (fc ≈ 600 Hz). by how much? We want to have 1 kHz signals to go through but reduce 60 Hz signals. so we need a highpass ﬁlter. The ampliﬁer output has a large amount of 60 Hz noise. i.e.67 × 10−4 60 Hz 1 + (ωc /ω60 )2 = 100 → 1 = ωc ≈ 10ω60 = 3751 rad/s RC Reasonable choices are R = 3. Spring 2006 39 . ECE65 Lecture Notes (F.3 dB).. we do not need to worry about the input impedance of our ﬁlter. The load for this ampliﬁer can be modeled as a 50 kΩ resistor. For this circuit: 1 Vo = H(jω) = Vi 1 − jωc /ω 1 ωc = RC Zi min = R Zo max = R As the output impedance of the inverting ampliﬁer circuit is “zero”. The output impedance of the ﬁlter is restricted by Zo max = R 50 kΩ → R ≤ 5 kΩ C + Inverting Amp. The impact on 1 kHz signal (ω1000 = 2000π rad/s) can be found from: H(jω = jω1000 ) = 1 1 + (ωc /ω1000 )2 = 1 1 + (3751/6283)2 = 0. We need to reduce the amplitude of noise by a factor of 10.1 → RC = 2.86 So the amplitude of 1 kHz signal is reduced by 14% (or by 1. H(jω = jω60 ) = Vo Vi = 1 1 + (ωc /ω60 )2 = 0. We have an ampliﬁer that ampliﬁes a 1 kHz signal from a detector. Design a ﬁrstorder passive ﬁlter which can be placed between the ampliﬁer and the load and does the job. Would this ﬁlter aﬀect the 1 kHz signal that we are interested in? If so. Najmabadi).
The tuner for an FM radio requires a bandpass ﬁlter with a central frequency of 100 MHz (frequency of a FM station) and a bandwidth of 2 MHz. we not: B = fu − fl = 2 MHz f0 = fu fL = 100 MHz Solution of the above two equations in two unknowns will give fl ≈ 99 MHz and fu ≈ 101 MHz. Najmabadi). To ﬁnd the cutoﬀ freqnecies.2 pF L = 2.2 × 10−12 R = 13. For this ﬁlter: ω0 = √ Q= 1 = 2π100 × 106 LC 2π100 × 106 L = = 50 R2 C 2π × 2 × 106 L + Vi − R C + Vo − ω0 = B Using a L = 1 µH inductor: 1 = 4π 2 1016 LC Choose: C = 2. Spring 2006 40 . 500 × 2.Problem 3. 500 R2 C → R2 = L 10−6 = = 182 → 2. the simplest ﬁlter will be an RLC ﬁlter as is shown. 500C 2.2 pF). b) What are its cutoﬀ frequencies? Because this is not a wideband ﬁlter. ECE65 Lecture Notes (F. a) Design such a ﬁlter.5 × 10−12 F Choose: R = 13 Ω (L = 1 µH and C = 2.5 Ω → 1 = 4π 2 1016 × 10− 6 C → C = 2.
We need a lowpass ﬁlter as it should allow lowfrequency signals (voice band) to go through while eliminating highfrequency signals (data band). ECE65 Lecture Notes (F.0 × 104 . The ﬁlter must meet the following speciﬁcations: a) For the voice band.509 → RC = 2. First. The commercial values then are C = 1 nF and R = 20 kΩ.891 → H(jf = 4 kHz) = 0. The transfer function of ﬁlters that satisfy this constraint is the curve labeled “1” in the ﬁgure and any transfer function located to the right of this curve (such as transfer function labeled “2”). and b) The transfer function should be as small as possible at 25 kHz. the low end of the data band. R + Vi −  H(j ω)  0 −1dB 2 1 f (kHz) 4 25 C + Vo − Second. we need the change in transfer function to be at most 1 dB for the frequency range of 04 kHz.0 × 10−5 1 = 7. Najmabadi). Spring 2006 41 .85 × 103 2πRC Choosing C = 1 nF. This requires that we choose the cutoﬀ frequency as small as possible. we have R = 2. the change in transfer function should be at most 1 dB.85 kHz 0.Problem 4. A telephone line carries both voice band (04 kHz) and data band (25 kHz to 1 MHz).509 fc = → 1 1 + (f /fc )2 fc = = 0. Therefore. the transfer function of our ﬁlter should be curve labeled “1” as it has the smallest possible value at 25 kHz: 20 log (H(jf = 4 kHz)) = −1 dB Using the expression for H(jω). we have: H(jf = 4 kHz) = f /fc = 0. the transfer function should be as small as possible at 25 kHz. The prototype of an RC lowpass ﬁlter is shown and its transfer function is: 1 1 H(jω) = = 1 + jω/ωc 1 + jf /fc The cutoﬀ frequency of the ﬁlter is not given and it should be found from the speciﬁcations.891 f = 7. Design a ﬁlter that lets the voice band through and rejects the data band.