FACULTY OF SCIENCES AND TECHNOLOGY

Electrical Engineering and Automation Department

Academic Year: 2024 – 2025
Master 1 Automation and System
Signal Processing Course Mr. Dahmane Ghenam

Chapter 2
Continuous-time filters (Analog filters)

2.1 Introduction
Filtering is an important area of signal processing. This chapter focuses on the design of
continuous time (CT) filters. We are particularly interested in the ideal filters and practical
filters that are categorized in four different categories (low-pass, high-pass, band-pass, and
band-stop). Filtering characteristics of a system are indicated by its response to sinusoids of
various frequencies varying from 0 to ∞. Such characteristics are called the frequency
response of the filter.

2.2 Ideal filters

Ideal filters have transfer functions such that the frequency components of the input signal that
fall within the pass-band are passed to the output without modification, whereas the frequency
components of the input signal that fall into the stop-band are completely eliminated from the
output signal. The ideal filter is not physically realizable because the impulse response ℎ(𝑡) of
this filter is non-causal which mean that the filter is non-causal system (unrealizable). As in
Figs. 2.2.1a 2.2.1b represent the ideal low-pass filter and its non-causal impulse response ℎ(𝑡)
(ℎ(𝑡) ≠ 0 for all 𝑡 < 0).

Fig. 2.2.1

2.2.1 Ideal lowpass filter

The transfer function 𝐻(𝜔) of an ideal lowpass filter is defined as follows:
 𝐴 |𝜔| ≤ 𝜔𝑐
𝐻(𝜔) = { (2.1)
0 |𝜔| > 𝜔𝑐

where 𝜔𝑐 is referred to as the cut-off frequency of the filter. The pass-band of the lowpass filter
is given by |𝜔| ≤ 𝜔𝑐 , while the stop-band of the lowpass filter is given by 𝜔𝑐 < |𝜔| < ∞. The
magnitude spectrum |𝐻(𝜔)| is plotted in the Fig. 2.2.2, and the phase frequency response
∠𝐻(𝜔) of an ideal lowpass filter is zero for all frequencies.

Fig. 2.2.2

2.2.2 Ideal highpass filter

The transfer function 𝐻(𝜔) of an ideal highpass filter is defined as follows:

0 |𝜔| ≤ 𝜔𝑐
𝐻(𝜔) = { (2.2)
𝐴 |𝜔| > 𝜔𝑐

The stop-band of the lowpass filter is given by |𝜔| ≤ 𝜔𝑐 , while the pass-band of the lowpass
filter is given by 𝜔𝑐 < |𝜔| < ∞. The magnitude spectrum |𝐻(𝜔)| is plotted in the Fig. 2.2.3,
and the phase frequency response ∠𝐻(𝜔) of an ideal lowpass filter is zero for all frequencies.

Fig. 2.2.3

2.2.3 Ideal bandpass filter

The transfer function 𝐻(𝜔) of an ideal bandpass filter is defined as follows:

𝐴 𝜔𝑐1 ≤ |𝜔| ≤ 𝜔𝑐2

𝐻(𝜔) = { (2.3)
0 |𝜔| < 𝜔𝑐1 ∨ |𝜔| > 𝜔𝑐2

where 𝜔𝑐1 and 𝜔𝑐2 are the cut-off frequency of the filter. The bandpass filter has a finite
bandwidth as it only allows a range of frequencies (𝜔𝑐1 ≤ 𝜔 ≤ 𝜔𝑐2 ). The magnitude spectrum
|𝐻(𝜔)| is plotted in the Fig. 2.2.4.
 Fig. 2.2.4

2.2.4 Ideal bandstop filter

The transfer function 𝐻(𝜔) of an ideal bandstop filter is defined as follows:

0 𝜔𝑐1 ≤ |𝜔| ≤ 𝜔𝑐2

𝐻(𝜔) = { (2.4)
𝐴 |𝜔| < 𝜔𝑐1 ∨ |𝜔| > 𝜔𝑐2

where 𝜔𝑐1 and 𝜔𝑐2 are the cut-off frequency of the filter. The ideal bandstop filter is the
converse of the ideal bandpass filter as it eliminates a certain range of frequencies (𝜔𝑐1 ≤ 𝜔 ≤
𝜔𝑐2 ) from the input signal. The magnitude spectrum |𝐻(𝜔)| is plotted in the Fig. 2.2.5.

Fig. 2.2.5

2.3 Non-ideal filters (practical filters)

An ideal filter has a passband (unity gain) and a stopband (zero gain) with a sudden transition
from the passband to the stopband. There is no transition band. For practical (or realizable)
filters, on the other hand, the transition from the passband to the stop band ( or vice versa) is
gradual and takes place over a finite band of frequencies.

2.3.1 First order low-pass filter

The general form for the transfer function of a first order low-pass filter is

𝜔𝑐 1
𝐻(𝑠) = 𝐺0 = 𝐺0 𝑠
𝑠 + 𝜔𝑐 1+𝜔
𝑐

Where

𝜔𝑐 : Cut-off frequency
𝐺0 : The gain of the filter at low frequency (𝜔 → 0) and which might be 1 or smaller (for a
passive circuit with a voltage divider) or have a magnitude greater than 1 for an active circuit.

The RC-circuit is an example of a first order passive low-pass filter shown in Fig. 2.3.1.

This filter have a negative pole (𝑠 = −𝜔𝑐 ) which is located in the left half s-plane (stable
system). See Fig. 2.3.2.

𝑹
𝑗𝜔

S-plane
𝑽𝒊𝒏 𝑪 𝑽𝒐

−𝜔𝑐 𝜎

Fig. 2.3.1 Fig. 2.3.2

The transfer function of the RC- circuit is given by:

1
𝐶𝑠 1
𝐻(𝑠) = =
1 1 + 𝑅𝐶𝑠
𝑅 + 𝐶𝑠

Let 𝑠 = 𝑗𝜔 then

1
1 𝑅𝐶
𝐻(𝑗𝜔) = =
1 + 𝑗𝑅𝐶𝜔 𝑗𝜔 + 1
𝑅𝐶

- The gain 𝐺0 = 1 ( the gain 𝐺0 is the gain 𝐺(𝜔) as 𝜔 → 0)

- The magnitude response 𝐺(𝜔) (gain) of the filter is determine as:

1 1
𝐺(𝜔) = |𝐻(𝑗𝜔)| = =
√1 + (𝑅𝐶)2 𝜔 2 𝜔 2
√1 + (
𝜔𝑐 )

- The corresponding logarithmic gain function 𝐺𝑑𝐵 (𝜔) is defined as

𝜔 2
𝐺𝑑𝐵 (𝜔) = 20 log(𝐺(𝜔)) = 20 log(|𝐻(𝑗𝜔)|) = −20log (√1 + ( ) )
𝜔𝑐

- The cut-off frequency is calculated as

𝐺0 1 𝟏
𝐺(𝜔𝑐 ) = |𝐻(𝑗𝜔𝑐 )| = = ⇒ 𝝎 = 𝝎𝒄 = , 𝐺0 = 1, gain at low frequency
√2 √2 𝑹𝑪

Or
 𝐺0
𝐺𝑑𝐵 (𝜔𝑐 ) = 20log(|𝐻(𝑗𝜔𝑐 )|) = 20log ( ) = −20log√2 = −3 dB
√2
- The phase response 𝜑(𝜔) is given by:
𝜔
𝜑(𝜔) = ∠𝐻(𝑗𝜔) = − tan−1 ( ) = − tan−1(𝑅𝐶𝜔)
𝜔𝑐

- The Fig. 2.3.3 a, b and c show the magnitude response linear scale plot, magnitude response
Bode plot and phase response plot, for the resistance 𝑅 = 500 Ω, the capacitance 𝐶 =
1⁄𝜋 μF.
- For 𝜔 = 𝜔𝑐 , 𝐺(𝜔𝑐 ) = 1⁄√2 ≅ 0.707
- 𝜔 = 𝜔𝑐 , 𝜑(𝜔𝑐 ) = −45°

(a)

-
(b)

(c)

Fig. 2.3.3
- The Fig. 2.3.4 shows the active circuit with inverting OP amp (Operational amplifier) for
the first order low pass filter.

We have:

1
𝑉𝑜 (𝑠) 𝑍2 (𝑠) 𝑅2 𝑅2 𝐶
𝐻(𝑠) = =− = (− )
𝑉𝑖 (𝑠) 𝑍1 (𝑠) 𝑅1 𝑠 + 1
𝑅2 𝐶
Z2(s)
C

Z1(s) R2
R1
-
Vi(s) + Vo(s)

Fig. 2.3.4

Clearly, this is the low pass filter with the fallowing parameters:

𝑅2 1
𝐺0 = − = −25 𝜔𝐶 = = 6250 rad⁄s
𝑅1 𝑅2 𝐶

- The Fig. 2.3.5 a and b show the magnitude response Bode plot and phase response plot,
for the resistance 𝑅1 = 1 KΩ, 𝑅2 = 25 KΩ and the capacitance 𝐶 = 6.4 nF.

(a) Fig. 2.3.5 (b)

2.3.2 First order High-pass filter

The general form for the transfer function of a first order High-pass filter is
 𝑠 𝐺0
𝐻(𝑠) = 𝐺0 =
𝑠 + 𝜔𝑐 1 + 𝜔𝑐
𝑠
Where

𝜔𝑐 : Cut-off frequency

𝐺0 : The gain of the filter at high frequency (𝜔 → ∞)

The RL-circuit is an example of a first order passive High-pass filter shown in Fig. 2.3.6.

This filter have one zero at origin and a negative pole (𝑠 = −𝜔𝑐 ) which is located in the left
half s-plane (stable system). See Fig. 2.3.7.

𝑗𝜔

S-plane

−𝜔𝑐 𝜎

Fig. 2.3.6 Fig. 2.3.7

The transfer function of the RL- circuit is given by:

𝐿𝑠 𝑠
𝐻(𝑠) = =
𝑅 + 𝐿𝑠 𝑠 + 𝑅
𝐿
- Clearly, 𝜔𝑐 = 𝑅 ⁄𝐿 and 𝐺0 = 1

Let 𝑠 = 𝑗𝜔 then

𝑗𝜔 𝑗𝜔
𝐻(𝑗𝜔) = =
𝑗𝜔 + 𝜔𝑐 𝑗𝜔 + 𝑅
𝐿

- The gain 𝐺0 = 1 ( the gain 𝐺0 is the gain 𝐺(𝜔) as 𝜔 → ∞)

- The magnitude response 𝐺(𝜔) (gain) of the filter is determine as:

𝜔 1
𝐺(𝜔) = |𝐻(𝑗𝜔)| = =
√𝜔 2 + (𝜔𝑐 )2 √1 + (𝜔𝑐 )
2
𝜔

- The corresponding logarithmic gain function 𝐺𝑑𝐵 (𝜔) is defined as

𝜔𝑐 2
𝐺𝑑𝐵 (𝜔) = 20 log(𝐺(𝜔)) = 20 log(|𝐻(𝑗𝜔)|) = −20log (√1 + ( ) )
𝜔
- The cut-off frequency is calculated as

𝐺0 1 𝑹
𝐺(𝜔𝑐 ) = |𝐻(𝑗𝜔𝑐 )| = = ⇒ 𝝎 = 𝝎𝒄 = , 𝐺0 = 1, gain at high frequency
√2 √2 𝑳

- The phase response 𝜑(𝜔) is given by:

𝜔
𝜑(𝜔) = ∠𝐻(𝑗𝜔) = ∠𝑗𝜔 − ∠(𝑗𝜔 + 𝜔𝑐 ) = 90° − tan−1 ( )
𝜔𝑐

- The Fig. 2.3.8 a and b show the magnitude response Bode plot and phase response plot, for
the resistance 𝑅 = 500 Ω, the inductance 𝐿 = 1⁄4𝜋 H.

(a) Fig. 2.3.8 (b)

- The Fig. 2.3.9 shows the active circuit with no-inverting OP amp (Operational amplifier)
for the first order high pass filter.

2.3.3 Butterworth Filter

An ideal filter has a passband (unity gain) and a stopband (zero gain) with a sudden transition
from the passband to the stopband. There is no transition band. For practical (or realizable)
filters, on the other hand, the transition from the passband to the stop band (or vice versa) is
gradual and takes place over a finite band of frequencies. We therefore define a stopband to be
a band over which the gain is below some small number 𝐺𝑠 , as illustrated in Fig. 2.3.9.
Similarly, we define a passband to be a band over which the gain is between 1 and some number
𝐺𝑝 (𝐺𝑝 < 1), as shown in Fig. 2.3.9. We have selected the passband gain of unity for
convenience. It could be any constant.
 Fig. 2.3.9

2.3.3.1 Design of the Butterworth lowpass filter

The amplitude response |𝐻(𝜔)| of an Nth-order Butterworth lowpass filter is given by

1
|𝐻(𝑗𝜔)| = (2.5)
𝜔 2𝑁
√1 + ( )
𝜔𝑐

Where 𝜔𝑐 is the cut-off frequency of the filter.

Normalized Butterworth Filter

In the design procedure, it proves most convenient to consider a normalized filter ℋ(𝑠), whose
cut-off frequency is 𝜔𝑐 = 1. For such a filter, the amplitude characteristic in Eq.(2.5) reduces
to

1
|ℋ(𝑗𝜔)| = (2.6)
√1 + (𝜔)2𝑁

The amplitude responses |ℋ(𝑗𝜔)| of normalized lowpass Butterworth filters are depicted in
Fig. 2.3.2 for various values of N.

We can prepare a table of normalized transfer functions ℋ(𝑠) that yield the frequency response
in Eq. (2.6) for various values of 𝑁. Once the normalized transfer function is obtained, we can
obtain the desired transfer function 𝐻(𝑠) for any value of 𝜔𝑐 by simple frequency scaling, where
we replace 𝑠 by 𝑠⁄𝜔𝑐 in the normalized ℋ(𝑠).
 |ℋ(𝑗𝜔)|

Fig. 2.3.10

 To determine the corresponding transfer function ℋ(𝑠), recall that ℋ(−𝑗𝜔) is the
complex conjugate of ℋ(𝑗𝜔). Therefore,

1
ℋ(𝑗𝜔)ℋ(−𝑗𝜔) = |ℋ(𝑗𝜔)|2 =
1 + 𝜔 2𝑁
Substituting 𝑠 = 𝑗𝜔 in this equation, we obtain

1
ℋ(𝑠)ℋ(−𝑠) =
𝑠 2𝑁
1 + (𝑗 )

The poles of ℋ(𝑠)ℋ(−𝑠) occur when

𝑠 2𝑁 = −(𝑗)2𝑁

In this result, we use the fact that −1 = 𝑒 𝑗𝜋(2𝑘−1) for integral values of 𝑘 and 𝑗 = 𝑒 𝑗𝜋⁄2 to
obtain

𝑠 2𝑁 = 𝑒 𝑗𝜋(2𝑘−1+𝑁) 𝑘 integer

This equation yields the poles of ℋ(𝑠)ℋ(−𝑠) as

𝑗𝜋
(2𝑘+𝑁−1)
𝑝𝑘 = 𝑒 2𝑁 𝑘 = 1,2,3, ⋯ ,2𝑁 (2.7)

Observe that all poles have a unit magnitude; that is, they are located on a unit circle in the 𝑠
plane separated by angle 𝜋⁄𝑁 as illustrated in Fig. 2.3.3 for odd and even 𝑁.

Since ℋ(𝑠) is stable and causal, its poles must lie in the LHP (Left Half Plane). The poles of
ℋ(−𝑠) are the mirror images of the poles of ℋ(𝑠) about the vertical axis. Hence, the poles of
ℋ(𝑠) are those in the LHP and the poles of ℋ(−𝑠) are those in the RHP in Fig.2.3.3.

The poles corresponding to ℋ(𝑠) are obtained by setting 𝑘 = 1,2,3, ⋯ , 𝑁 in Eq. (2.7); that is,
 Fig. 2.3.3 Butterworth poles and mirrors: (a) odd order (N = 5) and (b) even order (N = 6).

ℋ(𝑠) is given by

1
ℋ(𝑠) = (2.8)
(𝑠 − 𝑝1 )(𝑠 − 𝑝2 ) ⋯ (𝑠 − 𝑝𝑁 )

In general

1 1
ℋ(𝑠) = = 𝑁 (2.9)
𝐵𝑁 (𝑠) 𝑠 + 𝑎𝑁−1 𝑠 𝑁−1 + ⋯ 𝑎1 𝑠 + 1

Where 𝐵𝑁 (𝑠) is the Butterworth polynomial of the 𝑁th order. Table 2.1 shows the coefficients
𝑎1 , 𝑎2 ⋯ 𝑎𝑁−2 , 𝑎𝑁−1 for various values of 𝑁; Table 2.2 shows 𝐵𝑁 (𝑠) in factored form.

Table 2.1

Table 2.2
Frequency Scaling

Although Tables 2 .1 and 2 .2 are for normalized Butterworth filters with 𝜔𝑐 = 1, the results
can be extended to any value of 𝜔𝑐 by simply replacing 𝑠 by 𝑠⁄𝜔𝑐 . This step implies replacing
𝜔 by 𝜔⁄𝜔𝑐 in Eq. (2.6).

Determination of Butterworth Filter Order 𝑵

If 𝐺̂𝑥 is the dB gain of a lowpass Butterworth filter at 𝜔 = 𝜔𝑥 , then according to Eq. (2.5),
𝜔𝑥 2𝑁
̂
𝐺𝑥 = 20 log10 |𝐻(𝑗𝜔𝑥 )| = −10 log [1 + ( ) ]
𝜔𝑐

Substitution of the specifications in Fig. 2.3.9a (gains 𝐺̂𝑝 at 𝜔𝑝 and 𝐺̂𝑠 at 𝜔𝑠 ) in this equation
yields
𝜔𝑝 2𝑁
𝐺̂𝑝 = −10 log [1 + ( ) ] (2.10)
𝜔𝑐
𝜔𝑠 2𝑁
𝐺̂𝑠 = −10 log [1 + ( ) ] (2.11)
𝜔𝑐
Or
𝜔𝑝 2𝑁 ̂
( ) = 10−𝐺𝑝 ⁄10 − 1 (2.12)
𝜔𝑐
𝜔𝑠 2𝑁
( ) = 10−𝐺̂𝑠 ⁄10 − 1 (2.13)
𝜔𝑐
Dividing Eq. (2.13) by Eq. (2.12), we obtain
2𝑁
𝜔𝑠 10−𝐺̂𝑠 ⁄10 − 1
( ) = ̂
𝜔𝑝 10−𝐺𝑝 ⁄10 − 1
And

log[(10−𝐺̂𝑠 ⁄10 − 1)⁄10−𝐺̂𝑝 ⁄10 − 1]

𝑁= (2.14)
2 log(𝜔𝑠 ⁄𝜔𝑝 )
Also from Eq. (2.12),
𝜔𝑝
𝜔𝑐 = 1⁄2𝑁
(2.15)
̂
[10−𝐺𝑝 ⁄10 − 1]
Alternatively, from Eq. (1.13),
𝜔𝑠
𝜔𝑐 = 1⁄2𝑁
(2.16)
[10−𝐺̂𝑠 ⁄10 − 1]
Since 𝑁 is rounded up to an integer value, Eqs. (2.12) and (2.13) do not generally return the
same value of 𝜔𝑐 .