DIT, Kevin St Electric circuits Waed 3

Chapter 1
Basic filter configurations
An electrical filter is a circuit that modifies ac signals. It passes signals,
unattenuated, over a passband region, but attenuates frequencies in a stopband
region. There are five basic filter types:

♦ Low-pass filter,
♦ High-pass filter,
♦ Bandpass filter,
♦ Bandstop filter, and
♦ All-pass filter

Consider, for example, an ideal low-pass filter. A 1 Volt RMS, signal applied to
such a device, will emerge at the same level (1 V RMS or 0 dB) over the passband
region and 0 V (minus infinity dB) over the stopband region. Such a theoretical
filter device is called a BRICK-WALL FILTER. The response of an ideal low-
pass filter with the actual response superimposed is shown in figure 1.1.

Figure 1.1: Brick-wall filter response and actual response.

A low-pass filter passes all frequencies from dc (f = 0) up to the CUT-OFF

FREQUENCY ωp. Above this frequency, the signal is attenuated at a rate
determined by the filter order. In the high-pass filter, the reverse is true. All
frequencies below the cut-off frequency are attenuated, and all frequencies above
the cut-off are passed unattenuated. In reality, the signal is gradually attenuated
above dc. Because of the shape of the capacitive reactance with frequency graph,
the attenuation does not become apparent until the cut-off frequency ωp is reached.
The passband and stopband regions for low-pass and high-pass filters are shown
in Figures 1.2 and 1.3 respectively. Between the passband and stopband regions,
is a transition region, which can be reduced in width by increasing the filter order
n.

Copyright: Paul Tobin School of Electronics and Comms. Eng. 1

DIT, Kevin St Electric circuits Waed 3

Figure 1.2: Low-pass filter. Figure 1.3: High-pass filter.

The bandpass filter passes all frequencies over a desired band and attenuates
signals outside this band. The bandstop filter attenuates the signal over a band of
frequencies and outside this band, the signal is unattenuated. We have to specify,
two frequency pairs, for both band-pass and band-stop filters shown in Figures 1.4
and 1.5.

Figure 1.4: Bandpass filter. Figure 1.5: Bandstop filter.

The all-pass filter is not, in a strict sense, a filter but is used to modify the phase
response of circuits or signals. The response for such a filter is shown in Figure
1.6. The frequency response is flat over a large range of frequencies but has a
variable phase response.

Figure 1.6: Frequency response of an all-pass filter.

Copyright: Paul Tobin School of Electronics and Comms. Eng. 2

DIT, Kevin St Electric circuits Waed 3
The pole-zero plot of a digital all-pass filter is shown in Figure 1.7.

Figure 1.7: Pole-zero plot for the all pass filter.

Several techniques are used to implement the filtering process:

♦ Analogue LCR circuits,

♦ Active CR filters (Includes switched capacitor filters).
♦ Digital signal processing methods (FIR and IIR etc)
Each filtering technique has its advantages, but in this series of lectures, we deal
with active filter circuits only. Active filters have the advantage of requiring no
inductors, which can be large at very low frequencies and radiate magnetic fields.
However, active filters require dc power supplies and have a limited frequency
range (even with a good op-amp). The order of the filter will determine the roll-
off rate in the transition region i.e. first-order has a roll-off rate of - 20 dB/decade,
second-order is - 40 dB/decade etc. The order is also defined in terms of the
number of poles (second order has two poles).
However, unlike digital filters, active filters are normally restricted to tenth
order or less. Much higher order filtering is achieved using FIR and IIR digital
filtering techniques discussed in the other section of the course. An
APPROXIMATION FUNCTION TECHNIQUE uses mathematical functions to
approximate the filter response required. Several approximating functions exist, of
which the most commonly used ones are:
♦ Bessel,
♦ Butterworth,
♦ Chebychev, and
♦ Elliptic.
In this series of lectures, we will concern ourselves with the Butterworth and the
Chebychev loss functions, only. This is not to downplay the importance of the
other functions but similar analysis techniques can be applied to those functions.

Copyright: Paul Tobin School of Electronics and Comms. Eng. 3

DIT, Kevin St Electric circuits Waed 3
Some of these approximating functions will achieve better results for different
conditions and requirements. In certain circumstances, we will require a large
attenuation outside the pass-band and the phase response and step response might
not be of concern. Other requirements might demand a linear phase response over
a given range and could tolerate ripple in the pass-band. In audio circuits, a poor
transient or step response would make the audio sound hollow. LOSS
FUNCTION analysis is a hangover from passive filter analysis. This analysis
technique uses a magnitude loss function A($), which, when inverted, produces
the required transfer function. To illustrate the procedure, consider the first order
low-pass CR filter in Figure 1.8.

Figure 1.8 CR low-pass filter

Figure 1.9: Amplitude and phase response of the LPF CR circuit

The transfer function is written using the potential divider principle (see last years
notes on transfer functions):

Vo 1 1 1 / CR ωp
T .F = = = = = (1.1)
V in 1 + sCR 1 + j 2π fCR s + 1 / CR s +ω p

Copyright: Paul Tobin School of Electronics and Comms. Eng. 4

 DIT, Kevin St Electric circuits Waed 3
1 1 ω
T .F = ∠ − tan −1 ωCR = ∠ − tan −1 (1.2)
1+ω C R
2 2 2
ω ωp
1 + ( )2
ωp

The denominator of 1.1 is called an approximation function (Butterworth type).

Here a pole exists at s = - 1/CR. The amplitude and phase response of this network
is shown in Figure 1.9. The cut-off frequency ωp is calculated for the circuit
values C = 10 nF and R = 100 kΩ as:

1 1
ωp = = = 1000 rs-1
CR 10.10 −910 5

The attenuation at the stopband edge frequency of 10 krs-1 is 20 dB. This first
order filter will achieve this since it has a roll-off of -20 dB/decade. From 1 krs-1
to 10 krs-1, the attenuation is 20 dB with the attenuation at the passband edge
frequency, equal to -3 dB.

Exercise 1
Derive the high-pass filter transfer function and sketch, for the circuit values given
in the low-pass filter above, the amplitude response.

Approximating functions
In the following section we will show how the inverse of an APPROXIMATING
functions will produce a desired response such as the low -pass filter whose
frequency response is shown in Figure 1.9. A brick- wall response is not
achievable in practice but there are functions, which, if of sufficiently high order,
will achieve a good approximation to this response. The Butterworth and
Chebychev loss functions are all-pole loss functions and will be analyzed, in
detail, in the next section. Bessel and elliptic (Caur) are applied in a similar
manner but no analysis will be carried out here. The frequency response for
Chebychev and Butterworth loss functions, are shown in figure 1.10 and figure
1.11.

Copyright: Paul Tobin School of Electronics and Comms. Eng. 5

DIT, Kevin St Electric circuits Waed 3

Figure 1.10: Second-order Butterworth approximation function.

Figure 1.11: Second-order Chebychev loss function with 2 dB ripple.

Increasing the order of the Chebychev approximation function produces a result
shown in Figure 1.12 (note the change in roll-off over the region 0 to 1).

Copyright: Paul Tobin School of Electronics and Comms. Eng. 6

DIT, Kevin St Electric circuits Waed 3

Figure 1.12: Third-order Chebychev loss function with 2 dB ripple.

The approximation function is the inverse of the transfer function. Compare the
rate at which the second-order Butterworth and the Chebychev functions rise in
dB when the frequency is increased. The Chebychev function rises at a much
greater rate. We are using a function A(s) which, when inverted, will give the
required filter transfer function H(s), i.e.

1
H ( s) = (1.3)
A( s )
Taking the function used in figure 1.12 and inverting, yields the more familiar
filter response in figure 1.13. The roll-off rate of the Chebychev, third-order
function is greater than a third-order Butterworth function, but as you can see,
there is ripple or undulations present in the passband. This might cause problems
in certain audio applications. The elliptical filter also has ripple in the passband
but is present in the stopband region also.

Copyright: Paul Tobin School of Electronics and Comms. Eng. 7

DIT, Kevin St Electric circuits Waed 3

Figure 1.13: Chebychev third-order filter with 2 dB ripple.

The Butterworth and Chebychev loss functions are not primarily concerned with
the signal phase but that’s not to say that the phase response is unimportant. The
phase response is critical when considering complex signals such as a modulated
carrier. In this situation, it is critical that the phase response is linear so that each
frequency component in the signal will have equal time delays. The signal will be
distorted if each frequency component suffers a different delay. If the signal
comprises digital signals, then errors can occur due to a non-linear phase
response.

Figure 1.14: Group delay for normalised Butterworth functions.

Copyright: Paul Tobin School of Electronics and Comms. Eng. 8

DIT, Kevin St Electric circuits Waed 3
This delay is characterised by group delay, defined as, the rate of change of phase,
with a change in frequency. A graph of the group delay for normalised
Butterworth approximation functions plotted against the order n, is shown in
Figure 1.14. You can see that the GP worsens for higher order values of n.
The complete process for designing an active filter is shown in Figure 1.14(a)

Figure 1.14a: Overall active filter analysis

Copyright: Paul Tobin School of Electronics and Comms. Eng. 9