Filter Design

UNIT-4
 Filters
• Filters are required in most of the circuits in whole of the RF
range. In fact, filter is a two-port, reciprocal, passive, and
linear device, which attenuates heavily the unwanted signal
frequencies, while allowing transmission of wanted signal
frequencies.
All microwave filters are,
1. Made by using reactive elements only, i.e. capacitance and
inductance with sections of transmission lines having
distributive impedance, as lumped elements cannot be used
as microwave frequencies
2. Assumed to the loss less with a very small resistance.
3. Designed to operate between generator impedance (Zg)
and load impedance (ZL), with both resistive and normally
equal to 50 Ω.
• Normally, filters are of four types, namely low pass
(LS), high pass (HP), band pass (BP), and band stop
(BS).
• The low pass filter allows low-frequency signals to
be transmitted from its input and to the output with
little attenuation.
• This attenuation (called insertion loss IL) increases
significantly beyond cut-off frequency point. The
band pass filters allows the passing of signals to the
output with low attenuation (i.e. low insertion loss,
IL), between the given range of upper (fu) and lower
(fL) cut-off frequencies. The reverse is true in the
case of band stop filters.
• Insertion loss: The basic parameter of filter
design is insertion loss. The input power (Pi)
given to the filter may not be equal to the
output power (PL) from the filter to the load.
in resistive heating.
• Transmission phase change of signal (/t):
The phase shift of signal when transmitted through a
filter is different at various frequencies for the four
filters.
• Group delay (Tgd): It is the time it takes the
information (or signal) to traverse the filter network.
It is computed from the slope of the phase versus
frequency curve
• Band width (BW3dB): It is the difference between
the upper cut-off frequency (fu) and lower cut-off
frequencies (fL) in band pass and band stop filters.
These two cut-off frequencies are those frequencies
where the signal level has a difference of 3 dB with
reference to that at centre band frequency (fc)
• Ripple and ripple band amplitude
The flatness of signal in the pass band filter can be
quantified by specifying the ripple band amplitude in
terms of dB, which is 3 dB Shape factor (SF):This
describes the sharpness of the filter response and is
the ratio of 60 dB band width to 3 dB band width
• Quality factor (Q): The frequency selectivity of a
filter can be defined in terms of Q, the quality factor
as the ratio of average energy stored per cycle to the
energy lost per cycle at wo.
 RF Filter Design Techniques
• For getting the IL versus frequency profile of filters close
to the ideal ,various design techniques are available.
Only the three design techniques, e.g. the first order, and
the two specialised designs of Butterworth and
Chebyshev, which are with increasing level of
performances in terms of better slope factor (i.e.
sharpness of skirt). A comparison of the IL versus
frequency profiles of these three techniques is given in
Fig.,
 Design Techniques
• (a) First-order basic design techniques: This is the lowest
level of design of filter, consisting of only one module of L, C,
R component values of which are computed by ABCD
analyses.
• (b) Specialised filter design by insertion loss method: This is
done by using prototype low pass filter as the basic design. For
better performance with sharper profile of IL versus frequency,
we need to have more number of L, C components.
• For this, two design techniques are used:
• (i) Butterworth design for maximally flat or binomial.
• (ii) Chebyshev (or equal-ripple) design technique.
• The first design technique of Butterworth gives a flat profile of
IL versus frequency, while the Chebyshev design gives sharper
IL versus frequency profile but with unavoidable ripples
appearing in the pass frequency portion in all the filters (i.e.
 Specialised Filter Design by Insertion Loss Method

• Besides first-order filter design, the specialised

designs, i.e. Butterworth and Chebyshev designs, are
actually used in practice. These designs have sharper
roll-off IL versus frequency profile at cut-off.
• Before analysing these designs, let us study the power
loss ratio (PLR) for matched source and load:
• Choice of this polynomial ratio M/N decides
the type of response of IL versus frequency
profile. The practical and useful responses are:
• (i) Butterworth (called maximally flat)
• (ii) Chebyshev (called equal-ripple)
• (iii) Ecliptic function
• (iv) Linear phase.
• In the process of designing any of the four filters, the starting
point is the designing a prototype LPF, The reactant of this
LPF are in terms of a parameter called gn values as given in
the first column of Table. Here ‘n’ is the order (Number of
components of that LPF) choosen by designer.
• For better performance of filter profile, i.e. sharper IL versus
frequency, the number of L, C modules has to be larger. For
this, the number of components is defined by the order ‘n’ of
that filter.
• Also it has been observed that for the same number of
components, the Chebyshev filter gives fast roll-off of IL after
cut-off frequency, but at the cost of ripples in the IL versus
frequency . For computing the normalised component values,
called ‘g’ values of L and C components which are required in
the distributed form of a transmission line of our choice
(waveguide, coaxial or microstrip line), standard tables are
available. Higher is the order, the sharper is the IL profile.
• In both, the design techniques (Butterworth and Chebyshev)
following three steps are followed, for finally getting the de-
normalised values of L, C components from ‘g’ values and
finally realising them on the transmission line sections.
• (a) Choose the filter design order ‘n’: This choice is done by
the sharpness of the IL versus frequency profile expected, then
get the ‘g’ values of ‘L’ and ‘C’ of the filter from Table.1 for
Butterworth or Table.2 for Chebyshev.
• Next two steps are for scaling, i.e. impedance or frequency
transformation to the desired Z0 and fc values of our choice.
• (b) Impedance transformation for getting prototype of LPF
components: From the ‘gn’ values of the prototype LPF, of
order ‘n’, the normalised values of ‘L’ and ‘C’ (Lpn, Cpn) are
obtained using formulas of Table above for the normalised
frequency
Therefore,fn these
= f/fc prototype
as per given below:
values Ln, Cn are
valid for all frequencies.
Computing actual values of elements Ln,Cn from prototype LPF elements
Lpn,Cpn by Impedance transformation
• (c) Frequency transformation for getting actual
components of the four filter (LPF,HPF, BPS, BSF): Here
we choose the frequency and band width of our use, and
then by using the prototype values of Lpn, Cpn of order n
of LPF, we calculate the actual values
of inductance (Henry) and capacitance (Farad) by Table
10.4 for the required filter (may be LPF or HPF or BPF or
BSF). This is for order n, centre frequency (xc), and band
width (BW) chosen by designer.
• (d) Filter implementation: We have got the lumped
values of ‘Ln’ and ‘Cn’ of filter,which has to be realised
in the transmission line by line segment of different
structures (as shown in Figs. 10.5 and 10.6) as distributed
circuit elements.
 Butterworth Filter Design(Maximally Flat
or Binomial)

• This type of filter is known as maximally flat filter since no

ripple is permitted in its attenuation profile. For such low pass
filters, the insertion loss is found by the following polynomial:

where ‘a’ is a constant and we may select it to be a = 1, so that at (ω/ωc) =

1, the value of IL = 10 log (2) = 3 dB at cut-off frequency ω = ωc at fn = 1.
Possible realizations of π type and T type of LPF
• Two possible realisations of p type and T type of generic normalised low
pass filter are shown in Fig. 10.13. The element values of L, C
components in Fig. 10.13 are numbered as g0 (for generate) to gn + 1
(load). The elements in the circuit alternate between series inductance
and shunt capacitance.
• These elements are defined as:
g0 = Generator resistance in pi-type filters or generator conductance
T–type filters or inductance for series inductor or capacitance for shunt
capacitor
gn + 1 = Load resistance for p-type filter
or load conductance for T-type filter
gn = Inductance for series inductor of the filter or capacitance for shunt
capacitor of the filter
(n = 1, 2, 3, … N)
• For example, from Table 10.5 for n = 3,
• g1 = 1.0, g2 = 2.0, g3 = 1.0 and
• g3+1 = normalised load resistance or conductance = 1.
 Example Problem
a)A third-order Chebyshev low pass filter is to be designed with maximum allowed
ripple of 3 dB with cut-off frequency of 2.1 GHz on 50 Ω line.
(b) Extend this design for band pass filter with centre frequency of 2.5 GHz and ±10%
band width (i.e. 20%). (c) Draw the IL performance of each.
The fig shows the attenuation response for 3 dB chebyshev design LPF
• First we find the g values of Lpn and Cpn for LPF prototype
and then actual filter by impedance and frequency
transformation
• Step 1: Third-order filter g values: we use,
n = 3,

• Step 2: Prototype LPF by impedance transformation.

• Step 3: Frequency transformations
3a) getting actual values of L and C for LPF
from proto LPF values for fC = 2.1 GHz.
• For fc = 2.1 GHz; ωc = 2πfc = 13.188X109 rad/s.

BPF: Getting actual values of L and C of BPF from proto LPF

values for f0 = 2.5 GHz
Computing actual values of elements Ln,Cn from prototype LPF elements
Lpn,Cpn by Impedance transformation
 HPF Design
HPF: Actual values of L and C for HPF from prototype LPF values by Table 10.4.
The HPF circuit of order 3 (n = 3)
 BSF Design
• BSF: Actual values of L and C from proto
LPF .
 Problem 2
• Solve the above problem for Butterworth design. Here all the calculations
are same except gn values for n = 3. Butterworth filter designs are from
Table.