Radio Frequency Filter Design: Use of discrete elements and stubs

Table of Contents

Introduction........................................................................................................................pg 2

Theory..................................................................................................................................pg 3

Analysis of the Circuit..........................................................................................................pg 5

Design calculations...............................................................................................................pg 11

Simulations...........................................................................................................................pg 16

Analysis of Results................................................................................................................pg 21

Conclusions...........................................................................................................................pg 22

Comments.............................................................................................................................pg 23

Bibliography..........................................................................................................................pg 24

1 of 25
Radio Frequency Filter Design: Use of discrete elements and stubs

Introduction

In this report we will cover the basics in the developing a VHF (very high frequency) bandpass
filter used in RF (radio frequency) electronics. The report provides the user with a brief explanation of
what a bandpass filter is and its characteristics. We proceed to detailing the general mathematical
procedure that must be done in order to create a bandpass filter that meets a required set of
parameters. To prove such design we illustrate the result of the bandpass filter design utilizing the
frequency response of the filter. CAD (Computer Aided Design) tools such as MATLAB and Orcad
Pspice are used to illustrate the frequency response and check if the desired parameters are met.
Finally an analysis of such results is provided in order to instruct readers on what may have gone
wrong and why, in order to provide the reader with a valid explanation and solution.

2 of 25
Radio Frequency Filter Design: Use of discrete elements and stubs

Theory

A band-pass filter is a device that passes frequencies within a certain range and rejects
(attenuates) frequencies outside that range. An example of an analogue electronic band-pass filter is
an RLC circuit (a resistor–inductor–capacitor circuit).An ideal bandpass filter would have a
completely flat passband (e.g. with no gain/attenuation throughout) and would completely attenuate
all frequencies outside the passband.

Illustration 1: An ideal Low Pass filter.

Additionally, the transition out of the passband would be instantaneous in frequency. In

practice, no bandpass filter is ideal. The filter does not attenuate all frequencies outside the desired
frequency range completely; in particular, there is a region just outside the intended passband where
frequencies are attenuated, but not rejected. This is known as the filter roll-off, and it is usually
expressed in dB of attenuation per octave or decade of frequency. Generally, the design of a filter seeks
to make the roll-off as narrow as possible, thus allowing the filter to perform as close as possible to its
intended design. Often, this is achieved at the expense of pass-band or stop-band ripple.
The bandwidth of the filter is simply the difference between the upper and lower cutoff frequencies.
The shape factor is the ratio of bandwidths measured using two different attenuation values to

3 of 25
Radio Frequency Filter Design: Use of discrete elements and stubs

determine the cutoff frequency, e.g., a shape factor of 2:1 at 30/3 dB means the bandwidth measured
between frequencies at 30 dB attenuation is twice that measured between frequencies at 3 dB
attenuation.

Outside of electronics and signal processing, one example of the use of band-pass filters is in the
atmospheric sciences. It is common to band-pass filter recent meteorological data with a period range
of, for example, 3 to 10 days, so that only cyclones remain as fluctuations in the data fields.

Another method to developing filters is utilizing distributed elements. A distributed element

filter is an electronic filter designed for frequencies above the VHF band (30 to 300 MHz). At these
frequencies, the physical length of passive components is a significant fraction of the wavelength of the
operating frequency, and it becomes difficult to use the conventional lumped element model. The
distributed element model allows these components to be designed using transmission line theory
better suited for these frequencies.

The component forms used to make a distributed element filter work by creating reactive
elements that function, to first order, like lumped inductors and capacitors. This is accomplished by
creating a discontinuity along the transmission line which appear as a reactive impedance to a
wavefront travelling down the line.

The most noticeable difference between a distributed element filter and its lumped element
approximation is that the former will have multiple passband replicas of the lumped element
prototype passband because transmission line transfer characteristics repeat at harmonic intervals.
These spurious passbands are undesirable in most cases.

Illustration 2: Equivalent lumped parameter circuit.

4 of 25
Radio Frequency Filter Design: Use of discrete elements and stubs

Analysis of the Circuit

Assumptions

• The simulated results will be close but not exactly equal to the design specifications.
• A small margin of error is expected since there are many factors that affect the outcome, such
as:
▪ simulation program models
▪ prototype parameters interpolated values
• Distributed element simulation will not bet as accurate as other simulations due to the fact that
the prototype parameters and meant for discrete elements.

Mathematical Procedure and Equations

When designing radio frequency filters, the low pass filter serves as a starting model in which
the designer builds upon. The low pass filter or LP filter, as it will be referred to in this report from
now on, is the foundation of each design. All other types of filters such as high pass (HP), bandstop
(BS), bandpass (BP), peak filters; will derived from the LP filter.
Another important aspect of designing a filter is that the order of the filter determines the attenuation
characteristic of such. The higher the order of the filter, the quicker it will reach it's desired
attenuation, but the more elements it will require. A good designer uses only the necessary
components when developing a filter.

In order to make the task of developing a filter easier, standardization is used in all the catalog
tables used to determine the element values. This normalizes the values, this way you can use the same
characteristic curves and tables, no matter what the specifications of your design. An advantage of this
is that the designer can compare different filters for the same specifications. In this design we will
make use of the Butterworth filter. This type of filter provides a very flat response in the passband and
contains no ripple. The passband is the frequency range where the signal is not attenuated, at least not
intentionally. The Butterworth is considered to be a medium grade type of filter

5 of 25
Radio Frequency Filter Design: Use of discrete elements and stubs

The attenuation of the Butterworth filter is given by eq.(1):

; where w = the frequency at which the attenuation is desire

wC = the cutoff frequency at 3dB of the filter

n = the number of elements in the filter.

In order to develop a BP filter using the LP as a prototype we first need to determine how many
elements the design of the prototype LP will require.

Using eq. (2)

where BW is the bandwidth of the required value of attenuation

and BWC is the 3dB bandwidth of the bandpass filter.

we transform the BP requirements into LP requirements.

The value obtained is compared in the graph below in order to determine how many elements the
design needs.

6 of 25
Radio Frequency Filter Design: Use of discrete elements and stubs

Illustration 3: Attenuation characteristic of the

Butterworth filters.
The value obtained in eq. (2) is plotted in the graph above and the designer determines the minimum
elements he needs to reach the specified attenuation.

Once the number of elements is known the designer proceeds to use a normalized catalog of element
values, which depends on the number of elements required.

Since it will be impossible or not practical to display every

single value that can be performed. Often an interpolation is
needed in order to obtain more precise data to use in the
prototype. In the design we will proceed to explain we use
MATLAB in order to perform the interpolation.

Once the corresponding element values are known you

continue to build the LP prototype using the element values
provided. The resulting schematic should be just like the one
at the top of the table. Once you have the LP prototype
schematic we need to transform that prototype into a BP
filter.

Illustration 4: Butterworth LP element

values (normalized).

7 of 25
Radio Frequency Filter Design: Use of discrete elements and stubs

The transformation format is shown below:

Illustration 5: LP to BP circuit transformation.

To achieve the correct equivalent BP circuit you must add an inductive element of the same value
(normalized) in shunt to the capacitive element. This makes that branch to become parallel resonant.
The same is done with the series branch, adding a capacitor to the existing inductor of the same value
in series will make this branch series resonant. This process is justified in that at the center frequency
(fC), the series elements will produce a short circuit, and the parallel elements will produce an open
circuit. Thus letting the desired frequency pass without attenuation, except for the voltage drop due to
the loading effect.
Now we go on to scale the BP configuration. That is, denormalize the element values using the
following equations:
for the parallel resonant branches:

eq. (3)
eq.(4)

for the series resonant branches:

eq. (5)
eq. (6)

8 of 25
Radio Frequency Filter Design: Use of discrete elements and stubs

where R = final load impedance

B = the 3dB bandwidth of the final design
fO = center frequency of the design
LN = normalized inductor bandpass element values
CN = normalized capacitor bandpass element values
This finishes the design process of BP filter utilizing a LP prototype.

The classical way to to visualize the transfer function of the filter and view the frequency response is to
use Laplace. The transfer function H(s) for LP filters is:

a0 eq. (7)
H  s= N n−1
 s a n−1∗s a n−2∗s n−2.. . a 1∗s na 0

For BP filters we need only to make a substitute eq. (8) into eq. (7) and we obtain the transfer function
for a Bandpass filter.

w0 s w 0
s=    eq. (8)
BW w0 s

We will use CAD tool MATLAB to plot the transfer function of our BP filter design and compare it to
our specifications.

9 of 25
Radio Frequency Filter Design: Use of discrete elements and stubs

Another way to design RF filters is using stubs. A stub is a length of transmission line or waveguide
that is connected at one end only. The free end of the stub is either left open-circuit or (especially in
the case of waveguides) short-circuited. Neglecting transmission line losses, the input impedance of
the stub is purely reactive; either capacitive or inductive, depending on the electrical length of the stub,
and on whether it is open or short circuit. Stubs may thus be considered to be frequency-dependent
capacitors and frequency-dependent inductors.

In order to determine the length of the stub the following equations are used:

eq.(9)

ZSC is the short circuit impedance used to determine the length of capacitors.

eq. (10)

ZOC is the open circuit impedance used to determine the length of inductors.

Since we know the impedance of the discrete elements and we assume that Z 0 = 50 ohm, we can
determine the electrical length βl. Usually the length l is left in terms on the wavelength λ.

10 of 25
Radio Frequency Filter Design: Use of discrete elements and stubs

Design Calculations

The design of the following BP filter will have to meet these constraints:
1) The center frequency fC must be 800 Mhz.
2) The bandwidth at -3 dB must be 10% of fC
3) At 25% of fC the minimum attenuation must be at leas -23dB.
4) Rs = 50Ω & RL = 90Ω.

Using eq. 2:
BW/BWC = f/fC = (200/8) = 2.5
Using the Butterworth curves of illustration 3 we can see that we need at least 3 elements; n=3.
We demand a 3 element Butterworth LP prototype for the required attenuation.

Proceeding to use table in illustration 4 for a (Rs/RL) = 0.5555 in order to know the normalized values
of our 3 element LP prototype. Since 0.555 is nowhere on the table we need to interpolate. Using a non
linear interpolation will yield much better results. The interpolation was done using an algorithm in
MATLAB. The code is below:

clc;
clf;
fc=800*10^6
BW=80*10^6
Rl=90
%Interpolation algorithm

%N=3 Butterworth Low-Pass Prototype Element Values

x= [.9 .8 .7 .6 .5 .4 .3 .2 .1]
c1= [.808 .844 .915 1.023 1.181 1.425 1.838 2.669 5.167]
l2= [1.633 1.384 1.165 .965 .779 .604 .440 .284 .138]
c3= [1.599 1.926 2.277 2.702 3.261 4.064 5.363 7.910 15.455]

%normalized element values

C1n=interp1(x,c1,(50/90),'spline','extrap')
L2n=interp1(x,l2,(50/90),'spline','extrap')
C3n=interp1(x,c3,(50/90),'spline','extrap')

11 of 25
Radio Frequency Filter Design: Use of discrete elements and stubs

%element values of transformation after scaling

%where Rl is the load impedance, BW the bandwidth, fc the center

%frequency and Ln and Cn the normalized values of the elements.

%parallel-resonant branch
C1=(C1n/(2*pi*BW*Rl))
L1=((Rl*BW)/(2*pi*(fc^2)*C1n))
%series-resonant branch
C2=(BW/(2*pi*(fc^2)*L2n*Rl))
L2=((L2n*Rl)/(2*pi*BW))
%second parallel-resonant branch
C3=(C3n/(2*pi*BW*Rl))
L3=((Rl*BW)/(2*pi*(fc^2)*C3n))

this yields results:

C1N = 1.086
L2N = 0.8808
C3N = 2.9306

these are the normalized values of the LP prototype filter.

The schematic of the LP prototype is shown in illustration 6:

Illustration 6: The LP prototype circuit with normalized element values.

12 of 25
Radio Frequency Filter Design: Use of discrete elements and stubs

The next step once the prototype LP is finished is to transform the design from illustration 6 into a BP
filter using the method explained in the procedure section.

The resulting circuit is shown in illustration 7

Illustration 7: BP filter prototype with normalized values.

Notice that all that was done to alter the LP prototype was to add an opposite element of the same
value in series with the series elements and in shunt with the shunt elements. This will make our
circuit achieve complete resonance at the center frequency of 800 MHz.
Finally all that is needed is to de-normalize the elements or perform scaling using eq.(3) and eq.(4) for
the shunt branches and eq. (5) and eq. (6) for the series branch.
This produces element values:
C1 = 24 pF
L1 = 1.6487 nH
C2 = 0.2509 pF
L2 = 157.707 nH
C3 = 64.7805 pF
L3 = 610.965 pH

* See Appendix for detailed computations.

13 of 25
Radio Frequency Filter Design: Use of discrete elements and stubs

The resulting circuit is shown in illustration 8. This is the BP filter that should produce the
specifications that we wanted in our design.

Illustration 8: BP filter for 800 MHz fc; 80 MHz 3 dB BW and 200 MHz BW at -23dB attenuation.

Using eq. (9) and eq. (10) for the discrete elements we calculate the stub length; ending in open circuit
for capacitors and short circuit for inductors.

lC1 = 0.2238λ
lC2 = 0.01λ
lC3 = 0.24023λ

lL1 = 0.02641λ
lL2 = 0.239975λ
lL3 = 0.009763λ
Using the values in Pspice we can simulate the filter using stubs or distributed elements, terminated
either in short or open circuit, instead of discrete elements.
*See Appendix for detailed computations.
The consequent circuit is shown in illustration 9.

14 of 25
 Radio Frequency Filter Design: Use of discrete elements and stubs

Illustration 9: BP filter using stubs. The modeling of the stubs is made with transmission lines
terminated either in short circuit (inductors) or open circuit (capacitors).

The next section of the report includes all the simulations performed for the presented circuits.

15 of 25
 Radio Frequency Filter Design: Use of discrete elements and stubs

Simulations

MATLAB frequency response

Utilizing the frequency response transfer function eq.(7) and substituting eq. (8) in eq. (7) we
obtain the transfer function for the BP filter. Utilizing the transfer function we can plot in a
logarithmic scale the bode plot and obtain the filter transfer function. This gives us the level of
attenuation for different frequencies (in rad/s).
The MATLAB code that produces such plot is below:
clc;
clf;
s = tf('s');
wo=2*pi*800e6;
BW=2*pi*80e6;
S=((wo/BW)*((s/wo)+(wo/s)));
H = (1/((S)^3+2*(S)^2+2*(S)+1));
freq_resp =zpk(H)
w=logspace(8.3,11,10001);
bodemag(freq_resp,w)
bode(freq_resp,w)

Illustration 10: BP filter utilizing transfer function method in MATLAB. Notice that MATLAB normalizes the
transfer function.

16 of 25
 Radio Frequency Filter Design: Use of discrete elements and stubs

In illustration 10 we can confirm our design specifications. Dividing by 2π resulting in the frequency in
Hz (cycles/s).

To determine the 3 dB BW we substract fH – fL = (5.28 – 4.78)*109/(2π) = 79.577 Mhz. Which is very

close to the desire 80 Mhz bandwidth.
The same is done to determine the bandwidth at 23dB of attenuation.
(5.67-4.46)*109/(2π) = 192.57 Mhz.

Pspice frequency response

Utilizing Pspice the circuit in illustration 8 was simulated and the obtained results are shown
below.

Illustration 11: BP filter frequency response.

17 of 25
 Radio Frequency Filter Design: Use of discrete elements and stubs

A more detailed view can show us the points of interest in the graph.

Illustration 12: Detailed view of the frequency response of the BP filter design.

Now let's analyze the graph and see if it meets our expectations.
The 3 dB BW is (840.974 – 761.141)MHz = 79.833 MHz.

The BW at -23 dB attenuation is (915.689 – 699.040) MHz = 216.65 MHz

the voltage drop should be (Rl/(Rl + Rs)) = 90/(140) = 0.6428 V. In dB 20*log( 0.6428) = -3.8384 dB.
From this point we count – 3dB and obtain the the 3dB BW.

18 of 25
 Radio Frequency Filter Design: Use of discrete elements and stubs

We now simulate the design made using stubs shown in illustration 9.

The transfer function of the stub design:

Illustration 13: The stub design frequency response is very different to the discrete component response.
Nevertheless it shows the lowest level of attenuation around 800Mhz.
A more detailed view of the graph can illustrate what is happening around our center frequency.

Illustration 14: Detailed view of the distributed elements frequency response. The filter seems to be highly
selective. The level of attenuation is reached far more rapidly.

19 of 25
Radio Frequency Filter Design: Use of discrete elements and stubs

Analyzing the frequency response of the distributed element BP filter we determine the 3dB BW to be:
(802.693 – 796.891) MHz = 5.8 Mhz
the BW at -23dB is (808.15 – 786.053) MHz = 22.116 MHz

*The Appendix contains the simulation printouts for an easier reading.

20 of 25
Radio Frequency Filter Design: Use of discrete elements and stubs

Analysis of Results

Analysis of the MATLAB bode plot

The MATLAB bode plot showed very similar values to the expected results. A margin of error of
3.71% in the 200 MHz BW is an acceptable range, given that we are working with finite values.

Analysis of the Pspice simulations: discrete elements

The Pspice simulation utilizing discrete elements was a partial success since all of our
calculations were proven to be similar in the simulation except the attenuation at 200 MHz. The
attenuation at 200MHz was a little less than expected. In these results a decree of error is expected
since the prototype tables are perfect and we performed interpolation to obtain our normalized values
along with utilizing at most 5 significant figures in the element values. All this adds up and we obtain a
not so perfect design.

Analysis of the Pspice simulations: distributed elements

Utilizing stubs in the circuit instead of discrete elements, produced a filter that was much more
selective than the original one. The 3dB BW is reduced by more than 13 times the original.
This type of filter achieves the desired level of attenuation very rapidly, and has a narrow passband.
The filter almost resembles a peak filter, where only one desired frequency is desired. This type of
filter meets the requirements, but I might also attenuate a range of frequencies that we do not wish to
attenuate. The most obvious explanation to this is that the prototype used to developed the BP filter is
not supposed to be used to develop distributed element filters.

An explanation to the above results is found in the RF Circuit Design book: RF Circuit Design:
Theory and Applications by Reinhold Ludwig & Pavel Bretchko (ISBN 0-13-095323-7)
The books states that filter designs beyond 500 MHz are difficult to realize with discrete components
because the wavelength becomes comparable with the physical filter element dimension, resulting in
various losses severely degrading the circuit performance. When utilizing lumped component filters
Richard's transformations are used and Kuroda's identities.
Since we did not performed any of the above mathematics, this may have made our filter design
perform not as expected, yet it provides us with a simple and effective way to realize such filters.

21 of 25
Radio Frequency Filter Design: Use of discrete elements and stubs

Conclusions

This Filter design project introduces us to designing effective, yet demanding filters in a very
simple manner. We can meet the required specifications as long as our carrier frequency or operating
frequency range does not exceed 500 MHz. Our design exceeded above 500 MHz, at 800 MHz, for
which we did not meet the requirements, yet the error was not so big as to dismiss the filter. We
learned that there are other more precise techniques to utilizing stubs, than just simply finding their
electrical length. Since MATLAB does not consider any type of elements, only the ideal transfer
function, this was the closest simulation that met our requirements. In this report we explained how
one can develop high performance Butterworth filters utilizing a catalog of normalized tables that yield
very good results. RF filters and found in almost any kind of electronic device, be it analog or digital.
To be able to design one that meets your requirements is an integral part of any electrical engineer.

22 of 25
Radio Frequency Filter Design: Use of discrete elements and stubs

Comments

My experience developing this design was very satisfactory, since we implemented the
theoretical process discussed in class. The design process is pretty straightforward and easy to realize.
A very important argument is that above UHF (f > 500MHz), distributed elements filters are the
preferred choice, since discrete elements will not provide the user with an exact replica of what is
intended. Our design was affected by this constraint yet it was borderline with acceptable.
Please see Appendix for detailed calculations and more detailed simulation graphics.

23 of 25
Radio Frequency Filter Design: Use of discrete elements and stubs

Bibliography

1. Reinhold Ludwig (2000) RF Circuit Design and Applications. Upper

Saddle River, NJ: Prentice Hall.

2. www.wikipedia.org/Distributed_element_filter.htm

24 of 25
Radio Frequency Filter Design: Use of discrete elements and stubs

Appendix

This section contains the mathematical procedure done by hand and more detail graphical
representation of the simulations.

25 of 25