Chapter 6

Microwave Discrete and

Microstrip Filter Design
PathWave Advanced Design System (ADS)

Background
Microwave filters play an important role in any RF front end for the suppression of out of
band signals. In the lumped and distributed form, they are extensively used for both
commercial and military applications. A filter is a reactive network that passes a desired
band of frequencies while almost stopping all other bands of frequencies. The frequency
that separates the transmission band from the attenuation band is called the cutoff
frequency and denoted as fc. The attenuation of the filter is denoted in decibels or
nepers. A filter in general can have any number of pass bands separated by stop bands.
They are mainly classified into four common types – namely low-pass, high-pass,
bandpass, and band stop filters.

An ideal filter should have zero insertion loss in the pass band, infinite attenuation in the
stop band, and a linear phase response in the pass band. An ideal filter cannot be
realizable as the response of an ideal lowpass or band pass filter is a rectangular pulse
in the frequency domain. The art of filter design necessitates compromises with respect
to cutoff and roll off. There are basically three methods for filter synthesis. They are the
image parameter method, insertion loss method, and numerical synthesis. The image
parameter method is an old and crude method, whereas the numerical method of
synthesis is newer but cumbersome. The insertion loss method of filter design on the
other hand is the optimum and more popular method for higher frequency applications.
The filter design flow for insertion loss method is shown in Figure 1.

Find us at www.keysight.com Page 1

Figure 1. Filter design flow for insertion loss method

Since the characteristics of an ideal filter cannot be obtained, the goal of filter design is to approximate
the ideal requirements within an acceptable tolerance. There are four types of approximations – namely
Butterworth or maximally flat, Chebyshev, Bessel, and Elliptic approximations. For the prototype filters,
maximally flat or Butterworth provides the flattest pass band response for a given filter order. In the
Chebyshev method, sharper cutoff is achieved and the pass band response will have ripples of
amplitude 1+k2. Bessel approximations are based on the Bessel function, which provides sharper
cutoff, and Elliptic approximations results in pass band and stop band ripples. Depending on the
application and the cost, the approximations can be chosen. The optimum filter is the Chebyshev filter
with respect to response and the bill of materials. Filters can be designed both in the lumped and
distributed form using the above approximations.

Find us at www.keysight.com Page 2

Design of Microwave Filters
The first step in the design of microwave filters is to select a suitable approximation of the prototype
model based on the specifications.

Calculate the order of the filter from the necessary roll off as per the given specifications. The order can
be calculated as follows:

Butterworth Approximation:
𝜔′ 2N
LA (ω' ) = 10log10 {1+ε ( ) }
𝜔 𝑐

Where ε = {Antilog10 LA /10}-1 and LA = 3 dB for Butterworth

Chebyshev Approximation:
ω'
LA (ω') = 10log10 {1+εcos2 [ncos-1 ( )]} when ω' ≤ ω1 ' and
ω1 '

ω'
LA (ω') = 10log10 {1+εcosh2 [ncosh-1 ( )]} when ω' ≥ ω1 '
ω1 '

Where ωc is the angular cutoff frequency

ω' is the angular attenuation frequency

LA (ω') is the attenuation at ω'

N is the order of the filter

Where ε = {Antilog10 LAr / 10} - 1 and LAr is ripple in passband

The next step in the filter design is to calculate the prototype values of the filter depending on the type of
approximation. The prototype values for the Chebyshev and Butterworth approximations can be
calculated using the following equations:

Butterworth Approximation:

g0 = 1

gk = 2 sin {(2k - 1)π / 2n} where k = 1, 2, …, n and

gN+1 = 1

Where n is the order of the filter

Find us at www.keysight.com Page 3

Chebyshev Approximation:
LAr
β = ln (coth ) where LAr is ripple in the passband
17.37

β
γ = sinh ( )
2n

(2k - 1) π
ak = sin [ ] , k = 1, 2, 3, …, n
2n

kπ
bk = γ2 +sin2 ( ) , k = 1, 2, 3, …, n
n

2a1
g1 =
γ

4ak-1 ak
gk = , k = 2, 3, …, n
bk-1 gk-1

gn+1 = 1 for n odd

β
= coth2 ( ) for n even
4

After computing the prototype values, the prototype filter must be transformed with respect to frequency
and impedance to meet the specifications. The transformations can be done using the given equations
below.

For Lowpass Filter:

After impedance and frequency scaling:

' Ck
Ck =
R0 ωc

R0 Lk
L'k = where R0 =50 Ω
ωc

For the distributed design, the electrical length is given by:

Length of capacitance section (βlc ) : Ck Zl / R0

Length of inductance section (βll ) : Lk R0 / Zh

Where Zl is the low impedance value and Zh is the high impedance value

For bandpass filter:

After impedance and frequency scaling:

L1 Z0
L'1 =
ω0 Δ

' Δ
C1 =
L1 Z0 ω0

Δ Z0
L'2 =
ω0 C2

Find us at www.keysight.com Page 4

 ' C2
C2 =
Z0 Δ ω0

L3 Z0
L'3 =
ω0 Δ

' Δ
C3 =
L3 Z0 ω0

Where Δ is the fractional bandwidth Δ = (ω2 - ω1 ) / ω0

Calculating the electrical length of the distributed design of the bandpass filter will be discussed later in
this chapter.

Simulation of a Lumped and Distributed Lowpass Filter Using ADS

Typical Design
Cutoff Frequency (fc) : 2 GHz

Attenuation at f = 4 GHz : 30 dB (LA(ω))

Type of approximation : Butterworth

𝜔 2N
Order of the filter : LA (ω' ) = 10log10 {1+ε ( ) }
𝜔𝑐

Where ε = {Antilog10 LA /10} – 1

Substituting the values of LA(ω), ω, and ωc, the value of N is calculated to be 4.

Protype Values of the Lowpass Filter

The prototype values of the filter are calculated using the formulas given earlier:

g0 = 1

gk = 2 sin {(2k - 1)π / 2n} for k = 1, 2, …, n, and

gN+1 = 1

The prototype values for the given specifications of the filter are:

g1 = 0.7654 = C1

g2 = 1.8478 = L2

g3 = 1.8478 = C3

g4 = 0.7654 = L4

Find us at www.keysight.com Page 5

Lumped Model of the Filter
The lumped values of the lowpass filter after frequency and impedance scaling are given by the
formulas given earlier:
' Ck
Ck =
R0 ωc

R0 Lk
L'k = where R0 =50 Ω
ωc

The resulting lumped values are:

'
C1 = 1.218 pF

L'2 = 7.35 nH
'
C3 = 2.94 pF

L'4 = 3.046 nH

Distributed Model of the Filter

For the distributed design, the electrical length is given by:

Length of capacitance section (βlc ) : Ck Zl / R0

Length of inductance section (βll ) : Lk R0 / Zh

Where

Zl is the low impedance value

Zh is the high impedance value

R0 is the source and load impedance

ωc is the desired cutoff frequency

If we consider Zl = 10 Ω and Zh = 100 Ω, then

βlc1 = 0.1530

βli2 = 0.9239

βlc3 = 0.3695

βli4 = 0.3827

Find us at www.keysight.com Page 6

Since β = 2π / λ, the physical lengths are given by

lc1 = 1.680 mm

li2 = 10.145 mm

lc3 = 4.057 mm

li4 = 4.202 mm

Schematic Simulation Steps for Lumped Lowpass Filter

1. Open the Schematic window of ADS.

2. From the Lumped-Components library, select the appropriate components necessary for the lumped
element lowpass filter. As you place components on the schematic, enter the component values that
were calculated earlier.

Figure 2. Parts library for lumped elements

3. Connect the components with wire and terminate both ports of the lowpass filter using
terminations from the Simulation_S-Param library.

4. Place the S-Parameter simulation controller from the Simulation S-Param library and set its
parameters as:

Start = 0.1 GHz

Stop = 5 GHz

Number of Points = 101 (or Step Size = 49 MHz)

The design of the lumped element low pass filter is shown in Figure 3.

Find us at www.keysight.com Page 7

Figure 3. Schematic of the completed low pass filter

5. Simulate the circuit by clicking F7 or the simulation gear icon.

6. After the simulation is complete, ADS automatically opens the Data Display window, which shows the
results. If the Data Display window does not open, click Window > New Data Display. In the Data
Display window, select a rectangular plot. In the popup window, plot S(1,1) and S(2,1) as dB.

7. Click and insert a marker on S(2,1) trace around 2GHz to see the data display graph as shown in
Figure 4. The point at which S(2,1) is at -3 dB is the cutoff frequency. According to our design, this
should be around 2 GHz.

Figure 4. Graph showing the S(1,1) and S(2,1) of the lowpass filter

Find us at www.keysight.com Page 8

Results and Observations
The simulation of the lumped element model shows that the lowpass filter has a cutoff frequency of
about 2 GHz and has a gentle roll off, which is expected for a Butterworth filter.

Layout Simulation Steps for Distributed Low Pass Filter

Calculate the physical parameters of the distributed lowpass filter using the design procedure given
above. Since the length has already been calculated, the only parameter left to calculate is the width of
the Zl and Zh transmission lines for the stepped impedance low pass filter. As a reminder, Zl = 10 Ω and
Zh = 100 Ω, which means the low impedance line width is 24.7 mm and the high impedance line width is
0.66 mm. Both calculations were done with a dielectric constant of 4.6 and a thickness of 1.6 mm. For
simplicity, the low impedance and high impedance line widths are shown below:

wl = wc1 = wc3 = 24.7 mm

wh = wi2 = wi4 = 0.66 mm

Calculate the length and width of the 50 Ω line using Line Calc (Tools > Line Calc > Start Line Calc) in
the schematic window. Change the values for Substrate Parameters, Component Parameters, and
Electrical Characteristics (Z0 and E_Eff) to match Figure 5. Click Synthesize to generate the Physical
Characteristics.

The 50 Ω input and output transmission line characteristics should be:

Width = 2.93 mm

Length = 4.53 mm

Figure 5. Using Line Calc to calculate the parameters for 50 Ω line

Find us at www.keysight.com Page 9

Now that the parameters have all been determined, create a model of the stepped impedance lowpass
filter in the layout window. There are two ways to do this: using library components or by drawing
rectangles.

To create the model using library components, select the TLines-Microstrip library, and use the MLIN
components. The TLines-Microstrip library is shown in Figure 6. To create the model by drawing
rectangles, go to Insert > Rectangle.

Figure 6. TLines-Microstrip library

Ensure that each of the different sections are connected. Connect Pin 1 to the input and Pin 2 to the
output. The completed model of the stepped impedance (distributed) lowpass filter is shown in Figure 7.

Figure 7. Layout of the stepped impedance lowpass filter

Find us at www.keysight.com Page 10

Setup the EM simulation as described earlier in the EM simulation chapter. The following properties will
be used to define the substrate:

Material : FR_4

Er : 4.6

Height : 1.6 mm

Loss Tangent : 0.0023

Metal Thickness : 0.035 mm

Metal Conductivity : Cu (5.8E7 S/m)

The EM Setup window, showing the substrate definition, is shown in Figure 8. Make sure to enable
Edge Mesh (Options > Mesh > Edge Mesh) and select ‘Auto-determine edge width’.

Figure 8. EM Setup window showing the substrate definition

After setting up the simulation, press Simulate. As with earlier, create a rectangular plot showing S(1,1)
and S(2,1). This plot is shown in Figure 9.

Find us at www.keysight.com Page 11

Figure 9. Graph showing the S(1,1) and S(2,1) of the distributed lowpass filter

Notice that the 3 dB cutoff has shifted from 2 GHz to 1.68 GHz, as the theoretical calculations don’t
allow for accurate analysis of open-end effects and a sudden change in the impedance of the
transmission lines. In order to get closer to the 2 GHz cutoff frequency specification, the lines need to be
optimized. This optimization can be carried out using the Momentum simulator or by performing a
parametric sweep on the lengths of the capacitive and inductive lines.

Parametric EM Simulations in ADS

To begin parametric simulation on the layout, we need to define the variable parameters that will be
associated with the layout components. Click EM > Component > Parameters, as shown in Figure 10.

Find us at www.keysight.com Page 12

Figure 10. How to open parameter variable definitions

In the Design Parameters window, define variables for the length of capacitive and inductive lines. For
the default value, enter the lengths that were calculated earlier, and make sure to select the appropriate
unit. Set Type to Subnetwork, as these parameters will be associated with the microstrip library
components, which have parameterized artwork. If you were trying to parameterize the
polygon/rectangle-based components, you would select the Nominal/Perturbed method, which
requires additional information.

Figure 11. Design parameter definitions

After defining the parameters, double-click the respective components in the layout window and change
the length values to the corresponding variable names. Note that no units need to be defined here, as
the units are specified in the parameter definition. An example is in Figure 12.

Find us at www.keysight.com Page 13

Figure 12. Redefining the length parameters

After defining the parameter values for each transmission line, the next step is to create an EM model
and symbol that can then be used for parametric EM cosimulation. To create a parametric model and
symbol for the layout, go to EM > Component > Create EM Model and Symbol. Select both options in
the window that pops up.

Figure 13. Creating an EM model and symbol

Find us at www.keysight.com Page 14

After the EM model and symbol have been created, they will appear below the layout cell, as shown in
Figure 14.

Figure 14. EM model and symbol shown in the layout

Open a new schematic cell and drag and drop the emModel component to place it as a subcircuit. The
parameters defined in the layout view (from Figure 11) are the default values for the emModel component.
These values will be swept using the Parameter Sweep component in the ADS schematic, but the
Parameter Sweep is done on variables contained in a VAR block. Therefore, you will need to redefine the
default values in the VAR block and change the parameters for the emModel to use the variables in the
VAR block (i.e. set LC1=L1, which is defined the VAR block). In this case, the variables L1-L4 are defined
in the VAR block and assigned to the emModel component. This is shown in Figure 15.

For our initial sweep, add a Parameter Sweep component (from the Simulation-S_Param library). Open
the Parameter Sweep component and define Parameter to Sweep as L2 (first inductive line). Set the
sweep to go from 6.145 to 12.145 in steps of 1. Under the Simulations tab, define Simulation 1 as SP1.

Find us at www.keysight.com Page 15

Figure 15. Schematic for the Parameter Sweep

Click the Simulate button and plot S(1,1) and S(2,1) in the Data Display window. This shows how the
filter response changes with the length of the first inductive line. Add a line marker at 2 GHz to see the
S-parameters at this point. This is shown in Figure 16.

Figure 16. Plot showing the effect of the first inductive line length on the S -Parameters

Find us at www.keysight.com Page 16

From the plot, we can see that S(2,1) provides a 3 dB cutoff at 2 GHz for the first sweep value.
Therefore, 6.145 mm seems to be the best value for L2.

In the schematic, disable the Parameter Sweep, and set L2 = 6.145. Perform the simulation again to
see the filter response. The result is shown in Figure 17.

Figure 17. Plot showing the S-Parameters with the first inductive line length redefined

Results and Observations

Using theoretical values, the cutoff frequency for the distributed (stepped impedance) lowpass filter was
about 1.68 GHz, far below the specification of 2 GHz. Further optimization was needed to meet the
specification. By performing a parametric sweep, the first inductive line length was optimized. This
optimization showed significant impact to the 3 dB cutoff frequency by reducing the length of the first
inductive line. This raised the cutoff frequency to about 2 GHz. As this met the specification, no further
optimization was needed.

Simulation of a Lumped and Distributed Bandpass Filter Using ADS

Typical Design
Upper Cutoff Frequency (fc1) : 1.9 GHz

Lower Cutoff Frequency (fc2) : 2.1 GHz

Ripple in Passband : 0.5 dB

Order of the Filter :3

Type of Approximation : Chebyshev

Find us at www.keysight.com Page 17

Prototype Values of the Filter
The prototype values of the filter for a Chebyshev approximation are calculated using the formulas:
LAr
β = ln (coth ) where LAr is ripple in the passband
17.37

β
γ = sinh ( )
2n

(2k - 1) π
ak = sin [ ] , k = 1, 2, 3, …, n
2n

kπ
bk = γ2 +sin2 ( ) , k = 1, 2, 3, …, n
n

2a1
g1 =
γ

4ak-1 ak
gk = , k = 2, 3, …, n
bk-1 gk-1

gn+1 = 1 for n odd

β
= coth2 ( ) for n even
4

The prototype values for the given specifications of the filter are:

g1 = 1.5963 = C1

g2 = 1.0967 = L2

g3 = 1.5963 = C3

g4 = 1.0000

Lumped Model of the Filter

The lumped values of the bandpass filter after frequency and impedance scaling are given by:
L1 Z0
L'1 =
ω0 Δ

' Δ
C1 =
L1 Z0 ω0

Δ Z0
L'2 =
ω0 C2

' C2
C2 =
Z0 Δ ω0

L3 Z0
L'3 =
ω0 Δ

' Δ
C3 =
L3 Z0 ω0

Find us at www.keysight.com Page 18

 Where

Δ is the fractional bandwidth Δ = (ω2 - ω1 ) / ω0

Z0 = 50 Ω

The resulting lumped element values are calculated to be:

L'1 = 63 nH
'
C1 = 0.1004 pF

L'2 = 0.365 nH
'
C2 = 17.34 pF

L'3 = 63 nH
'
C3 = 0.1004 pF

The geometry of the lumped element bandpass filter is shown in Figure 18.

Figure 18. Geometry of lumped element bandpass filter

Distributed Model of the Filter

Calculate the value of J from the prototype values:
πΔ
Z0 J1 = √
2g1

πΔ
Z0 Jn = for n = 2, 3, …, N
2√gn-1 gn

πΔ
Z0 JN+1 = √
2gN gN+1

Find us at www.keysight.com Page 19

 Where

Δ is the fractional bandwidth Δ = (ω2 - ω1 ) / ω0

Z0 = 50 Ω

The values of odd and even mode impedances can be calculated as follows

Z0e = Z0 [1 + JZ0 + (JZ0 )2 ]

Z0o = Z0 [1 – JZ0 + (JZ0 )2 ]

Schematic Simulation Steps for the Lumped Bandpass Filter

Open a new schematic window and construct the lumped element bandpass filter, as shown in Figure
19. These values were calculated earlier and is the same geometry as the circuit in Figure 18. Setup the
S-Paramater simulation from 1.0 GHz to 3.0 GHz with steps of 5 MHz.

Figure 19. Lumped element bandpass filter schematic

Find us at www.keysight.com Page 20

Click the Simulate button and plot S(1,1) and S(2,1) in the data display window. Add markers for the
lower and upper cutoff frequencies. The results are shown in Figure 20.

Figure 20. Lumped element bandpass filter S-Parameters

Results and Observations

The results are satisfactory for the lumped element design. The simulation shows that the design is in
agreement with the specified lower and upper cutoff frequencies. The filter also has an appropriate roll
off for a Chebyshev filter. However, to get more accurate results, the circuit should be simulated and
optimized with the vendor component libraries. It may also be helpful to perform a yield analysis to
determine how the results are impacted by the tolerance of the lumped components.

Layout Simulation Steps for the Distributed Bandpass Filter

Calculate the even mode and odd mode impedance values (Z0e and Z0o) of the bandpass filter using the
design procedure given above. Synthesize the physical parameters (length, width, and spacing) for the
coupled lines for a substrate thickness of 1.6 mm and dielectric constant of 4.6. This can be done using
LineCalc. Make sure to change Component Type to MCLIN. This is shown in Figure 21.

Find us at www.keysight.com Page 21

The physical parameters of the coupled lines for the given values of Z0o and Z0e are shown below:

Substrate Parameters

Thickness: 1.6 mm

Dielectric constant: 4.6

Frequency: 2 GHz

Electrical length: 90 degrees

Section 1: Z0e = 70.61 Ω, Z0o = 39.24 Ω

Width = 2.290 mm

Spacing = 0.491 mm

Length = 20.723 mm

Section 2: Z0e = 56.64 Ω, Z0o = 44.77 Ω

Width = 2.803 mm

Spacing = 1.804 mm

Length = 20.255 mm

Section 3: Z0e = 56.64 Ω, Z0o = 44.77 Ω

Width = 2.803 mm

Spacing = 1.804 mm

Length = 20.255 mm

Section 4: Z0e = 70.61 Ω, Z0o = 39.24 Ω

Width = 2.290 mm

Spacing = 0.491 mm

Length = 20.723 mm

Find us at www.keysight.com Page 22

Figure 21. Using LineCalc to determine parameters

Calculate the width of the 50 Ω line using LineCalc. Make sure to reset Component Type to MLIN. The
length of the line is chosen to be 5 mm for simplicity. The parameters for the 50 Ω line should be:

Width = 2.918 mm

Length = 5 mm

Now that the parameters have been determined, create a model of the distributed bandpass filter in a
new layout window. The easiest way to do this is to use the library components. Select the TLines-
Microstrip library. Use the MCFIL components for the coupled line sections and the MLIN components
for the 50 Ω lines. Connect Pin 1 to the input and Pin 2 to the output. The finished layout is shown in
Figure 22.

Figure 22. Layout of the distributed bandpass filter

Find us at www.keysight.com Page 23

Setup the EM simulation using the procedure defined earlier for 1.6 mm FR4 dielectric. Define the
frequency plan as 1 GHz to 3 GHz with 101 number of points. Don’t forget to turn on Edge Mesh under
the Options > Mesh tab of the EM Setup window.

Once the simulation finishes, plot S(1,1) and S(2,1). Add markers to the -3dB bandwidth points to
determine the lower and upper cutoff frequencies. The plot is shown in Figure 23. Note that the markers
are not at -3 dB. That is because the passband ripple varies between approximately -1.0 dB and
-1.5 dB, which will slightly impact the definition of the lower and upper cutoff frequencies.

Figure 23. Plot of S-Parameters for distributed bandpass filter

Results and Observations

For the distributed design, the filter does not meet the specifications and the coupled lines need to be
optimized (length, width, spacing). The odd and even impedance values were calculated from standard
formulas using ideal assumptions and do not consider any electromagnetic or parasitic effects. The
optimization can be done using a variety of tools inside ADS.

Find us at www.keysight.com Page 24

 Conclusion
Congratulations! You have completed Microwave Discrete and Microstrip Filter Design. Check out more
examples at www.keysight.com/find/eesof-ads-rfmw-examples.

Learn more at: www.keysight.com

For more information on Keysight Technologies’ products, applications or services,
please contact your local Keysight office. The complete list is available at:
www.keysight.com/find/contactus

Find us at www.keysight.com Page 25

This information is subject to change without notice. © Keysight Technologies, 2016 - 2021, Published in USA, March 10, 2021, 5992-1624EN