EECS 142 Laboratory #2

Ladder Filter Design, Fabrication, and Measurement

A. M. Niknejad
Berkeley Wireless Research Center
University of California, Berkeley 2108 Allston Way, Suite 200
Berkeley, CA 94704-1302
September 14, 2010

1
 Z

L C
+ +
vs R vo
− −

Figure 1: A series RLC circuit.

1 Introduction
In this laboratory you will design, simulate, and build LC low-pass and band-pass filters. You
will characterize your fabricated filters using a Network Anayzer and an oscilloscope by mea-
suring the two-port parameters of the filter. The fabricated filter will be compared against
your simulations. The non-ideal effects of the components, such as loss and self-resonance,
as well as the board parasitic inductance, capacitance, and delay, will be highlighted in the
measurements.

1.1 Resonance
Because RLC circuits and resonance play such a crucial in filters, we begin with a review
of this topic. These circuits are simple enough to allow full analysis, and yet rich enough to
form the basis for most of the circuits we will study.

Series RLC Circuits

The RLC circuit shown in Fig. 1 is deceptively simple. The impedance seen by the source
is simply given by
 
1 1
Z = jωL + + R = R + jωL 1 − 2 (1)
jωC ω LC
1
The impedance is purely real at at the resonant frequency when =(Z) = 0, or ω = ± √LC .
At resonance the impedance takes on a minimal value. It’s worthwhile to investigate the
cause of resonance, or the cancellation of the reactive components due to the inductor and
capacitor. Since the inductor and capacitor voltages are always 180◦ out of phase, and one
reactance is dropping while the other is increasing, there is clearly always a frequency when
1
the magnitudes are equal. Thus resonance occurs when ωL = ωC . A phasor diagram, shown
in Fig. 2, shows this in detail.
So what’s the magic about this circuit? The first observation is that at resonance, the
voltage across the reactances can be larger, in fact much larger, than the voltage across the

2
 vL vL
vL vs
vR vR vR
vs
vs vC
vC
vC
ω < ω0 ω = ω0 ω > ω0
Figure 2: The phasor diagram of voltages in the series RLC circuit (a) below resonance, (b)
at resonance, and (c) beyond resonance.

resistors R. In other words, this circuit has voltage gain. Of course it does not have power
gain, for it is a passive circuit. The voltage across the inductor is given by
vs vs
vL = jω0 Li = jω0 L = jω0 L = jQ × vs (2)
Z(jω0 ) R
where we have defined a circuit Q factor at resonance as
ω0 L
Q= (3)
R
It’s easy to show that the same voltage multiplication occurs across the capacitor
1 1 vs 1 vs
vC = i= = = −jQ × vs (4)
jω0 C jω0 C Z(jω0 ) jω0 C R
This voltage multiplication property is the key feature of the circuit that allows it to be used
as an impedance transformer.
It’s important to distinguish this Q factor from the intrinsic Q of the inductor and
capacitor. For now, we assume the inductor and capacitor are ideal. We can re-write the Q
factor in several equivalent forms owing to the equality of the reactances at resonance
√ r
ω0 L 1 1 LC 1 L1 Z0
Q= = = = = (5)
R ω0 C R C R CR R
q
where we have defined the Z0 = CL as the characteristic impedance of the circuit.

1.1.1 Circuit Transfer Function

Let’s now examine the transfer function of the circuit
vo R
H(jω) = = 1 (6)
vs jωL + jωC +R

3
 jωRC
H(jω) = (7)
1− ω 2 LC + jωRC
Obviously, the circuit cannot conduct DC current, so there is a zero in the transfer function.
The denominator is a quadratic polynomial. It’s worthwhile to put it into a standard form
that quickly reveals important circuit parameters

jω R
L
H(jω) = 1
(8)
LC
+ (jω)2 + jω R
L

Using the definition of Q and ω0 for the circuit

jω ωQ0
H(jω) = (9)
ω02 + (jω)2 + j ωω
Q
0

Factoring the denominator with the assumption that Q > 12 gives us the complex poles of
the circuit r
± ω0 1
s =− ± jω0 1 − (10)
2Q 4Q2
The poles have a constant magnitude equal to the resonant frequency
v , ,!
u
u ω2 1
|s| = t 02 + ω02 1 − = ω0 (11)
4Q 4Q2

A root-locus plot of the poles as a function of Q appears in Fig. 3. As Q → ∞, the poles

move to the imaginary axis. In fact, the real part of the poles is inversely related to the Q
factor.

1.1.2 Circuit Bandwidth

As shown in Fig. 4, when we plot the magnitude of the transfer function, we see that the
selectivity of the circuit is also related inversely to the Q factor. In the limit that Q → ∞,
the circuit is infinitely selective and only allows signals at resonance ω0 to travel to the load.
Note that the peak gain in the circuit is always unity, regardless of Q, since at resonance the
L and C together disappear and effectively all the source voltage appears across the load.
The selectivity of the circuit lends itself well to filter applications. To characterize the
peakiness, let’s compute the frequency when the magnitude squared of the transfer function
drops by half   2
ω ωQ0 1
|H(jω)|2 = 2 = (12)
2

2
(ω02 − ω2) + ω ωQ0
This happens when
2
ω02 − ω 2

=1 (13)
ω0 ω/Q

4
 
 Q→∞

1


Q>

    
2
1
Q<
2
 








Figure 3: The root locus of the poles for a second-order transfer function as a function of Q.
The poles begin on the real axis for Q < 12 and become complex, tracing a semi-circle for
increasing Q.

H(jω0 ) = 1
1

0.8

Δω

0.6

Q=1
0.4

0.2
Q = 10
Q = 100
0.5 1 1.5 2 2.5 3 3.5 4
ω0

Figure 4: The transfer function of a series RLC circuit. The output voltage is taken at the
resistor terminals. Increasing Q leads to a more peaky response.

5
Solving the above equation yields four solutions, corresponding to two positive and two neg-
ative frequencies. The peakiness is characterized by the difference between these frequencies,
or the bandwidth, given by
ω0
∆ω = ω+ − ω− = (14)
Q
which shows that the normalized bandwidth is inversely proportional to the circuit Q
∆ω 1
= (15)
ω0 Q
You can also show that the resonance frequency is the geometric mean frequency of the 3 dB
frequencies
√
ω0 = ω+ ω− (16)

Energy Storage in RLC “Tank”

Let’s compute the ebb and flow of the energy at resonance. To begin, let’s assume that there
is negligible loss in the circuit. The energy across the inductor is given by

wL = 12 Li2 (t) = 12 LIM

2
cos2 ω0 t (17)

Likewise, the energy stored in the capacitor is given by

 Z 2
1 2 1 1
wC = 2 CvC (t) = 2 C i(τ )dτ (18)
C
Performing the integral leads to
2
IM
wC = 1
sin2 ω0 t (19)
2
ω02 C
The total energy stored in the circuit is the sum of these terms
 
1 2 2 1 2 2
ws = wL + wC = 2 IM L cos ω0 t + 2 sin ω0 t = 21 IM L (20)
ω0 C
which is a constant! This means that the reactive stored energy in the circuit does not
change and simply moves between capacitive energy and inductive energy. When the current
is maximum across the inductor, all the energy is in fact stored in the inductor
2
wL,max = ws = 12 IM L (21)

Likewise, the peak energy in the capacitor occurs when the current in the circuit drops to
zero
wC,max = ws = 21 VM2 C (22)
Now let’s re-introduce loss in the circuit. In each cycle, a resistor R will dissipate energy

2 2π
wd = P · T = 12 IM R· (23)
ω0

6
 Y

is L C R io

Figure 5: A parallel RLC circuit.

The ratio of the energy stored to the energy dissipated is thus

1
ws LI 2 ω0 L 1 Q
= 1 2 2 M2π = = (24)
wd I R
2 M ω0
R 2π 2π

This gives us the physical interpretation of the Quality Factor Q as 2π times the ratio of
energy stored per cycle to energy dissipated per cycle in an RLC circuit
ws
Q = 2π (25)
wd
We can now see that if Q  1, then an initial energy in the tank tends to slosh back and
forth for many cycles. In fact, we can see that in roughly Q cycles, the energy of the tank is
depleted.

Parallel RLC Circuits

The parallel RLC circuit shown in Fig. 5 is the dual of the series circuit. By “dual” we mean
that the role of voltage and current are interchanged. Hence the circuit is most naturally
probed with a current source is . In other words, the circuit has current gain as opposed
to voltage gain, and the admittance minimizes at resonance as opposed to the impedance.
Finally, the role of capacitance and inductance are also interchanged. In principle, therefore,
we don’t have to repeat all the detailed calculations we just performed for the series case,
but in practice it’s worthwhile exercise.
The admittance of the circuit is given by
 
1 1
Y = jωC + + G = G + jωC 1 − 2 (26)
jωL ω LC

which has the same form as Eq. 1. The resonant frequency also occurs when =(Y ) = 0, or
1
when ω = ω0 = ± √LC . Likewise, at resonance the admittance takes on a minimal value.
Equivalently, the impedance at resonance is maximum. This property makes the parallel

7
RLC circuit an important element in tuned amplifier loads. It’s also easy to show that at
resonance the circuit has a current gain of Q
is is
iC = jω0 Cvo = jω0 C = jω0 C = jQ × is (27)
Y (jω0 ) G
where we have defined the circuit Q factor at resonance by
ω0 C
Q= (28)
G
in complete analogy with Eq. 3. Likewise, the current gain through the inductor is also
easily derived
iL = −jQ × is (29)
The equivalent expressions for the circuit Q factor are given by the inverse of the relations
of Eq. 5
ω0 C R R R R
Q= = = 1 =q = (30)
G ω0 L √
LC
L L Z 0
C

The phase response of a resonant circuit is also related to the Q factor. For the parallel
RLC circuit the phase of the admittance is given by
1

!
ωC 1 − 2
6 Y (jω) = tan−1 ω LC
(31)
G

The rate of change of phase at resonance is given by


d6 Y (jω)  2Q
= (32)
dω 
ω0 ω0
A plot of the admittance phase as a function of frequency and Q is shown in Fig. 6.

1.1.3 Circuit Transfer Function

Given the duality of the series and parallel RLC circuits, it’s easy to deduce the behavior
of the circuit. Whereas the series RLC circuit acted as a filter and was only sensitive to
voltages near resonance ω0 , likewise the parallel RLC circuit is only sensitive to currents
near resonance
io vo G G
H(jω) = = = 1 (33)
is vo Y (jω) jωC + jωL +G
which can be put into the same canonical form as before
jω ωQ0
H(jω) = (34)
ω02 + (jω)2 + j ωω
Q
0

where we have appropriately re-defined the circuit Q to correspond the parallel RLC circuit.
Notice that the impedance of the circuit takes on the same form
1 1
Z(jω) = = 1 (35)
Y (jω) jωC + jωL +G

8
 100
Q = 100
75 Q = 10

Q=2
50
1
Q= 2
25

-25

-50

-75

0.2 0.5 1 2 5 10

ω/ω0

Figure 6: The phase of a second order admittance as function of frequency. The rate of
change of phase at resonance is proportional to the Q factor.

which can be simplified to

1
j ωω0 GQ
Z(jω) =  2 (36)
1 + jω
ω0
+ j ωω0 Q
At resonance, the real terms in the denominator cancel
R
jQ
Z(jω0 ) =  2 =R (37)
jω0 1
1+ +j Q
ω0
| {z }
=0

It’s not hard to see that this circuit has the same half power bandwidth as the series RLC
circuit, since the denominator has the same functional form
∆ω 1
= (38)
ω0 Q
A plot of this impedance versus frequency has the same form as Fig. 4 multiplied by the
resistance R.
Energy storage in a parallel RLC circuit is completely analogous to the series RLC case
and in fact the general equation relating circuit Q to energy storage and dissipation also
holds in the parallel RLC circuit.

2 The Many Faces of Q

As we have seen, in RLC circuits the most important parameter is the circuit Q and reso-
nance frequency ω0 . Not only do these parameters describe the circuit in a general way, but

9
 LPF HPF BPF BSF APF

Figure 7: Standard filter type include the (a) low-pass filter, (b) high-pass filter, (c) band-
pass filter, (d) band reject filter, and (e) all-pass filter.

they also give us immediate insight into the circuit behavior.

The Q factor can be computed several ways, depending on the application. For instance,
if the circuit is designed as a filter, then the most important Q relation is the half-power
bandwidth
ω0
Q= (39)
∆ω
We shall also find many applications where the phase selectivity of these circuits is of im-
portance. An example is a resonant oscillator where the noise of the system is rejected by
the tank based on the phase selectivity. In an oscillator any “excess phase” in the loop tends
to move the oscillator away from the natural resonant frequency. It is therefore desirable to
maximize the rate of change of phase of the circuit impedance as a function of frequency.
For the parallel RLC circuit we derived the phase of the admittance (Eq. 32) which gives us
another way to interpret and compute Q

ω0 d6 Y (jω)
Q= (40)
2 dω
For applications where the circuit is used as a voltage or current multiplier, the ratio of
reactive voltage (current) to real voltage (current) is most relevant. For a series case we
found
vL vC
Q= = (41)
vR vR
and for the parallel case
iL iC
Q= = (42)
iR iR
The last and one of the most important interpretations of Q is in the definition of energy,
relating the energy storage and losses in a RLC “tank” circuit. We can define the of Q a
circuit at frequency ω as the energy stored in the tank W divided by the rate of energy loss
dW dW
Q = W/ = ωW/ (43)
dφ dt

10
 Z1

Z2

(a) (b) (c)

Figure 8: (a) A simple RC low-pass filter. (b) A simple RC high-pass filter. (c) An arbitrary
filter viewed as a voltage divider.

(a) (b) (c) (d)

Figure 9: (a) An LC low-pass filter. (b) An LC high-pass filter. (c) An LC pass band filter.
(d) An LC band reject filter.

2.1 Filter Frequency Response

Filters are key building blocks in communication systems. Common filters that you are no
doubt familiar with include low-pass filters (LPF), high-pass filters (HPF), and band-pass
filters (BPF). Less common, but equally important, include band-reject filters and all-pass
filters. Common notation and ideal filter magnitude responses for these various filters are
shown in Fig. 7. The all-pass filter may seem useless at first, but is actually quite useful
when we examine the filter’s phase transfer characteristic. The filter can be used to equalize
the phase response of a distorted signal.
A LPF passes the lower frequencies to the output and attenuates the high frequency
components beyond the cut-off frequency. The simplest low-pass filter consist of an RC
circuit shown in Fig. 8a, which attenuates the signal at 20 dB/decade beyond the cutoff, a
rather gentle roll-off. To improve the roll-off, the LC filter shown in Fig. 9a can be used.
Intuitively, at low frequencies the inductor is a short and the capacitor is open, so the signal
is coupled to the output. At resonance the voltage across the inductor/capacitor equal,
which sets the 3-dB frequency. At high frequencies, though, the inductor reactance increases
and the capacitor reactance decreases, decoupling the input from the output and tending to
short the output signals to ground. The transfer function of this filter is given by analyzing
the circuit as a voltage divider (Fig. 8c), where Z1 = ZC ||RL and Z2 = jωL + RS

V2 ZC ||RL
= (44)
V1 ZC ||RL + jωL + RS

11
 passband ripple 3-dB frequency
gain or
insertion
loss
“skirt” roll-off

stopband attenuation

passband frequency

Figure 10: The desired filter response “mask” and the transfer characteristics of a filter.

Assume that R0 = RL = RS
1
= (45)
1 + (1 + jωL/R0 )(1 + jωR0 C)

which shows that the filter has two complex poles, causing a roll-off of 40 dB/dec. The
steepness and bandwidth of the filter near the cutoff frequency is controlled quality factor
Q of the transfer function.
This filter can be converted to a high-pass filter by interchanging the inductor and the
capacitor, which rejects the low frequencies due to the capacitor coupling and the inductor
shunting the signal to ground, and passes the high frequencies, as shown in Fig. 9b. How do
we realize a band-pass filter? If we view the circuit as an arbitrary voltage divider, notice
that in the passband of the filter the impedance Z2 is shorted and the impedance Z1 is
opened whereas the opposite occurs ideally in the stop-band. A short circuit is realized at
an arbitrary frequency using a series LC circuit and an open circuit is realized with a parallel
LC circuit. Putting these ideas together leads to Fig. 9c, a passband filter.
By the same argument, a band reject filter is realized by interchanging the role of the Z1
and Z2 , as shown in Fig. 9d. We see that simple LC resonant circuits are extremely versatile
and form the core of an entire family of filters. As we shall see this powerful insight can be
extended even further.
How do we select the various component values to realize a given filter response? Well,
the answer depends on what you are trying to achieve. A filter is typically characterized
by specifying the filter mask shown in Fig. 10. The mask is characterized by the following
parameters: corner frequency or 3-dB bandwidth, pass-band ripple, which measures how
much the in-band signal gain varies (which leads to distortion), the stop-band rejection, and
the steepness of the “skirt” of the filter (the transition region of the filter). The filter roll-off
is related to the filter order, or the number of poles. Other important metrics include the
group delay (see below), insertion loss of the filter (rather than the ripple), and the stop-band

12
 0 3.0E-9 0 3.0E-9

Mag. Resp. 2.4E-9 Mag. Resp. 2.4E-9

-20 -20

1.6E-9 1.6E-9

-40 -40
8.0E-10 8.0E-10
Group Delay Group Delay

-60 0.0 -60 0.0

0.0 0.5 1.0 1.5 2.0 2.5 3.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0

freq, GHz freq, GHz

(a) (b)

Figure 11: (a) A sharp roll-off filter has more than 40-dB attenuation at 2 GHz and poor
group delay (Chebyshev order 5) whereas (b) a soft roll-off filter has much smaller group
delay (Butterworth order 5).

ripple (only for certain filter types). Some filters have transmission zeros, which cause the
filter response to go to zero at specific frequencies in the stop-band. Insertion loss is related
to the fact that real inductors/capacitors have loss, and some of the input energy is absorbed
by the network and converted to heat.
An ideal filter would delay all frequency components of the signal by the same amount,
i.e. the filter would have a constant group delay
d6 H
τg = − (46)
dω
Any variation in the group delay leads to distortion since different components of the signal
arrive at different times. Notice that an ideal filter should therefore have a linear phase
response or a flat (constant) group delay. To see this, notice that a distortionless filter
should preserve the waveform shape, which means that the output can only be a scaled and
delayed version of the input signal (for the band of interest). The transfer function for such
a filter takes the form
H(s) = |H0 |e−sT (47)
where H0 is the scaling factor and T is the delay of the filter. It is therefore important for the
filter approximate this response in the band of interest, which means minimizing the ripple
in the amplitude response and realizing a linear phase response (or constant group delay).
In general there is a trade-off in the filter attenuation characteristics and the group
delay, which means that more out-of-band magnitude attenuation results in more phase
distortion in-band and vice-versa. In Fig. 11 we compare two filters which differ in their
attenuation rate; notice that the filter with higher attenuation has considerably worse group
delay variation. An all-pass filter can be used to compensate for the phase distortion of a
given filter, or to compensate for the phase distortion of a transmission medium (such as a
cable or device with poor frequency response).
A simple way to observe the distortion caused by the non-constant group delay is to
plot the step response of the filter. Since the step transition has high frequency components

13
 0 200

-20 100

-40 0

-60 -100

-80 -200
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0

freq, GHz freq, GHz

(a) (b)
2.5E-9 600

500
2.0E-9
400
vo, m V

1.5E-9
300
y

1.0E-9
200

5.0E-10 100

0.0 0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 0 2 4 6 8 10 12 14 16 18 20

freq, GHz time, nsec

(c) (d)

Figure 12: (a) The magnitude, (b) phase response, (c) group delay and (d) step response of
a filter.

14
 0

-20

dB ( S ( 2 , 1 ) )
dB ( S ( 1 , 1 ) )
-40

-60

-80
1E4 1E5 1E6 1E7 1E8 1E9

freq, Hz

Figure 13: The input reflection coefficient s11 and transfer coefficient s21 for a fifth-order
Chebyshev filter.

which must all arrive at the same instant, any deviation from a linear phase response leads
to distortion in the waveform, as shown in Fig. 12.
Filters are two-port elements and thus a full characterization requires the specification
of four complex parameters. If a filter is realized with only passive elements, then the two-
port is reciprocal and z12 = z21 . Many filters are also symmetric so z11 = z22 . In these
cases the filter is fully characterized by two complex frequency dependent parameters. Most
commonly filters are characterized in terms of their scattering parameters since this is how
the filters are measured. The input reflection coefficient, or s11 , is particulary important
since the energy transferred into the filter is given by
Pin = 1 − |Γ(ω)|2 = 1 − |s11 |2 (48)
which means in the pass-band we desire |s11 | ≈ 0 (no power reflected) whereas in the stop-
band we desire |s11 | ≈ 1, which means that all the incident power is reflected. If the filter
is realized with lossless or low-loss elements, then the input power is actually the power
delivered to the load. A filter characterized in this way is shown in Fig. 13, where the input
reflection s11 and transmission s21 are plotted versus frequency.
The filter transfer characteristics are measured or calculated through s21 . For an ideal
filter |s21 | ≈ 1 in the passband and |s21 | ≈ 0 in the stop-band. For any real filter, there is
some insertion loss due to the inevitable resistance in the components, and the magnitude
of |s21 | in the passband indicates this loss (under matched conditions).

2.2 Ladder Filters

The concept of a voltage divider can be extended, as shown in Fig. 14a, to even realize
higher out-of-band attenuation. The buffer is used to isolate the two filters and so the overall

15
 (a) (b)

Figure 14: (a) Cascading simple filters results in higher order roll-off characteristics. (b)
Cascaded filters without the buffer.

L2 L4 Ln−1

C1 C3 C5 Cn

(a)
L1 L3 Ln−1

C2 C4 Cn

(b)

Figure 15: (a) The canonical ladder low-pass filter of order n (odd). (b) The canonical ladder
filter of order n (even).

iin Z2 Va Z4

+ i2 i4 +
V1 Y1 i1 Y3 i3 Y5 i5 V2
− −

Figure 16: (a) An arbitrary fifth-order ladder filter structure.

16
transfer function is a cascade of the individual transfer functions, doubling the attenuation
from 40 dB/dec to 80 dB/dec, a significant improvement. In actual practice, the same effect
can be realized without the buffer, as shown in Fig. 14b, except now the transfer function
is more complicated but has the same order. All of the filters discussed up to this point
can in fact be extended in this fashion to realize higher order filters. The order of the
filter correspond to the number of “rungs” in the ladder filter, redrawn in standard form in
Fig. 15. This canonical ladder filter structure can be converted from low-pass to high-pass,
band-pass, or band-stop by the following simple transformations:

• LP → HP: L → C, C → L

• LP → BP: L → series LC, C → parallel LC

• LP → BS: L → parallel LC, C → series LC

The two-port parameters of an arbitrary ladder filter shown in Fig. 16 can be calculated
by noting that the input impedance is given by1

1
Z11 = (49)
1
y1 +
1
z2 +
1
y3 +
z4 + · · ·
To calculate the transfer characteristic Z21 , leave the output open-circuited and assume the
voltage V2 = 1V and find the input current step by step [1]. Note that i5 = y5 V2 = y5 and
i4 = i5 so we have
Va = i4 z4 + V2 = 1 + z4 y5 (50)
Repeating this calculation and using i2 = i3 + i4

i3 = y3 Va = y3 + y3 z4 y5 (51)

V1 = (i3 + i4 )z2 + Va = z2 y3 + z2 y3 z4 y5 + z2 y5 + z4 y5 + 1 (52)

The input current is given by I1 = y1 V1 + i2 , which leads to

y21 = y1 z2 y3 + y1 z2 y3 z4 y5 + y1 z2 y5 + y1 z4 y5 + y1 + y3 + y3 z4 y5 + y5 (53)

In practice carrying out the algebra is unnecessary since filters have been studied exten-
sively and canonical filter structures have been tabulated and are widely available [2]. Filter
design tools are also abundant on the web and through specialized software packages (such
as ADS). Nevertheless you are encouraged to play around with a few simple filters to gain
intuition before using the tools.

17
 RS > Ri Ri < RL
L2

L2 L2
C1 C3 C1 2 2 C3

(a) (b)

Figure 17: (a) A three section low-pass filter. (b) The series element is bisected to form two
“L” matching networks.

2.2.1 Impedance Matching

It is now clear that in a standard band-pass filter structure, all the LC component values
are chosen to resonate at the center frequency of the filter response, which constrains the
product of LC. The ratio of these components must be chosen carefully to produce the
desired filter properties. For instance, in order to obtain an impedance match, the input
impedance of the filter should equal the desired load impedance across the passband. If we
view a simple three section low-pass matching network as to front-to-back “L” matching
networks (bisecting the series element as shown in Fig. 17), then we view this as a classic
impedance down transformation by the factor (1 + Q22 )−1 followed by an up transformation
of (1 + Q21 ) resulting in an overall transformation of (at resonance)

1 + Q21
Zin = · RL (54)
1 + Q22

Which shows that the filter can amplify or attenuate the voltage, or in other words change
the impedance seen by the source. This impedance matching property of the filter is useful
if the source and load impedance are different. In many applications, though, these are the
same so ideally Zin = RL at resonance, which means we should choose the filter component
values to satisfy this constraint, or Q1 = Q2 . The overall transfer characteristics of the filter
come down to one free parameter, the quality factor Q.

2.3 Standard Filter Families

Filters have been thoroughly analyzed and classified into families of filters with specific
filter characteristics. These various filters trade-off passband ripple for sharper attenuation
or slightly worse group delay. The terminology behind these filters is widely known and
standardized (although the spelling Cheybchev varies) and the names of the filters derive
from the original mathematicians who studied the functions that underlie these filter transfer
characteristics.
1
In fact, this form if very useful for synthesis of a ladder filter of a given transfer function. Using long
division, a transfer function can be written in continued fraction form, and the element values are readily
calculated.

18
 |H(jω)|
0

-20

-40

N
=
N

3
N=

=5
-60

7
-80

0.5 1 5 10 50 100

f
f0

Figure 18: The voltage transfer characteristics of a Butterworth filter. The roll-off slope
varies with the filter order n.

2.3.1 Butterworth Filters

The simplest of the family of filters, the Butterworth filters are also known as “maximally
flat”, since the transfer function
 
 V2  1
 = p (55)
 V1  1 + (f /f0 )2n
has n zero derivatives at the origin. This means that the filter response remains as flat as
possible in the passband. As evident in Fig. 18, the order of the filter n determines the
stop-band rolloff. To realize such a filter requires exactly n elements in the ladder structure.
It is relatively easy to show that the poles of this filter lie uniformly on a half-circle in the
left-hand plane with radius ω0 .

2.3.2 Chebyshev Filters

The Chebyshev filter has a sharper roll-off compared to the Butterworth filter in the stop-
band, as shown in Fig. 19. The trade-off is that the Chebyshev filter introduces ripple in the
passband. The transfer function for this filter is given by
 
 V2  1
 = p (56)
 V1  1 + 2 Tn2 (f /f0 )
where Tn (x) is a Chebyshev polynomial of order n. This filter is realized with n elements.
The in-band ripple is controlled by adjusting the factor . For a given value of , the in-band

19
 |H(jω)|
0

-20

-40

-60

-80

0.5 1 5 10 50 100

f
f0

Figure 19: The voltage transfer characteristics of a Chebyshev and Butterworth filter. The
Chebyshev has faster roll-off and in-band ripple whereas the Butterworth filter has a flat
in-band response.

ripple is given by
1
ripple in dB = 20 log10 (57)
1 + 2
The Chebyshev polynomial is given by

Tn (x) = cosh(narccosh(x)) (58)

It is not too difficult to show that the poles of this filter are distributed on the left-hand side
of the s-plane along an ellipse[3].

2.3.3 Other Filter Families

In the literature you will encounter other filter families that display various other trade-offs.
For instance, the Bessel filter has a maximally flat group delay, which is ideal for applications
intolerant to phase distortion in the passband. Inverse Chebyshev filters have no ripple in
the passband but have ripple in the stop band instead, as shown in Fig. 20. These filters have
both poles and zeros in their transfer characteristics. Elliptical filters allow one to specify
passband and stopband ripple and achieve very good stop-band attenuation, but require
high-Q poles. For all filter families it’s important to remember that they are all realized by
using the ladder filter structure. Only the components values vary to change the transfer
function from one filter type to another.

20
 |H(jω)|
0

-10

-20

-30

-40

-50

0.2 0.5 1 2 5 10

f
f0

Figure 20: The voltage transfer characteristics of a type-II or Inverse Chebyshev filter. The
Inverse Chebyshev has good out-of-band rejection.

2.4 Filter Transformations

Most filters families are tabulated as low-pass filters and normalized for a cutoff frequency of
ωc = 1 radian/sec. Two types of filters can be used, one beginning with a shunt capacitor or
series inductor (Fig. 15). The component values are given as gn (Farads/Henrys), assuming
a source impedance of Rs = 1Ω and load impedance RL = 1Ω (odd-order filters). For
even-order filters, RL = 1/gn+1 .

2.4.1 Frequency and impedance scaling

The low-pass filter cutoff frequency can be scaled to ωc and scaled to a source impedance of
R0 by modifying gn in the following way
R0 gn
Ln = (59)
ωc
gn
Cn = (60)
R0 ωc

2.4.2 Low-pass to high-pass transformation

The frequency substitution −ωc /ω → ω 0 converts the filter prototype from low-pass to high-
pass. The new component values are given by
1
Cn0 = (61)
gn

21
 1
L0n = (62)
gn

2.4.3 Band-pass and band-stop transformation

 
ω0 ω ω0
The frequency substitution ω2 −ω1 ω0 − ω → ω 0 is used to tranform the low-pass filter to
a bandpass filter. The fractional bandwidth ∆ of the filter is given by
ω2 − ω1
∆= (63)
ω0
The center frequency is the geometric (not arithmetic) mean of the 3-dB frequencies
√
ω0 = ω1 ω2 (64)

Carrying out the arithmetic means that a series inductor is transformed into a series LC
circuit with component values
Ln R0
L0n = (65)
ω0 ∆
∆
Cn0 = (66)
ω0 Ln R0
and the shunt capacitors are transformed into a parallel LC circuit with components given
by
R0 ∆
L0n = (67)
ω0 Cn
Cn
Cn0 = (68)
ω0 R0 ∆
 −1
The inverse transformation ∆ ωω0 − ωω0 → ω 0 is used to realize a bandstop filter. The
series inductors are converted to parallel LC branches
R0 Ln ∆
L0n = (69)
ω0
1
Cn0 = (70)
ω0 Ln R0 ∆
whereas the shunt capacitors are converted into series LC branches
R0
L0n = (71)
ω0 Cn ∆
Cn ∆
Cn0 = (72)
ω0 R0

22
 Channel
Selectivity

Band

Power [dB]
Selectivity

Freq [Hz]

“In Band” “Out-of-Band”

Signals Interference
Desired Signal

Figure 21: A radio receiver must be able to discriminate between many undesired and
large interfering signals in order to “hear” a weak desired signal. Each arrow respresents a
modulated signal at a given frequency.

2.5 Filters in Communication Systems

In a wireless communication system filters are used to “pick a needle out of a haystack,”
or in other words to select the desired signal of interest in a sea of interfering signals. This
is shown schematically in Fig. 21, where the desired signal is at channel 2, or 910 MHz, in
a radio band that can support many different channels. The first filter in this example is
a fixed filter that selects all the channels of interest while rejecting “out of band” signals,
such as the multitude of in other frequency bands, such as UHF television up to 600 MHz
and other cellular and wireless communication signals in the 2-5 GHz spectrum. The second
filter is used to further isolate the desired channel from the rest of the signals to provide
channel selectivity. A difficult, but typical situation for a receiver is the “near-far” problem,
shown in Fig. 22, where a multitude of interferers appear around the desired signal which
is weak. If the receiver does not have sufficient sensitivity and linearity, these interfering
signals can jam a receiver.
Notice that the second narrow filter needs to have a variable center frequency unless we
use static channel assignments, which is very unrealistic in practice. Since in practice it is
much easier to build a high quality fixed center frequency filter rather than a tunable filter,
the most popular way to realize the same effect is to downconvert the signal of interest to
a fixed intermediate frequency (IF), where channel selection and blocker attenuation can be
performed. This is the basis of the superheterodyne receiver architecture, shown in Fig. 23.
Notice that there are three filters used in this architecture, an RF band select filter, an
image-reject filter, the importance of which will be highlighted later, and the a second IF
filter. In this lab you will design these filters, simulate them, and then build and test the
filters.
In a high-speed communication system using amplitude modulation, the time-domain

23
 IF/A
GC

Mix
er

RF

IF
/A
Sy

LC
PA

G
nt

Ta

C
VC

k n
LC

V
O

CO
RF

Tan
k
Syn
LL

M
PLL L
LLP P
PL

ix
e
th

r
PA
Mixer
Strong Nearby Jammers
PA

IF/AGC
LLP
PLL

VCO
Mixer
LC Tank
LNA
RF Synth

al IF/AGC
gn
Si
k
PLL

ea
VCO

W
nt LC Tank

sta
Di
RF Synth
d
s ire
De

r PA
ixe
M
P
PL
LL
L

O
VC
GC k
/A Ta
n
nth
IF LC
Sy
RF

Figure 22: A radio receiver must contend with a near-far scenario where the desired transmit-
ter is far away resulting in a weak received signal in the presence of many nearby unwanted
interferers.

24
 LNA ADC

VGA

VCO

Figure 23: A superheterodyne radio receiver architecture uses several band-pass, band-reject,
and low-pass filters.

ripple at the amplitude transitions leads to inter-symbol interference (ISI) and “eye closure”.
In Fig. 24 we compare the pulse response of a Bessel and Chebyshev filter. Note that in
both cases we lose the sharp edges due to filtering, but the Bessel filter does not ring, which
prevents “leakage” of one bit onto another.

3 Prelab
The filter designs in this section are probably easiest accomplished using a combination of
the filter-table web tools in the References section and CAD tools. As usual, however, any
combination of math and computation that works is fair game.

1. Design an LC low-pass ladder filter with a bandwidth of 100 MHz, in-band ripple of
1 dB, with 40 dB of attenuation at 200 MHz. Determine the required filter order and
the component values for the filter. The filter should be designed to interface to a
50Ω environment. You may use filter tables, software tools, or internet based tools to
design the filter. Make sure that the filter component values are realizable (in an 0603
footprint) and not unreasonably small (L < 0.5nH or C < .1pF) or large (L > 400nH
or C > 1nF).
2. Simulate your filter in SPICE (ADS or SpectreRF) and verify the performance. Include
plots of the magnitude and phase response on a log-scale. Identify the bandwidth, stop-
band rejection and the in-band ripple.
3. (Optional ) Simulate your filter using “Monte-Carlo” analysis by varying the compo-
nents by 20%. Depending on the simulator, you can do this by specifying each com-

25
 2.0

1.5
vin, V

1.0

0.5

0.0
0 5 10 15 20 25 30 35 40

time, nsec

1.2

1.0

0.8

0.6
v o, V

0.4

0.2

0.0

-0.2
0 5 10 15 20 25 30 35 40

time, nsec

Figure 24: Input pulse data waveform (top) and the filtered response (bottom) shows signif-
icant inter-symbol interference (ISI).

26
 ponent to be a random variable with the given mean value and standard deviation
(assuming Gaussian). The simulator will then run several hundred iterations of the
simulations while varying the component values according to the given statistics. Com-
ment.

4. Simulate the filter response but include the effect of finite component Q,the effect
of board parasitics and use the actual component values available in lab. Assume
that inductors with Q = 20 and capacitors with Q = 100 are available. How does
the component Q impact the filter response? How do the parasitics? How does the
component value mismatch?

5. Design an LC band-pass ladder filter with passband from 630 MHz to 780 MHz, in-band
ripple of 1 dB. Use the minimum filter order possible to realize a stop-band rejection
of 25 dB at 550 MHz and 860 MHz.

6. Simulate your filter in SPICE (ADS or SpectreRF) and verify the performance. Include
plots of the magnitude and phase response on a log-scale. Identify the bandwidth, stop-
band rejection and the in-band ripple.

7. Simulate your filter using Monte-Carlo analysis by varying the components by 20%.
Comment.

8. Simulate the filter response but include the effect of finite component Q,the effect
of board parasitics and use the actual component values available in lab. Assume
that inductors with Q = 20 and capacitors with Q = 100 are available. How does
the component Q impact the filter response? How do the parasitics? How does the
component value mismatch?

4 Experimental Work
4.1 PCB Manufacturing Information
All circuits will be fabricated using a simple two-layer printed circuit board (PCB). The
PCB consists of a dielectric material, usually FR4 (r = 4.4), with a thickness of 62 mils,
and two layers of Cu metal layer. The metal layers have a thickness of 34 µm. Normally you
would pattern the metal layers to produce your circuit but in the interest of time, the boards
have been prefabricated to take on a standard form. The backside of the board is a solid
ground plane. Connections to ground must travel through a “via” to reach the backside.
For this lab, the layout of the board is shown in Fig. 25. The input and output of the
board have footprints for SMA connectors which allow you to connect SMA cables. The
input and output microstrip transmission lines are interrupted periodically which allows you
to place components in series or in shunt in a ladder filter structure. Landing pads with vias
to ground also appear periodically to allow shunt components to be soldered to ground. To
solder components onto the board, connect the standard 0603 components available in lab2
in series or in shunt.
2
Components are classified according to their footprint size in mils.

27
 vias
SMA connector solder pad
footprint

component
footprint

Figure 25: The printed circuit board layout for a ladder filter structure.

shunt component series component

SMA series component

shunt component

(a) (b)

Figure 26: (a) An unfinished and (b) finished PCB layout.

4.2 In-Lab Procedure

1. Build low-pass LC filter using the provided PCB. The board layout, shown in Fig. 26,
supports a ladder filter structure. Note that each series pad can support two parallel
components. Components can be soldered to ground on either side of the board (the
top or bottom).

2. Using the Network Analyzer, measure the magnitude and phase response of the fil-
ter. Adjust the start-stop frequencies and the resolution to clearly capture the filter
response on a log scale. You should ensure that the input leads of the SMA have been
de-embedded from the measurement by recalling the appropriate saved settings on
the analyzer or by using the port-extension procedure. Save the measured S11 (input
reflection) and S21 (transmission) of the filter as an ’S2P’ file.
Ensure that these measurements are not too different from simulation; you should be
able to use theory and computation to accurately predict the performance of your filter.
The goals for this part of the problem are to have a filter which meets specifications
– which is especially possible at this frequency – and to have models of the parasitics
which explain the differences from the original simulation.

3. Optional Measure the time-domain step response of the filter on an oscilloscope. Save
a plot of the step response.

4. Build the LC band-pass filter using the same procedure as above.

28
 5. Repeat all the measurements on the passband filter. If the measured performance
is very different from what you expected, then you have not properly modeled the
parasitics of the components. Pay close attention to the component parasitics measured
in the first lab or re-characterize the filter circuit board using zero-ohm resistors and
test capacitors and inductors.
As before, the goal of this exercise is to achieve good performance with this filter and
to have an accurate, experimentally verified model of the parasitics that agrees with
the measurements.

6. Optional Measure the time-domain step response of the filter on an oscilloscope. Save
a plot of the step response.

7. Optional Measure the frequency response of the actual components of your design on
the VNA. Test fixtures with series/shunt elements and shorts (via) are provided. These
structures can be used to estimate the component parasitics as well as the board level
parasitics (impedance of the via to ground).

8. Optional : If time permits, build an LC band-pass filter using a board with wide or
narrow traces to see observe performance with an unmatched transmission line.

5 Post Laboratory
1. For both filters, calculate the filter voltage transfer characteristic V2 /V1 and explain
how this is related to S21 .

2. Compute 1 − |S21 |2 and compare it to |S11 |2 as a measure of the “excess” loss in both
filters. Where is the excess loss the greatest?

3. Calculate the group delay for the filters in the passband and up to double the passband
frequency in the lowpass filter and up to half the lower pass frequency and double the
high pass frequency in the band pass filter.

4. Calculate the insertion loss and input impedance of the filters in the pass-band.

5. Compare measurement and simulation results by overlapping the measured S11 and S21
of your filter with the simulations. Explain differences keeping board-level parasitics
and other effects under consideration. Try to get the best match possible between your
models and your measurements.

6. In the design of a communication system, which filter non-ideality would you consider
to be the most important? Explain.

7. Comment on the accuracy of your design. What was necessary to change about your
models and measurements to get good accuracy between them. Make sure that any
parasitics that are used in this explanation have measurements confirming their value.

29
 8. Consider different board layouts employing either wider or narrower traces. Explain
how each layout should affect the overall response of the filter. Optional : Do measure-
ment results confirm your intuition.

9. How much time did you spend on this lab? Any feedback is appreciated.

6 References
I found the following references useful in reviewing material for this lab.There is a good
lecture on filters in EECS 247 [4]. Hagen’s book [5] has a nice and easy chapter on filters. The
website http://www-users.cs.york.ac.uk/~fisher/lcfilter/ uses the design procedure
from Hagen for generating standard ladder filters. The definite reference for filter tables is
the handbook by Matthaie and co-authors [2]. Many other online tools exists for commercial
programs such as MathCAD, Mathematica, and matlab, in addition to stand-alone programs.
Check online and for links in the RFCafe website. You can also use the filter synthesis
package that comes with the ADS program. For background reading I used Wikipedia [6],
which is a valuable and evolving source of information. To learn how to design higher
frequency filters and distributed element filters using the insertion loss method, consult a
book on Microwave Circuits such as Pozar [7]. If you are interested in the deep theory behind
filters, read a book such as Network Theory and Filter Design [1].
In summary
http://www-users.cs.york.ac.uk/~fisher/cgi-bin/lcfilter and
http://www.wa4dsy.net/filter/filterdesign.html
are good sources for ladder filter L and C values and help with design.
Information on how to run statistical simulations in ADS (which you will have to do for
lab 3) is available at
http://edocs.soco.agilent.com/display/ads2009/Using+Statistical+Design

References
[1] V. K. Aatre, Network Theory and Filter Design, 1986, New Age Publishers.

[2] G. Matthaei, L. Young, Microwave Filters, Impedance-Matching Networks, and Cou-

pling Structures, 1980, Artech House.

[3] wikipedia.org, “Chebyshev filter”.

[4] B. Boser,EECS 247 Lecture Notes, University of California, Berkeley.

[5] J. Hagen, Radio-Frequency Electronics: Circuits and Applications, 1996, Cambridge

University Press.

[6] wikipedia.org, “Electronic filter”.

[7] D. Pozar, Microwave Engineering, 1998, Wiley.

30