EE133 – Winter 2002

Cookbook Filter Guide

Welcome to the Cookbook Filter Guide!
Don’t have enough time to spice out that perfect filter before Aunt Thelma comes down for dinner?
Well this handout is for you! The following pages detail a fast set of steps towards the design and
creation of passive filters for practical use in communications and signal processing. Amaze your
friends and neighbors as you do twice the amount of work in half the time!

Here’s a brief overview of the steps:

® Specify your filter type

® Implement a low-pass version of your filter
® Transform it to what you really want (high-pass, bandpass, bandstop)
® Simulate and Iterate

® Step 1: What Filter Do I Want?

This is where you have to do most of your thinking. It’s no good to cook up a nice steak dinner
if Auntie Thelma turned vegetarian last year. Here are the specifications that are most often
quoted with filters:

Filter Type (Low-Pass, High-Pass, Band-Pass, Band-Stop)

Center frequency (rad/s or Hz)
Bandwidth (rad/s or Hz)
Cut-off/Roll-off rate (dB)
Minimum Attenuation Required in Stopband
Input Impedance (ohms)
Output Impedance (ohms)
Overshoot in Step Response
Ringing in Step Response

We will be creating filters through a method known as the insertion loss method. Insertion Loss,
or the Power Loss Ratio PLR.is defined as:
PowerSource 1
PLR = =
PowerDelivered 1− Γ(ω )
2

Γ is our famliar reflection coefficient. It turns out that this is expressible in the following form of :
M(ω 2 )
PLR =1 +
N(ω 2)
Where M and N are two real polynomials. Which simply means that we can define an arbitrary filter
response and use this formula to match it to real components, thus allowing us to make it physically
realizable. There are several standard filter responses, each with their own advanatages and drawbacks

Standard Filter Responses

• Butterworth (AKA Maximally Flat or Binomial) Filters
Butterworth filters are general purpose filters. Another common name for them is
a maximally flat filter, which refers to the relatively flat magnitude response in the pass-
band. Attenuation is –3dB at the design cutoff frequency with a –20dB/decade roll off per
pole above the cutoff frequency.
1
EE133 – Winter 2002
Cookbook Filter Guide

• Chebyshev (AKA equal ripple magnitude) Filter

Chebyshev filters have a steeper attenuation above the cutoff frequency, but at the
expense of amplitude “ripples” in the pass-band. For a given number of poles, a steeper
cutoff can be achieved by allowing more pass-band ripple. The cutoff frequency is
defined at the point at which the response falls below the ripple band of the pass-band.

• Bessel (AKA Linear Phase or Maximally Flat Time Delay) Filters

Sometimes a design requires a filter to have a linear phase in order to avoid signal
distortion. In general a good phase output (i.e. linear) always comes at the expense of a
good magnitude response (i.e. fast attenuation).

• Elliptic Filters
For the previous filters, as the frequency gets progressively further from the
center frequency the attenuation increases. Sometimes a design only requires a minimum
attenuation in the stop-band. This relaxes some constraints on the response which allow
a better cutoff rate. However, this filter has ripples in both the passband and stop-band.

® Step 2: Prototyping a Low-Pass Design

Having finally specified your filter, it’s time to prototype a low-pass version of your filter.
Although it seems counterintuitive to spend time on a filter that doesn’t even necessarily have the pass-
band/stop-band characteristics of your desired filter, it will become apparent that there is relation
between the values derived for a low-pass situation and the other filter types.

As an example we will design a low-pass filter for a source impedance of 50 ohm, a cut-off frequency of
1MHz and which requires a minimum attenuation of 40dB at 10MHz.

Determine the type of filter and N, the order of the filter:

First we have to determine which of the filter types we want to use. Do we care about
having linear phase? Or is maxmium cut-off attenuation the critical factor? Once that’s
been done, we can then determine the order of the filter necessary to fit the required
attenuation spec. Usually, we refer to a graph like the one below showing the attenuation
characteristics for various N versus normalized frequency for a particular filter type.

In this case, we decide that a flat ma gnitude response in the most appropriate. So then we
look at the graph.The definition of normalized frequency is:
w
f norm = −1
wc
we find that wc=1MHz wo =10MHz which means the normalized frequency is 9. Looking
on the graph we see that N=2 will easily satisfy our requirements for 40dB of attenuation.

2
EE133 – Winter 2002
Cookbook Filter Guide

Design a Normalized Low-Pass Filter using a Table:

Once that is done, we can now design a second order prototype filter for a source
impedance of 1 ohm, a cut-off frequency of 1 rad/sec. As shown in the figure below, we use one
of two equivalent ladder circuits. Note the way the element values are numbered, with g0 at the
generator to gN+1 at the load.

How to read this chart:

g o = generator resistance
or a
generator conductance

gk=
inductance for series inductors
or a
capacitance for shunt capacitors

g N+1 =
load resistance if g N is a shunt C
or a
load conductance if g N is a series L

A key point is that the components

alternate between shunt and series.
Note that during out prototyping,
inductors are always series, capacitors
are always shunt. The only difference
is whether or not the first element is
series or shunt.

3
EE133 – Winter 2002
Cookbook Filter Guide

To design a filter of a particular response (i.e. Butterworth, or Bessel) there is a unique ratio of
components to be used. These ratios are usually kept is handy tables like the one below.

Element Values for Butterworth (Maximally Flat) Low-Pass Filter Prototypes

(go =1,wc=1,N=1 to 10)

N g1 g2 g4 g5 g6 g7 g8 g9 g10 g11
g3
1 2.0000 1.0000
2 1.4142 1.4142 1.0000
3 1.0000 2.0000 1.0000 1.0000
4 0.7654 1.8478 1.8478 0.7654 1.0000
5 0.6180 1.6180 2.0000 1.6180 0.6180 1.0000
6 0.5176 1.4142 1.9318 1.9318 1.4142 0.5176 1.0000
7 0.4450 1.2470 1.8019 2.0000 1.8019 1.2470 0.4450 1.0000
8 0.3902 1.1111 1.6629 1.9615 1.9615 1.6629 1.1111 0.3902 1.0000
9 0.3473 1.0000 1.5321 1.8794 2.0000 1.8794 1.5321 1.0000 0.3473 1.0000
10 0.3129 0.9080 1.4142 1.7820 1.9754 1.9754 1.7820 1.4142 0.9080 0.3129 1.0000

For a second order, we see that g1 and g2 must both equal 1.4142 and g3 , the load, must equal the go , the
source resistance. So the following circuit will get us a low pass filter centered about 1 rad/sec.

® Step 3: Transforming the Low-Pass Prototype to the Desired Filter

All we have to do now is transform the prototype into something we can really use. The tables assume a
set impedance Rs=Rl=1 ohm (except for Chebyshev filters, where the source and load resistance differ
by a set amount). In addition, the filter’s cutoff frequency is 1 rad/sec, a fairly useless value in general.
So we need to scale for impedance and frequency.

Let’s start with impedance scaling. To take this into account, we need to multiply the impedances of the
prototype by the desired Rs. Since the prototype input impedance is always 1 ohm, we can simply
multiply by the ratio. Rl naturally needs to scale proportionally, while all capacitors need to be divided
by Rs and all inductors need to be multiplied by Rs.

Now that we have these original values, we need to swap components in order to obtain the correct filter
type. Below are the substitutions.

4
EE133 – Winter 2002
Cookbook Filter Guide

w1 − w2
∆= is your friend
w0
∆ is a term to describe the sharpness of the filter. If you look at it really hard, you’ll see it’s actually the
inverse of Q. In filter design the center frequency is not chosen as the arithmetic mean, but the
geometric mean. In other words:
w1 + w2
wo ≠
2
wo = w1 w2

This allows for the math to work out such that straight substitutions are allowable. What this means is
that often times, you’ll pick your bandwidth and just go for a center frequency in that area. Of course,
you could work out the math to get exactly the center frequency you want, but sometimes, you really
don’t want to think that hard about a minor filter.

With that all said here are the transformations. The formulas on the next page were derived by
comparing the frequency dependent impedances and mapping the cutoff frequency to different
situations. In each one, the Lk or Ck from your prototype is coverted into a new Lk ’ or Ck ’. A derivation
of these formulas can by found in Posner’s Microwave Transmission. NOTE: These formulas
ALREADY take into account the scaling for impedances.

5
EE133 – Winter 2002
Cookbook Filter Guide
TRANSFORMATION TABLE

Low-Pass

1
L' = RL
k w s k
c
Simple impedance and frequency scaling.
1 Ck
C' =
k w R
c s
High-Pass

1 1
C' =
k w R L
c s k
Note the swapping of capacitors and inductors:
1 Rs
L' =
k w C
c k
Bandpass

1 Rs Lk
L'Lk =
wo ∆
Series inductor Lk transforms to a series LC circuit defined by:
1 ∆
C'Lk =
wo RsL k

1 Rs ∆
L'Ck =
wo Ck
Shunt capacitor Ck transforms to a parallel LC circuit with element values:
1 Ck
C'Ck =
wo Rs∆
Bandstop

1
L'Lk = ∆R L
wo s k
Series inductor Lk transforms to a parallel LC circuit defined by:
1 1
C'Lk =
wo ∆Rs L k
1 Rs
L'Ck =
wo ∆Ck
Shunt capacitor Ck transforms to a series LC circuit with element values:
1 ∆Ck
C'Ck =
wo Rs

In our case, we can see that we need to multiply our L value by Rs/wc and our C value by 1/wc Rs.
L=11.25uH and C=4.50nF. Now we test this in HSPICE or Matlab and see if our response is good
enough for our purposes. If not, we iterate, making educated tweaks. Note that we’ve mainly designed
for the frequency domain in this methodology, which is fairly straightforward. Time-domain
characterstics, such as ringing, or overshoot, are often equally important but the methods for dealing
with them are completely different. But that’s for another handout to cover…

6
EE133 – Winter 2002
Cookbook Filter Guide
Element Values for Butterworth (Maximally Flat) Low-Pass Filter Prototypes
(go=1,wc =1,N=1 to 10)
N g1 g2 g3 g4 g5 g6 g7 g8 g9 g10 g11
1 2.0000 1.0000
2 1.4142 1.4142 1.0000
3 1.0000 2.0000 1.0000 1.0000
4 0.7654 1.8478 1.8478 0.7654 1.0000
5 0.6180 1.6180 2.0000 1.6180 0.6180 1.0000
6 0.5176 1.4142 1.9318 1.9318 1.4142 0.5176 1.0000
7 0.4450 1.2470 1.8019 2.0000 1.8019 1.2470 0.4450 1.0000
8 0.3902 1.1111 1.6629 1.9615 1.9615 1.6629 1.1111 0.3902 1.0000
9 0.3473 1.0000 1.5321 1.8794 2.0000 1.8794 1.5321 1.0000 0.3473 1.0000
10 0.3129 0.9080 1.4142 1.7820 1.9754 1.9754 1.7820 1.4142 0.9080 0.3129 1.0000

Element Values for Bessel (Maximally Flat Time Delay) Low-Pass Filter Prototypes (go=1,wc =1,N=1 to 10)
N g1 g2 g3 g4 g5 g6 g7 g8 g9 g10 g11
1 2.0000 1.0000
2 1.5774 0.4226 1.0000
3 1.2550 0.5528 0.1922 1.0000
4 1.0598 0.5116 0.3181 0.1104 1.0000
5 0.9303 0.4577 0.3312 0.2090 0.0718 1.0000
6 0.8377 0.4116 0.3158 0.2364 0.1480 0.0505 1.0000
7 0.7677 0.3744 0.2944 0.2378 0.1778 0.1104 0.0375 1.0000
8 0.7125 0.3446 0.2735 0.2297 0.1867 0.1387 0.0855 0.0289 1.0000
9 0.6678 0.3203 0.2547 0.2184 0.1859 0.1506 0.1111 0.0682 0.0230 1.0000
10 0.6305 0.3002 0.2384 0.2066 0.1808 0.1539 0.1240 0.0911 0.0557 0.0187 1.0000

Element Values for Chebyshev (Equal-Ripple) Low-Pass Filter Prototypes (go=1,wc =1,N=1 to 10)
0.5dB Ripple Band
N g1 g2 g3 g4 g5 g6 g7 g8 g9 g10 g11
1 0.6986 1.0000
2 1.4029 0.7071 1.9841
3 1.5963 1.0967 1.5963 1.0000
4 1.6703 1.1926 2.3661 0.8419 1.9841
5 1.7058 1.2296 2.5408 1.2296 1.7058 1.0000
6 1.7254 1.2479 2.6064 1.3137 2.4758 0.8696 1.9841
7 1.7372 1.2583 2.6381 1.3444 2.6381 1.2583 1.7372 1.0000
8 1.7451 1.2647 2.6564 1.3590 2.6964 1.3389 2.5093 0.8796 1.9841
9 1.7504 1.2690 2.6678 1.3673 2.7239 1.3673 2.6678 1.2690 1.7504 1.0000
10 1.7543 1.2721 2.6754 1.3725 2.7392 1.3806 2.7231 1.3485 2.5239 0.8842 1.9841

3.0dB Ripple Band

N g1 g2 g3 g4 g5 g6 g7 g8 g9 g10 g11
1 1.9953 1.0000
2 3.1013 0.5339 5.8095
3 3.3487 0.7117 3.3487 1.0000
4 3.4289 0.7483 4.3471 0.5920 5.8095
5 3.4817 0.7618 4.5381 0.7618 3.4817 1.0000
6 3.5045 0.7685 4.6061 0.7929 4.4641 0.6033 5.8095
7 3.5182 0.7723 4.6386 0.8039 4.6386 0.7723 3.5182 1.0000
8 3.5277 0.7745 4.6575 0.8089 4.6990 0.8018 4.4990 0.6073 5.8095
9 3.5340 0.7760 4.6692 0.8118 4.7272 0.8118 4.6692 0.7760 3.5340 1.0000
10 3.5384 0.7771 4.6768 0.8136 4.7425 0.8164 4.7260 0.8051 4.5142 0.6091 5.8095

Please note that these tables assume the following:

2
wc =
LC
Which means that your filters will have a smaller bandwidth than you usually expect if you define 3dB
as your cutoff frequency. You may easy compensate this by just roughly doubling wc

7
Butterworth normalized values

 For filter having a load impedance at least 10 times the RS, the load is considered infinite and the last component
must be a capacitor.
 For filter having RL less than Rs/10, it is considered as zero and the last component must be an series inductor

Normalized Butterworth element values, Rs=∞ or Rs=0

If the load impedance is close to the source impedance (within a factor of 0.1 or ten times), any configuration can be used
depending on the filter order

Normalized Butterworth element values, Rs=Rl=1

Chebyshev response

0.01 dB Chebyshev attenuation values filter order

0.1 dB Chebyshev attenuation values filter order

0.25 dB Chebyshev attenuation values filter order

0.5 dB Chebyshev attenuation values filter order

1 dB Chebyshev attenuation values filter order

Normalized Chebyshev element values, 0.01 dB ripple. Equal load.

Normalized Chebyshev element values, 0.1 dB ripple. Equal load.

Normalized Chebyshev element values, 0.25 dB ripple. Equal load.

Normalized Chebyshev element values, 0.5 dB ripple. Equal load.

Normalized Chebyshev element values, 1 dB ripple. Equal load.

NOTE: If Rs=RL is not possible to design even order filters

Load impedance limits for even-order Chebyshev filters

Normalized Chebyshev element values, 0.01 dB ripple. Rs=∞ or 0

Normalized Chebyshev element values, 0.1 dB ripple. Rs=∞ or 0

Normalized Chebyshev element values, 0.25 dB ripple. Rs=∞ or 0

Normalized Chebyshev element values, 0.5 dB ripple. Rs=∞ or 0

Normalized Chebyshev element values, 1 dB ripple. Rs=∞ or 0

Bessel response

Bessel attenuation values filter order

Rs≥1

Rs≤1

If Rs=RL the ladder can begin with eiter a capacitor or an inductor. Minimum number of inductors is preferred.

Bessel components cannot be calculated using formulaes. It should be used the transfer function and fractional division.
Bessel LC Values Rs=∞ or Rs=0

Bessel LC Values Rs=1

Bessel LC Values Rs=2

Bessel LC Values Rs=3

Bessel LC Values Rs=4

Bessel LC Values Rs=8