EE247 - Lecture 2

Filters
• Material covered today: – Nomenclature
– Filter specifications
• Quality factor
• Frequency characteristics
• Group delay
– Filter types
• Butterworth
• Chebyshev I
• Chebyshev II
• Elliptic
• Bessel
– Group delay comparison example

EECS 247 Lecture 2: Filters © © 2005 H.K. Page 1

EECS 247 Lecture 2: Filters ©

 Filter Specifications
Frequency characteristics (lowpass filter):
– Passband ripple (Rpass)
– Cutoff frequency or -3dB frequency
• – Stopband rejection
– Passband gain
Phase characteristics:
– Group delay
SNR (Dynamic range)
• SNDR (Signal to Noise+Distortion ratio)
Linearity measures: IM3 (intermodulation distortion), HD3
• (harmonic distortion), IIP3 or OIP3 (Input-referred or outputreferred
• third order intercept point)
• Power/pole & Area/pole

• © 2005 H.K. Page 4

Nomenclature
Filter Types

Provide frequency selectivity Phase shaping

or equalization

EECS 247 Lecture 2: Filters © © 2005 H.K. Page 3

EECS 247 Lecture 2: Filters ©

 •
Quality Factor (Q)

The term Quality Factor (Q) has different definitions:

– Component quality factor (inductor & capacitor
Q)
• – Pole quality factor
– Bandpass filter quality factor

Next 3 slides clarifies each

© 2005 H.K. Page 6

Lowpass Filter Frequency Characteristics

H ( jw )
Passband Ripple (Rpass) f−3dB
H (0)
Passband 3dB
Transition
Gain H ( jw ) Band
Stopband
Rejection

H ( jw )
fc f
0
fstop
Frequency (Hz)
x 10

Passband Stopband
Frequency

EECS 247 Lecture 2: Filters © © 2005 H.K. Page 5

EECS 247 Lecture 2: Filters ©

 Component Quality Factor (Q)

• For any component with a transfer function:

H j( w)= R(w)+1jX (w)

• Inductor
Inductor Q: & Capacitor Quality Factor

YL= Rs+1j Lw
•
Capacitor Q : QL=wRsL Rs L

ZC = 1 +1jwC
Rp
Rp

C
QC =wCRp

© 2005 H.K. Page 8

• Quality factor is defined as:

Q= XR((ww)) → Average Power

DissipationEnergy Stored

EECS 247 Lecture 2: Filters © © 2005 H.K. Page 7

EECS 247 Lecture 2: Filters ©

 Pole Quality Factor
jw
s-Plane

wx wP

sx s

wx
QPole =
2s x

EECS 247 Lecture 2: Filters © © 2005 H.K. Page 9

EECS 247 Lecture 2: Filters ©

 What is Group Delay?
• Consider a continuous time filter with s-domain transfer function
G(s): jq(w) G(jw) ” ‰G(jw)‰e

• Let us apply a signal to the filter input composed of sum of two

sinewaves at slightly different frequencies (∆ω<<ω):

vIN(t) = A1sin(wt) + A2sin[(w+Dw) t]

• The filter output is:

vOUT(t) = A1 ‰G(jw)‰sin[wt+q(w)] +

A2 ‰G[ j(w+Dw)]‰sin[(w+Dw)t+ q(w+Dw)]

EECS 247 Lecture 2: Filters © © 2005 H.K. Page 11

EECS 247 Lecture 2: Filters ©

 What is Group Delay?
Signal Magnitude and Phase Impairment

vOUT(t) = A1 ‰G(jw)‰sin {w [t + q(ww) ]} +

{
+ A2 ‰G[ j(w+Dw)]‰sin (w+Dw) t + q(ww) + ddq(ww) - [ (
q(ww) Dww) ]}

What is Group Delay?

Signal Magnitude and Phase Impairment

Phase distortion is avoided only if:

dq(w) q(w)
-
dw w =0
•
Clearly, if θ(ω)=kω, k a constant, no phase distortion
This type of filter phase response is called “linear phase”
Phase shift varies linearly with frequency τGR ≡ -
• dθ(ω)/dω is called the “group delay” and also has units
• of time. For a linear phase filter τGR ≡τPD =k τGR=τPD
implies linear phase
•
Note: Filters with θ(ω)=kω+c are also called linear phase filters, but
they’re not free of phase distortion

• © 2005 H.K. Page 14

• If the second term in the phase of the 2nd sin wave is non-zero, then
the filter’s output at frequency ω+∆ω is time-shifted differently than
the filter’s output at frequency ω
“Phase distortion”
• If the second term is zero, then the filter’s output at frequency
ω+∆ω and the output at frequency ω are each delayed in time by -
θ(ω)/ω
EECS 247 Lecture 2: Filters ©
• τPD ≡ -θ(ω)/ω is called the “phase delay” and has units of time EECS

247 Lecture 2: Filters © © 2005 H.K. Page 13

EECS 247 Lecture 2: Filters ©

 What is Group Delay?
Signal Magnitude and Phase Impairment

• If τGR=τPD No phase distortion vOUT(t) = A1 ‰G(jw)‰sin

[w(t - t )] +
GR

Summary
Group Delay

• Phase delay is defined as:

τPD ≡ -θ(ω)/ω [ time]
•
Group delay is defined as : τGR ≡ -
dθ(ω)/dω [time]
•
If θ(ω)=kω, k a constant, no phase distortion
•
For a linear phase filter τGR ≡τPD =k

© 2005 H.K. Page 16

+ A2 ‰G[ j(w+Dw)]‰
sin
[(w+Dw) (t - t )]
GR

• If alsoG( jw)=G[ j(w+Dw)] for all input frequencies

within the signal-band, vOUT is a scaled, time-shifted
replica of the input, with no “signal magnitude distortion” :

EECS 247 Lecture 2: Filters ©

• In most cases neither of these conditions are realizable

exactly EECS 247 Lecture 2: Filters © © 2005 H.K. Page 15

EECS 247 Lecture 2: Filters ©

 Filter Types
Lowpass Butterworth
Filter 0

Magnitude (dB)
-20
• Maximally flat amplitude within
the filter passband -40

-60

NormalizedGroup Delay
dN H( jw) 0 5

Phase (degrees)
=0
-200 3
dw w=0
-400 1
0 1 2
• Moderate phase distortion
Normalized Frequency
Example: 5th Order Butterworth filter

EECS 247 Lecture 2: Filters © © 2005 H.K. Page 17

EECS 247 Lecture 2: Filters ©

 Filter Types
Chebyshev I Lowpass Filter

• Chebyshev I filter
– Ripple in the passband
– Sharper transition band
compared to Butterworth
– Poorer group delay
– As more ripple is allowed in
the passband:
• Sharper transition band
• Poorer phase response

Normalized Frequency
Example: 5th Order Chebyshev filter

EECS 247 Lecture 2: Filters © © 2005 H.K. Page 19

EECS 247 Lecture 2: Filters ©

 Filter Types
Chebyshev II Lowpass

• Chebyshev II filter Bode

Diagram

Magnitude (dB)
0
– Ripple in stopband
-20
– Sharper transition
band compared to -40
Butterworth -60
– Passband phase 0
more linear

Phase (deg)
-90
compared to -180
Chebyshev I
-270
-360
0 0.5 1 1.5 2
Frequency [Hz]

Example: 5th Order Chebyshev II filter

EECS 247 Lecture 2: Filters © © 2005 H.K. Page 21

EECS 247 Lecture 2: Filters ©

 Filter Types
Elliptic Lowpass Filter
0

• Elliptic filter

Magnitude (dB)
-20
– Ripple in passband
– Ripple in the stopband -40

– Sharper transition band

-60
compared to Butterworth & both
0
Chebyshevs

Phase (degrees)
– Poorest phase response
-200

-400
0 1 2
Normalized Frequency
Example: 5th Order Elliptic filter

EECS 247 Lecture 2: Filters © © 2005 H.K. Page 23

EECS 247 Lecture 2: Filters ©

 Filter Types
Bessel Lowpass Filter

• Bessel
– All poles
– Maximally flat group delay
– Poor amplitude attenuation
– Poles outside unit circle
(s-plane)
– Relatively low Q poles

EECS 247 Lecture 2: Filters © © 2005 H.K. Page 25

EECS 247 Lecture 2: Filters ©

 Filter Types
Comparison of Various Type LPF Magnitude Response

Magnitude dB
)
-20

(
-40

-60

0 1 2

Magnitude (dB)
Normalized Frequency

Bessel
All 5th order filters with
same corner freq. Butterworth
Chebyshev I
Chebyshev II
Elliptic

EECS 247 Lecture 2: Filters © © 2005 H.K. Page 27

EECS 247 Lecture 2: Filters ©

 •
• Group Delay Comparison
Example

Lowpass filter with 100kHz corner frequency

Chebyshev I versus Bessel
– Both filters 4th order- same -3dB point
– Passband ripple of 1dB allowed for Chebyshev I

© 2005 H.K. Page 30

Comparison of Various LPF Groupdelay

EECS 247 Lecture 2: Filters ©

5
28
Bessel
Chebyshev I
0.5dB Passband Ripple
 1
12
10 Butterworth

4
1

Ref: A. Zverev, Handbook of filter synthesis, Wiley, 1967.

EECS 247 Lecture 2: Filters © © 2005 H.K. Page 29

EECS 247 Lecture 2: Filters ©

 Magnitude Response

EECS 247 Lecture 2: Filters © © 2005 H.K. Page 31

Phase Response

-50

-100
Phase [degrees]

-150

-200

-250

-300
4th Order Chebychev1
4th Order Bessel
-350
0 0.5 1 1.5 2
5
Frequency [Hz] x 10

EECS 247 Lecture 2: Filters ©© 2005 H.K. Page 32

 Group Delay

EECS 247 Lecture 2: Filters © © 2005 H.K. Page 33

EECS 247 Lecture 2: Filters ©

 Step Response
1.4
4th Order Chebychev1
4th Order Bessel
1.2

Amplitude
0.8

0.6

0.4

0.2

0
0 0.5 1 1.5 2
-5
Time (sec) x 10

EECS 247 Lecture 2: Filters © © 2005 H.K. Page 35

EECS 247 Lecture 2: Filters ©

1.5
Input
1 Output
0.5

1.5

-0.5

0.5

-1

-1.5
1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2
x 10
-4
-0.5

Pulse Broadening -1

Bessel versus -1.5

1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2

Chebyshev
-4
x 10

8th order Bessel 4th order Chebyshev I

Chebyshev has more pulse broadening compared to Bessel More ISI

EECS 247 Lecture 2: Filters © © 2005 H.K. Page 37

EECS 247 Lecture 2: Filters ©

 Measure of Signal Degradation
Eye Diagram

• Eye diagram is a useful graphical illustration for signal

degradation
• Consists of many overlaid traces of a signal using an oscilloscope
where the symbol timing serves as the scope trigger
• It is a visual summary of all possible intersymbol interference
waveforms
– The vertical opening immunity to noise
– Horizontal opening timing jitter

EECS 247 Lecture 2: Filters © © 2005 H.K. Page 39

EECS 247 Lecture 2: Filters ©

 Eye Diagram

EECS 247 Lecture 2: Filters ©

 Chebyshev versus Bessel

EECS 247 Lecture 2: Filters ©

0.8

0.6

0.4

0.2

-0.2

-0.4

-0.6

EECS 247 Lecture 2: Filters © © 2005 H.K. Page 41

EECS 247 Lecture 2: Filters ©

 Summary
Filter Types
– Filters with high signal attenuation per pole _
poor phase response
– For a given signal attenuation requirement of
preserving constant groupdelay Higher
order filter
• In the case of passive filters _ higher component count
• Case of integrated active filters _ higher chip area &
power dissipation
– In cases where filter is followed by ADC and DSP
• Possible to digitally correct for phase non-linearities
incurred by the analog circuitry by using phase equalizers

© 2005 H.K. Page 44

Eye Diagrams

1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2
Time x 10-5 Time x 10-5

Random data maximum symbol rate 1/100kHz Filter with

constant group delay More open eye Lower BER (bit-error-rate)

EECS 247 Lecture 2: Filters © © 2005 H.K. Page 43

EECS 247 Lecture 2: Filters ©

 RLC Filters

•Bandpass filter: R Vo

Vin L C

VVino =s2+wQRCoss+wo2

wo =1 LC

Q=woRC=LwRo

Singularities: Complex conjugate poles + zeros and zero & infinity

EECS 247 Lecture 2: Filters © © 2005 H.K. Page 45

RLC Filters

•Design a bandpass filter with:

R Vo
•Center frequency of 1kHz
•Q of 20 V in L C

•Assume that the inductor has series R resulting in an

inductor Q of 40
•What is the effect of finite inductor Q on the overall Q?

EECS 247 Lecture 2: Filters © Page 46

© 2005 H.K.
 RLC Filters
Effect of Finite Component Q
1 1 1
= +
Q filt Qideal Qind.
filt Q=20 (ideal L)
Q=13.3 (Qind.=40)

eComponent Q must be much higher compared to

desired filter Q
EECS 247 Lecture 2: Filters © © 2005 H.K. Page 47

RLC Filters

R Vo

Vin L C

Question:
Can RLC filters be integrated on-chip?

EECS 247 Lecture 2: Filters © 2005©

H.K. Page 48
 Monolithic Inductors

•
Monolithic LC Filters
Monolithic inductor in CMOS tech.
– L<10nH with Q<7
Max. capacitor size (based on realistic chip area)
– C< 10pF

cLC filters in the monolithic form feasible:

- Frequency >500MHz
- Only low quality factor filters

Learn more in EE242

© 2005 H.K. Page 50

Feasible Quality Factor & Value

c Feasible monolithic inductor in CMOS tech. <10nH with Q <7

vRef: “Radio Frequency Filters”, Lawrence Larson; Mead workshop presentation 1999

EECS 247 Lecture 2: Filters © © 2005 H.K. Page 49

EECS 247 Lecture 2: Filters ©

 Monolithic Filters
• Desirable to integrate filters with critical frequencies <<
500MHz • Per previous slide LC filters not a practical
option in the integrated form for non-RF frequencies
• Good alternative:

cActive filters built without the need for inductors

EECS 247 Lecture 2: Filters © © 2005 H.K. Page 51