L2_f05
-
Upload
mohammed-rasheduzzaman-al-amin -
Category
Documents
-
view
218 -
download
0
Transcript of L2_f05
-
8/10/2019 L2_f05
1/26
EECS 247 Lecture 2: Filters 2005 H.K. Pag e 1
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. Pag e 2
Filters
Filters Provide frequency selectivity and/or phase shaping
nV
FilteroutV
( )jH
( )jH
-
8/10/2019 L2_f05
2/26
-
8/10/2019 L2_f05
3/26
EECS 247 Lecture 2: Filters 2005 H.K. Pag e 5
0
x 10Frequency (Hz)
Lowpass Filter Frequency Characteristics
( )jH
( )jH
( )0H
Passband Ripple (Rpass)
TransitionBand
cPassband
Passband
Gain
s topfStopband
Frequency
StopbandRejection
f
( )jH
3dBf
dB3
EECS 247 Lecture 2: Filters 2005 H.K. Pag e 6
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
-
8/10/2019 L2_f05
4/26
EECS 247 Lecture 2: Filters 2005 H.K. Pag e 7
Component Quality Factor (Q)
For any component with a transfer function:
Quality factor is defined as:
( )( ) ( )
( )( )
Energy StoredAverage Powe r Diss ipation
1H jR jX
XQ
R
=+
=
EECS 247 Lecture 2: Filters 2005 H.K. Pag e 8
Inductor & Capacitor Quality Factor
Inductor Q :
Capacitor Q :
Rs LL L1 LY Q
Rs j L Rs
= =+
Rp
C
C C1Z Q CRp1 jRp
C
= =+
-
8/10/2019 L2_f05
5/26
EECS 247 Lecture 2: Filters 2005 H.K. Pag e 9
Pole Quality Factor
x
j
P
xPo lex
Q2
=
s-Plane
EECS 247 Lecture 2: Filters 2005 H.K. Page 10
Bandpass Filter Quality Factor (Q)
-20
-15
-10
-5
0
0.1 1 10f1 fcenter f2
0
-3dB
f = f2 - f1
( )H jf
Frequency
Magnitude(dB)
Q= fcenter/f
-
8/10/2019 L2_f05
6/26
EECS 247 Lecture 2: Filters 2005 H.K. Page 11
Consider a continuous time filter with s-domain transfer function G(s):
Let us apply a signal to the filter input composed of sum of two
sinewaves at slightly different frequencies (
-
8/10/2019 L2_f05
7/26
EECS 247 Lecture 2: Filters 2005 H.K. Page 13
What is Group Delay?
Signal Magnitude and Phase Impairment
{ ]}[vOUT(t) = A1 G(j) sin t +()
+
{ ]}[+ A2 G[j(+)]sin (+) t +d()
d()
+()
-( )
If the second term in the phase of the 2nd sin wave is non-zero, then the
filters output at frequency + is time-shifted differently than thefilters output at frequency Phase distortion
If the second term is zero, then the filters output at frequency +
and the output at frequency are each delayed in time by -()/ PD -()/ is called the phase delay and has units of time
EECS 247 Lecture 2: Filters 2005 H.K. Page 14
Phase distortion is avoided only if:
Clearly, if ()=k, k a constant, no phase distortion
This type of filter phase response is called linear phase
Phase shift varies linearly with frequency
GR -d()/d is called the group delay and also has unitsof time. For a linear phase filter GR PD =k
GR= PD implies linear phase
Note: Filters with()=k+c are also called linear phase filters, buttheyre not free of phase distortion
What is Group Delay?Signal Magnitude and Phase Impairment
d()d
()- = 0
-
8/10/2019 L2_f05
8/26
EECS 247 Lecture 2: Filters 2005 H.K. Page 15
What is Group Delay?
Signal Magnitude and Phase Impairment
IfGR= PD No phase distortion
[ )](vOUT(t) = A1 G(j) sin t -GR +
[+ A2 G[j(+)] sin (+) )]( t -GR
If alsoG( j)=G[ j(+)] for all input frequencies withinthe signal-band, vOUT is a scaled, time-shifted replica of the
input, with no signal magnitude distortion :
In most cases neither of these conditions are realizable exactly
EECS 247 Lecture 2: Filters 2005 H.K. Page 16
Phase delay is defined as:
PD -()/ [ time]
Group delay is defined as :
GR -d()/d [time]
If ()=k, k a constant, no phase distortion
For a linear phase filter GR PD =k
SummaryGroup Delay
-
8/10/2019 L2_f05
9/26
EECS 247 Lecture 2: Filters 2005 H.K. Page 17
Filter Types
Lowpass Butterworth Filter
Maximally flat amplitude
within the filter passband
Moderate phase distortion
-60
-40
-20
0
Magnitude(dB)
1 2
-400
-200
Normalized Frequency
Phase(degrees)
5
3
1
N
ormalizedGroupDelay
0
0
Example: 5th Order Butterworth filter
N
0
d H( j )0
d
=
=
EECS 247 Lecture 2: Filters 2005 H.K. Page 18
Lowpass Butterworth Filter
All poles
Poles located on the unit
circle with equal angless-plane
j
Example: 5th Order Butterworth Filter
-
8/10/2019 L2_f05
10/26
EECS 247 Lecture 2: Filters 2005 H.K. Page 19
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
1 2
-40
-20
0
Normalized Frequency
Magnitude(dB)
-400
-200
0
Phase(degrees)
0
Example: 5th Order Chebyshev filter
35
0 No
rmalizedGroupDelay
EECS 247 Lecture 2: Filters 2005 H.K. Page 20
Chebyshev I Lowpass Filter Characteristics
All poles
Poles located on an ellipse
inside the unit circle
Allowing more ripple in the
passband:
_Narrower transition band
_Sharper cut-off
_Higher pole Q
_Poorer phase response
Example: 5th Order Chebyshev I Filter
s-planej
Chebyshev I LPF 3dB passband ripple
Chebyshev I LPF 0.1dB passband ripple
-
8/10/2019 L2_f05
11/26
EECS 247 Lecture 2: Filters 2005 H.K. Page 21
Bode Diagram
Frequency [Hz]
Phase(deg)
Magnitude(dB)
0 0.5 1 1.5 2
-360
-270
-180
-90
0
-60
-40
-20
0
Filter Types
Chebyshev II Lowpass
Chebyshev II filter
Ripple in stopband
Sharper transition
band compared to
Butterworth
Passband phase
more linear
compared to
Chebyshev I
Example: 5th Order Chebyshev II filter
EECS 247 Lecture 2: Filters 2005 H.K. Page 22
Filter Types
Chebyshev II Lowpass
Example:
5th Order
Chebyshev II Filter
s-plane
j
Both poles & zeros
No. of poles n
No. of zeros n-1
Poles located both inside
& outside of the unit circle
Zeros located onj axis
Ripple in the stopband
only
poles
zeros
-
8/10/2019 L2_f05
12/26
EECS 247 Lecture 2: Filters 2005 H.K. Page 23
Filter Types
Elliptic Lowpass Filter
Elliptic fi lter
Ripple in passband
Ripple in the stopband
Sharper transition band
compared to Butterworth &
both Chebyshevs
Poorest phase response
Magnitude(dB)
Example: 5th Order Elliptic filter
-60
1 2Normalized Frequency
0-400
-200
0
Phase(degrees)
-40
-20
0
EECS 247 Lecture 2: Filters 2005 H.K. Page 24
Filter Types
Elliptic Lowpass Filter
Example: 5th Order Elliptic Filter
s-plane
j
Pole
Zero
Both poles & zeros
No. of poles n
No. of zeros n-1
Zeros located onj axis
Sharp cut -off
_Narrower transition
band
_Pole Q higher
compared to the
previous filters
-
8/10/2019 L2_f05
13/26
EECS 247 Lecture 2: Filters 2005 H.K. Page 25
Filter Types
Bessel Lowpass Filter
s-planej
Bessel
All poles
Maximally flat group
delay
Poor amplitude
attenuation
Poles outside unit circle
(s-plane)
Relatively low Q poles
Example: 5th Order Bessel filter
Pole
EECS 247 Lecture 2: Filters 2005 H.K. Page 26
Magnitude Response of a Bessel Filter as aFunction of Filter Order (n)
Normalized Frequency
Magnitude[dB]
0.1 100-100
-90
-80
-70
-60
-50
-40
-30
-20
-10
0
n=1
2
3
4
5
76
Filte
rOrder
Increased
1 10
-
8/10/2019 L2_f05
14/26
EECS 247 Lecture 2: Filters 2005 H.K. Page 27
Filter Types
Comparison of Various Type LPF Magnitude Response
-60
-40
-20
0
Normalized Frequency
Magnitude(dB)
1 20
BesselButterworth
Chebyshev I
Chebyshev II
Elliptic
All 5th order filters with same corner freq.
Magnitude(dB)
EECS 247 Lecture 2: Filters 2005 H.K. Page 28
Filter Types
Comparison of Various LPF Singularities
s-plane
j
Poles Bessel
Poles Butterworth
Poles Elliptic
Zeros Elliptic
Poles Chebyshev I 0.1dB
-
8/10/2019 L2_f05
15/26
EECS 247 Lecture 2: Filters 2005 H.K. Page 29
Comparison of Various LPF Groupdelay
Bessel
Butterworth
Chebyshev I
0.5dB Passband Ripple
Ref: A. Zverev,Handbook of filter synthesis, Wiley, 1967.
1
12
1
28
1
1
10
5
4
EECS 247 Lecture 2: Filters 2005 H.K. Page 30
Group Delay Comparison
Example
Lowpass filter with 100kHz corner frequency
Chebyshev I versus Bessel
Both filters 4th order- same -3dBpoint
Passband ripple of 1dBallowed for Chebyshev I
-
8/10/2019 L2_f05
16/26
EECS 247 Lecture 2: Filters 2005 H.K. Page 31
Magnitude Response
Bode Magnitude Diagram
Frequency [Hz]
Magnitude(dB)
104
105
106
-70
-60
-50
-40
-30
-20
-10
0
4th Order Chebychev 14th Order Bessel
EECS 247 Lecture 2: Filters 2005 H.K. Page 32
Phase Response
0 0.5 1 1.5 2
x 105
-350
-300
-250
-200
-150
-100
-50
0
Frequency [Hz]
Phase[degrees]
4th Order Chebychev 14th Order Bessel
-
8/10/2019 L2_f05
17/26
EECS 247 Lecture 2: Filters 2005 H.K. Page 33
Group Delay
104
105
106
0
2
4
6
8
10
12
14
Frequency [Hz]
GroupDelay[s
]
4th Ord. Chebychev 14th Ord. Bessel
EECS 247 Lecture 2: Filters 2005 H.K. Page 34
Normalized Group Delay
104
105
106
0
0.5
1
1.5
2
2.5
3
Frequency [Hz]
GroupDelay[normalized]
4th Ord. Chebychev 14th Ord. Bessel
-
8/10/2019 L2_f05
18/26
EECS 247 Lecture 2: Filters 2005 H.K. Page 35
Step Response
Time (sec)
Amplitude
0 0.5 1 1.5 2
x 10-5
0
0.2
0.4
0.6
0.8
1
1.2
1.44th Order Chebychev 14th Order Bessel
EECS 247 Lecture 2: Filters 2005 H.K. Page 36
Intersymbol Interference (ISI)ISI Broadening of pulses resulting in interference between successive transmitted
pulses
Example: Simple RC filter
-
8/10/2019 L2_f05
19/26
EECS 247 Lecture 2: Filters 2005 H.K. Page 37
Pulse Broadening
Bessel versus Chebyshev
1.1 1.2 1. 3 1 .4 1. 5 1 .6 1.7 1. 8 1. 9 2
x 10-4
-1.5
-1
-0.5
0
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
-1.5
-1
-0.5
0
0.5
1
1.5
8th order Bessel 4th order Chebyshev I
Chebyshev has more pulse broadening compared to Bessel More ISI
InputOutput
EECS 247 Lecture 2: Filters 2005 H.K. Page 38
0 0.2 0.4 0.6 0.8 1 1.2 1.4
x 10-4
-1.5
-1
-0.5
0
0.5
1
1.5
0 0.2 0.4 0.6 0.8 1 1.2 1.4
x 10-4
-1.5
-1
-0.5
0
0.5
1
1.5
0 0.2 0.4 0.6 0.8 1 1.2 1.4
x 10-4
-1.5
-1
-0.5
0
0.5
1
1.5
1111011111001010000100010111101110001001
1111011111001010000100010111101110001001 1111011111001010000100010111101110001001
Response to Random Data
Chebyshev versus Bessel
4th order Bessel 4th order Chebyshev I
Input Signal:
Symbol rate 1/130kHz
-
8/10/2019 L2_f05
20/26
EECS 247 Lecture 2: Filters 2005 H.K. Page 39
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 40
Measure of Signal Degradation
Eye Diagram
Random data with symbol rates: 1/50kHz
1/100kHz
1/130kHz
Frequency [Hz]
Magnitude(dB)
2 4 6 8 10 12 14
x 104
-10
-9
-8
-7
-6
-5
-4
-3
-2
-1
0
BesselChebychev 1
2 4 6 8 10 12 14
x 104
0
0.5
1
1.5
2
2.5
3
Frequency [Hz]
GroupDelay[normalized]
4th Ord. Chebychev 14th Ord. Bessel
-
8/10/2019 L2_f05
21/26
EECS 247 Lecture 2: Filters 2005 H.K. Page 41
1 1.1 1.2 1.3 1.4 1.5 1.6 1.7
x 10-5
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
Time
4th Order Chebychev
0.8 0.9 1 1.1 1.2 1.3 1.4 1.5
x 10-5
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
Time
4th Order Bessel
0.6 0.8 1 1.2 1.4 1.6
x 10-5
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
Time
Input Signal
Eye Diagram
Chebyshev versus Bessel
4th order Chebyshev I4th order Bessel
Input Signal
Random data
Symbol rate:1/ 130kHz
EECS 247 Lecture 2: Filters 2005 H.K. Page 42
Eye Diagrams
2.5 3 3.5 4 4.5 5
x 10
-5
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
Time
4th Order Chebychev
2.5 3 3.5 4 4.5 5
x 10
-5
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
Time
4th Order Bessel
100%
Eye openin g
70%
Eye openin g
Random data maximum power symbol rate 1/50kHz
-
8/10/2019 L2_f05
22/26
EECS 247 Lecture 2: Filters 2005 H.K. Page 43
Eye Diagrams
Filter with constant group delay More open eye Lower BER (bit-error-rate)
Random data maximum symbol rate 1/100kHz
1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2
x 10-5
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
Time
4th Order Bessel
1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2
x 10-5
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
Time
4th Order Chebychev
80%
Eye opening
40%
Eye openin g
EECS 247 Lecture 2: Filters 2005 H.K. Page 44
Summary
Filter Types
Filters with high signal attenuation per pole _poor phase response
For a given signal attenuation requirement ofpreserving constant groupdelay Higher orderfilter 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
-
8/10/2019 L2_f05
23/26
EECS 247 Lecture 2: Filters 2005 H.K. Page 45
RLC Filters
Bandpass filter:
Singularities: Complex conjugate poles + zeros and zero & infinity
o
so RC
2 2in oQ
o
oo
VV s s
1 LC
RQ RCL
=+ +
=
= =
oVR
CLnV
EECS 247 Lecture 2: Filters 2005 H.K. Page 46
RLC Filters
Design a bandpass filter with:
Center frequency of 1kHz
Q of 20
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?
oVR
CLnV
-
8/10/2019 L2_f05
24/26
EECS 247 Lecture 2: Filters 2005 H.K. Page 47
RLC Filters
Effect of Finite Component Q
ideal f i l t ind.f i l t
1 1 1Q QQ
= +
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 48
RLC Filters
Question:
Can RLC filters be integrated on-chip?
oVR
CLnV
-
8/10/2019 L2_f05
25/26
EECS 247 Lecture 2: Filters 2005 H.K. Page 49
Monolithic Inductors
Feasible Quality Factor & Value
vRef: Radio Frequency Filters, Lawrence Larson; Mead workshop presentation 1999
c Feasible monolithic inductor in CMOS tech.
-
8/10/2019 L2_f05
26/26
EECS 247 Lecture 2: Filters 2005 H.K. Page 51
Monolithic Filters
Desirable to integrate filters with critical frequencies