Sei sulla pagina 1di 12

IV.

Active filters
1. Frequency response with RC filter and LC filter
2. Active filters: an overview
Frequency domain
Time domain
. Filter types
!utterworth and Che"yshev filters
!essel filter
Filter comparison
1. Frequency response with RC filter and LC filter
Repetition
Let us remind simple# ordinary RC filter:
R
C
$
out
$
in
Fi%.1. &imple RC filter.
The output depends on the input:
( )
2
1
2 2 2
1
1
C R
V V
in out
+

.
'e assume R(const and C(const and frequency chan%es its value from ) to infinity .
Therefor from the a"ove formula results:
The filter is not an amplifier# it will rather weaken signal.
he !iggest value of output $
out
($
input
is o"taina"le for (). For any *)# $
out
+$
input
.
The frequency for which the term
1 RC
is called "d# frequency.
'hat does the "d# frequency mean, Let us chec- it up.
( )
2
1
1 1
1
2
1

+

input
out
V
V
2)1.11.)1 1 Renata /alic-a
And the same in deci"els:
dB dB . dB . * log
V
V
log
input
out
3 010 3 3010 0 10
2
1
20 20
0n one octave 1octave $eans twice a frequency2 the output drops to half. The simple RC
filter has a fall off(3d! per octave: 3d!4octave. Fi%. 2 shows characteristic of RC filter.
) .
)
.12d! ).25
).5
.3d!
.d!
)d! 1
14 2 ().6)6
$out4$in
%entle slope
3d!4octave s-irt
d! 7 77(27
1 octave
Fi%.2. Characteristic of simple RC filter.
&uch filters are sufficient only if si%nals "ein% re8ected "y the filter are situated far from the
desired pass."and:
pass
re8ect
input frequency
to the
output
Fi%.. Lower frequencies that are a"le to pass from input to output of the filter while hi%her
frequencies are re8ected.
$ery often however# frequency of undesired distur"ances is very close to the useful si%nal. 0n
that case# an appro9imation of the ideal :!rick%wall& frequency response is required
1Fi%.;.2.
2)1.11.)1 2 Renata /alic-a
Fi%.;. :!ric-.wall: frequency response.
'ithin
f , f f 2 0
0
the input volta%e $
in
appears on the output# and $
out
($
in
.
!eyond f
)
# 1f*f
)
2 the output should "e equal to ).
&imple RC filter does not ena"le to %et the desired shape from Fi%.;. <ow to %et closer to the
ideal, =ne possi"le way is to increase num"er of "asic RC section from Fi%.1. Frequency
response of multisection RC filters is shown in Fi%. 5.
) f
d!
f ) 1 f4f
d!
the ori%inal the normali>ed
Fi%.5. Frequency response of multisection RC filters. A and ! are linear plots# while C is
lo%arithmic. ! and C responses have "een normali>ed 1scaled2 for d! attenuation at unit
frequency. ?oint 2 on hori>ontal a9is for A means 8ust f(2<># point 2 for ! means f(2@f
d!
.
?oint 1.) for C means lo%1f4f
d!
2(1# therefore f4f
d!
(1) and f(1)@f
d!
.
2)1.11.)1 Renata /alic-a
To compare characteristics closer they are ad8usted in such a way that f
d!
for every individual
filter is assumed to equal to 1.
As it results from the fi%ure# when more sections are added# the d! point moves down in
frequency. For n(2 1see A2 the filter is more :low.pass: then it is for n(1: f
d!#n(2
+f
d!#n(1
.
&imple RC filter %ives 3d!4octave# 2 such sections %ive 2@3(12d!4octave# sections
1Ad!4octave and so on. !ut this simple cascadin% causes that the input impedance of the ne9t
section loads seriously the output of previous section. =utput of multisection filter is %ettin%
smaller# step "y step# and finally it "ecomes to small to "e useful. The final falloff is steeper
1see C2 "ut the -nee is still soft 1see !2 and does not remind :"ric-.wall: characteristic. 'e
mi%ht conclude: many soft knees do not a hard knee make.
Let us set to%ether the assets and the draw"ac-s of cascaded RC filter:
cons pros
0nput resistance loads the previous section
de%radin% the response.
Bany soft -nees do not a hard -nee ma-e.
&-irt is steeper. For n sections the final
falloff(n@3d!4octave.
Fi%.3. A %ood passive LC "andpass filter 1inductances are in m< and capacitances are in pF2.
!ottom: measured response of the circuit.
Including L ele$ent into desi%n 1see Fi%.32 "rin%s some ma%ic# which can not "e performed
without L# "ut there are also serious draw"ac-s:
2)1.11.)1 ; Renata /alic-a
The hi%her the requirements 1remem"er desired wall."ric- shape2 the more co$ple' the
filter# therefore it "ecomes !ulky and e'pensive.
The response is sli%htly degraded.
L elements are :lossy: that means they are not pure inductance# "ut in fact can "e
represent as R# L and C elements in series.
L element is non%linear# which means it is not constant for different value of current and
frequency.
Conclusion(
0nductors as filter elements leave $uch to !e desired. 'hat we really need is a way to o"tain
inductorless filter with characteristic of LC filter. The solution is# so called# active filter. 0t
uses R# C and op.amp.
). Active filters( an overview
There are different -inds of active filters:
*hen !and is criterion# the followin% are "ein% distin%uished: hi%h.pass# low.pass#
"and.pass and "and.re8ect.
f
f
f
f
low pass
"and pass
pass
pass
pass
pass
"and
re8ect
pass
pass
hi%h pass
*hen shape is criterion# the followin% are "ein% distin%uished: with ma9imal flat
characteristic# with ma9imal steepness of characteristic and with minimal distortion in
output si%nal.
As we want to desi%n inductorless filter we need an element which is a"le to mimic 1(to
simulate2 inductor. There are 2 interestin% elements:
2)1.11.)1 5 Renata /alic-a
1. Ce%ative 0mpedance Converter C0C.
2. Dyrator.
+egative I$pedance Converter +IC
0t converts impedance to its ne%ative 1opposition2. This element converts C to its ne%ative that
is to L:
C
L , L j
C
j
C
j
to converted is
C
j
C j
z
C
2 2
NIC
1

1
z
1 1

.
ima%inary a9is
i$pedance
C
j
2
1

voltage
>1C
real a9is
current
>2C
C
j
1
1

ima%inary a9is
i$pedance
voltage
>1L
real a9is
current
>2L
L j
1

L j
2

For type C element:


current %oes "efore volta%e
the "i%%er frequency the smaller impedance
For type L element:
volta%e %oes "efore current
the "i%%er frequency the "i%%er impedance
To sum up:
C
j
C
j
2
NIC
1
z

causes that volta%e %oes "efore current# like it is for L.


!ut the "i%%er frequency %ives the smaller impedance# not like it is for L.
,yrator
This element converts C into a true L:
2 2
GYR
z
1 1
CR L , L j CR j to converted is
C
j
C j
z
C

To sum up: L j CR j
2
GYR
z causes that volta%e %oes "efore current# like it is for L.
The "i%%er frequency the "i%%er impedance again like it is for L.
This element# which consists of R# C and op.amp and is called an active filter# is a"le to
mimic any inductance in a filter. The field of active filters is currently very successful. 0t is
possi"le to o"tain almost any desired form 1shape2 of filter characteristic.
Frequency do$ain
Let us concentrate on frequency characteristics of filters. There are of them shown in Fi%. 6.
2)1.11.)1 3 Renata /alic-a
Fi%.6. Filter characteristics versus frequency.
1Pass band, stop band, ripple band is sometime written as passband, stopband,
rippleband. Whih is orret! No one "nows. It is #p to $o#.%
Pass band E it is ran%e of frequencies# which are not seriously dumped "y the filter. The end
of the pass"and is usually 1"ut not necessarily# another definitions are allowed# see ripple
band2 defined as the f
d!
.
Cut-off frequency E allocates the ri%ht ed%e of the pass"and.
Ripple band E 1ripple ( very small wave2 it is the ran%e of %ain within which those very small
waves contain. The pass "and may "e defined "y such a frequency f
c
for which the
characteristic leaves ripple "and.
Transition region 1skirt2 E it starts from f
c
and ends at the "e%innin% of stop band.
Stop band E the "and of stron%# serious attenuation. The definition says: stop "and is the "and
in which attenuation is E;)d! 1or E5)d!# E3)d! etc. and it depends on particular situation#
filter application2.
Phase shift versus frequency. 0t should "e linear for low.pas filter. 0f not E the output will
contain distortions. This means "ad quality of noisy output si%nal.
Time delay is closely related with phase shift. 0t should "e constant and it is constant as lon%
as phase shift is linear.
0ma%ine you have a tape recorder which time delay characteristic is not constant. 0ma%ine that
the hi%her frequency the "i%%er time delay.
input si%nal is a %roup of notes
a2time delay is constant
f1 f2 f f; input si%nal
delayed output si%nal#
f1 f2 f f;
The input and the output sequence of notes are the same.
input si%nal is a %roup of notes
"2time delay is not constant# is "i%%er for "i%%er frequencies
f1 f2 f f; input si%nal
delayed output si%nal#
f1 f2 f f;
The output sequence of notes is different from the input sequence of notes.
i$e do$ain
2)1.11.)1 6 Renata /alic-a
Filters are also characteri>ed in time domain 1Fi%.A2. 0t is important when input has
form of step or pulse:
&TF?
?GL&F
Fi%.A. Filter characteristic versus time. Ri%ht: step response.
Rise time
r
t
E it is the time when response reaches H)I of its final value.
Settling time
s
t
. it is time required to set output within specified levels 1for instance t5I of
the final value2 and stay there.
Overshoot E it is undesired feature. 0t shows how "i% is the first ripple. The fi%ure shows 15I
overshoot. 0ts hei%ht is 15I of the final value.
Ringing E all the small and "i% ripples# waves# undesired distur"ances.
All the a"ove parameters seam to "e self e9planatory 1with a little help of the fi%ure2.
". Filter types
There are 1;2 main criterions when we tal- of filter quality:
The ultimate rate of falloff. 0t is usually n@3d!4octave.
The flatness of pass"and 1it tal-s of the ripples2.
Jistortion caused "y nonlinear phase shift 1that is "y not constant time delay2.
=ne may also care a"out time domain parameters: rise time# overshoot# settlin% time# etc.
Let us assume that we have chosen low.pass filter with falloff 1steepness2(3d!4octave. That
means we have quite complicated filter# which consists of 3 elementary sections. 'e need 1
capacitor# or 1 inductor# for each section# so the required rate of falloff of filter response
determines the comple9ity of the filter.
<avin% the filter# which consists of n sections we call it :n%pole filter:. The name :pole:
comes from a method of analysis# which is "eyond the area of our interest.
=nce we have decided the falloff# we can opti$i-e the flatness "ut we can do that at the
e'pense of the falloff steepness. !y allowin% some ripples we can %et steeper s-irt.
2)1.11.)1 A Renata /alic-a
There are availa"le classic filters# "ein% standards. The most important are of them:
The #utterworth filter E ma9imally flat pass"and characteristic.
The Che!yshev filter E the steepest transition from pass"and to stop"and.
The #essel E ma9imally flat 1constant2 time delay.
All the filters are availa"le in low pass# hi%h pass or "and pass version.
#utterworth and Che!yshev filters
Fi%.H. Cormali>ed low.pass !utterworth filter characteristics. Cote the improved steepness for
the hi%her order filters.
he #utterworth filter produces the flattest pass!and response# at the e9pense of steepness
in the transition re%ion from pass"and to stop"and. 0t has poor phase characteristic 1see
Fi%.12. The output depends on the input as follows:
2
1
2
1
1
1
1
]
1

,
_

n
c
in
out
f
f
V
V
#
where n is the order of filter 1num"er of poles# individual sections2# f
c
it is d! frequency.
As it results from Fi%.H# increasin% n flattens the pass"and and steepens the falloff.
0n most applications we allow some small ripples 1let us say 1d! ripples2 throu%hout the
pass"and.
he Che!yshev filter is an alternative: "y allowin% some pass"and ripples it %ives sharper
-nee. The output depends on the input as follows:
2
1
2 2
1
1
1
1
]
1

,
_

c
n
in
out
f
f
C
V
V
#
2)1.11.)1 H Renata /alic-a
where C
n
is the Che"yshev polynomial of the first -ind of de%ree n# is a constant that sets
the pass"and ripples. Li-e the !utterworth# the Che"yshev has rather poor phase
characteristic.
Lo%arithmic scale of amplitude response and
normali>ed frequency.
Linear scale of amplitude response and
normali>ed frequency.
Fi%.1). Comparison of some common 3.pole low.pass filters. The same filters are plotted on
"oth linear and lo%arithmic scales.
As you can see "oth# the !utterworth and the Che"yshev# are tremendous improvements over
an ordinary 3.pole RC filter.
0t should "e notice# that active filters constructed with R and C of finite tolerance 1accuracy is
usually 5I# 1)I# sometimes is less2 will deviate from the theoretically calculated response.
This means that in fact real !utterworth will produce some pass "and ripples anyway. Fi%.11
shows the worst case variation in resistor and capacitor value on filter response.
Fi%.11. The effect of R and C tolerance on active filter characteristic.
From this point of view# the Che"yshev is very rational desi%n. 0t is called an equiripple
filter.
2)1.11.)1 1) Renata /alic-a
&uppose that we need a filter with flatness to ).1d! within the pass"and and 2)d! attenuation
at a frequency 25I "eyond the top of the pass"and. 0t will require a 1H.pole !utterworth and
only A.pole Che"yshev. Fi%. 12 shows how we specify filter frequency response %raphically.
Fi%.12. &pecifyin% filter frequency response parameters.
Allowa"le ran%e of filter %ain E ripples . passMAX pass passMIN
< <
.
The minimum frequency at which the response leaves the pass"and .
cutoff
f
.
The ma9imum frequency at which the response enters the stop"and . stop
f
.
The minimum attenuation in the stop "and . stop

.
#essel filter
0n situation when the shape of output wave is paramount# a linear phase filter 1constant time
delay filter2 is desired. A filter whose phase shift varies linearly with frequency is equivalent
to a constant time delay for si%nals within the pass "and. That means the waveform is not
distorted. The #essel 1called also the Thomson2 has ma9imally flat time delay.
2)1.11.)1 11 Renata /alic-a
Fi%.1. Comparison of time delays for 3.pole !essel and !utterworth low pass filters. The
e9cellent time.domain performance of the !essel minimi>es waveform distortion.
The poor time delay performance of the !utterworth %ives nondesired effects such as
overshoot when driven with pulse si%nals. The price we pay for the !essel7s constancy of time
delay is less steep an amplitude response 1see Fi%.1)2.
Filter co$parison
Fi%.1; %ives more information a"out time.domain performance for the Che"yshev# the
!utterworth and the !essel.
Fi%.1;. A step response comparison for 3.pole low.pass filters normali>ed for d! attenuation
at 1<>.
The !essel is a very desira"le filter where performance in time domain is important. The
Che"yshev# with its hi%hly desira"le amplitude versus frequency characteristic has the poorest
time domain performance of the . The !utterworth is in "etween in time domain and
frequency properties.
2)1.11.)1 12 Renata /alic-a

Potrebbero piacerti anche