feedback properties · 2008. 4. 17. · good feedback loop design ⇒ high loop (pk ) gain and high...
TRANSCRIPT
2008 Spring ME854 - GGZ Page 1Performance Specification and Limitations
Performance Specification and LimitationsPerformance Specification and Limitations
Feedback PropertiesFeedback Properties
systemfeedback standard following heConsider t
++ K P +
+
r -
u
di
up
dy
n
).()()()()( :Note
)()()( ,))(()(
to fromy sensitivitinput the Definely.respective
outputplant at the loop thebreaking ),()()(
and plant, theinput to at the loop thebreaking ),()()(
functions transfer loopoutput andinput Define
i
1
i
susPsKsdsu
sdsSsusLIsS
ud
sKsPsL
sPsKsL
pip
ipi
pi
o
i
−=
=+=
=
=
−
ies?uncertaint of face theinfeedback of benefits theachieve How to
2008 Spring ME854 - GGZ Page 2Performance Specification and Limitations
Performance Specification and LimitationsPerformance Specification and Limitations
++K P
+
+
r-
udi
up
dy
n
ly.respective matrics, difference returnoutput andinput
called are )( and )( . since ofary complementa is :Note
))()(()()(
))()(()()(
are matricesy sensiticitary complementoutput andinput The
)()()( ,))(()(
to fromy sensitivitoutput theDefine
1
1
1
sLIsLISITST
sLIsLsSIsT
sLIsLsSIsT
sdsSsysLIsS
yd
oi
oooo
iiii
ooo
++−=
+=−=
+=−=
=+=
−
−
−
iiooP
iioo
iooo
oioo
dSdKSnrKSu
dTdKSnrKSu
PdSnTdrSyr
dSPdSnrTy
+−−=
−−−=
−+−=−
++−=
)(
)(
)(
)(
thenstable, internally is system loop closed theIf
Feedback PropertiesFeedback Properties
2008 Spring ME854 - GGZ Page 3Performance Specification and Limitations
Performance Specification and LimitationsPerformance Specification and Limitations
For plant output y++
K P+
+
r-
udi
up
dy
n
iiooP
iioo
iooo
oioo
dSdKSnrKSu
dTdKSnrKSu
PdSnTdrSyr
dSPdSnrTy
+−−=
−−−=
−+−=−
++−=
)(
)(
)(
)(
directly.output plant effacts which reference theimpact todirect has noiseSensor :Note
)())(()(
)(
1))(()(
reduce torequiresoutput plant at the rejection edisturbanc Good
)( reduce rejection (s) edisturbancoutput Plant
)()( reduce rejection (s) edisturbancinput Plant
1
1
rn
dPSPPKIPS
dPKI
PKIS
sSd
sPsSd
iio
o
o
oi
↓⇒=+=
↓⇒+
=+=
⇒
⇒
−
−
σσσ
σσσ
directly.input plant effacts which reference theimpact todirect has noiseSensor :Note
)(
1))(()(
)())(()(
reduce torequiresinput plant at therejection edisturbanc Good
)()( reducerejection (s) edisturbancoutput Plant
)( reduce rejection (s) edisturbancinput Plant
1
1
rn
dKPI
KPIS
dKSKKPIKS
sKsSd
sSd
ii
oi
i
ii
↓⇒+
=+=
↓⇒=+=
⇒
⇒
−
−
σσσ
σσσ
For plant input up
Feedback PropertiesFeedback Properties
2008 Spring ME854 - GGZ Page 4Performance Specification and Limitations
Performance Specification and LimitationsPerformance Specification and Limitations
++K P
+
+
r-
udi
up
dy
n
)(
1)())(()( 1)(or 1)(
)(
1)())(()( 1)(or 1)(
invertible are and Suppose
1)( 1 )(
1)( 1 )(
imply that equations These
1)( if ,1)(
1)(
1)(
1
1)( if ,1)(
1)(
1)(
1 Then
1)()(1)(
1)()(1)( Note
1-1
1-1
PPPKIKKSKPPK
KKPPKIPSKPPK
KP
KPS
PKS
KPσKPσ
SKPσ
PKσPKσ
SPKσ
KPKPIKPσ
PKPKIPKσ
o
o
i
o
i
o
σσσσσσ
σσσσσσ
σσ
σσ
σ
σ
σσ
σσ
=≈+=⇒>>>>
=≈+=⇒>>>>
>>⇔<<
>>⇔<<
>−
≤≤+
>−
≤≤+
+≤+≤−
+≤+≤−
−
−
Feedback PropertiesFeedback Properties
2008 Spring ME854 - GGZ Page 5Performance Specification and Limitations
Performance Specification and LimitationsPerformance Specification and Limitations
++K P
+
+
r-
udi
up
dy
n
)(reject 1 )(gain controller Large
)(reject 1)()(gain loopoutput Large of eperformanc Good
dominated is frequency eBetween th
dominated is frequency eBetween th
4444444444 34444444444 21
4444444444 84444444444 76
d
d
ii
p
i
dP
dKPLu
>>
>>=
σ
σσperformance
)(reject 1 )(gain controller Large
)(reject 1)()(gain loopoutput Large of eperformanc Good
:Summary
dominated is frequency eBetween th
dominated is frequency eBetween th
4444444444 34444444444 21
4444444444 84444444444 76
id
d
i
o
dK
dPKLy
>>
>>=
σ
σσperformance
input.plant at rejection edisturbanc goodimply not doesoutput at rejection
edisturbanc good s,other word in ;)( meany mecessarilnot does )(
samll Hence,diagnoal. are and both unless general, in that Note
io
io
SS
KPSS
σσ
≠
disturbance
disturbance
gain)(controller high and gain)(loop Highdesign loopfeedback Good KPK⇒
Feedback PropertiesFeedback Properties
2008 Spring ME854 - GGZ Page 6Performance Specification and Limitations
Performance Specification and LimitationsPerformance Specification and Limitations
++K P
+
+
r-
udi
up
dy
n
dominated is frequency at small )( gain loop small )(or
requires general in whichpoles, plan halfrighyt no has
)det()det()))(det(())(det(
if stable is system perturbed The ).0 whenstable is system theis,(that stable
nomial is system loop closed theand stable with)( toperturbed is model
plant that theAssuming y.uncertaint model todue high
yarbitraril made benot can gain loop ereality th inBut
∆⇒∆∆
∆++=+
∆++=∆++
=∆
∆∆+
PKTT
TIPKIPKI
PKIPKIPKII
PI
oo
o
To
σσ
48476
frequency highat gain loop Low
requiresit gSuppressin frequency. highat t significan is noisesensor Typically,
gain loop highat ,)(
that Recall. noisesensor consider Now
o
oioo
T
nryPKdSPdSnrTy
n
−≈++−=
Feedback PropertiesFeedback Properties
2008 Spring ME854 - GGZ Page 7Performance Specification and Limitations
Performance Specification and LimitationsPerformance Specification and Limitations
++K P
+
+
r-
udi
up
dy
n
frequencyhigh at low relatively be shall ))((or ))(( gains loop Therefore,
1)]([
1)]([
1))(( if have We.invertible is Assuming
)()()(
since )saturationactuator (leading controlhigh ly unreasonab causemay
)1))(( while1))((or 1))(((
ofbandwidth theoutside gains loop large hand,other In the
1
1
jωLjωL
jPjP
jPP
ddnrPdTdnrKSdTdnrKSu
u
jP jωLjωLP
io
iiiiiio
io
σσ
ωσωσ
ωσ
ωσσσ
>>=
<<
−−−≈−−−=−−−=
<<>>>>
−
−
small is gain loop the whenlarge not too )( keep todesirable isit Therefore,
)()(
sincelow is gain loop wherefrequency at the small be shall gain
controller the,saturationactuator avoid To ).( gain controller heConsider t
K
dnrKdTdnrKSu
K
iio
σ
σ
−−≈−−−=
Feedback PropertiesFeedback Properties
2008 Spring ME854 - GGZ Page 8Performance Specification and Limitations
Performance Specification and LimitationsPerformance Specification and Limitations
++K P
+
+
r-
udi
up
dy
n
1)( ),( ),( >>KKPPK σσσ
MK
KPPK
≤
<<
)(
1)( ),(
σ
σσ
Feedback PropertiesFeedback Properties
2008 Spring ME854 - GGZ Page 9Performance Specification and Limitations
Performance Specification and LimitationsPerformance Specification and Limitations
ionsspecificat definefor used are )( functions ghtingdesign wei control futureFor
,1
,1)(
where
1)()(
:ionspecificat thedescribe
ofunction t weightinga usecan oneOr .))()(1()( where
,)(
,)(
becan ion specificat thesystemscalar afor example,For functions.
transferloop closed some of in termsor functionsy sensitivitry complement
and/or functionsy sensitivit on the specified becan
systemfeedback a of objectives eperformanc that theRecall
0
0
1
0
0
e
e
oe
o
o
o
W
/M
/εjW
jSjW
jKjPjS
MjS
jS
>∀
≤∀=
≤
+=
>∀≤
≤∀≤
−
ωω
ωωω
ωω
ωωω
ωωω
ωωεω
Weighted Weighted HH22 and and HH∞∞ PerformancePerformance
++K P
+
+
r-
udi
up
dy
n
2008 Spring ME854 - GGZ Page 10Performance Specification and Limitations
Performance Specification and LimitationsPerformance Specification and Limitations
Weighted Weighted HH22 and and HH∞∞ PerformancePerformance
++ K P +
+
-
n~
nW
eWey
n
dW
d~
d
uW
id~
id
pu
u
u~
iW
rWr
Standard feedback configuration with weights
Weighting matrices are used for following purposes:
1. Sensor/actuator unit conversion, signal scaling over channels and frequencies
2. Wi and Wd may be chosen to reflect frequency contents of di and d
3. Wn may be used to model frequency content of sensor noise n
4. We can be use to scale the output requirements (shape of output sensitivity)
5. Wu may be used for providing control and actuator requirements
6. Wr is an option to be used with non-unity feedback systems
2008 Spring ME854 - GGZ Page 11Performance Specification and Limitations
Performance Specification and LimitationsPerformance Specification and Limitations
Weighted Weighted HH22 and and HH∞∞ PerformancePerformance
K P-
n~
nW
eWey
n
dW
d~
d
uW
id~
id
pu
u
u~
iW
rWr
It is very essential that some appropriate weighting
matrices be used in order to utilize the optimalcontrol theory to be discussed in this class (i.e., H2 and H∞ theory).
As part of the controller design process, a very important step is to select weighting
matrices We, Wd, and Wu, and possibly Wn, Wi, and Wr. Note that selection of these
weighting matrices is not trivial and during the design process, these weighting matrices are
often used as controller design “tuning parameters”.
Control design is a process of selecting a controller K(s) such that some certain
weighted signals are made small.
2008 Spring ME854 - GGZ Page 12Performance Specification and Limitations
Performance Specification and LimitationsPerformance Specification and Limitations
Weighted Weighted HH22 and and HH∞∞ PerformancePerformance
K P-
n~
nW
eWey
n
dW
d~
d
uW
id~
id
pu
u
u~
iW
rWr
HHHH2222 Performance
2
2
2
2
22
2
2
2
2
20
22
2
}~{
functioncost
following theusemay design control realisticA ).actuator(s saturatemay that
gain controlhigh very toleadsfunction cost above ofon minimizati controllerA
norm an }{}{
~ edisturbanc the todue
error theofenergy expected theminimize weIf n.expectatio thedenotes where
)(
and
)()(~
i.e., direction, random with impulsean as modeled
elyapproximat becan ~
edisturbanc that theAssume
=+
⇐==
=
=
∫∞
∗
dou
doe
doe
WKSW
WSWueE
H
HWSWdteEeE
de
E
IE
ttd
d
ρρ
ηη
ηδ
2008 Spring ME854 - GGZ Page 13Performance Specification and Limitations
Performance Specification and LimitationsPerformance Specification and Limitations
Weighted Weighted HH22 and and HH∞∞ PerformancePerformance
K P-
n~
nW
eWey
n
dW
d~
d
uW
id~
id
pu
u
u~
iW
rWr
HHHH2222 Performance – a constrained optimization problem
.matrix weightingspecial a with controller
an is problem thisosolution t The t.requiremen eperformancoutput given a to
subjectenergy expected control theminimize todinterprete becan problem This
}){(
constraintenergy expectedoutput an subject to
norm an }{}{
matrix eightingconstant w a with control
weighteda ofenergy expected theminimize weIf n.expectatio thedenotes where
)~~
(
by definedmatrix covariance its with process random
mean zero a is ~
edisturbanc that theassume Now
2
22
2
2
2
20
22
2
2
2
e
ydo
dou
u
d
W
H
WSyyE
HWKSWdtuEuE
Wu
E
WddE
d
σσ ≤=
⇐==
=
∗
∞
∗
∫
2008 Spring ME854 - GGZ Page 14Performance Specification and Limitations
Performance Specification and LimitationsPerformance Specification and Limitations
Weighted Weighted HH22 and and HH∞∞ PerformancePerformance
K P-
n~
nW
eWey
n
dW
d~
d
uW
id~
id
pu
u
u~
iW
rWr
HHHH∞∞∞∞ Performance
yuncertaint modeling of toleranceii)
error sensor and edisturbanc of tradeoffi)
:control on sLimitation 2H
2
2
2
22
21
~
21
~
21
~
}~{sup
criterion mixeda or with
~sup
constraintenergy control subject to
sup
:functioncost
following theg Minimizinproblem. control following heconsider tNow
2
2
2
2
2
2
∞≤
∞≤
∞≤
∞∞
=+
≤=
=
dou
doe
d
dou
d
doe
d
WKSW
WSWue
WKSWu
WSWe
HH
ρρ
γ
2008 Spring ME854 - GGZ Page 15Performance Specification and Limitations
Performance Specification and LimitationsPerformance Specification and Limitations
Weighted Weighted HH22 and and HH∞∞ PerformancePerformance
K P-
n~
nW
eWey
n
dW
d~
d
uW
id~
id
pu
u
u~
iW
rWr
HHHH∞∞∞∞ Performance
problem.y sensitivit-mixed thecalledusually is problem design
This matrices. scalingy uncertaintdependent -frequency theare and where
}{sup
functioncost a weighted minimize can Or we
21
2
21
2
2
2
21
2
2
∞
∞≤
=+
H
WW
WKSW
WSWue
o
doe
d ρρ
Note: for a scalar system, the H∞ norm minimization problem can be viewed as
minimizing the maximum magnitude of the system’s steady-state response
under the worst case sinusoidal inputs.
This leads to the next optimization problem …..
2008 Spring ME854 - GGZ Page 16Performance Specification and Limitations
Performance Specification and LimitationsPerformance Specification and Limitations
Weighted Weighted HH22 and and HH∞∞ PerformancePerformance
K P-
n~
nW
eWey
n
dW
d~
d
uW
id~
id
pu
u
u~
iW
rWr
LLLL2222 to LLLL∞∞∞∞ Performance
et.given targ a than less is
magnitudeoutput case worst thesuch that size)actuator (minimum magnitude
controlmaximun theminimize 1, than less norm with edisturbanc allfor
:follows as dinterprete becan problem This
)(sup
constraintoutput an subject to
over norm an )(sup
norm weightedfollowing theMinimizing
2
22
1
2
1
2
2
2
2
L
ty
LtuW
L
y
d
u
d
σ≤
ℜ⇐
∞≤
∞∞≤
∞
Fortunately, this control problem is equivalent to the constrained H2 optimal
control design problem shown before.
2008 Spring ME854 - GGZ Page 17Performance Specification and Limitations
Performance Specification and LimitationsPerformance Specification and Limitations
Selection of Weighting FunctionsSelection of Weighting Functions
An Ad Hoc Process
10 , :shootover
4
: timesettling
8.03.0 , 16.26.0
: timerise
input step afor that know we theory,control classic From
)2(
systemorder second standard aconsider
,function weightinga choose To t.significan are and wheresfrequencie low
typicallys,frequencie of rang aover small keep want to webefore, discussed As
)1( ,)(
iserror tracking thebefore, discussed
as left.on shown systemfeedback SISOan Consider
21
2
1
<<=
≈
≤≤+
≈
+==
−=−+−=−=
−−
−
ξ
ξω
ξω
ξ
ξωω
ξ
πξ
eM
t
t
ssPK L
Wdr
S
PKSSPdTndrSyr e
p
n
s
n
r
n
n
e
i
++K P
+
+
r-
udi
up
dy
n
2008 Spring ME854 - GGZ Page 18Performance Specification and Limitations
Performance Specification and LimitationsPerformance Specification and Limitations
Selection of Weighting FunctionsSelection of Weighting Functions
22
n
2
)2(
1
1
consider function,y sensitivit the torelated is eperformanc theSince
along. of functiona isshoot over b)
and response system a) :pointskey Two
nn
n
ss
ss
L S
ωξωξω
ξ
ω
++
+=
+=
∝++
K P+
+
r-
udi
up
dy
n
n
s
s
nb
M
jSSM
ω
αωω
ξα
αξα
ξαα
ω
ω
=
++=
+−
+=
==
=
∞
max
2
2222
22
max
815.05.0
4)1(
4
)(:let and
2/
bandwidth loop Closed
05.0=ξ
10.0=ξ
20.0=ξ
80.0=ξ
00.1=ξ
50.0=ξ
2008 Spring ME854 - GGZ Page 19Performance Specification and Limitations
Performance Specification and LimitationsPerformance Specification and Limitations
Selection of Weighting FunctionsSelection of Weighting Functions
⇒
−=⇒= −
−
maximal findcan one , Using
)ln(
)ln(
and between
iprelationsh theusing minimal desired calculatecan We
),( tsrequiremendomain given timefor Now
max_min
1
min
,
2
s
P
Pp
P
psr
M
M
MeM
M
Mtt
ξ
πξ
ξ
ξ
ξ
πξ
++K P
+
+
r-
udi
up
dy
n
s
MsW
ωjsMs
sS
tt
bse
bs
r
ω
ωω
ω
ξ
+=
⇒
∀=+
≤
/
lyequivalentOr
responese. ry timesatisfacto provide will
,/
(s)
function y sensitivit
a Therefore, . requiredbandwidth minimal the
calculatecan one , and desired , using Then,
b
smin
2008 Spring ME854 - GGZ Page 20Performance Specification and Limitations
Performance Specification and LimitationsPerformance Specification and Limitations
Selection of Weighting FunctionsSelection of Weighting Functions
k
kb
bM
s
e
e
b
bMs
e
e
e
e
e
sW
W
sW
W
WeSW
SW
SWS
ks
s
+
+=
⇒+
+=
≤≤
∞<=
=
∞
∞
εω
ω
ε
εω
ω
ω
ε
follows as modified becan
required isansition steeper tr a If t.requiremen
eperformanc esatisfy th to suitable a choose
alwayscan wepurpose, practicalfor Therefore,
becan of choice possible One
1 s.t. )0(
such that matrix weightinga choose togoing are weperfectly,
input step the trackingof Instead hniques.design tec optimalour by handled benot
canit and (unstable) axisimaginary on pole a has function y transfersensitivit
weighted that theNote . and 0)0(that
implieshich infinity w is 0at weighting theSince
1
++K P
+
+
r-
udi
up
dy
n
2008 Spring ME854 - GGZ Page 21Performance Specification and Limitations
Performance Specification and LimitationsPerformance Specification and Limitations
Selection of Weighting FunctionsSelection of Weighting Functions
⇒+
+=
+=
⇒
−−−=
bcu
ubcu
u
bc
ubcu
u
bc
u
i
e
u
s
MsW
W
MsW
W
M
KS
TddnrKSu
W
W
ωεω
ε
ωω
ω
/
ghtingproper wei afor introduced is
an system stable CL of limitation toDue
/
be would candidate aHence,
sfrequencie
noisesensor and )(bandwidth controllerby limited isgain frequency High ii)
frequency lowin gain largefairly actuator theoflimit saturation and
andeffort control ofcost by the limited is sfrequencie lowin of Magnitude i)
:signal controlin slimitation following heConsider t
)(
equation signal control thegconsiderinby ofthat
similar to is weightingcontrol ofselection The
u
++K P
+
+
r-
udi
up
dy
n
uε
2008 Spring ME854 - GGZ Page 22Performance Specification and Limitations
Performance Specification and LimitationsPerformance Specification and Limitations
BodyBody’’s Gain and Phase Relations Gain and Phase Relation
++K P
+
+
r-
udi
up
dy
n
small toobenot shall 2
)(sin4)(1)(1
differencereturn hand,other On the
Margin Phase )( )(ln
slope crossBody
Therefore,
time)ofmost frequency crossing @ (negativeplot
Body of slop theis )(ln
and from deviates
en quickly wh decreases 2
cothln that Note
ln2
cothln and )ln(: where
2cothln
ln1)(
henfunction t transfer loop phase minimal stable a be )(Let
2
11
0
2/2/
2/2/
0
0
ccc
c
c
jLjLjL
jLd
jLd
d
jLd
ee
ee
dd
LdjL
sL
ωπωω
ωπυ
ω
υ
ωω
ωυ
υωωυ
υυ
υπω
υυ
υυ
∠+=+=+
∠+∝
−
+==
=∠
−−
−
−
∞
∞−∫
υ-3 -2 -1 0 1 2 3
0
1
2
3
4
5
6
7
8
2cothln
υ
2008 Spring ME854 - GGZ Page 23Performance Specification and Limitations
Performance Specification and LimitationsPerformance Specification and Limitations
BodyBody’’s Gain and Phase Relations Gain and Phase Relation
++K P
+
+
r-
udi
up
dy
n
20db).-(or 1- angreater th be toslopover cross theto
wanteNormally w small). (or too deep toobenot should
)( of slopover cross thesummary,a as Therefore, ωjL
υ
(-2) 40db- slop
12 margin phase with1060
10)( o
42
5
≈
≈++
=ss
sL
-60
-40
-20
0
20
40
Mag
nitud
e (
dB
)
100
101
102
103
104
-180
-135
-90
-45
0
Pha
se
(d
eg
)
Bode Diagram
Frequency (rad/sec)
-30
-20
-10
0
10
20
30
Mag
nitud
e (
dB
)
101
102
103
104
-135
-90
-45
0
Pha
se
(d
eg
)
Bode Diagram
Frequency (rad/sec )
(-1) db 20- slop
78 margin phase with1060
10500)( o
42
5
≈
≈++
+=
ss
ssL
2008 Spring ME854 - GGZ Page 24Performance Specification and Limitations
Performance Specification and LimitationsPerformance Specification and Limitations
BodyBody’’s Gain and Phase Relations Gain and Phase Relation
++K P
+
+
r-
udi
up
dy
n
Effect of non-minimal
phase transfer function
( )
lag pahse additional scontribute function trasfer loop of zeros phase minimal-Non
28,53,90:/
Define
2cothln
)(ln1
2cothln
)(ln1
)()(
Therefore, .)()( and phase minimal and stable is where
)()(
as factorized be can )(
then),( of zeros plan halfright be ,Let
4/,2/,0
001
1 0
00
1 0
00
1 0
000
2
2
1
1
21
0
⇒
−−−=+
+−∠=
+
+−∠+=
+
+−∠+=
+
+−∠+∠=∠
=
+
+−
+
+−
+
+−=
=
∞
∞−=
∞
∞−=
=
∫ ∑
∫ ∑
∏
ooo
L
L
zzz
k
i i
i
k
i i
imp
k
i i
imp
mpmp
mp
k
k
k
zj
zjz
zj
zjd
d
jLd
zj
zjd
d
jLd
zj
zjjLjL
sLsLL
sLzs
zs
zs
zs
zs
zssL
sL
sLz,zz
ωωω
ωφ
ωω
υυ
υ
ω
π
ωω
υυ
υ
ω
π
ωω
ωω
2008 Spring ME854 - GGZ Page 25Performance Specification and Limitations
Performance Specification and LimitationsPerformance Specification and Limitations
BodyBody’’s Sensitivity Integrals Sensitivity Integral
++K P
+
+
r-
udi
up
dy
n
effect. bedWater frequeniesother at onebelow kept isit if one
exceeds functiony sensitivit theof magnitude the
over which rangefrequency a is thereTherefore,
0)(ln
tosimplifies integral thestable, is )( If
)Re()(ln
holds integraly sensitivit s Bode'following the then),( of poles plane half-right
open be ,let and zeros thanpoles more two
leastat withfunction transfer loop-open thebe )(Let
0
10
21
⇒
=
=
∫
∑∫
∞
=
∞
ωω
ωω
djS
sL
pdjS
sL
p,pp
sL
m
i
i
mL
)( ωjS
+
−
1
ω
constants. given some are 0 and 0, ,0 where
),[ ,1~)(
example, For s.frequencie specific
a above small be function transfer loop open requires often sconstraint Bandwidth
h
1
>>>
∞∈∀<≤≤+
βω
ωωεω
ω β
h
hh
M
MjL
2008 Spring ME854 - GGZ Page 26Performance Specification and Limitations
Performance Specification and LimitationsPerformance Specification and Limitations
BodyBody’’s Sensitivity Integrals Sensitivity Integral
++K P
+
+
r-
udi
up
dy
n
[ ]( )
[ ]( )
.)(over t significan be couldy Sensitivit
~11
)(max
have We 0. stable, is )( that whenNote
.)Re(
where
~11
)(max
Then,
left. theonplot see low, tohigh from variesgain
loop whensfrequencie tworepresent and Let
)(
,
1
)(
,
hl
lh
lh
m
i i
lh
hlh
l
hl
lh
hlh
l
hl
jS
sL
pπ
ejS
ωω
εε
ω
α
ωωα
εε
ω
ωω
ωωβωωω
ω
ωωω
ωωβωωω
ω
α
ωωω
−⇒
−
≥
=
−=
−
≥
−−
∈
=
−−
∈
∑
-60
-40
-20
0
20
40
Ma
gn
itud
e (
dB
)
100
101
102
103
104
-180
-135
-90
-45
0
Ph
as
e (
de
g)
Bode Diagram
Frequency (rad/sec)
lωhω
2008 Spring ME854 - GGZ Page 27Performance Specification and Limitations
Performance Specification and LimitationsPerformance Specification and Limitations
Analyticity ConstraintsAnalyticity Constraints
++K P
+
+
r-
udi
up
dy
n
0. ) Re( withany for )()()()()(
theoremmodules maximumby Hence,
)()( ,)()(
have westable, are )( and )( Assume
)( and )(
Define
.conditions sufficientonly are e that thesNote .conditions ion)interpolat
(ory analyticit thesesatisfyingby guaranteed isstability internal that theNote
,2,1 ,0)( ,1)(
and,
,2,1 ,1)( ,0)(
Then, stable. is system
loop closed that theSuppose ).( of zeros and poles
plan halfright open thebe , and ,Let
11
11
11
11
2121
>≥=
∈∈
+=+=
+
−=
+
−=
===
===
−
∞
−
∞
∞−
∞−
−−
==∏∏
zzzSzBsSsBsS
HsTsBHsSsB
LILTLIS
zs
zssB
ps
pssB
kizTzS
mipTpS
sL
z,zzp,pp
pp
zp
k
j j
j
z
m
i i
ip
ii
ii
km
L
L
LL
2008 Spring ME854 - GGZ Page 28Performance Specification and Limitations
Performance Specification and LimitationsPerformance Specification and Limitations
Analyticity ConstraintsAnalyticity Constraints
++K P
+
+
r-
udi
up
dy
n
zero. plane hand-rightthan
smallermuch bemust system loop-closed theofbandwidth that indicates This
1. stable, is )( if where
)1
()1
(1
:/
Then zero.plan halpright a is and ,1 ,)( suppose Now
)()()(
thenstable, is such that weight a be suppose problem, edFor weight
)()(
have we,Similarily
)()(
then );( of zeroplan halfright a be Let
1
/
1
1
1
1
1
=
−≈−−
≤⇒=+
−≤
+
+
≤=
−
+≥
−
+=≥
−
+=≥
∏
∏
∏
∏
=
∞++
=∞
=
−
∞
=
−
∞
α
ααβε
αεωω
εωω
sL
Mz
M
zω
pz
pz
z
Mz
zSWsW
pz
pzzWsSsW
SWW
zp
zppBsT
pz
pzzBsS
sLz
s
m
i s
b
i
i
b
bs
es
Ms
e
m
i i
iee
ee
k
j j
j
z
m
i i
ip
b
bs