EE 508 Lect 20 Fall 2020
EE 508 Lect 20 Fall 2020
EE 508 Lect 20 Fall 2020
Lecture 20
Sensitivity Functions
• Comparison of Filter Structures
• Performance Prediction
What causes the dramatic differences in performance between these two structures?
How can the performance of different structures be compared in general?
C1
R1 R2 VOUT
K
VIN C2
Review from last time
1
R1R2C1C2
T s =K
1 1 1-K 1 Equal R, Equal C, Q=10 Pole
s2 +s + + +
R1C1 R2C1 R2C2 R1R2C1C2 Locus vs GBN
R3
R1 R2 VOUT
-K
VIN R4
C1 C2
1
R1R2C1C2
T s = -K
1 R1 1 1 C2 1+ R1 R3 1+K + R1 R 4 1+ R 2 R3 + R 2 R1
s2 +s 1+ + + 1+ +
1 1
R C R 3 R C
4 2 R 2 2
C C1 R1R 2 C1C 2
Review from last time
Effects of GB on poles of KRC and -KRC Lowpass Filters
Im
Over-order
pole
Re
Actual “desired” poles
Desired poles
dF f k
dxi
Sxi
F i1 xiN
XN
Review from last time
Consider:
R VOUT
1
T s =
1+RCs
VIN C
ω0
T s =
s + ω0
1
ω0 =
RC
Review from last time
Proof:
S S f
xi
Xi i
xi Sfxi i
Xi i xi
S Xi i
i
xi Xi
xi
i 1 xi It is often the case that functions of interest are
S Xi i
xi i Xi i of the form expressed in the hypothesis of the
Xi theorem, and in these cases the previous claim is
correct
S Xi i
xi i
Review from last time
are dependent only upon the circuit architecture and not dependent
upon the components and and the right terms are dependent only upon
the components and not dependent upon the architecture
This observation is useful for comparing the performance of two or more circuits
where the function f shares this property
Review from last time
Summed Sensitivity
m
S S f
xi
i 1
Not very useful because sum can be small even when individual
sensitivities are large
Schoeffler Sensitivity
m
S f
xi
i 1
Strictly heuristic but does differentiate circuits with low sensitivities from those
with high sensitivities
Review from last time
R R
S f
All resistors
i
C
All capacitors
SCf i
OA
sf
All op amps
i
Review from last time
Homogeniety (defn)
VIN C1 C2
Q Q Q Q
Determine the passive Q sensitivities SR1
S R2
S C1
S C2
1
ω0 = Q=
R1R2C1C2
R1R2C1C2 R1C1+R1C2 +R2C2
SRQ 2 • 1
R1C1+R1C2 +R2C2
2
1
Q
Example R1 V1 R2 VOUT
VIN C1 C2
Q Q Q Q
Determine the passive Q sensitivities S
R1
S R2
S C1
S C2
1
R C +R C +R2C1 R1R2C1C2 R2C1C2 - C1+C2 R1R2C1C2 R
-1/2 1/2
1 1 1 2
S
Q 2 • 1
R1C1+R1C2 +R2C2
R1 2
Q
1 R1 C1 +C2
SRQ
1
2 R1C1 +R1C2 +R2C2
R1 V1 R2 VOUT
Example
VIN C1 C2
1 R1 C1 +C2 1 R 2C2
SRQ SRQ
1
2 R1C1 +R1C2 +R2C2 2
2 R1C1 +R1C2 +R 2C2
1 R1C1 1 C2 R1 R 2
S
Q SQC
C1
2 R1C1 +R1C2 +R2C2
2
2 R1C1 +R1C2 +R 2C2
Verify k2 k1
Q
S
Q
Ci
=0 SR =0
i=1 i
i=1
Could have saved considerable effort in calculations by using these theorems after
SRQ1
and SQC 1
were calculated
Corollary 3: If all op amps in an RC
active filter are ideal and there are k1
resistors and k2 capacitors and if pk is any
pole and zh is any zero, then
k2
k1
S
pk
= 1 S
pk
Ci
= 1
Ri i=1
i=1
and k2
= 1
k1
= 1
zh
S
zh
S Ci
Ri i=1
i=1
Corollary 3: If all op amps in an RC
active filter are ideal and there are k1
resistors and k2 capacitors and if pk is any
pole and zh is any zero, then
k1 k2
S
pk
Ri
= 1 S
pk
Ci
= 1
i=1 i=1
and
k2
= 1
k1
= 1
zh
S
zh
S Ci
Ri i=1
i=1
Proof:
It was shown that scaling the frequency dependent elements by a factor η divides
the pole (or zero) by η
Thus roots (poles and zeros) are homogeneous of order -1 in the frequency
scaling elements
Proof:
Thus roots (poles and zeros) are homogeneous of order -1 in the frequency
scaling elements
(For more generality, assume k3 inductors)
k2 k3
S
p
Ci
+ S = 1
p
Li
(1)
i=1 i=1
Since impedance scaling does not affects the poles, they are homogenous of
order 0 in the impedances
k1 k2 k3
p p p
R 1/C L
S + S + S =0 (2)
i=1 i i=1 i i=1 i
Since there are no inductors in an active RC network, is follows from (1) that
k2
S 1
C
p
i
i=1
And then from (2) and the theorem about sensitivity to reciprocals that
k1
S 1
R
p
i
i=1
Corollary 4: If all op amps in an RC
active filter are ideal and there are k1
resistors and k2 capacitors and if ZIN is any
input impedance of the network, then
k1 k2
ZIN ZIN
S Ri
- S Ci
=1
i=1 i=1
Claim: If op amps in the filters
considered previously are not ideal but are
modeled by a gain A(s)=1/(s), then all
previous summed sensitivities developed for
ideal op amps hold provided they are
evaluated at the nominal value of =0
Sensitivity Analysis
If a closed-form expression for a function f
is obtained, a straightforward but tedious
analysis can be used to obtain the
sensitivity of the function to any
components f x
f
Sx = •
x f
m i m
Consider: ai s (s-zi )
T s = i=0
n i
=K i=1
n
bi s (s-pi )
i=0 i=1
Closed-form expressions for T(s), T(jω), |T(jω)|, T jω , ai, bi, can be
readily obtained
Sensitivity Analysis
If a closed-form expression for a function f is
obtained, a straightforward but tedious analysis
can be used to obtain the sensitivity of the
function to any components
f x
S = • f
x f
x
Consider: m i m
ai s (s-zi )
T s = i=0n i
=K i=1
n
bi s (s-pi )
i=0 i=1
C1
R1 R2 VOUT
K
VIN C2
K0
R1R2C1C2
T s =
1
s2 +s
1 1-K 0 + 1 K s s2 +s 1 + 1 + 1 + 1
+ + 0 R C R C R C R R C C
1 1
R C R C
2 1 R 2 2
C R R C C
1 2 1 2 1 1 2 1 2 2 1 2 1 2
Consider R1
K0
R2C1C2
T s =
1
R1s2 +s +R1
1 1-K 0 + 1 K s R s2 +s 1 +R 1 +R 1 + 1
+R1 0 1 C 1 1 R CC
1
C R C
2 1 R 2 2
C R C C
2 1 2 1 R C
2 1 R 2 2
C 2 1 2
K0
R C C R1 0
T s = 2 1 2
1
1
1 1 2 1 1-K 0 +K s s2 +s 1 + 1
s K 0 s s + 1
+R s +s + 0 R C R C
1 C R C C
2 1 2 1C R C C
2 1 2 2 1
R C R 2 2
C 2 1 2 2
Example of Bilinear Property
C1
VIN K0
C2
T s = R2C2
3-K 0 1 2 3 1
s2 +s + 2 2 K 0 s s +s + 2 2
RC R C RC R C
K0
T s =
R2 C2s2 K 0 sC2 +R sC 3-K 0 3K 0C s2 + 1+K 0 s
dz dp
i i
Note: Do not need to find expressions for the poles or the zeros to find the pole
and zero sensitivities !
Note: Do need the poles or zeros but they will generally be known by design
Note: Will make minor modifications for extreme values for x (i.e. τ for op amps)
Root Sensitivities
Theorem: If pi is any simple pole of T(s), then
x D1 pi
Sx
pi
pi d D p i
dp
i
D s =D0 s +xD1 s
By definition of a pole,
D pi =0
\ D pi =D0 pi +xD1 pi 0
Root Sensitivities
\ D pi =D0 pi +xD1 pi
Differentiating this expression implicitly WRT x, we obtain
pi D1 pi
x D pi
p
i
Root Sensitivities
pi D1 pi
x D pi
p
i
Finally, from the definition of sensitivity,
x pi x D1 pi
S pi
pi x pi D pi
x
p
i
Root Sensitivities
x pi x D1 pi
S pi
pi x pi D pi
x
p
i
Observation: Although the sensitivity expression is readily
obtainable, direction information about the pole movement is
obscured because the derivative is multiplied by the quantity pi
which is often complex. Usually will use either
pi
sxpi
or
x
x pi x D1 pi
Spi
pi x pi D pi
x
p
i
which preserve direction information when working with pole or
zero sensitivity analysis.
Root Sensitivities
Summary: Pole (or zero) locations due to component
variations can be approximated with simple analytical
calculations without obtaining parametric expressions for
the poles (or zeros).
D1 pi
s
pi
D s D0 s x D1 s
D pi
x and
p
i p iN
Alternately, x pi
pi pi Sx
x
Example: Determine SRp i
1
for the +KRC Lowpass Filter for equal R, equal C
C1
N0 s +xN1 s
T s =
R1 R2 VOUT
D0 s +xD1 s
K
x pi x D p
VIN C2 Spx i
1 i
K s
K0 pi x pi D pi
1+K 0 s p
i
K0
R1R2C1C2
T s =
1
s2 +s
1 1-K 0 + 1 K s s2 +s 1 + 1 + 1 + 1
+ + 0 R C R C R C R R C C
R1C1 R2C1 R2C2 R1R2C1C2 1 1 2 1 2 2 1 2 1 2
evaluate at τ=0
K0
R2C1C2
T s =
1 1 1 2 1 1-K 0
s + +R1 s +s +
C1 R2C1C2 R2C1C2 R2C1 R2C2
Example: Determine SRp i
1
for the +KRC Lowpass Filter for equal R, equal C
C1
N0 s +xN1 s
T s =
R1 R2 VOUT
D0 s +xD1 s
K
x pi x D p
Spx i
1 i
pi D pi
VIN C2
K0
pi x
K s p
1+K 0 s i
K0
R2C1C2 1
D1 s =s2 +s
1-K0
T s = +
1 1 1 2 1 1-K 0 R2C1 R2C2
s + +R1 s +s +
C1 R2C1C2 R2C1C2 R2C1 R2C2
1
D s = s
1 1 2 1 1-K 0 R s2 +s ω0 +ω2
+ 1
+R s +s + 1 Q 0
1C R C C
2 1 2 R 2 1 2
C C
2 1
R C R 2 2
C
1
p2 +p +
1-K 0
x pi 1 R2C1 R2C2
SRp
1
pi x pi ω0
i
2p
Q
Example: Determine SRp i
1
for the +KRC Lowpass Filter for equal R, equal C
N0 s +xN1 s
C1
T s =
D0 s +xD1 s
R1 R2 VOUT T
K
x pi x D p
Spx i
1 i
pi x pi D pi
VIN C2
K0 p
K s
1+K 0 s i
1
p2 +p +
1-K 0 K0
x pi 1 R2C1 R2C2 T s =
R1R 2C1C2
SRp 1 1-K 0 +
1
pi x pi ω0 s2 +s
1 1
i
2p
+ +
R1C1 R 2C1 R 2C2 R1R 2C1C2
Q
1
+p
1 1
p2 +p
1 1-K0 + 1 =0
+ +
x pi 1 R1R2C1C2 R1C1 R1C1 R2C1 R2C2 R1R2C1C2
SRp
1
pi x pi ω0
i
2p 1 1-K0 = - 1 1
Q p2 +p + -p
1 R2C1 R2C2 R1R2C1C2 R1C1
ω02 + p
x pi 1 R1C1
SRp
1
pi x ω0 ω0
i
2p
Q
Example: Determine SRp i
1
for the +KRC Lowpass Filter for equal R, equal C
C1
1
R1 R2 VOUT ω02 + p
K x pi 1 R1C1
Spx
pi x ω0
i
ω0
VIN C2 i
2p
K0 Q
K s
1+K 0 s
1
For equal R, equal C ω0 =
RC
ω0 ω0
ω0 - 1-4Q2
x pi 1 ω02 + pω0 SRp 2Q 2Q
Sp
pi x ω0 ω0 ω
R1 1
0 1-4Q2
i
2p Q
Q
x pi ω0 + p 1 1
Q- 1-4Q2
SRp
1
pi x ω0 SRp 2 2
2p 1-4Q2
1
Q
Example: Determine SRp i
1
for the +KRC Lowpass Filter for equal R, equal C
C1
x pi
Spx i
R1 R2 VOUT
pi x
K
For equal R, equal C
VIN C2
K0 1 1
K s Q- 1-4Q2
1+K 0 s SRp 2 2
1-4Q2
1
x
p i p i Spx
i
x
R1
pi ω0 0.5 0.5 j
R1
Example: Determine SRp i
1
for the +KRC Lowpass Filter for equal R, equal C
C1
x pi
Spx i
R1 R2 VOUT pi x
K
For equal R, equal C
VIN C2
K0
K s
1+K 0 s
R1
For high Q pi ω0 0.5 0.5 j
R1
Could we have assumed equal R equal C before calculation?
No ! Analysis would not apply (not bilinear)
Results would obscure effects of variations in individual components
Was this a lot of work for such a simple result?
Yes ! But it is parametric and still only took maybe 20 minutes
But it needs to be done only once for this structure
Can do for each of the elements
What is the value of this result?
Understand how components affect performance of this circuit
Compare performance of different circuits for architecture selection
Transfer Function Sensitivities
T s T jω
S
x S x
s=jω
T jω T jω θ T jω
S x S x jθS θ
x
where
S T jω
x =Re S T jω
x
S θ
x
1
= Im
θ
S T jω
x
Transfer Function Sensitivities
a s i
i
N s
If T(s) is expressed as T s = i=0
n
D s
b s
i=0
i
i
m n
ais Si ai
bisiSbxi
STx s i 0
x
then - i 0
N s D s
N0 s xN1 s
If T(s) is expressed as T s =
D0 s xD1 s
T jω
ωC ω
Want ωC x
S ωC
x
x ωC
Band-edge Sensitivities
T jω
ωC ω
T jω
ω
ωC ω
Proof:
Observe
T jω T jω
ω ω
T jω
T jω T jω x x
ω x ω ω
x
Band-edge Sensitivities
T jω
T jω
T jω T jω x
T jω
x ω
ω x ω ω ωC ω
x
T jω
ω x
x T jω T jω
ω
S
Sωx Tx jω
T jω x
Sω
ω x T jω ω
x T jω ω x
ω
T jω SxT jω ω=ω
Sωx C
T jω
C
T jω x S
ω
ω x x T jω ω=ωC
x ω T jω ω
ω T jω
Sensitivity Comparisons
Consider 5 second-order lowpass filters
(all can realize same T(s) within a gain factor)
C1
R L
VOUT R1 R2
C VIN K
VIN VOUT
C2
(a) (b)
C1
R1 RQ
C1 C2 R4
R3 R1 R2 R0 R2 R3
C2
VIN VIN
VOUT
R3
R1 R2 R4 R5
VOUT
VIN R5
C1 C2 -K = -
R4
-KRC Lowpass
(e)
For all 5 structures, will have same transfer function within a gain factor
Kω02
T s =
2 ω0
s +s +ω02
Q
a) – Passive RLC
R L
VOUT
VIN C
1
VOUT
T s = = LC
VIN 2 R 1
s +s +
L LC
1 1 L
ω0 = Q=
LC R C
b) + KRC (a Sallen and Key filter)
C1
R1 R2
VOUT
VIN C2
R5
R4
R5
K = 1+
R4
K
R1R2C1C2
T s =
1 R1C1 R 2 C2 R1C2 R1C1 1
s +s
2
+ + K +
R1R2C1C2 R 2 C2 R1C1 R2C1 R2C2 R1R2C1C2
1
Q=
1 R1C1 R2C2 R1C2 R1C1
ω0 = + + K
R1R2C1C2 R C R C R C R C
2 2 1 1 2 1 2 2
Case b1 : Equal R, Equal C
R1 = R2 = R C1 = C 2 = C
1 1 Kω02
ω0 =
RC K=3- T s =
Q ω
s2 +s 0 +ω02
Q
Case b2 : Equal R, K=1
1 C1
R1 = R2 = R Q=
2 C2
c) Bridged T Feedback C1
R3 R1 R2
C2
VIN
VOUT
1
R1R3C1C2
T s =
C 1 R1 R2 R1R2 1
s +s
2 2
+ + +
R3
C1 R1R 2C1C2 R1 R3 R1R2C1C2
1 1
ω0 = Q=
R1R2C1C2 C2 R1 R R1R2
+ 2+
C1 R3 R1 R3
If R1=R2=R3=R and C2=9Q2C1
1
9Q2R2C12
T s =
1 1
s2 +s 2 +
3Q RC1 9Q2R2C12
d) 2 integrator loop
R1 RQ
C1 R4
R2 C2
R0 R3
VIN
VOUT
R4 1
R4 1
T s = -
R3 R0R 2C1C2
0 =
1 R4 1 R3 R0R2C1C2
s2 +s +
RQC2 R3 R0R 2C1C2
RQ C2
Q=
R0R 2 C1
For: R0 = R1 = R2 = R C1 = C 2 = C R3 = R4
1
2 2
T s = - R C
1 1
s2 +s + 2 2
Q RC
R C
1
ω0 =
RQ =QR RC
d) - KRC (a Sallen and Key filter)
R3
R1 R2 R4 R5
VOUT
VIN R5
C1 C2 -K = -
R4
K
R1R2C1C2
T s = -
R 1 C2 1 1 1+ R1 R3 1 K + R1 R 4 1 R2 R3 R2 R1
s2 +s 1+ 1 1+ +
R 3 1 1
R C C1 R 2 2
C R 4 2
C R1R2C1C2
5+K 0
Often R1=R2=R3=R4=R, C1=C2=C Q=
5
Stay Safe and Stay Healthy !
End of Lecture 20