1
vs. f
H(ejωT)
3. SC realization
(1) H0(z)= z +1
z − 0.9063
CD
Vin |
|
+ |
-1/CE |
C |
-1 |
|
|
1-Z |
+ |
S |
(1+Z-1) |
|
2 |
|
CS/2
fp
0 f fp
H0(z)=- |
Cs / 2 |
× |
z +1 |
|
z −CE /(CD +CE ) |
|
CD +CE |
Cs=2 arbitrarily chosen φ1=>CE=0.9063 , CD=0.0937
φ2
Vout
(2) H1(z)= |
|
z2 |
+C z +1 |
Q1= |
(a 2 |
+b |
2 ) |
12 |
|
Low-Q |
|
|
1 |
|
|
|
1 |
1 |
|
|
0.99 |
(1/ f |
)z2 |
+ (e |
/ f |
)z +1 |
|
|
2 |
a1 |
|
|
|
|
|
|
|
|
|
|
1 |
|
1 |
1 |
|
|
|
|
|
|
|
|
|
|
The SCF is shown on P.14-21.
The component values are: C1"=a0=1, C1'=a2-a0=0,
C2=C3=
b1 +b2 +1 =
(e1 +1) / f1 +1 0.12436,
14-61 CHUNG-YU WU
C1=(a0+a1+a2)/C3 0.30358,
C4=b2-1=1/f1-1 0.12939, CA=CB=1.
(3) H2(z)= |
|
|
z2 +C |
2 |
z +1 |
Q2= |
(a 2 |
+b |
2 ) |
12 |
|
4.33 =>High-Q |
|
|
|
|
|
2 |
2 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
( 1 |
f |
)z2 + (e2 f2 )z +1 |
|
|
2 |
|
a2 |
|
|
|
|
|
|
|
|
|
|
|
|
|
2 |
|
|
|
|
|
|
|
|
|
|
|
|
The SCF is shown on P.14-23. |
|
|
|
|
|
|
|
|
|
The component values are: |
|
|
|
|
|
|
|
|
|
C1"= a2 / b2 |
= f2 0.96845 |
C1 '= (a1 −a0 ) / b2 c3 |
= 0, |
C2 = C3 =
(1+b1 +b2 ) / b2 =
f2 +e2 +1 0.13795,
C1=(a0+a1+a2)/b2C3=(2+c2)f2/C3 0.58645, C4=(1-1/b2)/C3=(1-f2)/C3 0.22873.
(4)Overall SCF
* Ho (low-pass linear section) is placed first
=>High-frequency out-of-band signals and input noise can be attenuated. The antialiasing filter preceding the SCF has a lower requirement.
* H2 (high-Q section) is placed to the center=>good signal-to-noise ratio
|
|
|
|
CD |
φ1 |
|
|
|
|
C2 |
|
φ1 |
|
CS/2 |
CE |
φ |
|
|
|
|
C4 |
|
φ2 |
|
|
|
|
2 |
|
|
|
|
|
|
V |
φ |
C |
S |
φ |
|
C |
|
|
|
|
CB |
|
1 |
|
2 |
φ |
φ |
C |
φ |
C3 |
|
|
in |
|
|
|
|
|
|
|
|
A1 |
2 |
1 |
2 |
A |
φ2 |
|
|
|
φ2 |
|
|
|
A2 |
1 |
|
|
|
|
φ1 |
|
φ1 |
φ1 |
|
φ2 |
A3 |
A |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
φ1 |
|
|
|
|
|
|
|
|
|
C1" |
|
|
|
|
SECTION 1 |
|
SECTION 2 (HIGH Q) |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
C2 |
|
φ1 |
|
|
|
|
C |
|
C4 |
|
φ |
|
φ2 |
φ2 |
C1 |
A |
|
|
CB |
2 |
|
φ2 |
φ1 |
C3 |
|
|
A |
|
|
φ2 |
|
|
|
A4 |
|
|
|
|
|
|
|
φ2 |
A5 |
Vout |
|
|
|
|
|
|
|
φ1 |
φ1 |
φ2 |
|
|
|
φ1 |
|
|
|
C1"
Thought the same procedures, we have CA 17.666,CB 7.7286 C1 1.9926,C2 =1
C3 2.1116,C4 = 2
C1" 6.3085
5 Final Design
Cmin is chosen as 0.5pF => C=1
op amp: gain 70dB bandwidth 3 MHz passband sensitivity to capacitance variation
§14-12.2 Bilinear Ladder SCF Design
1. The same filter specification.
Elliptic ladder filter is chosen(fifth-order). The result is
14-62 CHUNG-YU WU
4.Scaling
1)Vp1=occurs at dc where H0(1)=-CS/CD=-21.345
(1)We want an overall passband gain of 1. =>Ho(1)→-1 => CD=CS=2, CE 19.345, C1"20.672, C2 12.518
(Multiplying all capacitors connected or switched to the output node of op-amp A1 by 21.345)
(2)All capacitors at the input node of A1 should be scaled so that the
smallest (Cs |
2 |
) equals 1. (O.K.) |
|
|
2)Vp2 (peak output voltage of op-amp A2) occurs around fp2=1.10kHz
(1)Vp2 177.05 for Vin=1
Reducing Vp1/Vin to 1
=>CA and C3 are multiplied by 177.05=> CA 177.05, C3 24.424.
(2) Vp3 180.80 at 1.07kHz
=>CB,C2, and C4 are multiplied by 180.80=>CB 180.80, C2 24.941, C4 41.354.
(3)Minimize total capacitance=>C1, C2, C4, and CA at the input node of op-amp A2 are scaled to make C1=1
=>C1=1, C2 1.9926, C4 3.3036, CA 14.144
(4)Similarly, C1"=1, C3 1.1815, CB 8.7466. (The input of A3)
3)Vp4 503.57 and Vp5 230.14
0.2dB/1%
14-63 CHUNG-YU WU
|
RS |
L2 |
L4 |
|
|
|
|
|
|
|
Vin |
C1 |
C2 C3 |
C4 |
C5 |
RL |
Normalized component values:
Rs=RL=1 C1=0.85535 C2=0.15367 L2=1.20763 C3=1.48438 C4=0.46265 L4=0.89794 C5=0.63702
ŵap=1 rad/s
2. Frequency prewarping and denormalization
ω |
|
2 |
tan |
ωpT |
= 2 f |
|
tan |
πf p |
6291.4667rad / s |
|
2 |
|
|
ap |
|
T |
|
c |
|
fc |
Multiplying each resistor by z0, each inductor by L0=z0/ωap, and each capacitor by C0= 1z0 ωap. 50Ω
Usually choose z0=real source and termination resistance 100Ω 600Ω
Here, C0=1 is chosen => z0= |
1 |
and L0= |
1 |
|
ω |
ap |
ω |
|
2 |
|
|
|
|
|
|
|
ap |
We have the denormalized element values as:
C1=0.85535, C2=0.15367, L2=1.20763×Lo=3.05090×10-8,
C3=1.48438, C4=0.46265, L4=0.89794×Lo=2.26851×10-8,
C5=0.63702, Rs=RL=z0=1.58945×10-4
3. SC realization
Using the exact design technique of SC ladder filter (Section 14-10), the state equations are
-V = − |
|
|
1 |
( |
1 |
(V |
−V ) − I |
|
+ sC' |
V ), |
|
|
|
|
|
2 |
1 |
|
|
|
|
|
|
|
|
|
in |
1 |
2 |
3 |
|
|
|
|
|
sC'1 RS |
|
|
|
|
-I2=-( |
|
|
1 |
|
− sCL2 ) (V1 V3), |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
sL2 |
|
|
|
|
V3= |
|
1 |
|
|
(−I2 − sC'2 V1 − sC4 'V5 + I4 ), |
|
sC'3 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
14-65 |
|
4.Scaling |
|
|
|
|
CHUNG-YU WU |
|
|
|
|
|
|
|
Vin=1V, we have: |
Vp1 0.92V, |
|
C1=1.00000 |
C05=1.14172 |
|
A1: CA,C2,C21, and C02 |
|
|
Vp2 |
34V, |
|
C2=1.83854 |
C06=1.52861 |
|
multiplied by Vp1 |
|
|
Vp3 |
0.764V, |
C3=2.00000 |
C07=2.02212 |
|
A2: C3,C01, and C03 |
|
Vp4 |
28.86V, |
CA=13.87171 |
C08=1.00000 |
|
multiplied by Vp2 |
|
Vp5 |
0.5V, |
|
C01=1.77112 |
CE=8.27441 |
|
|
|
|
|
|
|
|
C02=1.20275 |
CD=14.43078 |
|
|
|
|
|
C03=1.00000 |
CL=1.00000 |
|
|
|
|
|
C04=1.00000 |
C41=2.09575 |
|
|
|
|
|
CB=11.11901 |
C42=5.67396 |
|
for dynamic range scaling |
|
|
|
Cc=14.46156 |
C21=1.29480 |
|
|
|
|
|
C22=1.90667 |
|
minimum-capacitance scaling: |
Cs |
|
=13.87171 |
|
CA |
2 |
|
|
|
|
|
|
|
5.Final design
Cmin , OP amp: 70dB 3 MHz
=>Passband ripple: 0.06dB minimum stopband loss 39.5dB
Maximum sensitivity: 0.05dB %
§14-12.3 LDI Ladder SCF Design
1. LCR prototype circuit
Fifth-order elliptic LC ladder filter with the same lowpass specifications.
2. Frequency prewarping and denormalization ωap ωp (for simplicity)
|
z0=1Ω => C0= 1 |
3 |
) |
F, L0= |
1 |
H |
|
(2π103 ) |
|
(2π10 |
|
|
|
The denormalized element values:
Rs=1Ω,
C1=136.13318µF, C4=73.633034µF, C2=24.45734µF, L4=142.91159µH, L2=192.20028µH, C5=101.38488µH, C3=236.24641µF, RL=1 Ω.
State equations: |
|
−V1 +Vin + sC V |
|
|
|
-V = |
|
1 |
|
( |
− I |
|
), |
s(C |
+C |
) |
|
1 |
|
Rs |
2 3 |
|
2 |
|
|
1 |
2 |
|
|
|
|
|
|
|
-I2= V3 −V1 , |
|
|
|
|
sL2 |
|
|
|
|
|
|
|
1 |
|
|
|
|
V3= − s(C2 +C3 +C4 ) ( − I2 − sC2V1 − sC4V5 + I4 ), |
I4= V3 −V5 , |
|
|
|
|
|
sL4 |
|
|
|
|
|
-V5= − |
s(C4 |
1 |
(I4 |
+ sC4V3 |
− |
V5 ). |
|
+C5 ) |
|
|
|
RL |
14-66 CHUNG-YU WU
3.SCF design
The flow diagram is shown on P.14-? whereas the active-RC circuit is given on P.14-?.
The SCF is shown on P.14-? where T=20µs is chosen and the component values are
C1+C2=160.59µF, |
C2+C3+C4=334.34µF, |
CS= |
|
T |
=20µF, |
C4+C5=175.018µF, |
|
|
|
|
|
Rs |
|
|
|
C= |
T |
|
=20µF, |
CL= |
T |
=20µf |
|
|
|
1 |
|
|
|
|
RL |
4.Scaling
Dynamic range scaling with Vpi listed: followed by minimum-capacitance scaling
Element values: |
SCF: |
C1=8.03214, |
|
C3=12.97271, |
C1A=1, |
C3A=1.08390, |
C1B=1.07930, C3B=1, |
C1C=1.29263, C3C=1.02540, |
C1D=1.13212, C3D=1.66885, |
C2=13.42236, C4=15.76379, |
C2A=1.08053, C4A=1.71203, |
C2B=1, |
|
C4B=1, |
|
|
C5=8.75121, |
|
|
C5A=2.20614, |
|
|
C5B=1, |
|
|
C5C=6.29664. |
5.Final design
Cmin Passband ripple: 0.095dB>0.044dB Minimum stopband loss: 40.5dB OP amp: 70dB, 3MHz
Maximum passband sensitivity: 0.08dB
%
for A1:Vp1=0.927 V at 1.182kHz. for A2:Vp2=1.198 V at 1.121 kHz. for A3:Vp3=0.857V at 1.061 kHz. for A4:Vp4=1.105 V at 1.061 kHz. for A5:Vp5=0.501 V at 967 kHz.
14-67 CHUNG-YU WU
§14-13 Nonideal Effects in Switched-Capacitor Filters
1. Switch Turn-On Resistance
The turn-On resistance of a MOSFET can be written as
|
Ron= |
|
uco |
|
1 |
|
|
2 |
w |
(VG S |
−VT ) |
|
φ |
2 |
L |
|
|
|
|
|
Vin |
|
|
|
C |
* Nonlinear behavior |
|
|
|
|
|
|
signal voltage -Vss |
|
|
|
|
|
|
|
|
|
|
The Ron effect on the simple SC integrator:
|
|
|
|
C2 |
|
|
φ 1 |
φ 2 |
|
|
|
|
|
φ1 |
|
Vin |
|
- |
|
|
C1 |
+ |
Vout |
|
|
|
|
|
φ2 |
|
|
|
|
t=nT t=(n+1)T
C2
φ1
R1 R2
-φ2
|
+ |
V1 |
C1 |
+ |
Vout |
|
|
t=nT |
nT+T/2 (n+1)T |
|
- Vin |
|
|
|
|
At t=nT , V1(t)=V1(nT)=Vin(nT)(1-e −T
2R1C1 )
Assume φ1 |
and φ2 are activated for T |
2 |
. |
|
|
|
|
|
|
|
|
|
|
∆Q (nT+ |
T |
) =C1V1(nT )(1-e −T / 2R2C1 )=[Vout(nT+T)-Vout(nT)]C2 |
2 |
Let R1=R2=R |
|
|
|
|
=> H(z)= |
−(1 −e−T 2 RC1 )2 |
C C |
|
|
|
|
|
z −1 |
1 |
2 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Ideal: |
|
|
H(z)=- |
C1 C2 |
|
|
|
|
|
|
|
z −1 |
|
|
|
|
|
|
|
|
|
|
|
|
|
14-68 CHUNG-YU WU
Error: ε=1-(1-e −T / 2RC1 )2 2e −T / 2RC1
Usually ε<0.1%(cap. ratio error) is acceptable.
2e −T / 2RC1 ≤10−4
|
=> |
RC1 |
=RC1fc ≤ |
1 |
0.05 |
|
T |
2ln 20000 |
|
|
|
|
or RC1 ≤ 20T (= 201fc)
fc=500KHz, C1=5pF R ≤ 20KΩ ; fc=100MHz, C1=2pF, R ≤ 250Ω?
2.Clock Feedthrough Noise
*All switches directly connected to the integrating node generate clock feedthrough noises.
φ2
n


T
* All clock feedthrough noises are proportional to the sampling frequency. They may have a dc component.
* As soon as the clock feedthrough error voltage does not
14-69 CHUNG-YU WU
saturate the OP AMP, it can be eliminated at the output by reconstruction filters(LPF).
* The dc component cause offset voltage problems.
3.Junction Leakage
* Worst-case (100°C or 125°C) leakage at the integrating node: ~10 nA/mil2 5µm×5µm junction => 400 pA leakage
*fs, max is about 25KHz in this case to avoid significant errors.
*The leakage cause dc offset voltages.
4.DC offset Voltage of the OP AMP
+
pratical op amp
C2
φ2 φ2
-
C1 +
+
-
Voff
+
+ ideal op amp
-
Voff Voff = 5~20mv
Vout
* Vout=(1+ C1 ) Voff
C2
*Integrator-based design may have a dc offset problem if no other negative feedback paths exist.
*Too-low-frequency operation is not good.
C2
5. Finite Gain of the OP AMP. |
|
φ2 |
|
|
|
φ2 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Vout(nT)=Vc2(nT)- |
|
1 |
Vout(nT) |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
- |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
A |
|
|
|
C1 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
O |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Vout |
C2[Vc2(nT)-Vc2(nT-T)] |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
+ |
|
|
|
|
|
|
φ1 |
|
φ1 |
|
|
|
|
|
|
|
|
|
|
|
A0 |
+C1[Vin(nT)+ |
1 |
|
Vout(nT)]=0 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
A |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
O |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
V |
(z) |
|
|
|
|
−(C |
C |
)[1 |
+ (1 +C |
C |
) / A |
]−1 z |
|
|
|
|
=>H(z)= |
|
out |
|
= |
|
1 |
2 |
|
1 |
2 |
|
O |
|
|
|
|
|
|
Vin (z) |
|
z −(1 + 1 |
AO |
) /[1 + (1 +C |
C ) / A ] |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
2 |
|
O |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|