239

Reference:C:\Documents and Settings\Murmann\My Documents\lib\Mathcad\defaults.mcd

2-Stage OTA Design Example ("Simple Approach")

Cf

Cs CL

- +
Vsd Vid Vod
+ -
Cs
CL

Cf

M4a,b

M1a,b

M3a,b M2a,b

(Miller compensation, neutralization caps and CMFB not shown)

240

Technology Data
2
VDD := 3V γ :=
3

gm/ID, gm/Cgg lookup tables:

Reference:C:\Documents and Settings\Murmann\My Documents\teaching\ee214_autumn07\otaexample\gm

Design Objectives

fc := 200MHz PM := 75deg DR := 72 (dB) G := 2 ε s := 0.5%

Given Component Values

Cs
Cs := 400fF CL := 200fF Cf :=
G

Design Decision: Channel lengths

Return factor estimate assuming Cgg1 is ~equal to sum of feedback caps

Cf
Cgg1 := Cs + Cf β est := β est = 0.167
Cf + Cs + Cgg1

1 3
avo := avo = 1.2 × 10
ε s⋅ β est

For simplicity, assume that each stage contributes same gain (not necessarily accurate/optimal)

avostage := avo avostage = 34.641

For simplicity, assume that intrinsic gain of both signal path devices is 2x stage gain (not
necessarily accurate/optimal). Intrinsic gain requirements:

gmro := 2avostage gmro = 69.282

肯定是先从增益着手，选取L
Based on gmro design charts, we decide:

L1 := 0.55μm L2 := 0.4μm

For simplicity, choose same lengths for pmos/nmos loads (not necessarily optimal)

L3 := L2 L4 := L1
 241

Design Decision: gm/ID

S S
gmID1 := 10 gmID2 := 10
A A

Design Decision: Relative gm/ID of active loads, output swing

Ratio of gm/ID in load devices vs. signal path (often want g1,2 to be <1, to minimize noise and
Cjunction; however, very small g1,2 may impose DR limitations)

gmID3 gmID4
g 31 = g 31 := 1 g 42 = g 42 := 1
gmID1 gmID2

Allocated Vdsmin for M2 and M4 and resulting output swing:

VDSmin := 500mV Vodmax := VDD − 2 ⋅ VDSmin (Differential peak amplitude)

从动态范围入手，相当于从噪声入手，比噪声有用
Step 1: Calculate Cc based on DR spec

(
C2 := CL + 1 − β est ⋅ Cf ) C2 = 366.667 fF Total stage 2 load

2
0.5⋅ Vodmax
Ntot := Ntot = 355.234 μV
DR
10
10

⎡ 1 k B ⋅ Tr k B ⋅ Tr ⎤
Ntot = 2 ⋅ ⎢ ⋅ (
⋅ γ ⋅ 1 + g 31 +
CL ⎣
) ( )
⋅ ⎡γ ⋅ 1 + g 42 + 1⎤ ⎥
⎦
⎣ β Cc ⎦

( )
1
⋅ k B⋅ Tr⋅ γ ⋅ 1 + g 31
β est
Cc := 2 ⋅ Cc = 0.659 pF
k B ⋅ Tr
Ntot −
C2 ⎣ ( )
⋅ ⎡γ ⋅ 1 + g 42 + 1⎤
⎦
242 两个思路：先得到补偿网络Cc，就能算出gm；
如果先从电流入手，则是从gm推出Cc；
其实就是先算补偿网络还是先算前馈电路的问题

Step 2: Calculate transconductances

1
g m1 := ⋅ 2 ⋅ π ⋅ fc⋅ Cc g m1 = 4.966 mS
β est

k := tan⎛⎜ π ⋅ ⎞ k = 3.732
PM
⎟
⎝ 180deg ⎠

g m2 := k ⋅ 2 ⋅ π ⋅ fc⋅ C2 (approx.) g m2 = 1.72 mS

Step 3: Find ID, W

g m1 g m2
ID1 := ID2 := ID1 = 0.497 mA ID2 = 0.172 mA
gmID1 gmID2

( ) ( )
A A
IDW1 := pidw L1 , gmID1 IDW2 := nidw L2 , gmID2 IDW1 = 1.947 IDW2 = 8.893
m m

ID1 ID2
W1 := W2 := W1 = 255.015 μm W2 = 19.336 μm
IDW1 IDW2

Calculate the width ratios (w31 = W3/W1) for the chosen gm/ID

w31 :=
(
pidw L1 , gmID1 ) w31 = 0.219
(
nidw L3 , gmID1 ⋅ g 31 )

w42 :=
(
nidw L2 , gmID2 ) w42 = 4.567
(
pidw L4 , gmID2 ⋅ g 42 )
W3 := W1 ⋅ w31 W4 := W2 ⋅ w42 W3 = 55.835 μm W4 = 88.313 μm

55 88
= 18.333 = 14.667
3 6
 243

OTA circuit for simulation (feedback network, CMFB, Cc, etc. not shown)

85uA 85uA 1020uA 170uA 170uA

M4c M4a,b
15/0.55 15/0.55
15/0.55
M=12 M=2
M4d

M1a,b
255/0.55

18/0.4 18/0.4
M3d M3c
M3a,b
18/0.4 M2a,b
M=6 19/0.4

Simulation results (first run, without any tweaking):

f c=130.75MHz, PM=55.78deg, T0=245

Magnitude [dB]

80
60
40
20
0
-20
-40 -2 0 2 4
10 10 10 10
f [MHz]
Phase [degrees]

0
-50
-100
-150
-2 0 2 4
10 10 10 10
f [MHz]
244

PSD [V2/Hz]

-20
10
5 10
10 10
f [Hz]
Sqrt(Integral) [μVrms]

Integral=417.87uVrms, DR=70.59dB (for Vodmax=2.00V)

500

0 5 10
10 10
f [Hz]

Loop crossover and phase margin are way off! Why? Look at .op output:
离得很远
gm1 = 5.12mS
gm2 = 1.88mS

Not bad. What's wrong?

Junction caps are fairly large and comparable to other caps. E.g. Cdb1~300fF; this will
significantly impact nondominant pole.

Cgg1=688fF, means that beta is smaller than what we had budgetd above (beta,est)

....

Bottom line: Hard to get "reasonable" matching between hand calculations and Spice using
simplified expressions and ignoring junctions caps. Let's fix this...
 245

subckt x1 x1 x1 x1 x1 x1
element 1:m1a 1:m1b 1:m2a 1:m2b 1:m3d 1:m3c
model 0:pch214 0:pch214 0:nch214 0:nch214 0:nch214 0:nch214
region Saturati Saturati Saturati Saturati Saturati Saturati
id -515.7219u -515.7219u 209.9559u 209.9559u 85.0000u 100.5239u
ibs 0. 0. 0. 0. 0. 0.
ibd 0. 0. 0. 0. 0. 0.
vgs -1.0731 -1.0731 795.4179m 795.4179m 726.2152m 726.2152m
vds -1.2776 -1.2776 1.4843 1.4843 726.2152m 1.9455
vbs 926.9464m 926.9464m 0. 0. 0. 0.
vth -924.8901m -924.8901m 594.7855m 594.7855m 602.4640m 590.0494m
vdsat -186.1804m -186.1804m 162.1420m 162.1420m 116.5494m 123.5860m
vod -148.1634m -148.1634m 200.6324m 200.6324m 123.7513m 136.1659m
beta 31.1696m 31.1696m 11.1159m 11.1159m 10.5036m 10.4989m
gam eff 458.3707m 458.3707m 894.1238m 894.1238m 894.1238m 894.1238m
gm 5.1229m 5.1229m 1.8862m 1.8862m 1.1167m 1.2474m
gds 67.8360u 67.8360u 20.4147u 20.4147u 14.0027u 12.6163u
gmb 903.7574u 903.7574u 495.7571u 495.7571u 298.8398u 328.6908u
cdtot 419.4667f 419.4667f 25.7526f 25.7526f 28.4889f 22.8775f
cgtot 681.3669f 681.3669f 37.5110f 37.5110f 34.9915f 35.1430f
cstot 938.5513f 938.5513f 69.9117f 69.9117f 65.4864f 65.6493f
cbtot 767.3791f 767.3791f 69.2340f 69.2340f 69.7566f 64.1621f
cgs 528.6546f 528.6546f 27.3328f 27.3328f 25.2053f 25.3129f
cgd 122.2292f 122.2292f 4.3896f 4.3896f 4.1602f 4.1588f

subckt x1 x1 x1 x1 x1 x1
element 1:m3b 1:m3a 1:m4d 1:m4c 1:m4b 1:m4a
model 0:nch214 0:nch214 0:pch214 0:pch214 0:pch214 0:pch214
region Saturati Saturati Saturati Saturati Saturati Saturati
id 515.7219u 515.7219u -100.5239u -1.1886m -209.9559u -209.9559u
ibs 0. 0. 0. 0. 0. 0.
ibd 0. 0. 0. 0. 0. 0.
vgs 726.2152m 726.2152m -1.0545 -1.0545 -1.0545 -1.0545
vds 795.4179m 795.4179m -1.0545 -926.9464m -1.5157 -1.5157
vbs 0. 0. 0. 0. 0. 0.
vth 601.7593m 601.7593m -739.0005m -740.0639m -735.1556m -735.1556m
vdsat 116.9465m 116.9465m -311.3437m -310.5103m -314.3549m -314.3549m
vod 124.4559m 124.4559m -315.5066m -314.4432m -319.3516m -319.3516m
beta 63.0200m 63.0200m 1.8139m 21.7668m 3.6278m 3.6278m
gam eff 894.1238m 894.1238m 472.5709m 472.5709m 472.5709m 472.5709m
gm 6.7553m 6.7553m 559.8132u 6.6300m 1.1578m 1.1578m
gds 81.4843u 81.4843u 10.9880u 147.0550u 17.4548u 17.4548u
gmb 1.8061m 1.8061m 127.5074u 1.5107m 263.4207u 263.4207u
cdtot 167.9230f 167.9230f 29.4059f 361.5686f 54.5446f 54.5446f
cgtot 210.0083f 210.0083f 39.8665f 478.3955f 79.7341f 79.7341f
cstot 392.9864f 392.9864f 66.0755f 792.9036f 132.1528f 132.1528f
cbtot 415.5359f 415.5359f 62.0693f 753.5347f 119.8678f 119.8678f
cgs 151.2783f 151.2783f 30.0686f 360.8146f 60.1421f 60.1421f
cgd 24.9604f 24.9604f 7.0195f 84.2342f 14.0390f 14.0390f
246

Reference:C:\Documents and Settings\Murmann\My Documents\lib\Mathcad\defaults.mcd

2-Stage OTA Design Example ("Accurate Approach")

Cf

Cs CL

- +
Vsd Vid Vod
+ -
Cs
CL

Cf

M4a,b

M1a,b

M3a,b M2a,b

(Miller compensation, neutralization caps and CMFB not shown)

 247

Technology Data
2
VDD := 3V γ := Lmin := 0.35μm
3
k dbn := 0.65 k dbp := 0.8 (approximate ratios of Cdb/Cgg for minimum length device)

gm/ID, gm/Cgg lookup tables:

Reference:C:\Documents and Settings\Murmann\My Documents\teaching\ee214_autumn07\otaexample\gm

Design Objectives

fc := 200MHz PM := 75deg DR := 72 (dB) G := 2 ε s := 0.5%

Given Component Values

Cs
Cs := 400fF CL := 200fF Cf :=
G

Design Decision: Channel lengths 这一步其实可以用lambda曲线，不用intrinsic gain

Return factor estimate assuming Cgg1 is ~equal to sum of feedback caps

Cf
Cgg1 := Cs + Cf β est := β est = 0.167
Cf + Cs + Cgg1

1 3
avo := avo = 1.2 × 10
ε s⋅ β est

For simplicity, assume that each stage contributes same gain (not necessarily accurate/optimal)

avostage := avo avostage = 34.641

For simplicity, assume that intrinsic gain of both signal path devices is 2x stage gain (not
necessarily accurate/optimal). Intrinsic gain requirements: 设计余量

gmro := 2avostage gmro = 69.282

Based on gmro design charts, we decide:

L1 := 0.55μm L2 := 0.4μm
248

For simplicity, choose same lengths for pmos/nmos loads (not necessarily optimal)

L3 := L2 L4 := L1

Design Decision: Relative gm/ID of active loads, output swing 同一个branch的Vov相同，matching

Ratio of gm/ID in load devices vs. signal path (often want g1,2 to be <1, to minimize noise and
Cjunction; however, very small g1,2 may impose DR limitations)

gmID3 gmID4
g 31 = g 31 := 1 g 42 = g 42 := 1
gmID1 gmID2

Allocated Vdsmin for M2 and M4 and resulting output swing:

VDSmin := 500mV Vodmax := VDD − 2 ⋅ VDSmin (Differential peak amplitude)

Usually have to leave margins for PVT and gain drop at swing. Depending on final design outcome
(optimum gm/ID), it may be worth re-visiting this assumtion.

ITERATION

Adjust following parameters to minimize total current computed in step 4

Cgg1
c1 = (good starting point: c1=1) c1 := 1 设初值
Cs + Cf
Cgg2
c2 = (good starting point: c2=1) c2 := 1
CL

Step 1: Estimate/calculate return factor and all capacitances

先算所有的cap
Cf
β := β = 0.167
(Cf + Cs)⋅ (1 + c1)
(
Cgg1 := c1 ⋅ Cs + Cf ) Cgg1 = 600 fF

Cgg2 := c2 ⋅ CL Cgg2 = 200 fF

 249

Lmin
Cdb1 := k dbp ⋅ Cgg1 ⋅ Cdb1 = 305.455 fF
L1

To find Cdb3, we estimate the width ratio w31 = W3/W1, assuming gm/ID1=10S/A (not known at
this point; also won't matter much...)

pidw⎛⎜ L1 , 10
S⎞
⎟
w31 :=
⎝ A⎠
w31 = 0.219 Cdb3 := w31⋅ Cdb1 ⋅
k dbn
Cdb3 = 54.338 fF
nidw⎛⎜ L3 , 10 ⋅ g 31⎞⎟
S k dbp
⎝ A ⎠

Lmin
Cdb2 := k dbn ⋅ Cgg2 ⋅ Cdb2 = 113.75 fF
L2

nidw⎛⎜ L2 , 10
S⎞
⎟
w42 :=
⎝ A⎠
w42 = 4.567 Cdb4 := w42⋅ Cdb2 ⋅
k dbp
Cdb4 = 639.428 fF
pidw⎛⎜ L4 , 10 ⋅ g 42⎞⎟
S k dbn
⎝ A ⎠

C1 := Cdb1 + Cdb3 + Cgg2 C1 = 559.793 fF Total stage 1 load

C2 := CL + Cdb2 + Cdb4 + ( 1 − β ) ⋅ Cf
3
C2 = 1.12 × 10 fF Total stage 2 load

Step 2: Calculate Cc based on DR spec

根据噪声（或者说DR）算补偿电容
2
0.5⋅ Vodmax
Ntot := Ntot = 355.234 μV
DR
10
10

⎡ 1 k B ⋅ Tr k B ⋅ Tr ⎤
Ntot = 2 ⋅ ⎢ ⋅ (
⋅ γ ⋅ 1 + g 31 + ) (
⋅ ⎡γ ⋅ 1 + g 42 + 1⎤ ⎥
⎣ ⎦ )
⎣ β Cc CL
⎦

( )
1
⋅ k B⋅ Tr⋅ γ ⋅ 1 + g 31
β
Cc := 2 ⋅ Cc = 0.56 pF
k B ⋅ Tr
Ntot −
C2 ⎣ ( )
⋅ ⎡γ ⋅ 1 + g 42 + 1⎤
⎦

(Note: for low DR designs, Cc is not neccesarily set by noise and becomes part of the optimization
process (e.g. add variable c3=cc/CL to the iteration loop)
250

Step 3: Calculate transconductances and transit frequencies 根据GBW计算gm1，再根据相位裕度计算

非主极点与主极点的倍数关系，从而计算
1
g m1 := ⋅ 2 ⋅ π ⋅ fc⋅ Cc g m1 = 4.222 mS 出gm2，有了两个跨导值，根据先前的电
β
容计算两个MOS管的本征频率
k := tan⎛⎜ π ⋅ ⎞ k = 3.732
PM
⎟
⎝ 180deg ⎠

⎛ C2⋅ C1 ⎞
g m2 := k ⋅ 2 ⋅ π ⋅ fc⋅ ⎜ + C1 + C2⎟ g m2 = 13.128 mS
⎝ Cc
⎠
1 g m1
fT1 := ⋅ fT1 = 1.12 GHz
2 ⋅ π Cgg1

1 g m2
fT2 := ⋅ fT2 = 10.447 GHz
2 ⋅ π Cgg2
根据计算出的本征频率来确定对应的gm/id，
再计算id
Step 4: Find gm/ID, ID

( ) ( )
1 1
gmID1 := pgmid L1 , fT1 gmID2 := ngmid L2 , fT2 gmID1 = 10.404 gmID2 = 6.21
V V
g m1 g m2
ID1 := ID2 := ID1 = 0.406 mA ID2 = 2.114 mA
gmID1 gmID2

IDtotal := ID1 + ID2 IDtotal = 2.52 mA

This design (with c1=1, c2=1 -> IDtotal=2.52mA) will certainly work, but there's lots of room for
power optimization. With the above script, it is quite easy to iteratively step c1 and c2 up/down to
search for a power minimum.

For c1=1, c2=1, it is interesting to note that there is lots of self-loading (The Cdb's present a large
fraction of the caps that set fc and fp2). Hence, going to smaller c1 and/or c2 may help lower
power.

Another try with c1=0.5, c2=1 yields IDtotal=1.819mA. Making the pmos smaller reduces its
contributed capacitance faster than gm/ID drops; beta also improves for small c1. Hence, there is
a net reduction in power.

Similarly, using yet another try with c1=0.5, c2=0.5 yields IDtotal=1.497mA.

Manual iterations are very useful for developing intuition; but it is also possible to automate the
search process using an optimization function:
 251

Cf
( )
f c1 , c2 := β←
(Cf + Cs)⋅ (1 + c1)
Cgg1 ← c1 ⋅ ( Cs + Cf)

Cgg2 ← c2 ⋅ CL

Lmin
Cdb1 ← k dbp ⋅ Cgg1 ⋅
L1

pidw⎛⎜ L1 , 10
S⎞
⎟
Cdb3 ←
⎝ A⎠
⋅ Cdb1 ⋅
k dbn

nidw⎛⎜ L3 , 10 ⋅ g 31⎞⎟
S k dbp
⎝ A ⎠
Lmin
Cdb2 ← k dbn ⋅ Cgg2 ⋅
L2

nidw⎛⎜ L2 , 10
S⎞
⎟
Cdb4 ←
⎝ A⎠
⋅ Cdb2 ⋅
k dbp

pidw⎛⎜ L4 , 10 ⋅ g 42⎞⎟
S k dbn
⎝ A ⎠
C1 ← Cdb1 + Cdb3 + Cgg2

C2 ← CL + Cdb2 + Cdb4 + ( 1 − β ) ⋅ Cf

( )
1
⋅ k B⋅ Tr⋅ γ ⋅ 1 + g 31
β
Cc ← 2 ⋅
k B ⋅ Tr
Ntot −
C2 ⎣ (
⋅ ⎡γ ⋅ 1 + g 42 + 1⎤ ) ⎦
1
g m1 ← ⋅ 2 ⋅ π ⋅ fc⋅ Cc
β

k ← tan⎛⎜ π ⋅ ⎞
PM
⎟
⎝ 180 ⋅ deg ⎠
⎛ C2 ⋅ C1 ⎞
g m2 ← k ⋅ 2 ⋅ π ⋅ fc⋅ ⎜ + C1 + C2 ⎟
⎝ Cc ⎠
1 g m1
fT1 ← ⋅
2 ⋅ π Cgg1

1 g m2
fT2 ← ⋅
2 ⋅ π Cgg2

gmID1 ← if ⎛⎜ pgmid L1 , fT1 > 4 ⋅ , pgmid L1 , fT1 , 0.1⋅ ⎞⎟

( ) ( )
1 1
⎝ V V⎠
⎛ 1 1⎞
252

gmID2 ← if ⎛⎜ ngmid L2 , fT2 > 4 ⋅

1⎞
( ) ( )
1
, ngmid L2 , fT2 , 0.1⋅ ⎟
⎝ V V⎠
g m1 g m2
IDtotal ← +
gmID1 gmID2
IDtotal

优化目标是c1 c2，也即cgg，ft
Initial guess for optimization c1 := 1 c2 := 1
Given c1 > 0 c2 > 0

(
Copt := Minimize f , c1 , c2 )

⎛ 0.145 ⎞ f⎛ Copt , Copt ⎞ = 0.989 mA

Copt = ⎜ ⎟ ⎝ 0 1⎠
⎝ 0.379 ⎠

M := CreateMesh( f , 0.01 , 0.8 , 0.01 , 0.8 , 40 , 40)

M
Comments:
- Shallow power minimum for small c1 and c2
- the optimum is close to the "steep cliff" imposed by limiting gm/ID to practical values >4S/A in the
objective function (in the power minimum, gm/ID1=4.7S/A, gm/ID2=7.7S/A)
 253

Re-calculate components with optimizer result:

c1 := Copt c2 := Copt
0 1

Estimate/calculate return factor and all capacitances

Cf
β := β = 0.291
(Cf + Cs)⋅ (1 + c1)
(
Cgg1 := c1 ⋅ Cs + Cf ) Cgg1 = 87.269 fF

Cgg2 := c2 ⋅ CL Cgg2 = 75.833 fF

Lmin
Cdb1 := k dbp ⋅ Cgg1 ⋅ Cdb1 = 44.428 fF
L1

pidw⎛⎜ L1 , 10
S⎞
⎟
w31 :=
⎝ A⎠
w31 = 0.219 Cdb3 := w31⋅ Cdb1 ⋅
k dbn
Cdb3 = 7.903 fF
nidw⎛⎜ L3 , 10 ⋅ g 31⎞⎟
S k dbp
⎝ A ⎠

Lmin
Cdb2 := k dbn ⋅ Cgg2 ⋅ Cdb2 = 43.13 fF
L2

nidw⎛⎜ L2 , 10
S⎞
⎟
w42 :=
⎝ A⎠
w42 = 4.567 Cdb4 := w42⋅ Cdb2 ⋅
k dbp
Cdb4 = 242.447 fF
pidw⎛⎜ L4 , 10 ⋅ g 42⎞⎟
S k dbn
⎝ A ⎠

C1 := Cdb1 + Cdb3 + Cgg2 C1 = 128.164 fF Total stage 1 load

C2 := CL + Cdb2 + Cdb4 + ( 1 − β ) ⋅ Cf C2 = 627.376 fF Total stage 2 load

Step 2: Calculate Cc based on DR spec

( )
1
⋅ k B⋅ Tr⋅ γ ⋅ 1 + g 31
β
Cc := 2 ⋅ Cc = 0.34 pF
k B ⋅ Tr
Ntot −
C2 ⎣ ( )
⋅ ⎡γ ⋅ 1 + g 42 + 1⎤
⎦
254

Step 3: Calculate transconductances and transit frequencies

1
g m1 := ⋅ 2 ⋅ π ⋅ fc⋅ Cc g m1 = 1.469 mS
β

k := tan⎛⎜ π ⋅ ⎞
PM
⎟ k = 3.732
⎝ 180deg ⎠

⎛ C2⋅ C1 ⎞
g m2 := k ⋅ 2 ⋅ π ⋅ fc⋅ ⎜ + C1 + C2⎟ g m2 = 4.652 mS
⎝ Cc ⎠
1 g m1
fT1 := ⋅ fT1 = 2.679 GHz
2 ⋅ π Cgg1

1 g m2
fT2 := ⋅ fT2 = 9.763 GHz
2 ⋅ π Cgg2

Step 4: Find gm/ID, ID

( ) ( )
1 1
gmID1 := pgmid L1 , fT1 gmID2 := ngmid L2 , fT2 gmID1 = 4.639 gmID2 = 6.916
V V
g m1 g m2
ID1 := ID2 := ID1 = 0.317 mA ID2 = 0.673 mA
gmID1 gmID2

Step 5: Calculate device widths

( ) ( )
A A
IDW1 := pidw L1 , gmID1 IDW2 := nidw L2 , gmID2 IDW1 = 9.535 IDW2 = 17.356
m m

ID1 ID2
W1 := W2 := W1 = 33.208 μm W2 = 38.755 μm
IDW1 IDW2

W3 := W1 ⋅ w31 W4 := W2 ⋅ w42 W3 = 7.271 μm W4 = 177.009 μm

 255

OTA circuit for simulation (feedback network, CMFB, Cc, etc. not shown)

50uA 100uA 700uA 700uA 700uA

M4c M4a,b
25/0.55 25/0.55
25/0.55
M=7 M=7
M4d

M1a,b
33/0.55

M=2
1/0.4 1/0.4
M3d M3c
M3a,b
1/0.4 M2a,b
M=7 39/0.4

Simulation results (first run, without any tweaking):

f c=181.34MHz, PM=75.68deg, T0=298

Magnitude [dB]

80
60
40
20
0
-20
-40 -2 0 2 4
10 10 10 10
f [MHz]
Phase [degrees]

0
-50
-100
-150
-2 0 2 4
10 10 10 10
f [MHz]
256

PSD [V2/Hz]

-20
10
5 10
10 10
f [Hz]
Sqrt(Integral) [μVrms]

Integral=351.77uVrms, DR=72.09dB (for Vodmax=2.00V)

400

200

0 5 10
10 10
f [Hz]

Very close to specs!

- fc is about 10% lower than expected. This is mostly due to the dominant pole approximation; the
second pole pulls fc to lower frequencies. Also, Cgd of M2 adds additional compensation
capacitance, which also reduces fc.

- PM and DR are essentially right on target.

The discrepancies, in general, can be resolved in two ways: (1) Spice tweaking (OK for small
changes), (2) Re-visit above calculations and improve assumtions and equations. E.g. factor in the
expected error from the dominant pole approximation.

An advantage of the presented methodology is that most, if not all discrepancies/errors can be
tracked down by comparing the Spice component values (gm, gm/ID, Cdb, ... from .op) with those
used in the optimization routine.

Close inspection of the .op values below reveals that most small signal parameters calculated
above agree with Spice to within 10-20%.
 257

subckt x1 x1 x1 x1 x1 x1
element 1:m1a 1:m1b 1:m2a 1:m2b 1:m3d 1:m3c
model 0:pch214 0:pch214 0:nch214 0:nch214 0:nch214 0:nch214
region Saturati Saturati Saturati Saturati Saturati Saturati
id -344.6871u -344.6871u 772.2910u 772.2910u 50.0000u 104.2676u
ibs 0. 0. 0. 0. 0. 0.
ibd 0. 0. 0. 0. 0. 0.
vgs -1.2816 -1.2816 871.7327m 871.7327m 1.0988 1.0988
vds -1.4099 -1.4099 1.4986 1.4986 1.0988 2.0205
vbs 718.4015m 718.4015m 0. 0. 0. 0.
vth -885.7614m -885.7614m 595.0075m 595.0075m 584.3535m 574.9695m
vdsat -384.5490m -384.5490m 205.6017m 205.6017m 304.2131m 307.9633m
vod -395.8371m -395.8371m 276.7252m 276.7252m 514.4807m 523.8647m
beta 3.9883m 3.9883m 23.0145m 23.0145m 531.3387u 1.0629m
gam eff 461.1502m 461.1502m 894.1238m 894.1238m 894.1238m 894.1238m
gm 1.5295m 1.5295m 4.9575m 4.9575m 161.8125u 331.7534u
gds 27.5794u 27.5794u 56.8387u 56.8387u 2.9301u 4.0365u
gmb 282.3301u 282.3301u 1.2918m 1.2918m 42.9472u 86.8132u
cdtot 55.0457f 55.0457f 52.2020f 52.2020f 2.0215f 3.4641f
cgtot 88.4114f 88.4114f 77.6753f 77.6753f 1.7924f 3.5848f
cstot 126.0118f 126.0118f 143.1825f 143.1825f 4.4518f 8.8976f
cbtot 104.3431f 104.3431f 140.4000f 140.4000f 5.1705f 9.7628f
cgs 68.7789f 68.7789f 56.7937f 56.7937f 1.3590f 2.7114f
cgd 15.6526f 15.6526f 9.0726f 9.0726f 207.4045a 414.8090a

subckt x1 x1 x1 x1 x1 x1
element 1:m3b 1:m3a 1:m4d 1:m4c 1:m4b 1:m4a
model 0:nch214 0:nch214 0:pch214 0:pch214 0:pch214 0:pch214
region Saturati Saturati Saturati Saturati Saturati Saturati
id 344.6871u 344.6871u -104.2676u -703.2478u -772.2910u -772.2910u
ibs 0. 0. 0. 0. 0. 0.
ibd 0. 0. 0. 0. 0. 0.
vgs 1.0988 1.0988 -979.5312m -979.5312m -979.5312m -979.5312m
vds 871.7327m 871.7327m -979.5312m -718.4015m -1.5014 -1.5014
vbs 0. 0. 0. 0. 0. 0.
vth 586.6658m 586.6658m -739.8673m -742.0442m -735.5167m -735.5167m
vdsat 303.2846m 303.2846m -251.4877m -249.7596m -254.9415m -254.9415m
vod 512.1684m 512.1684m -239.6639m -237.4870m -244.0145m -244.0145m
beta 3.7191m 3.7191m 3.0506m 21.3547m 21.3541m 21.3541m
gam eff 894.1238m 894.1238m 472.5709m 472.5709m 472.5709m 472.5709m
gm 1.1179m 1.1179m 741.4965u 5.0238m 5.4115m 5.4115m
gds 27.2459u 27.2459u 13.0870u 115.7219u 74.4836u 74.4836u
gmb 297.8439u 297.8439u 170.1165u 1.1533m 1.2402m 1.2402m
cdtot 14.9146f 14.9146f 49.3740f 364.8318f 316.8690f 316.8690f
cgtot 12.5467f 12.5467f 67.0284f 469.1707f 469.2508f 469.2508f
cstot 31.1678f 31.1678f 109.9526f 769.6267f 769.7437f 769.7437f
cbtot 36.9572f 36.9572f 103.2448f 741.9384f 693.9432f 693.9432f
cgs 9.5188f 9.5188f 50.4655f 353.2113f 353.3468f 353.3468f
cgd 1.4518f 1.4518f 11.8133f 82.6932f 82.6926f 82.6926f