Chapter 15 Digital CMOS Circuits
¾ 15.1 General Considerations
¾ 15.2 CMOS Inverter
¾ 15.3 CMOS NOR and NAND Gates
Chapter Outline
Following three important questions will be addressed (1) What limits the speed of a digital gate?
(2) How much power does a gate consume while running at a certain speed?
(3) How much “noise” can a gate tolerate while producing a valid output?
Inverter Characteristic
X = A
¾ An inverter outputs a logical “1” when the input is a logical “0”
and vice versa.
Examples 15.1 & 15.2: NMOS Inverter
1 2
, : ( )
in TH out in TH out DD D D DD 2 n ox D in TH
V V V V V V V I R V C W R V V
μ L
> > − = − = − −
2
, :
1 [2( ) ]
2
in TH out in TH
out DD D D DD n ox D in TH out out
V V V V V
V V I R V C W R V V V V
μ L
> ≤ −
= − = − − −
2 2
1 1
( ) ( DD )
out in TH in TH
n ox D n ox D n ox D
V V V V V V
W W W
C R C R C R
L L L
μ μ μ
= − + − − + −
: saturation, : linear
DS GS TH DS GS TH
V ≥V −V V <V −V
:
in TH out DD
V ≤V V =V
Vout is at the lowest when Vin is at VDD.
If we neglect the second term in the square brackets, then
This is equivalent to viewing M1 as a resistor of value
Example 15.1 & 15.2: NMOS Inverter (cont’d)
2
[( ) ]
2
out min DD D D max
out min
DD D n ox DD TH out min
V V R I
W V
V R C V V V
μ L
, ,
, ,
= −
= − − −
1
1
[1 ( )]
1 ( ) [ ( )]
n ox DD TH
DD
out min DD
n ox D DD TH D n ox DD TH
C W V V
V L
V V
W W
C R V V R C V V
L L
μ
μ μ
−
, −
+ −
≈ =
+ − + −
1
1 [ ( )( )]
on n ox DD TH
R = μ C W L V/ −V −
¾ The CS stage resembles a voltage divider between RD and Ron1 when M1 is in deep triode region. It produces VDD when M1 is off.
2
[( ) ]
2
@ linear region
DS
D n ox GS TH DS
V
I C W V V V
μ L
= − −
∵
Transition Region Gain
¾ Ideally, the VTC of an inverter has infinite transition region gain.
However, practically the gain is finite.
Infinite Transition Region Gain Finite Transition Region Gain
Example 15.3: Gain at Transition Region
¾ Transition Region: 50 mV
¾ Supply voltage: 1.8V
1.8 36
v 0.05
A = =
V
0– V
2: Transition Region
Logical Level Degradation
¾ Since real power buses have losses, the power supply levels at two different locations will be different. This will result in
logical level degradation.
Example 15.4: Logical Level Degradation
5A 25mΩ 125mV Δ = V × =
Supply A=1.8V Supply B=1.675V
If inverter Inv1 produces a logical ONE given by the local value of VDD, determine the degradation as sensed by inverter Inv2.
The Effects of Level Degradation and Finite Gain
¾ In conjunction with finite transition gain, logical level
degradation in succeeding gates will reduce the output swings of gates.
Example 15.5: Small-Signal Gain Variation of NMOS Inverter
¾ the small-signal gain is the largest in the transition region.
Sketch the small-signal voltage gain for the characteristic shown in Fig.15.4 as a function of Vin.
Fig.15.4
Example 15.6: Small-Signal Gain Above Unity
¾ The transition region at the input spans a range narrower than 0 to VDD.
Prove that the magnitude of the small-signal gain obtained in Example 15.5 must exceed unity at some point.
Noise Margin
¾ Noise margin is the amount of input logic level degradation that a gate can handle before the small-signal gain becomes -1.
VDD
VIL
VIN VOUT
VDD
VIH
VIN VOUT
VDD
When Vin is equal to VIL, M1 is in saturation region. VDS > VGS - VTH 1.Using the small-signal gain :
2.Using differentiation :
Example 15.7: NMOS Inverter Noise Margin
( ) 1
m D n ox IL TH D
g R C W V V R μ L
= − =
1
L IL TH
n ox D
NM V V
C W R μ L
= = +
( ) 1
out
n ox IL TH D
in
V W
C V V R
V μ L
∂ = − − = −
∂
As Vin drives M1 into the triode region (or linear region),
( )
21 2
out DD 2 n ox D in TH out out
V V C W R V V V V
μ L ⎡ ⎤
= − ⎣ − − ⎦
1 2 2( ) 2
2
out out out
n ox D out in TH out
in in in
V W V V
C R V V V V
V μ L ⎡ V V ⎤
∂∂ = − ⎢⎣ + − ∂∂ − ∂∂ ⎥⎦ with V∂ out/∂Vin = −1 1
2 2
in TH
out
n ox D
V V
V W
C R
μ L
= + − Vin =VIH NMH =VDD −VIH
1 2
( )
out DD 2 n ox in TH D
V V C W V V R
μ L
= − −
saturation region
Example 15.8: Minimum V
out1 1
1 1
0.05
19 19 [ ( )]
on
out min DD DD
D on
D on n ox DD TH
V R V V
R R
R R C W V V
μ L
,
−
≈ ≤
+
≥ = ⋅ −
The output low level of an NMOS inverter is always degraded. Derive a relationship to guarantee that this degradation remains below 0.05VDD.
Example 15.9: Dynamic Behavior of NMOS Inverter
¾ Since digital circuits operate with large signals and experience nonlinearity, the concept of transfer function is no longer meaningful. Therefore, we must resort to time-domain analysis to evaluate the speed of a gate.
(0 ) ( ) (0 )
out DD n ox D DD TH out D
V V C W R V V V R
μ L
− = − − −
( ) (0 ) [ (0 )] 1 exp , 0
out out DD out
D L
V t V V V t t
R C
− − ⎛ − ⎞
= + − ⎜ − ⎟ >
⎝ ⎠
0 95 DD out(0 ) [ DD out(0 )] 1 exp 95% D L
V V V V T
R C
− − ⎛ − ⎞
. = + − ⎜ − ⎟
⎝ ⎠ 95
ln 0 05
(0 )
DD
% D L
DD out
T R C V
V V −
→ = − .
− Assuming (0 )VDD −Vout − ≈VDD, T95% ≈3R CD L
(0 )
1 ( )
DD out
n ox D DD TH
V V
C W R V V μ L
→ − =
+ −
0
DD out out
L D
V V dV
R C dt
− − + = out out DD 0
L D L D
dV V V
dt + C R −C R =
/ D L t R C
Vout = +A Be−
Rise/Fall Time and Delay
TR: the time required for the output to go from 10% of VDD to 90 % of VDD
TF: the time required for the output to go from 90% of VDD to 10 % of VDD delay: the difference
between the time points at which the input
and the output cross 0.5 VDD
Example 15.10: Time Constant
Assuming a 5% degradation in the output low level, determine the time constant at node X when VX goes from low to high.
Assuming CX ≈WLCox and RD =19Ron1,
2
19
( )
19
( )
D X ox
n ox DD TH
n DD TH
R C WLC
C W V V L
L V V
τ μ
μ
= = ⋅
−
= −
Example 15.11: Interconnect Capacitance
Wire Capacitance per Mircon: 50 aF/µm, (1aF=1x10-18F) Total Interconnect Capacitance: 15000×50×10-18 =750 fF
Equivalent to 640 MOS FETs with W=0.5µm, L=0.18µm, Cox =13.5fF/µm2 What is the interconnect capacitance driven by Inv1?
1.17 fF
GS ox
C ≈ WLC =
∵
CICC
GS640
IC GS
C
C ≈
Power-Delay Product
¾ The power delay product of an NMOS Inverter can be loosely thought of as the amount of energy the gate uses in each
switching event.
2
1 1
2
1 2
1
Power
2
( ) ( ) ( ) with
( ) ( ) ( )
, since typically .
PHL PLH
DD DD
D DD D X D X D
D on D on
DD
D DD D X D X
D on
DD X D on
T T
PDP
V V
I V R C R C I
R R R R
I V R C V R C
R R
V C R R
= ⋅ +
≈ ⋅ = ⋅ =
+ +
≈ ⋅ = ⋅
+
≈
PDP has dimension of
energy, i.e. it indicates how much energy is consumed for a logical operation
Example 15.12: Power-Delay Product
3
DD2 oxPDP = V WLC
( )( )
3
3
PLH D X
D DD D X
T R C
PDP I V R C
≈
=
Assuming TPLH is roughly equal to three time constants, determine the power-delay product for the low-to-high transitions at node X
TPLH: Propagation delay time for the low-to-high transition
Summary
¾ Digital CMOS circuits account for more than 80% of the semiconductor market.
¾ The speed, power dissipation, and noise immunity of digital gates are critical parameters
¾ The input/output characteristic of a gate reveals its immunity to noise or degraded logical levels
¾ Noise margin is defined as the voltage degradation on the high or low levels that places the signal at the unit-gain point of the input/output characteristic
¾ The speed of gates is given by the drive capability of the
transistors and the capacitances contributed by transistors and interconnecting wires
¾ The power-speed trade-off of gates is quantified by the power- delay product
Drawbacks of NMOS Inverter
¾ Because of constant RD, NMOS inverter consumes static power even when there is no switching.
¾ RD presents a tradeoff between speed and power dissipation.
Risetime limitation due to RD
Static power consumption during output level low
2
/
DD D
V R
Improved Inverter Topology
¾ A better alternative would probably have been an “intelligent”
pullup device that turns on when M1 is off and vice versa.
Improved Fall Time
¾ This improved inverter topology decreases fall time since all of the current from M1 is available to discharge the capacitor.
CMOS Inverter
¾ A circuit realization of this improved inverter topology is the CMOS inverter shown above.
¾ The NMOS/PMOS pair complement each other to produce the desired effects.
Full swing or rail-to-rail swing
CMOS Inverter
¾ A circuit realization of this improved inverter topology is the CMOS inverter shown above.
¾ The NMOS/PMOS pair complement each other to produce the desired effects.
Full swing or rail-to-rail swing
I
( / )W L 1
( / )W L 2
CMOS Inverter Small-Signal Model
(
1 2)(
1||
2)
out
m m O O
in
v g g r r
v = − +
¾ When both M1 and M2 are in saturation, the small-signal gain is shown above.
2 0
vin v
− + =
1 0
vin v
− + =
1 2
vin = =v v
Vin
VDD
Operation of PMOS Device
Gate
p+ D p+ S
n-type substrate VDD
VDD 0 V 0 V
GS G S DD
0
V = V − V = − V −
DS D S DD
0
V = V − V = − V −
Vin= -VDD
VD=-VDD
Gate
p+ D p+ S
n-type substrate -VDD
-VDD 0 V
0 V
GS G S
0
DDV = V − V = − V
DS D S
0
DDV = V − V = − V
In saturation region ( )
DS GS TH
DS GS TH
DS GS TH
V V V
V V V
V V V
− ≥ − − −
≤ +
≤ +
> 0
Voltage Transfer Curve of CMOS Inverter
1 2 out DD
Region 1: M is off and M is on. V =V .
1 2 in TH1
out in TH2
Region 2: M is in saturation and M is in triode region ( + ).
Valid only when V > V + |V |.
V ≈V Δ
(
1)
2(
2)( ) ( )
21 2
1 2 | |
2
This quadratic equation can be solved in terms of ( ).
n ox in TH p ox DD in TH DD out DD out
DD out out in
W W
C V V C V V V V V V V
L L
V V V f V
μ ⎛⎜⎝ ⎞⎟⎠ − = μ ⎛⎜⎝ ⎞ ⎡⎟ ⎣⎠ − − − − − ⎤⎦
− → =
out in TH2 2
V V + |V | = → M is about to exit the triode region
Voltage Transfer Curve of CMOS Inverter
1 2
in TH1 out in TH2
Region 3: M and M are in saturation. Apperas as vertical line assuming no channel-length modulation. Valid when V - V ≤ V ≤ V + |V |.
( ) ( )
( )
2 2
1 2
1 2
1 2
1 2
1 2
1 1
Solving | | ,
2 2
| |
n ox in TH p ox DD in TH
n TH p DD TH
in
n p
W W
C V V C V V V
L L
W W
V V V
L L
V
W W
L L
μ μ
μ μ
μ μ
⎛ ⎞ − = ⎛ ⎞ − −
⎜ ⎟ ⎜ ⎟
⎝ ⎠ ⎝ ⎠
⎛ ⎞ ⋅ + ⎛ ⎞ ⋅ −
⎜ ⎟ ⎜ ⎟
⎝ ⎠ ⎝ ⎠
= ⎛⎜⎝ ⎞⎟⎠ + ⎛⎜⎝ ⎞⎟⎠
Voltage Transfer Curve of CMOS Inverter
1
2 out in TH1
Region 4: Similar to Region 2. M is in triode region and M is in saturation. Valid only when V V - V .≤
(
1)
2(
2)
21 2
1 2 | |
2
This quadratic equation can be solved in terms of ( ).
n ox in TH out out p ox DD in TH
DD out out in
W W
C V V V V C V V V
L L
V V V f V
μ ⎛⎜⎝ ⎞⎟⎠ ⎡⎣ − − ⎤⎦ = μ ⎛⎜⎝ ⎞⎟⎠ − −
− → =
1 2 out
Region 5: M is on and M is off. V =0.
Example 15.14: Switching Threshold
The switching threshold or the “trip point” of the inverter is when Vout equals Vin.
Determine a relationship between (W/L)1 and (W/L)2 that sets the trip point of the CMOS inverter to VDD/2, thus providing a “symmetric” VTC
2 2
1 1 2 2
1 2
1 1
2 2 2 2
DD DD DD DD
n ox TH p ox TH
V V V V
W W
C V C V
L L
μ ⎛⎜⎝ ⎞⎟⎠ ⎛⎜⎝ − ⎞ ⎛⎟ ⎜⎠ ⎝ +λ ⎞⎟⎠ = μ ⎛⎜⎝ ⎞⎟⎠ ⎛⎜⎝ − | |⎞ ⎛⎟ ⎜⎠ ⎝ +λ ⎞⎟⎠
1
1 2
2
Assuming 1 1 ,
2 2
DD DD p
n
V V W
W
λ λ μ
μ
⎛ ⎞ ⎛ ⎞
+ ⎜ ⎟ ≈ + ⎜ ⎟ ≈
⎝ ⎠ ⎝ ⎠
n 2.5 p
μ ≈ μ
2 2.5 1
W ≈ W
Example 15.15: VTC
¾ As the PMOS device is made stronger, NMOS device requires higher input voltage to establish ID1=ID2. Thus, the VTC is
shifted to the right.
W
2Noise Margins
VIL is the low-level input voltage at which (δVout/ δVin) = -1
( ) ( )( ) ( )
( )
( ) ( )
2 2
1 2
1 2
in
1 1
2 2
In Region 2,
1 2 | |
2
Differentiating both sides with respect to V ,
2 2 | |
n ox in TH p ox DD in TH DD out DD out
n ox in TH
p ox DD out DD in TH
W W
C V V C V V V V V V V
L L
C W V V
L
C W V V V V V
L
μ μ
μ μ δ
⎛ ⎞ − = ⎛ ⎞ ⎡ − − − − − ⎤
⎜ ⎟ ⎜ ⎟ ⎣ ⎦
⎝ ⎠ ⎝ ⎠
⎛ ⎞ − =
⎜ ⎟
⎝ ⎠
⎛ ⎞ − − − − −
⎜ ⎟
⎝ ⎠
( )
(
1) (
2)
1 2
2 With 1,
2 | |
out out
DD out
in in
out in
n in TH p out in TH DD
V V
V V
V V
V V
W W
V V V V V V
L L
δ
δ δ
δ δ
μ μ
⎡ ⎤
+ −
⎢ ⎥
⎣ ⎦
= −
⎛ ⎞ − = ⎛ ⎞ − − −
⎜ ⎟ ⎜ ⎟
⎝ ⎠ ⎝ ⎠
Noise Margins
( ) ( )
( ) ( )( ) ( )
1 2
1 2
2 2
1 2
2 out
Assuming , we must solve
2 | | and
2 | |
2
Deleting V , we get 0, where A= (1 1)( 3), 8
1( 4
n p
in TH out in TH DD
in TH DD in TH DD out DD out
in in
W W
a L L
a V V V V V V
a V V V V V V V V V
A V B V C a a
B a
μ ⎛ ⎞ μ ⎛ ⎞
= ⎜⎝ ⎟⎠ ⎜⎝ ⎟⎠
− = − − −
⎡ ⎤
− = ⎣ − − − − − ⎦
⋅ + ⋅ + = − +
= + 1 2 2 2 1 2 12
2 1 2 2
1 2 1 2
1
3)( | |), 1[3( | |) 2 ( | |) ( 4) ].
8
1 ( 3)( | |)
4 4
2 2
1 1
( 3)( | |) ( 3)( | |)
4 2
2 1( 1)( 3) 8
2 (
DD TH TH DD TH TH DD TH TH
DD TH TH
IL
DD TH TH DD TH TH
DD TH
V aV V C V V aV V V a a V
B a a V V V
B B AC
V A A
a V aV V a a V V V
a a a V V
− − = − − + − − +
− ± + − −
− ± −
= =
− + − − ± + − −
=
⋅ − +
= − | 2 |) ( 1 | 2 |)
( 1) 3 1
TH DD TH TH
V V aV V
a a a
− − − −
− + −
Noise Margins
( )
( )
1 2 1 2 12
1 2
2 , where .
1 3 1
Assuming symmetry, ( 1, | | ),
3 1
8 4 .
n
DD TH TH DD TH TH
IL L
p
TH TH TH
IL L DD TH
W
a V V V V aV V L
V NM a
a W
a a
L
a V V V
V NM V V
μ μ
⎛ ⎞
⎜ ⎟
− − − − ⎝ ⎠
= = − =
− ⎛ ⎞
− +
⎜ ⎟
⎝ ⎠
= = =
= = +
Noise Margins
VIH is the high-level input voltage at which (δVout/ δVin) = -1.
( )
( )
1 2 2 1
1/ , ,
1 2 1 2
2
1 1 3 1
TH TH TH TH
IH DD IL a a V V V V
DD TH TH DD TH TH
V V V
a V V V V aV V
a a a
→ → →
= −
− − − −
= −
− + −
1 2
1 2
Assuming symmetry, | | and ,
5 1
8 4
3 1
8 4
TH TH TH n p
IH DD TH
H DD IH DD TH
W W
V V V
L L
V V V
NM V V V V
μ ⎛ ⎞ μ ⎛ ⎞
= = ⎜⎝ ⎟⎠ = ⎜⎝ ⎟⎠
= −
= − = +
Example 15.17: Noise Margins of Ideal Inverter
, ,
2
DD H ideal L ideal
NM = NM = V
3 1
8 4
L H DD TH
NM = NM = V + V 1
2 V
DD>
1
TH
2
DDV < V
∵
VIL
VIN
VIH
VIN
VDD
H DD IH
NM =V −V
L IL
NM =V
Example 15.18: Floating Output
1 2
/ 2 / 2
TH DD
TH DD
V V
V V
>
>
¾ When Vin=VDD/2, M1 and M2 will both be off and the output floats.
Charging Dynamics of CMOS Inverter
¾ As Vout is initially charged high, the charging is linear since M2 is in saturation. However, as M2 enters the triode region the charge rate becomes sublinear.
Example 15.19: Charging Current
¾ The current of M2 is initially constant as M2 is in saturation.
However as M2 enters the triode region, its current decreases.
In saturation region
DS GS TH
V ≤V + V 2 2
At boundary
TH DD 0 DD TH
V −V ≤ −V + V
Example 15.20: Variation of Output Waveform
¾ As the PMOS size is increased, the output exhibits a faster transition.
Discharging Dynamics of CMOS Inverter
¾ Similar to the charging dynamics, the discharge is linear when M1 is in saturation and becomes sublinear as M1 enters the
triode region.
In saturation region
In linear region
Rise Delay
( )
( )
2
2
2 2
2
2 2
1 2
2 2 2
Initially, M is in saturation,
| |
| |
( ) ,only up to | | .
2 | |
Thus,
D p ox DD TH
D
out out TH
L
TH L
PLH
p ox DD TH
I C W V V
L
V t I t V V
C
V C
T W
C V V
L
μ
μ
⎛ ⎞
= ⎜⎝ ⎟⎠ −
= =
= ⎛⎜⎝ ⎞⎟⎠ −
VDD 0.5VDD
Vin Vout
Rise Delay
( ) ( ) ( )
( ) ( ) ( )
2
2
2 2
2
2 2
2
Thereafter operates in the triode region,
| |
1 2
2
1 2 2
out
D L
out
p ox DD TH DD out DD out L
out ox
p
DD TH DD out DD out L
M I C dV
dt
dV
C W V V V V V V C
L dt
dV C W
C L dt
V V V V V V
μ
μ
=
⎛ ⎞ ⎡ − − − − ⎤ =
⎜ ⎟ ⎣ ⎦
⎝ ⎠
⎛ ⎞
= ⎜⎝ ⎟⎠
− − − −
( )
2
2 2
2 2
2 2
1 2
2 2
2
2
Integrating from to / 2,
ln 3 4 ln 3 4
,
2 ln 3 4
out TH DD
TH TH
L
PLH on L
DD DD
p ox DD TH
PLH PLH PLH
TH TH
on L
DD TH DD
V V V
V V
T C R C
W V V
C V V
L Thus
T T T
V V
R C
V V V
μ
=
⎛ ⎞ ⎛ ⎞
= ⎛⎜⎝ ⎞⎟⎠ − ⎜⎝ − ⎟⎠= ⎜⎝ − ⎟⎠
= +
⎡ ⎛ ⎞⎤
= ⎢⎢⎣ − + ⎜⎝ − ⎟⎠⎥⎥⎦
DD out
V −V =u du 2 1ln u au u = a a u
− −
∫
Fall Delay
Fall Time Delay
1 1
1 1 1
1 1
1
1
Similarly,
2 ln 3 4
2 ln 3 4
L TH TH
PHL
DD TH DD
n ox DD TH
TH TH
on L
DD TH DD
C V V
T W V V V
C V V
L
V V
R C
V V V
μ
⎡ ⎛ ⎞⎤
= ⎛⎜⎝ ⎞⎟ ⎣⎠ ⎡ − ⎤⎦ ⎢⎣ − + ⎜⎝ − ⎟⎠⎥⎦
⎡ ⎛ ⎞⎤
= ⎢⎣ − + ⎜⎝ − ⎟⎠⎥⎦