VLSI Design
Chapter 5
CMOS Circuit and Logic Design
Jin-Fu Li
Chapter 5 CMOS Circuit and Logic
Design
•
•
•
•
•
•
CMOS Logic Gate Design
Physical Design of Logic Gates
CMOS Logic Structures
Clocking Strategies
I/O Structures
Low-Power Design
National Central University
EE613 VLSI Design
2
Logic Gate Design Issues
•
Hierarchical design
− Architecture level
− RTL/logic gate level
− Circuit level
− Layout level
•
Critical paths – the path with the longest delay
that require attention to timing details
•
The number of Fanins and Fanouts affects the
performance of the circuits
National Central University
EE613 VLSI Design
3
Concept of Fanin and Fanout
•
Fanin
− The fanin of any complex gate is defined as the number of
inputs of this gate
•
Fanout
− The fanout of a complex gate is defined as the number of
driven inputs attached to the output of this gate
N
N
Fanout=N
National Central University
Fanin=N
EE613 VLSI Design
4
Logic Gate Design – NAND Gate
t dr =
Rp
n
( mnC
d
+ C r + kC g )
•
Rp = the effective resistance of p-device in a minimumsized inverter
•
•
•
•
•
n = width multiplier for p-devices in this gate
•
Cr = routing capacitance
k = the fanout
m = fanin of gate
Cg = gate capacitance of a minimum-sized inverter
Cd = source/drain capacitance of a minimum-sized
inverter
National Central University
EE613 VLSI Design
5
Logic Gate Design – Fanins and Fanouts
m=3, k=4
Cr
mnCd
National Central University
kCg
EE613 VLSI Design
6
Logic Gate Design – NAND Gate Rise Time
t dr =
=
=
Rp
( mnC
n
Rp
( mnrC
n
R pC g
n
+ C r + kC g )
d
g
+ q ( k ) C g + kC g )
( mnr + q ( k ) + k )
= R g C g mr +
R pC g
n
q (k ) +
R pC g
n
k
Separate delay into internal delay and external delay caused by fanouts
t dr = tint − r + k × toutput − r
tint − r = R p C g mr
toutput − r =
National Central University
R pC g
n
q(k )
(1 +
)
k
EE613 VLSI Design
7
Logic Gate Design – NAND Gate Fall Time
t df
Rn
= m
( mnrC
n
= R n C g m r + mk
2
= t int − f + k × t output
g
+ q ( k ) C g + kC g )
RnC g
n
(1 +
q (k )
)
k
− f
We want the rise time to be equal to the fall time
t dr = t df
Rp
Rn
( mnrC g + q ( k )C g + kC g ) = m
( mnrC g + q ( k )C g + kC g )
n
n
R p = mR n
Hence we must design Wp = Wn , thus the delay time is
m
t df = t dr
Rn
=
( m 2 nrC g + mq ( k )C g + mkC g )
n
National Central University
EE613 VLSI Design
8
Typical CMOS NAND & NOR Delays
Rn − nand
(m × 4 × 0.005 + C L )
4
4 × t f − nand
t f − nand = m
Assume :
n=4
kC g = C L
Rn − nand =
q(k ) = 0
m(0.02 × m + kC g )
rC g = Cd = 0.005 pf
A
B
C
D
A
B
C
D
National Central University
Delay (ns)
Delay (ns)
10.0 ns
ND4-Fall 10.0 ns
NR4-Fall
0.0 0.25 0.5 0.75 1.0
Capacitive load (pf)
EE613 VLSI Design
NR4-Rise
ND4-Rise
0.0 0.25 0.5 0.75 1.0
Capacitive load (pf)
9
Logic Gate Design – Gate Delays
NAND- and NOR-Gates Delays Measured with SPICE
GATE
tinternal-f
(ns)
INV
ND2
ND3
ND4
ND8
NR2
NR3
NR4
NR8
.08
.2
.41
.68
2.44
.135
.14
.145
.19
National Central University
toutput-f
(ns/pf)
1.7
3.1
4.4
5.7
10.98
1.75
1.83
1.88
1.8
EE613 VLSI Design
tinternal-r
(ns)
.08
.15
.2
.25
.38
.25
.52
.9
3.35
toutput-r
(ns/pf)
2.1
2.1
2.1
2.1
2.2
4.1
6.2
8.2
16.4
10
Logic Gate Design – Efficient Resistance
Efficient Resistance Value for a
Typical 1u CMOS Process
National Central University
GATE
Rn ( Ω)
INV
ND2
ND3
ND4
NR2
NR3
NR4
7.1K
6.3K
6.0K
5.9K
7.3K
7.4K
7.5K
EE613 VLSI Design
Rp ( Ω)
8.5K
8.6K
8.7K
8.8K
8.4K
8.4K
8.4K
11
Logic Gate Design – 8-Input AND Gate
A
CB
D
E
F
G
H
A
B
C
D
E
F
G
H
A
B
C
D
E
F
G
H
National Central University
CL
Approach 1
CL
Approach 2
Approach 3
CL
EE613 VLSI Design
12
Logic Gate Design – 8-Input AND Gate
Comparison of Approaches to Designing an 8-Input AND Gate
Approach
Delay
Stage 1ns
Delay
Stage 2ns
1
ND8->INV
2.82
ND8
falling
3.37
INV
rising
6.2
(6.5)
2
ND4->NR2
.88
ND4
falling
4.36
NR2
rising
5.24
(5.26)
3
ND2->NR2
ND2->INV
.31
ND2
falling
.4
NR2
rising
National Central University
Delay
Stage 3ns
.31
ND2
falling
EE613 VLSI Design
Delay
Stage 4ns
2.17
INV
rising
Total Delay
(SPICE) ns
3.19
(3.46)
13
Basic Physical Design
•
•
•
•
•
•
•
•
Gates: Inverter, NAND, and NOR
Complex Gates
Standard Cells
Gate Array
Sea of Gates
Layout Optimization
Transmission Gates
2-Input Multiplexer
National Central University
EE613 VLSI Design
14
Physical Design – CMOS Inverter
Vdd
Vdd
a
z
a
z
Vss
Vss
National Central University
EE613 VLSI Design
15
Physical Design – NAND Gate
Vdd
Vdd
z
z
a
b
a
b
Vss
National Central University
EE613 VLSI Design
Vss
16
Physical Design – NOR Gate
Vdd
a
b
Vdd
z
z
Vss
Vss
b
National Central University
EE613 VLSI Design
a
17
Physical Design – Complex Gates
•
All complex gates can be designed using a single
row of N-transistors and a single row of Ptransistors, aligned at common gate connections
•
Design procedure
− Draw two dual graphs to P transistor tree and N
transistor tree
− Find all Euler paths that cover the graph
− Find a P and an N Euler path that have identical
labeling
− If not found, break the gate in the minimum
numbers of places to achieve step 3
National Central University
EE613 VLSI Design
18
Physical Design – Complex Gates
VDD
C
D
Z
I1
D
C
B
I3
I1
I2
A
Z
C
D
National Central University
A
I3
Vss
B
I2
B
A
Z
EE613 VLSI Design
19
Physical Design – Complex Gates
D
C
D
C
B
B
A
A
Vdd
z
Vss
A
National Central University
B
C
EE613 VLSI Design
D
20
Physical Design – XNOR Gate (1)
A
B
Z’
Z
Vdd
Z’
z
A
Z’
B
A
Vss
A
National Central University
B
Z’
B
EE613 VLSI Design
21
Physical Design – XNOR Gate (2)
A
B
Z
Vdd
Vss
z
B
A
National Central University
EE613 VLSI Design
22
Physical Design – Automated Approach
A
B
C
D
E
Vdd
A
E
D
C
E
B
Vss
D
E
C
A
B
D
E
C
A
B
Vdd
P
N
Vss
National Central University
EE613 VLSI Design
23
Physical Design – Standard-Cell Approach
WVdd
Wp
Dnp
Wn
a
National Central University
b
c
d
EE613 VLSI Design
z
WVss
24
Physical Design – Standard-Cell Layout
Vdd
Vdd
Vss
a
b
National Central University
c
z
EE613 VLSI Design
Vss
a
b
c
z
25
Physical Design – Gate Array Layout (1)
Vdd
Vss
National Central University
EE613 VLSI Design
26
Physical Design – Gate Array Layout (2)
Vdd
Gate array cells
Routing channels
Vss
National Central University
EE613 VLSI Design
27
Physical Design – Sea-of-Gate Layout
well contacts
Vdd supply
P-transistors
poly gates
N-transistors
Vss supply
substrate contacts
National Central University
EE613 VLSI Design
28
Physical Design – Sea-of-Gate (NAND3)
a
a
b
a
National Central University
b
c
z
c
b
z
c
EE613 VLSI Design
29
Physical Design – CMOS Layout Guidelines
•
Run VDD and VSS in metal at the top and bottom of
the cell
•
•
Run a vertical poly line for each gate input
•
Place n-gate segments close to VSS and p-gate
segments close to VDD
•
Connection to complete the logic gate should be
made in poly, metal, or, where appropriate, in
diffusion
Order the poly gate signals to allow the maximal
connection between transistors via abutting
source-drain connection.
National Central University
EE613 VLSI Design
30
Physical Design – Improvement in Density
•
Better use of routing layers – routes can occurs
over cells
•
•
•
More “merged” source-drain connections
More usage of “white” space in sparse gates
Use of optimum device sizes – the use of smaller
devices leads to smaller layouts
National Central University
EE613 VLSI Design
31
Physical Design – Layout Optimization
Vdd
clk
F
A<0>
F
A<1>
A<2>
A<3>
Vss
clk A<3> A<2> A<1> A<0>
National Central University
EE613 VLSI Design
32
Physical Design – Layout Optimization
2
A
A
B
C
D
Z
B
C
D
1
Vdd
Right
Wrong
Z
Vss
A
National Central University
B
C
D
EE613 VLSI Design
A
B
C
D
33
Physical Design – Transmission Gate
National Central University
EE613 VLSI Design
34
Physical Design – Transmission Gate
National Central University
EE613 VLSI Design
35
Physical Design – 2-Input Multiplexer
z
c
a
z
-c
z
a
b
b
c
c
National Central University
-c
EE613 VLSI Design
36
CMOS Logic – Pseudo-nMOS Logic
VDD
βp
F
βn
A
VL
Time
β n (VDD − VTn )VOL =
for
βn
2
(VDD − | VTp |) 2
VTn = −VTp = VT
βp
VOL =
(VDD − VT )
2βn
National Central University
EE613 VLSI Design
37
CMOS Logic – Dynamic CMOS Logic
Z
N-logic
Block
inputs
evaluate
clk
precharge
clk
clk
clk
Z=(A+B).C
C
A
National Central University
Y=ABC
B
B
clk
A
C
clk
EE613 VLSI Design
38
CMOS Logic – Dynamic CMOS Logic
clk=1
A
C
1
B
C1
C
C2
1
0
A
C2
C1
C
charge sharing model
clk=1
CVDD = (C + C1 + C2 )VA
C
VA =
VDD
C + C1 + C2
If for example
National Central University
C1 = C2 = 0.5C
then this voltage would be VDD/2
EE613 VLSI Design
39
CMOS Logic – Dynamic CMOS Logic
clock
N1
N2
N1
inputs
N Logic
N Logic
T d1
Erroneous State
N2
clock
T d2
National Central University
EE613 VLSI Design
40
CMOS Logic – Clocked CMOS Logic
-clk
Z
clk
A
B
C
D
E
National Central University
EE613 VLSI Design
41
CMOS Logic – Pass-Transistor Logic
A
-A
-B
A
-B
-A
OUT
B
A
OUT
OUT
B
B
A
Complementary
National Central University
Single-polarity
EE613 VLSI Design
Cross-coupled
42
CMOS Logic – CMOS Domino Logic
Basic gate
Z
A
B
C
D
E
clk
National Central University
EE613 VLSI Design
43
CMOS Logic – CMOS Domino Logic
weak p device
Z
inputs
N-logic
Block
clk
Z
inputs
clk
Static version
National Central University
N-logic
Block
Latched version
EE613 VLSI Design
44
CMOS Logic – CMOS Domino Logic
clk
clk
N-logic
A5
C2
A4
C3
A3
C4
A2
C5
A1
C6
A0
C7
National Central University
C1
N-logic
N-logic
N-logic
EE613 VLSI Design
45
CMOS Logic – NP Domino Logic
clk
-clk
clk
to futher P blocks
inputs
stable
during
clk=1
N-logic
P-logic
other P blocks
other N blocks
clk
N-logic
other N blocks
other P blocks
-clk
clk
to futher P blocks
inputs
stable
during
clk=1
N-logic
P-logic
other P blocks
National Central University
N-logic
other N blocks
EE613 VLSI Design
46
CMOS Logic–Advantages of Dynamic Logic
•
•
•
Smaller area than fully static gates
Smaller parasitic capacitance, hence higher speed
Glitch free operation if designed carefully
National Central University
EE613 VLSI Design
47
CMOS Logic – CVSL
F
-F
Q
-Q
Differential
Inputs
nMOS
Combinational
Network
a
d
e
b
Basic version
National Central University
-d
-e
c -b
-c
-a
A particular function
EE613 VLSI Design
48
CMOS Logic – CVSL
(abcd)=(0000)
clock
-Q
clock
-Q
Differential
Inputs
Q
nMOS
Combinational
Network
Q
-d
d
-d
d
c
-c
c
-c
b
-b
b
-b
a
-a
clock
clock
Clocked version
National Central University
A 4-way XOR gate
EE613 VLSI Design
49
Clocking Strategies – Clocked Systems
outputs
inputs
Combinational Logic
current
state
bits
next
state
bits
Q D
Q D
Q D
clock
A simple finite state machine
National Central University
EE613 VLSI Design
50
Clocking Strategies – Clocked Systems
10 ns
10 ns
C1
inputs
C2
outputs
D Q
D Q
D Q
D Q
D Q
D Q
C1
C2
outputs
inputs
D Q
D Q
D Q
A pipeline system
National Central University
EE613 VLSI Design
51
Clocking Strategies – Latches and Reg.
Cycle time Tc
clock
Setup Time (Ts)
Data
Hold Time (Th)
Q
Clock-to-Q Delay (Tq)
National Central University
EE613 VLSI Design
52
Clocking Strategies – Latches
D
clk
0
1
Q
D
S
clk
Q
clk
0
D
1
Q
S
clk
D
Q
National Central University
EE613 VLSI Design
53
Clocking Strategies – Registers
clk
D
0
1
QM
S
clk
0
1
D
Q
S
clk
QM
Q
master
National Central University
slave
EE613 VLSI Design
54
Clocking Strategies – Registers
clk=0
clk=1
master
National Central University
slave
EE613 VLSI Design
55
Clocking Strategies – Registers
D
Q
clk
clk
National Central University
EE613 VLSI Design
56
Clocking Strategies – JK Registers
J
K
A
D
K
J
B
Q
clk
QN
Q
QN
-clk
clk
J=K=0; Q=D
JN=KN=1; A=QN, B=1; D=AN=Q
J K clk Q QN
J=0;K=1
0
0
1
1
KN=0,JN=1; A=1, B=1; D=0
J=1; K=0
KN=1, JN=0; A=QN, B=Q; D=1;
0
1
0
1
Q
0
1
QN
QN
1
0
Q
J=1; K=1
KN=0, JN=0; A=1, B=Q; D=QN
National Central University
EE613 VLSI Design
57
Clocking Strategies – System Timing
Reg.
A
Tq
Combinational Logic
Td
Ts
Reg.
B
Latch
A
Tq
Combinational Logic
Td
Ts
Latch
B
clock
clock
A
Latch
A Tq
B
Combinational Logic Ts
Tda
Latch
B
Combinational Logic
Tdb
Latch
C
clock
National Central University
EE613 VLSI Design
58
Clocking Strategies – System Timing
Tda < Tc1 − Tqa − Tsb
Tqa : the clock-to-Q time of latch A
Tsb
: the setup time of latch B
Similarly,
Tdb < Tc 0 − Tqb − Tsc
Finally,
Tc = Tda + Tdb + 2[Tq + Ts ]
National Central University
EE613 VLSI Design
59
Clocking Strategies – Setup & Hold Time
Td
D Q
Din Pad
φin
φin
Din
Pad
Tφ
φ
t = t−
t = t+
For an ideal DFF,
= t − , then Q should be high
If Din becomes to low when t = t + , then Q still is high
If Din is high when t
National Central University
EE613 VLSI Design
60
Clocking Strategies – Setup & Hold Time
Tφ
φ
Tφ
Td
Td
D
When Td
When
> Tφ , Din should become high earlier and Q can become high
Td < Tφ , Din should retain at high longer and Q can be still at high
Tφ
Din
D
TS Tφ
Td
Din
D
Td
Th = Td − Tφ
TS = Td − Tφ
National Central University
Th
EE613 VLSI Design
61
Clocking Strategies – Setup & Hold Time
Td2
D Q
q1
Tdq
Logic
D Q
Tdl
M1
clk
d2
M2
Tdc
delay
delay
T c1
T c2
1. When Tdc>Tdq+Tdl, M2 latches
the New data
2. When Tdq+Tdl-Tdc>TC , M2
latches Old data twice
Therefore, 0<Tdq+Tdl-Tdc<TC
clk
T c1
Td2 Old data
T c2
New data
New Data
Tdc
National Central University
TC
EE613 VLSI Design
62
Clocking Strategies – D Register
-clk
clk
clk
-clk
clk
-Q
D
-clk
clk
-clk
Q
clk
-clk
-clk
D
-clk
clk
clk
clk
National Central University
Q
-clk
EE613 VLSI Design
63
Clocking Strategies – Clock Skew
-clk
clk
Feedthrough condition
D
clk-in
Q
clk
clk
Buffers Necessary for
Large Loads
-clk
National Central University
EE613 VLSI Design
-clk
64
Clocking Strategies–Skew Clock Pipeline
CL2
(9ns)
(5ns)
FF
(5ns)
CL3
FF
FF
FF
CL1
clk3
clk2
-2ns
0ns
-2ns
clk1
clk
A
7ns B
C
D
clk
clk1
clk2
clk3
National Central University
A
B
C
A
D
B
A
EE613 VLSI Design
C
B
D
C
65
Clocking Strategies – Latches
-clk
Q
D
clk
1.
Low area cost
2.
Driving capability of D must override the feedback inverter
clk
D
-clk
Q
-clk
clk
National Central University
EE613 VLSI Design
66
Clocking Strategies – Latches
Vdd
clk
-clk
D
D
Q
-clk
Q
clk
clk
-clk
Vss
National Central University
EE613 VLSI Design
67
Clocking Strategies – DETDFF
clk
D
-D
clk
Q1
Q1
-Q1
clk
Latch 1
National Central University
EE613 VLSI Design
68
Clocking Strategies – DETDFF
clk
Q2
-Q2
clk
Q2
D
-D
clk
Latch 2
National Central University
EE613 VLSI Design
69
Clocking Strategies – DETDFF
Latch 2
Q2
D
-Q2
Q
-Q
-Q1
Q1
clk
Latch 1
clk
National Central University
Latch 1 enabled
Q2=-Q2=low
Latch 2 enabled
Q1=-Q1=high
EE613 VLSI Design
70
Clocking Strategies – Register
Asynchronously resettable register
-clk
-reset
Q
clk
-clk
-clk
clk
clk
-clk
D
-clk
clk
-reset
Q
National Central University
EE613 VLSI Design
71
Clocking Strategies – Register
Asynchronously settable and resettable register
-clk
-reset
Q
clk
-clk
-clk
clk
D
-clk
clk
-clk
-set
National Central University
EE613 VLSI Design
72
Clocking Strategies – Dynamic Registers
Dynamic single clock latches
clk
clk
-Q
D
D
-clk
D
clk
-clk
-clk
Dynamic single clock registers
-clk
clk
-Q
D
-clk
National Central University
Q
D
clk
-clk
-clk
clk
Q
clk
EE613 VLSI Design
73
Clocking Strategies – Single Clock
clock
Logic
L1
L2
Logic
L2 opaque
clock
L1 opaque
L1 transparent
L2 transparent
National Central University
EE613 VLSI Design
74
Dynamic Latches – Single-Phase Clocking
Clock active high latch
D
Clock active high latch with buffer
D
X
Q
CLK
Dn
CLK
Xn
Qn
0
H
1
0
1
H
0
1
1
L
Xn-1
Qn-1
0
L
1
Qn-1
National Central University
CLK
EE613 VLSI Design
X
-Q
75
Dynamic Latches – Single-Phase Clocking
Clock active low latch
Clock active low latch with buffer
D
D
CLK
CLK
Q
X
Dn
CLK
Xn
Qn
0
L
1
0
1
L
0
1
1
H
Xn-1
Qn-1
0
H
1
Qn-1
National Central University
EE613 VLSI Design
X
-Q
76
Dynamic Latches – Single-Phase Clocking
Clock active high latch without
feedback
Clock active low latch without
feedback
D
D
X
CLK
Q
CLK
X
Q
Assume that the capacitance of node X
is 0.002pF and the leakage current I is
1nA.
Therefore, T=CV/I=0.002pFx5V/1nA=100us.
That is, the latch needs to be refreshed each 100us.
Otherwise, the output Q will become high.
National Central University
EE613 VLSI Design
77
Dynamic Registers – TSPC
Positive edge trigger register
CLK
D
B
CLK
-Q
A
D
A
B
-Q
tf
tr
The value of the hold time of this flip flop is
close to zero.
National Central University
EE613 VLSI Design
78
Phase Locked Loop Clock Techniques
•
PLL for synchronization
clock
clock
chip
chip
clock pad
PLL
clock pad
clock route
clock route
dclk
dclk
output pad
output pad
dclk+dpad
clock
dclk
dclk+dpad
clock
T1
T1=Input buffer delay
+routing RC delay
T2
data out
National Central University
dclk
T2=Clock-to-Q delay
+output buffer delay
T2
data out
EE613 VLSI Design
79
Phase Locked Loop – Clock Multiplying
Clock-multiplying PLL
Synchronize data transfer between chips
clock
chip
PLL
/4
clock pad
clock
clock route
clock
PLL
PLL
bus
dclk
system clock
output pad
dclk+dpad
clock
Synchronize the output enable signals
1.
Reduce tristate fights
2.
Improve overall timing
dclk
National Central University
EE613 VLSI Design
80
Typical Phase Locked Loop
Programmable
Frequency divider
(/n)
U
Phase Detector
reference clock fn
Charge Pump
Filter
D
VCO
Vc
nxfn
Vdd
Low-pass filter
U
Vc
D
National Central University
EE613 VLSI Design
81
Phase Locked Loop – Phase Detector
clkext
S
Q
U
Q
D
R
S
clk
R
S
clk
Q
UP
National Central University
NOP
clkext
R
EE613 VLSI Design
clk
DN
clkext
82
Phase Locked Loop – Charge Pump
•
Charge pump circuits
Vrefp
Pref
U
U
Out
Out
D
D
Vrefn
Nref
Biased by current mirror
The output current of the charge pump
can be adjusted through the control of
the current mirror.
National Central University
EE613 VLSI Design
83
Phase Locked Loop – Low-Pass Filter
•
Simple implementation of low-pass filter
In
Out
TG
NC1
•
•
NC2
The two capacitors C1 and C2 are in the order of tens of pF
The capacitor C2 is added in parallel to the simple RC lowpass filter to form a second order filter
− The stability of the system is maintained even with the
process variation of these on-chip components
•
Note that these capacitors can occupy a large portion of
the PLL
National Central University
EE613 VLSI Design
84
Phase Locked Loop – VCO
Current-starved inverter type VCO
Delay cell
I
I
I
fVCO
V
Control voltage
V-I converter
Odd number of stages
Voltage-Controlled Delay Line (VCDL) type VCO
tin
tin+t
Control voltage
National Central University
EE613 VLSI Design
85
Phase Locked Loop
U
Low-pass filter
Phase Detector
Vc
VCO
fout
D
fin
fout
fin
D
National Central University
EE613 VLSI Design
86
Phase Locked Loop – Programmable VCO
Delay cell
Delay cell
Delay cell
V-I converter
Shift register
VC
Generated clock
National Central University
EE613 VLSI Design
87
Single-Phase Logic – NP Domino Logic
•
NP-Domino Logic
− Allow pipelined system architecture
clk
-clk
nMOS
pMOS
Logic
Logic
-clk
clk
The circuit performs precharge-discharge operation when clock is low,
and all stage evaluate output levels when the clock is high.
clk section
National Central University
EE613 VLSI Design
88
Single-Phase Logic – NP-Domino Logic
•
-clk section
-clk
clk
nMOS
pMOS
Logic
Logic
clk
-clk
The circuit performs precharge-discharge operation when clock is high,
and all stage evaluate output levels when the clock is low.
National Central University
EE613 VLSI Design
89
Single-Phase Logic – NP-Domino Logic
•
A pipelined NP-Domino CMOS system
clk
A
-clk
section
clk
B
C
section
section
clk
A
B
C
National Central University
a0
a2
a1
b0
b2
b1
c0
EE613 VLSI Design
b1
b2
90
Single-Phase Logic – Clock Skew
•
Uses of clock skew to extend clock cycle (not
recommended)
Td2
Logic
clock
delay
Tc1
clock
Tc1
Td2
old data
new data
Td2 < Tc1
National Central University
EE613 VLSI Design
91
Single-Phase Logic – Avoiding Clock Skew
•
Lock-up Latch
Lock-up latch
Logic
clock
•
delay
Contra-data-direction clock
Logic
delay
National Central University
EE613 VLSI Design
clock
92
Two-Phase Clocking
•
Dynamic register
-ph1
-ph2
D
Q
ph1
ph2
ph1
ph2
ph1=1,ph2=0
C1
C2
C1
C2
ph1=0,ph2=1
National Central University
EE613 VLSI Design
93
Two-Phase Clocking
•
Failure due to clock skew
ph1
ph2
ph1=1,ph2=1
C1
National Central University
EE613 VLSI Design
C2
94
Two-Phase Clocking
•
Two-phase registers with single-polarity clocks
ph1
ph1
National Central University
ph2
ph2
EE613 VLSI Design
95
Clock Distribution
•
In a large CMOS chip, clock distribution is a serious
problem
−
−
−
−
−
−
•
Vdd=5V
Creg=2000pF (20K register bits @ 0.1pF)
Tclk=10ns
Trise/fall=1ns
Ipeak=Cdv/dt=(2000px5)/1n=10A
Pd=CVdd2f=2000px25x100=5W
Methods for reducing the values of Ipeak and Pd
− Reduce C
− Interleaving the rise/fall time
National Central University
EE613 VLSI Design
96
Clock Distribution
•
Clocking is a floorplanning problem because clock delay
varies with position on the chip
•
Ways to improve clock distribution
− Physical design
∗ Make clock delays more even
∗ At least more predictable
− Circuit design
∗ Minimizing delays using several stages of drivers
•
Two most common types of physical clocking networks
− H tree
− Balanced tree
National Central University
EE613 VLSI Design
97
Clocking Distribution – H Tree
clock
National Central University
EE613 VLSI Design
98
Clocking Distribution – Balanced Tree
clock
National Central University
EE613 VLSI Design
99
Clocking Distribution – Reducing Power
•
Techniques used to reduce the high dynamic
power dissipation
− Use a low capacitance clock routing line such as
metal3. This layer of metal can be, for example,
dedicated to clock distribution only
− Using low-swing drivers at the top level of the tree
or in intermediate levels
National Central University
EE613 VLSI Design
100
Clocking Distribution – Half-Swing Driver
Vdd
C1
clkp
C2
CA
-clkp
Vout
clkn
-clkn
C3
CB
C4
Gnd
Clock
National Central University
EE613 VLSI Design
101
I/O Structures – Pads
•
Types of pads
−
−
−
−
•
•
Vdd, Vss pad
Input pad (ESD)
Output pad (driver)
I/O pad (ESD+driver)
All pads need guard ring for latch-up protection
Core-limited pad & pad-limited pad
Core-limited pad
Pad-limited pad
PAD
PAD
I/O circuitry
I/O circuitry
National Central University
EE613 VLSI Design
102
Input Pads – ESD Protection
Input pad without ESD protection
Assume I=10uA, Cg=0.03pF, and t=1us
The voltage that appears on the gate is about 330volts
PAD
Input pad with ESD protection
PAD
National Central University
EE613 VLSI Design
103
I/O Pads – Tristate & Bidirectional Pads
Tristate pad
output-enable
OE
P
OE
OUT
PAD
N
data
D
N
P
OUT
0
X
0
1
Z
1
0
1
1
0
1
1
0
0
1
D
Bidirectional pad
PAD
National Central University
EE613 VLSI Design
104
Input Pads – Schmitt Trigger Circuit
Transfer characteristic of Schmitt trigger
Vout
VDD
Vin
VT-
VT+
VDD
1.
Hysteresis voltage VH=VT+-VT-
2.
When the input is rising, it switches when Vin=VT+
3.
When the input is falling, it switches when Vin=VT-
National Central University
EE613 VLSI Design
105
Input Pads – Schmitt Trigger Circuit
Voltage waveforms for slow input
Vout
VDD
Vin
VT+
VTTime
Schmitt trigger turns a signal with a very slow transition into a signal with a sharp
transition
National Central University
EE613 VLSI Design
106
Input Pads – Schmitt Trigger Circuit
A CMOS version of the Schmitt trigger
VDD
P1
VFP
P3
P2
Vin
Vout
N2
VFN
N3
N1
1. When the input is rising, the VGS of the transistor N2 is given by
2. When
Vin = VT +
3. Then VFN
VGS 2 = Vin − VFN
, N2 enters in conduction mode which means VGS2
= VTn
= VT + − VTn
National Central University
EE613 VLSI Design
107