3391ch6.pdf

(311 KB) Pobierz
Microelectronics, Second Edition
6
Integrated Circuit
Design
6.1 Introduction
.........................................
6 - 1
6.2 An Overview of the IC Design Process
................
6 -2
6.3 General Considerations in IC Design
.................
6 -2
6.4 Design of Small-Scale and Medium-Scale Integrated
Circuits
6 -6
..........................
6 -12
.....................................
6 -14
6.7 Gate Arrays
..........................................
6 -15
6.8 Standard Cells
.......................................
6 - 16
6.9 Programmable Logic Devices
.........................
6 - 16
6.10 Reducing Propagation Delays
Samuel O. Agbo
.........................
6 -19
•
Eugene D. Fabricius
6.11 Output Buffers
.......................................
6 -22
6.1 Introduction
Integrated circuits (ICs) are classified according to their levels of complexity: small-scale integration (SSI),
medium-scale integration (MSI), large-scale integration (LSI) and very large-scale integration (VLSI).
They are also classified according to the technology employed for their fabrication (bipolar, N metal
oxide semiconductor (NMOS), complementary metal oxide semiconductor (CMOS), etc.). The design
of integrated circuits needs to be addressed at the SSI, MSI, LSI, and VLSI levels. Digital SSI and MSI
typically consist of gates and combinations of gates. Design of digital SSI and MSI is presented in Sec. 6.3,
and consists largely of the design of standard gates. These standard gates are designed to have large noise
margins, large fan out, and large load current capability, in order to maximize their versatility.
In principle, the basic gates are sufficient for the design of any digital integrated circuit, no matter
how complex. In practice, modifications are necessary in the basic gates and MSI circuits like flip-flops,
registers, adders, etc., when such circuits are to be employed in LSI or VLSI design. For example, circuits
to be interconnected on the same chip can be designed with lower noise margins, reduced load driving
capability, and smaller logic swing. The resulting benefits are lower power consumption, greater circuit
density, and improved reliability. On the other hand, several methodologies have emerged in LSI and VLSI
design that are not based on interconnections or modification of SSI and MSI circuits. Both approaches
to LSI and VLSI design are presented in the following sections.
6 - 1
© 2006 by Taylor & Francis Group, LLC
..............................................
752745954.011.png
6 -2
Microelectronics
6.2 An Overview of the IC Design Process
The effort required for the design of an integrated circuit depends on the complexity of the circuit. The
requirement may range from several days effort for a single designer to several months work for a team of
designers. Custom design of complex integrated circuits is the most demanding. By contrast, semicustom
design of LSI and VLSI that utilize preexisting designs, such as standard cells and gate arrays, requires less
design effort.
IC design is performed at many different levels
and Fig. 6.1 is a nonunique depiction of these lev-
els. Level 1 presents the design in terms of subsys-
tems (standard cells, gate arrays, custom subcircuits,
etc.) and their interconnections. Design of the sys-
tem layout begins with the floor plan of level 3. It
does not involve the layout of individual transistors
and devices, but is concerned with the geometric ar-
rangement and interconnection of the subsystems.
Level 4 involves the circuit design of the subsystems.
Levels 2 and 5 involve system and subcircuit simula-
tions, respectively, which may lead to modifications
in levels 1 and/or 4.
Discussion here will focus primarily on the sys-
tem design of level 1 and the subsystem circuit de-
sign of level 4. Lumped under the fabrication
process of level 7 are many tasks, such as mask
generation, process simulation, wafer fabrication,
testing, etc. Broadly speaking, floor plan generation
is a part of layout. For large ICs, layout design is
often relevant to system and circuit design.
SYSTEM SPECIFICATIONS
BLOCK DIAGRAM OF SYSTEM AND SUBSYSTEMS
SYSTEM SIMULATIONS
FLOOR PLAN
CIRCUIT DESIGN OF SUBSYSTEMS
CIRCUIT SIMULATIONS
LAYOUT OF SUBSYSTEMS
THE FABRICATION PROCESS
FIGURE 6.1 Different levels in IC design.
6.3 General Considerations in IC Design
Layout or actual arrangement of components on an IC is a design process involving design tradeoffs.
Circuit design often influences layout and vice versa, especially in LSI and VLSI. Thus, it is helpful to
outline some general considerations in IC design:
Circuit components : Chip area is crucial in IC design. The usual IC components are transistors,
resistors, and capacitors. Inductors are uneconomical in area requirements and are hardly ever
used in ICs except in certain microwave applications. Transistors require small chip areas and are
heavily utilized in ICs. Resistor area requirements increase with resistor values. IC resistors generally
range between 50
and 100 k
. Capacitors are area intensive and tend to be limited to 100 pF.
Isolation regions : Usually, different components would be placed in different isolation regions. The
number of isolation regions, however, should be minimized in the interest of chip area economy by
placing more than one component in an isolation region whenever possible. For example, several
resistors could share one isolation region.
Design rules : Geometric design rules that specify minimum dimensions for features and separation
between features must be followed for a given IC process.
Power dissipation : Chip layout should allow for adequate power dissipation and avoid overheating
or development of hot spots on the chip. In low-power circuits such as CMOS ICs, device size is
determined by lithographic constraints. In circuits with appreciable power dissipation, device size
is determined by thermal constraints and may be much larger than lithographic constraints would
allow.
© 2006 by Taylor & Francis Group, LLC
752745954.012.png 752745954.013.png
Integrated Circuit Design
6 -3
Device size determination from power-density con-
siderations is illustrated with the aid of Fig. 6.2. The
device resides close to the surface of the substrate.
Thus, heat flow is approximately one-dimensional,
from the device to the substrate, although it is ac-
tually a three-dimensional flow. Assuming an infi-
nite heat sink of ambient temperature T A , substrate
thickness
DEVICE
SUBSTRATE
4 X
dQ
dt
, device of
surface area A , and temperature T A +
α
T , the rate
of heat flow toward the heat sink, d Q
/
dt , is given by
HEAT SINK
dQ
dt
= α
A
T
(6.1)
X
The power density or power dissipation per unit
area of the device is
FIGURE 6.2 Heat flow in an IC device.
P
A = α
T
(6.2)
X
Device Scaling
The trend in IC design, especially VLSI, is progressively toward smaller sizes. Scaling techniques permit
the design of circuits that can be shrunk as technological developments allow progressively smaller sizes.
Two basic approaches to scaling are full scaling and constant voltage scaling.
Constant voltage (CV) scaling : All device dimensions, both surface and vertical, are reduced by the
same scaling factor S , but voltages are not scaled. Voltages are maintained at levels compatible with
transistor-transistor logic (TTL) supply voltages and logic levels.
Scaling of device dimensions has implications for other device parameters. Full scaling tends to maintain
constant electric field strength and, hence, parameters that are no worse off, as device dimensions are
reduced, but does not ensure TTL voltage compatibility. Table 6.1 compares effect on device parameters
TABLE 6.1 Effect of Full Scaling and Constant Voltage Scaling on IC
Device Parameters
Parameter
Full Scaling
CV Scaling
Channel length, L
1 / S
1 / S
Channel width, W
1 / S
1 / S
Oxide thickness, t OX
1
/
S
1
/
S
Supply voltage, V DD
1
/
S
1
Threshold voltage, V TD
1
/
S
1
Oxide capacitances, C OX , C SW , C F OX
S
S
Gate capacitance, C g
= C OX WL
1 / S
1 / S
Transconductances, K N , K p
S
S
Current, I D
1 / S
1 / S
Power dissipation per dense, P
1
/
S 2
S
Power dissipation per unit area, P
/
A
1
S 3
Packing density
S 2
S 2
Propagation delay, t p
1 / S
1 / S 2
Power-delay product, Pt p
1
/
S 3
1
/
S
© 2006 by Taylor & Francis Group, LLC
X , thermal conductivity
Full scaling : All device dimensions, both surface and vertical, and all voltages are reduced by the
same scaling factor S .
752745954.014.png 752745954.001.png 752745954.002.png 752745954.003.png 752745954.004.png 752745954.005.png 752745954.006.png 752745954.007.png 752745954.008.png
6 -4
Microelectronics
of the two scaling approaches. A compromise that is often adopted is to use full scaling for all internal
circuits and maintain TTL voltage compatibility at the chip input/output (I/O) pins.
Although manyscalingrelationshipsarecommon to MOS fieldeffecttransistors(MOSFETs)andbipolar
ICs (Keyes, 1975), the scaling relationships of Table 6.1 more strictly applies to MOSFETs. Bipolar doping
levels, unlike MOSFETs, are not limited by oxide breakdown. Thus, in principle, miniaturization can be
carried further in bipolar processing technology. However, bipolar scaling is more complex. One reason
is that the junction voltages required to turn on a bipolar junction transistor (BJT) does not scale down
with dimensions.
Geometric Design Rules
m process. Design rules should protect
against fatal errors such as short circuits due to excessive misalignment of features, or open circuits due to
too narrow a metal or polysilicon conductive path.
Generalized design rules that are portable between processes and scalable to facilitate adaptation to
shrinking minimum geometries as processes evolve are desirable. Other advantages of generalized design
rules include increased design efficiency due to fewer levels and fewer rules, automatic translation to final
layout, layout-rule and electrical-rule checking, simulation, verification, etc. The Mead-Conway approach
(1980) to generalized design rules is to define a scalable and process-dependent parameter, lambda (
µ
m process may not be appropriate for a 0.5-
µ
),
as the maximum misalignment of a feature from its intended position on a wafer or half the maximum
misalignment of two features on different mask layers. Table 6.2 shows a version of the Mead-Conway
scalable design rules for NMOS (Fabricius, 1990).
CMOS ICs utilize both NMOS and PMOS devices. Starting with a p -substrate, the NMOS would be
fabricated on the p -substrate and the PMOS in an n -well, and vice versa. With the addition of an n -well,
p -well or twin tub process, CMOS fabrication is similar to that for NMOS, although more complex.
Table 6.3 (Fabricius, 1990) shows the Mead-Conway scalable CMOS design rules. The dimensions are
given in multiples of
λ
λ
and the rules are specified by the MOS Implementation System (MOSIS) of the
TABLE 6.2 MOS Implementation System (MOSIS) NMOS Design Rules
Mask Level
Feature
Size (times λ )
N + Diffusion
diffusion width
2
diffusion spacing
3
Implant mask
implant-gate overlap
2
implant to gate spacing
1.5
Buried contact mask
contact to active device
2
overlap with diffusion
1
contact to poly spacing
2
contact to diffusion spacing
2
Poly mask
poly width
2
poly spacing
2
poly to diffusion spacing
1
gate extension beyond diffusion
2
diffusion to poly edge
2
Contact mask
contact width
2
contact-diffusion overlap
1
contact-poly overlap
1
contact to contact spacing
2
contact to channel
2
contact-metal overlap
1
Metal mask
metal width
3
metal spacing
3
© 2006 by Taylor & Francis Group, LLC
Design rules specify minimum device dimensions, minimum separation between features and maximum
misalignment of features on an IC. Such rules tend to be process and equipment dependent. For example,
a design rule for a 2-
752745954.009.png
Integrated Circuit Design
6 -5
TABLE 6.3 MOSIS Portable CMOS Design Rules
Mask Level
Feature
Size (times
λ
)
n -well and p -well
well width
6
well to well spacing
6
n + , p + active diffusion or implant
active width
3
active to active spacing
3
source/drain to well edge
6
substrate/well contact
3
active to well edge
3
Poly mask
poly width or spacing
2
gate overlap of active
2
active overlap of gate
2
field poly to active
1
p -select, n -select
select-space (overlap) to (of) channel
3
select-space (overlap) to (of) active
2
select-space (overlap) to (of) contact
1
Simpler contact
contact size
2
×
2
to poly
active overlap of contact
2
contact to contact spacing
2
contact to gate spacing
2
Denser contact
contact size
2
×
2
to poly
poly overlap of contact
1
contact spacing on same poly
2
contact spacing on different poly
5
contact to non-contact poly
4
space to active short run
2
space to active long run
3
Simpler contact
contact size
2
×
2
to active
active overlap of contact
2
contact to contact spacing
2
contact to gate spacing
2
Denser contact
contact size
2
×
2
to active
active overlap of contact
1
contact spacing on same active
2
contact spacing on different active
6
contact to different active
5
contact to gate spacing
2
contact to field poly short run
2
contact to field poly long run
3
Metal 1
width
3
metal 1 to metal 1
3
overlap of contact to poly
1
overlap of contact to active
1
via
size
2
×
2
via-to-via separation
2
metal 1/via overlap
1
space to poly or active edge
2
via to contact spacing
2
Metal 2
width
3
metal 2 to metal 2 spacing
4
metal overlap of via
1
Overglass
bonding pad (with metal 2 undercut)
100
×
100
µ
m
probe pad
75
×
75
µ
m
pad to glass edge
6
µ
m
© 2006 by Taylor & Francis Group, LLC
752745954.010.png
Zgłoś jeśli naruszono regulamin