You are on page 1of 5

Proceediigs of the 1985ACM Computer Science Conference-Agendafor ComputingResearch:

TheChallengefor Creativity,1985March 12-14

VLSI DESIGN PROCESS

Vishwani D. Agrawal
Samuel H. C. Poon

AT &T Bell Laboratories


Murray Hill, New Jersey 07974

ABSTRACT- This paper presents a review of the computer- on to the physical design phase. The physical design
aided design of VLSI devices. For illustrative purposes, the mainly consists of mask layout. After layout, the design is
design methodology is modeled as an idealized two-level process. further verified for the effects of capacitance and resistance
In the first level, known as the functional level, the requirements due to routing and interconnection. Again, the physical
of the device are converted into logic description and tests are design and the verification form a tight iterative loop. If
generated. The second level consists of the physical design the design requirements are satisfied after the physical
involving layout. In each level, our model represents the design as
layout, the design, in the form of mask and test data, is
a feedback process with verification closing the loop on synthesis
or layout. While the future design methodology might be closer
passed on to manufacturing.
to this idealization, today’s design process differs significantly In order to minimize the design cycle, the design issues, as
depending upon the specific design environment. much as possible, must be resolved in the functional design
phase. For example, the design cycle would be lengthened
if the design requirements cannot be satisfied during the
1. INTRODUCTION physical layout primarily due to difficult to meet timing
The primary objective of VLSI design is to produce the requirements. Then the functional design would need
details of a VLSI device so that it can be physically modification and the physical design will have to be
realized. The main activities of the design process are repeated. During functional design, the CAD tools that
functional design, test generation, and physical design. help designers, are the tools for floor planning, and
The solutions of complex design problems have relied on accurate timing and testability analyses.
experienced experts and on computer-aids. However, with To elaborate upon this concept further, let us consider the
the tremendous growth in the complexity of VLSI, a interaction between the functional design and the physical
greater use of computer aids can be anticipated. design. In Fig. 1, which illustrates our model, the
In this paper, a systems approach is used to model the feedback between the design levels occurs in two ways.
design process. This model emphasizes tighter feedback The feedback, that is totally contained within a design
loops to minimize the time of design cycle. A variety of phase, provides efficient verification. However, the feedback
computer-aided design (CAD) tools that support the between the phases could lengthen the design process. For
design activities, are described. Unfortunately, a real example, certain timing problems discovered during layout
design may rarely follow the idealized model. verification may not be correctable on the layout. These
will require modifications in the functional design.
2. VLSI DESIGN CYCLE
3. COMPUTER-AIDED DESIGN
The design activities can be divided into two phases:
functional and physical. The functional phase consists of As shown in Fig. 1, the design is accomplished in two
synthesis and testability. The iterations between synthesis steps. The functional design, is supported by computer-
and verification ensure that the design will meet the aided engineering (CAE) tools. The second step, physical
requirements. Similarly, tests are verified through fault design, is supported by a variety of computer-aided design
simulation. Then the logic design and test data are passed (CAD) took.
Figure 2 shows how the computer-aids would fit in an ideal
Permissionto Copy without fee all or part of this material is granted design process. The various computer aids [ll used in this
provided that the copies are not madeor distributed for direct process are described below.
commercialadvantage,the ACM copyright noticeand the title of the
publication and its date appear, and notice is given that copying is by Requirements Analysis. The design begins with an
permission of the Association for Computing Machinery. To copy analysis of the requirements consisting of functional and
otherwise, or to republish, requires a fee and/or specific permission.
physical (package, power, cost, reliability, etc.)
specifications. The analysis is aimed at determining the
completeness and feasibility of the specifications. This
Q 1985 ACM 0-89791-150-4/85/003/0074 $00.75

74
k’roceedmgs of the 1985 ACM Computer Science Conference-Agenda for Computing Research: The Challenge for Creativity, 1985 March 12-14

DESIGN \ MANU;;;CTJRING -+

LOGIC
I DESIGN AND ’ MASK AND VLSI
REQUIREMENTS SYNTHESIS TEST DATA TEST DATA WAFER DEVICES
p AND TEST C LAYOUT w FABRICATION b
4 GENERATION - -b + AND PACKAGING -

- VERIFICATION * - VERIFICATION +- - TESTING *


a -
TIMING, TESTABILITY. TIMING, TESTABILITY,
QUALITY, RELIABILITY QUALITY. RELIABILITY
AND YIELD PROBLEMS AND YIELD PROBLEMS
Fig. 1 VLSI design and manufacturing. The three part process would, in an ideal case, be supported
by computer-aided engineering (CAE), computer-aided design (CAD), and computer-aided
manufacturing (CAM). The feedback shown in heavier lines must be reduced or eliminated for
an efficient design cycle.

VERIFICATION
4k2”KLENTS

t
ARCHITECTURE

1
AUDITS I TESTABILITY, DESIGN CAPTURE
VERIFICATION
SIMULATION FLOOR -FLANNING
TJMING ANALYSIS // \\

\ 1 SYNTHESIS 1 1 SYNTHESIS 1 1 SYNTHESIS 1 1 c&+&s

\ 1 CHIP MOC
GENERATION

VERIFICATION

,FJl+“LCIRCUIT
--I EXTRACTION I I

DESIGN AND 4 I
TEST DATA

Fig. 2 The activities of an ideal design process.

75
Proceedings of the 1985ACM Computer Science Conference-Agenda for Computing Research:The Challenge for Creativity, 1985March 12-14

analysis is commonly carried out manually by experts One such system, known as PLEX (111, uses the C
although some computer aids are possible [2]. programming language. Synthesis is performed for a fixed
Algorithms. A conceptual design produces algorithms and predetermined architecture. PLEX produces the layout
which when implemented in hardware will meet the of a special purpose microprocessor to implement the chip
functional requirements. Complex algorithms can be function. Another silicon compiler is MacPitts [121,
simulated and validated using computers. Designers write developed at MIT.
programs for this verification as there are no general Layout of chips, in the absence of CAD tools would have
purpose C.4D tools. Conceptual design of a chip is been the most laborious task. Fortunately, many layout
essentially a manual process. tools are available. Basic layout problems are placement
Architecture. The algorithms are implemented by means and interconnection 1131. Placement produces a detailed
of an interconnection of functional blocks. Typical floor plan. Often, placement is done manually and
functional blocks may be arithmetic logic unit (ALU), interconnection (or routing), automatically. However, some
memory (RAM, ROM), control unit (finite state automated placement and routing systems are in use for
machine), bus, etc. Architectural design is also a manual standard cell designs 1141, 1151. Most device (transistor)
process and its correctness must be verified. Although level layout systems [lo], 1161 are interactive and are used
simulation at this point is not very common, some CAD for layout of standard cells and super-cells (few hundred
tools have been developed. The chip is modeled in a high transistors). Large memory chips are an exception. These
level language like C or Pascal. Simulators like AIDE 131 chips contain thousands of transistors, but interactive
allow simulation of a large number of situations to cover layout tools arc still useful because the chip has a
the functional specifications. Chip stimuli (vectors) and the repetitive structure.
expected responses for logic verification are also generated. Other layout tools are layout editors 1171, design rule
Floor planning is an approximate layout of the chip checkers and circuit extractors 1181, [191. Layout editors
geometry. Sizes of functional blocks are estimated and are special purpose editors which allow probing and
they are placed on the plan. Technology (NMOS, CMOS, modification of mask description language. Layout design
bipolar, and feature size) is taken into account. Floor rule checkers audit mask description for design rules
planning provides initial information on chip size, speed, (minimum feature separation, etc.). Circuit extractors
yield, and cost. These are matched with the chip perform reverse synthesis. They convert the mask
requirements. Problems will lead to changes in architecture description into transistor level description which can be
and/or algorithms. simulated or compared with the prelayout description for
verification.
Synthesis. The design from architecture to layout is
carried out in several stages. The functional blocks in VeriJication is the process which ensures correctness of
architecture are transformed into logic (Boolean gate) design after it has been synthesized. For verification,
design, logic design is transformed into circuit (transistor normally a functional, logic or transistor level model of
or MOS device) design, and finally, the design is iaid out circuit is created. This model describes the circuit in
as a set of geometrical masks. In order to analyze and hardware description languages. Dynamic behavior is
verify the design descriptions by computers we need verified through simulation. Input stimuli and expected
hardware description languages. Hardware description responses are produced from requirements or from
languages (HDL) serve the same purpose in hardware architectural simulation. Simulators are implementations
design as the programming languages in software. Various of algorithms which compute circuit activity. Logic
HDLs describe the circuit at behavior (or register- simulators treat the circuit as interconnection of Boolean
transfer), logic, device, and mask levels 141. Synthesis gates [201. MOS simulators can simulate circuits with
refers to translation of circuit description either to physical timing information (device characteristics, parasitics,
(mask) or closer to physical level. There is no general routing delays, etc.) 1211, [221 or with transistors as ideal
synthesis procedure. Logic synthesis traditionally has been switches 1231. Circuit analysis programs 1241 represent
a manual process. There are, however, CAD tools available the most accurate tools. Because of the detailed analysis
for combinational logic minimization [51, 161. Using these they perform these programs are expensive to use. They
it is possible to synthesize control logic in the form of finite are normally used for simulating critical timing paths of
state machines and programmable logic arrays (PLAs) [7]. the chip.
Another popular style of synthesis uses standard cells [8], Other forms of verification consist of static timing analysis
[9]. Standard cells are general purpose logic blocks with a 1251, and circuit comparison (topological isomorphism) of
complexity of 30 to 40 MOS transistors and can be very extracted description with the verified prelayout description
effectively laid out by interactive CAD tools 1101. Cell 1261. Other methods of verification which are being
libraries (files cuntaining logic, device and mask level developed are based on symbolic simulation and theorem
data) are used for synthesis, verification and layout. proving techniques 1271.
More recent form of synthesis tools, still in the The errors discovered during verification must be analyzed,
development stage, are the silicon compilers. These and their causes, located and removed. These
convert the high level functional description directly into modifications often impact the high-level design also.
layout. Algorithms are, of course, devised manually. They Backward interactions of this type are the main reasons for
arc then coded in a programming style hardware language. lengthy design schedules. It is beneficial to review the

76
Fkceedmgs of the 1985ACM Computer Science Conference-Agenda for ComputingResearch:
TheChallengefor Creativity,1985March 12-14

design at various stages for possible problems of Procedure for Boolean Function,” Proc. 2lst Des.
manufacturability, reliability, verifiability, testability and Auto. Conf, Albuquerque, N.M., June 1984, pp.
design updatability. Such audits are mostly manual but 699-702.
there is scope for knowledge-based expert systems in this I71 P. Agrawal and M. J. Meyer, “Automation in the
area 1281. Design of Finite State Machines,” VLSI Design, Vol.
Testing. A chip, designed and verified, can not be V, pp. 74-84, September 1984.
produced unless tests are available. Processed silicon 181 N. J. Elias and A. W. Wetzel, “The IC Module
wafers are tested on automatic test equipment (ATE). Compiler, A VLSI System Design Aid,” Proc. 20th
Each chip on a wafer is tested under the control of a test Des. Auto. Conf., Miami Beach, FL, June 1983, pp.
program which applies enough stimuli to the chip so as to 46-49.
test all (or most) internal nodes. Test stimuli (in the form
of binary vectors) are generated as part of the design 191 J. Dussault, C-C Liaw, and M. M. Tong, “A High
process. These stimuli are different from verification Level Synthesis Tool for MOS Chip Design,” Proc.
stimuli used in simulation. Unlike verification stimuli, test 21st Des. Auto. Conf., Albuquerque, N.M., June
1984, pp. 308-314.
stimuli have a fault coverage requirement. Tests are
expected to cover a high percentage (typically, 90 to 100 [lo] A. D. Lopez and H-F S. Law, “A Dense Gate Matrix
percent) of stuck-type faults [291. A fault simulator I301 Layout Method for MOS VLSI,” IEEE J. Sol. State
is normally used for evaluating the fault coverage of tests. Circuits, Vol. ED-27, pp. 1671.1675.
It is difficult to guarantee high fault coverage for very I1 11 M. R. Buric, et al, “Plex: Automatically Generated
large chips unless testability is considered in the early Microcomputer Layouts,” Proc. IEEE Int. Conf
phases of design. Testability analysis tools 1311 provide Comp. Des., Port Chester, N.Y., October 1983, pp.
useful information about the testability of logic design. 181-184.
The concepts of design for testability take a step further in I121 R. Southard, “MacPitts: An Approach to Silicon
defining the design rules which will result in testable Compilation,” Computer, Vol. 16, December 1983.
design such that the whole process of test generation can
be completely automated I321. Of course, the cost of I131 J. Soukup, “Circuit Layout,” Proc. IEEE, Vol. 69,
testability is a slight amount (10 to 20 percent) of added pp. 1281-1304, October 1981.
hardware. This added hardware is justified for very large I141 G. Persky, D. N. Deutsch, and D. G. Schweikert,
chips where high fault coverage (and high product quality) “LTX - A Minicomputer based System for
can not be otherwise guaranteed. Automated LSI Layout,” J. Des. Auto. Fault
Tolerant Comp., Vol. 1, pp. 217-255, May 1977.
4. CONCLUSION
I151 A. E. Dunlop, V. D. Agrawal, D. N. Deutsch, M. F.
In this paper we have presented the VLSI design as an Jukl, P. Kozak, and M. Wiesel, “Chip Layout
idealized model. CAD tools, that support various activities Optimization using Critical Path Weighting,” Proc.
of design, are reviewed. While today’s design process is Des. Auto. Conf., Albuquerque, N.M., June 1984,
far from being ideal, the model gives some insight into the pp. 133-l 36.
future.
1161 N. Weste, “Virtual Grid Symbolic Layout,” Proc.
18th Des. Auto. Conf, Nashville, TN, June 1981, pp.
REFERENCES 225-233.
I171 D. H. Potter, J. D. Tauke, and A. A. Yiannoulos,
111 V. D. Agrawal,“Computer-Aids in VLSI Design,” “Parameterized Procedural Module Generation for
International Conference on Computers, Systems & Physical Integrated Circuit Design,” Proc. IEEE Int.
Signal Processing, Bangalore, India, December 1O- Conf Comp. Des., Port Chester, N.Y., Oct.-Nov.,
12, 1984. 1983, pp. 290-293.
121 M. Dorfman and R. F. Flynn, “Arts - An Automated [18l P. A, Swartz, B. R. Chawla, T. R. Luczejko, K.
Requirements Traceability System,” J. Syst. Mednick, and H. K. Gummel, “HCAP - A
Software, Vol. 4, pp. 63-74, April 1984. Topological Analysis Program for IC Mask
[3] D. J. Ellenberger and Y. W. Ng, “AIDE - A Tool for Artwork,” Proc. IEEE Int. Conf Comp. Des., Port
Computer Architecture Design,” Proc. 18th Des. Chester, N.Y., Oct.-Nov., 1983, pp. 298-301.
Auto. Conf. Nashville, TN, June 1981, pp. 796-803. I191 T. G. Szymanski and C. J. Van Wyk, “Space
141 S. G. Shiva, “Computer Hardware Description Efficient Algorithms for VLSI Artwork Analysis,”
Languages - A Tutorial,” Proc. IEEE, Vol. 67, pp. Proc. 20th Des. Auto. Conf, Miami Beach, FL, June
1605-1615, December 1979. 1983, pp. 734-739.
151 S. J. Hong, R. G. Cain, and D. L. Ostapko, “MINI: I201 S. G. Chappell, C. H. Elmendorf, and L. D.
A Heuristic Approach for Logic Minimization,” IBM Schmidt, “LAMP: Logic-Circuit Simulators,” Bell
J. Res. Dev., Vol. 18, pp. 443-458, September 1974. Syst. Tech. J., Vol. 53, pp. 1451-1476, October 1974.
161 N. N. Biswas, “Computer Aided Minimization I211 B. R. Chawla, H. K. Gummel, and P. Kozak,

77
Proceedings of the 1985ACM Computer Science Conference-Agenda for Computing Research:The Challenge for Creativity, 1985March 12-14

“MOTIS - An MOS Timing Simulator,” IEEE [271 A. E. Ruehli and G. S. Ditlow, “Circuit Analysis,
Trans. Cir. Syst., Vol. CAS-22, pp. 901-910, Logic Simulation, and Design Verification for VLSI,”
December 1975. Proc. IEEE, Vol. 71, pp. 34-48, January 1983.
[22] V. D. Agrawal, A. K. Bose, P. Kozak, H. N. Nham, 1281 V. E. Kelly, “The CRITTER System - Automated
and E. Pacas-Skewes, “Mixed-Mode Simulation in Critiquing of Digital Circuit Designs,” Proc. 2Zst
the MOTIS System,” J. Digital Systems, Vol. V, pp. Des. Auto. Conf., Albuquerque, N.M., June 1984,
383-400, Winter 1981. pp. 419-425.
1231 R. E. Bryant, “A Switch-Level Model and Simulator [29] S. C. Seth and V. D. Agrawal, “Automated Chip
for MOS Digital Systems,” IEEE Trans. Comput., Testing,” IEEE Spectrum, to be published.
Vol. C-33, pp. 160-177, February 1984.
1301 P. Gael and P. R. Moorby, “Fault Simulation
1241 L. W. Nagel, “ADVICE for circuit simulation,” Proc. Techniques for VLSI Circuits,” VLSI Design, Vol. V,
Znt. Symp. Cir. and Syst., Houston, TX, April 28-30, pp. 22-26, July 1984.
1980, pp. 28-30.
[31] D. M. Singer, “Testability Analysis of MOS VLSI
[251 V. D. Agrawal, “Synchronous Path Delay Analysis in Circuits,” Proc. Znt. Test Co&, Philadelphia, PA,
MOS Circuit Simulators,” Proc. Des. Auto. ConjI, October 16-18, 1984, pp. 690-696.
Las Vegas, Nevada, June 1982, pp. 629-635.
1321 V. D. Agrawal, S. K. Jain, and D. M. Singer, “A
[261 S. McCabe, “A Program for Hierarchical CAD System for Design for Testability,” PZSZ
Verification of VLSI Layouts,” VLSZ Design, Vol. V, Design, Vol. V, pp. 46-54, October 1984.
pp. 70-76, August 1984.

78

You might also like