Nof S.Y. Springer Handbook of Automation

Подождите немного. Документ загружается.

Computer-Aided Design, Computer-Aided Engineering, and Visualization 37.6 3-D Animation Production Process 645

37.6 3-D Animation Production Process

Computer animations and simulations are commonly

used in the engineering design process to visualize

movement of parts, determine possible interferences of

parts, and to simulate design analysis attributes such

as ﬂuid and thermal dynamics. The 3-D model ﬁles of

most CAD systems can be converted into a format that

can be used as input into popular animation software

programs, such as Maya and 3ds Max. Once the ﬁles

have been input into the animation software program,

the animation process can begin. The animation process

can be quite complex, depending on the level of realism

necessary for the design visualization. This section will

describe the steps necessary to create design animations

from CAD models.

The 3-D animation production process can be di-

vided into three main phases:

•

Concept development and preproduction

(Sect.37.6.1)

•

Production (Sect.37.6.2)

•

Postproduction and delivery (Sect.37.6.3).

37.6.1 Concept Development

and Preproduction

Several key activities take place during this phase,

including story development and visual design, pro-

duction planning, storyboarding, soundtrack recording,

animation timing, and production of the animatic.

Every animation tells a story, “...you need not to

have characters to have a story ...”[37.9]; for example

Fig. 37.5 An example of preliminary storyboard illustrating the futuristic assembly process of a Boeing 787 (courtesy of

Purdue University, with permission from N. Adamo-Villani, C. Miller)

an architectural walkthrough or a medical visualization

has a story in the sense that the events progress in a log-

ical and effectively developed way. Story development

begins with a premise – an idea in written form [37.10].

When the premise is approved, it is expanded into an

outline or treatment – a scene-by-scene description of

the animation – and the treatment is ﬂeshed out into

afullscript. Visual design is carried out concurrently

to story development. It is during this conceptual stage

that the visual style of the animation is deﬁned, charac-

ter/object/environment design is ﬁnalized and approved

for production, and the story idea is translated into a vi-

sual representation – the storyboard.

A storyboard is a sequence of images – panels –

and textual descriptions describing the story, design, ac-

tion, pacing, sound track, effects, camera angles/moves,

and editing of the animation. Figure 37.5 shows an ex-

ample of a preliminary storyboard. Animation timing

is the process of pacing the action on the storyboard

panels in order to tell the story in a clear and effec-

tive manner. The most common method of timing is

the creation of a story reel or animatic. The creation

of the animatic [37.11]: “... is essentially the process

of combining the sound track with the storyboard to

pace out the sequence”. In addition to scanned-in sto-

ryboard panels with digitized soundtrack, the animatic

can include simulated camera moves and rough mo-

tion of characters and objects. The main purpose of

the animatic is to show the ﬂow of the story by block-

ing the timing of the individual shots and deﬁning the

transitions between them. It provides an opportunity to

Part D 37.6

646 Part D Automation Design: Theory and Methods for Integration

experiment with different cinematic solutions and to

visualize whether the ﬁnal animation makes sense as

a ﬁlmic narrative.

37.6.2 Production

The production phase includes the following activities:

3-D modeling, texturing, rigging, animation, camera

setup, lighting, and rendering.

Modeling

“Modeling is the spatial description and placement

of objects, characters, environments and scenes with

a computer system” [37.12]. In general, 3-D models for

animation are produced using one of four approaches:

surface modeling, particle-systemmodeling, procedural

modeling or digitizing techniques.

In surface modeling surfaces are created using

spline, polygon or subdivision surfaces modeling meth-

ods. A spline model consists of one or several patches,

i.e., surfaces generated from two spline curves. Dif-

ferent splines generate different types of patches; the

majority of spline models used in 3-D animation con-

sist ofnonuniform rational B-splines (NURBS) patches,

which are generated from NURBS curves. A polygonal

model consists of ﬂat polygons, i. e., multisided objects

composed of edges, vertices,andfaces;asubdivision

surface results from repeatedly reﬁning a polygonal

mesh to create a progressively ﬁner mesh. Each sub-

division step reﬁnes a submesh into a supermesh by

inserting more vertices. In this way several levels of de-

tail are created, allowing highly detailed modeling in

isolated areas.

Common techniques used to create surface models

include: lathe, extrude,loft,andBoolean operations;for

instance, a polygonal mesh or a NURBS surface can

be created by drawing a curve in space and rotating it

around an axis (lathe or revolve); or by drawing a curve

and pushing it straight back in space (extrude); or by

connecting a series of contour curves (loft). Boolean

operators allow for combination of surfaces in vari-

ous ways to produce a single piece of geometry. Three

Boolean operations are commonly used in 3-D model-

ing for animation: addition or union, subtraction,and

intersection. The addition operation combines two sur-

faces into a single, uniﬁed surface; the subtraction

operation takes away from one object the space occu-

pied by another object, and the intersection operation

produces an object consisting of only those parts shared

by two objects, for instance, overlapping parts.

Particle-system modeling is an approach used to

represent phenomena as ﬁre, snow, clouds, smoke, etc.

which do not have a stable and well-deﬁned shape.

Such phenomena would be very difﬁcult to model with

surface or solid modeling techniques because they are

composed of large amounts of molecule-sized particles

rather than discerniblesurfaces. In particle-systemmod-

eling, the animator creates a system of particles, i. e.,

graphical primitives such as points or lines, and de-

ﬁnes the particles’ physical attributes. These attributes

control how the particles move, how they interact with

the environment, and how they are rendered. Dynam-

ics ﬁelds can also be used to control the particles’

motion.

Procedural modeling includes a number of tech-

niques to create 3-D models from sets of rules.

L-systems, fractals,andgenerative modeling are exam-

ples ofprocedural modelingtechniques sincethey apply

algorithms for producing scenes; for instance, a terrain

model can be produced byplotting anequation of fractal

mathematics that recursively subdivides and displaces

a patch.

When a physical model of an object already exists,

it is possible to create a corresponding 3-D model using

various digitizing methods. Examples of digitizing tools

include 3-D digitizing pens and laser contour scanners.

Each time the tip of a 3-D pen touches the surface of the

object to be digitized, the locationof a point isrecorded.

In this way it is possible to compile a list of 3-D co-

ordinates that represent key points on the surface. The

3-D modeling software uses these points to build the

corresponding digital mesh, which is often a polygonal

surface. In laser contour scanning the physical object is

placed on a turntable, a laser beam is projected onto its

surface, and the distance the beam travels to the object

is recorded. After each 360

◦

rotation a contour curve

is produced and the beam is lowered a bit. When all

contour curves have been generated, the 3-D software

builds a lofted surface.

Surface models can be saved to a variety of for-

mats. Some ﬁle formats are exclusive to speciﬁc

software packages (proprietary formats), while oth-

ers are portable, which means they can be exchanged

among different programs. The two most common

portable formats are “.obj” (short for object) introduced

by Alias for high-end computer animation and visual

effects productions, and the drawing interchange for-

mat (DXF) developed by Autodesk and widely used

to exchange models between CAD and 3-D animation

programs.

Part D 37.6

Computer-Aided Design, Computer-Aided Engineering, and Visualization 37.6 3-D Animation Production Process 647

Texturing

Texturing is the process of deﬁning certain characteris-

tics of a3-D surface suchas color, shininess,reﬂectivity,

transparency, incandescence, translucence, and smooth-

ness. Frequently, these characteristics or parameters are

treated as a single set called a shader or a computer

graphics (CG) material.Ashader parameter can be as-

signed a single value (for, example the color parameter

can be assigned an RGB value of 255,0,0; in this case

the entire surface is red), or the value can vary across

the surface. Two-dimensional (2-D) texture mapping is

a method of varying the texture parameter values across

the surface using 2-D images. For example a 2-D pic-

ture can be applied to a 3-D surface to produce a certain

color or transparency pattern. The 2-D picture can be

a digital photo, a scanned image, an image produced

with a 2-D paint program, or it can be generated pro-

cedurally. In general, the application of the 2-D image

to the 3-D surface can be implemented in two ways:

by projecting the image onto the surface (projection

mapping) or by stretching it across the surface (pa-

rameterized mapping). In certain situations 2-D texture

mapping does not produce realistic texture effects; for

instance, a virtual block of marble rendered using 2-D

texture techniques will appear to be wrapped in marble-

patterned paper, rather than made of marble. To solve

this problem it is possible to use another technique

called solid texture mapping. The idea behind this tech-

nique is that you create a virtual volume of texture and

you immerse your object in that volume [37.9]. Many

3-D software packages allow for creation of procedural

3-D textures. Figure 37.6 shows a rendering produced

using a variety of texturing techniques.

Fig. 37.6 A 3-D image produced using a variety of texture

maps: color, transparency, bump, reﬂection, and translu-

cence (courtesy of Purdue University, with permission

from N. Adamo-Villani)

Rigging

Rigging is the process of setting up a 3-D object or

character for animation. Common rigging techniques

include: forward kinematics (FK)andinverse kinemat-

ics (IK), hierarchical models, skeletal systems, limits,

and constraints. When the 3-D object/character to be

animated is made of multiple segments, the segments

(or nodes) can be organized in an FK hierarchical

model.InanFK hierarchy a node just below another

node is called a child, while the node just above is

called a parent, and the ﬂow of transformations goes

from parent to child. In order to animate an FK hi-

erarchical model, each node needs to be selected and

transformed individually to attain a certain pose. This

process can become complicated and tedious when the

model is very elaborate; for example, imagine a situ-

ation in which the animator needs to place the hand

of a 3-D human-like character on a particular object.

With an FK model, the animator has to ﬁrst rotate

the shoulder, then the lower arm, then the wrist, hand

and ﬁngers, working from the top of the hierarchy

down. He cannot select the hand and place it on the

object because the other parts of the arm will not fol-

low, as they are parents of the hand. This problem can

be solved by creating an inverse kinematics model in

which the transformations travel upward through the

hierarchy, for instance, from the hand to the shoul-

der. In this case the animator can place the hand on

the object and have the other segments (lower arm

and upper arm) follow the motion. An IK model is

also called an IK chain and each node is referred to

as a link. The ﬁrst link of the chain is called the

root of the chain and the end point of the last link

is called the effector (as it affects the positions of all

the other links, i. e., it effects the IK solution). In prin-

ciple, each link in a chain can rotate any number of

degrees around any axis. While these unrestricted ro-

tations may be appropriate in certain situations, they

are not likely to produce realistic results when the IK

system is applied, for instance, to a human or ani-

mal model. This is due to the fact that human and

animal joints have rotational limits; for example, the

knee joint cannot bend beyond ≈180

◦

. To solve this

problem it is common to set up limits and constraints.

Limits can be deﬁned for any of a model’s three ba-

sic transformations (translation, rotation and scale); in

addition, the transformations can be constrained (i.e.,

associated) to the transformations of other objects. Po-

sition or point, rotation or orient,anddirection or

aim are common types of constraints used in 3-D ani-

mation.

Part D 37.6

648 Part D Automation Design: Theory and Methods for Integration

Many 3-D animation packages allow for creation of

both FK and IK models. In general, to rig complex char-

acters or objects the animator creates a skeletal system,

i.e., a hierarchical model composed of joints connected

by bones. The segments that make up the 3-D char-

acter/object are parented to the joints of the skeleton

and the skeleton can function as an FK or IK model,

with the possibility to switch between the two modes

of operation during the animation process. If the 3-D

object/character is supposed to deform during motion,

the 3-D geometry can be attached (or skinned)tothe

skeletal joints; in this case the skeleton functions as

a deformation system.

Animation

Common 3-D animation techniques include: keyframe,

motion path, physically-based,andmotion capture ani-

mation.Inkeyframe animation, the animator sets key

values to various objects’ parameters and saves these

values at particular points in the timeframe; this pro-

cess is called setting keyframes. After the animator

has deﬁned the keyframes, the 3-D software interpo-

lates the values of the object’s parameters between the

keyframes. To gain more control of the interpolation,

a parameter curve editor is available in the major-

ity of 3-D animation packages. The parameter curve

editor shows a graphical representation of the varia-

tion of a parameter’s value over time (the animation

curve. The animation curves are Bézier curves whose

control points are the actual keyframes. Two tangent

handles (vectors) are available at each control point (or

keyframe) and allow the animator to manipulate the

shape of the curve and thus the interpolation.

Motion path animation is used when the object to

be animated needs to follow a well-deﬁned path (for in-

stance, a train moving along the tracks). In this case the

animator draws the path, attaches the object to the path,

and deﬁnes the number of frames required to reach the

end of the path. In addition, the animator has the abil-

ity to control the rate of motion along the path, and the

orientation of the object. The main advantage of motion

path over keyframe animation is that, if the path is mod-

iﬁed, the animation of the object that follows it updates

to the path’s changes.

Physically based animation is based on dynamics

methods and is used to generate physically accurate

simulations. Several steps are required to set up a dy-

namic simulation including deﬁnition of the objects’

physical attributes (i.e., mass, initial velocity, elastic-

ity, etc.); modeling of dynamic forces acting upon the

objects; and deﬁnition of collisions. A dynamic simu-

lation can be baked before the animation is rendered;

baking is the process of generating an animation curve

for each parameter of an object whose change over time

is caused by the dynamic simulation.

Motion capture animation, also referred to as per-

formance animation or digital puppetry [37.13]:

... involves measuring an actor’s (or object) posi-

tion andorientation in physicalspace andrecording

that information in a computer-usable form.

In general the position or orientation of the actor is

measured bya collectionof inputdevices (optical mark-

ers or sensors) attached to the actor’s body. Each input

device has three degrees of freedom (DOF) and pro-

duces 3-D rotational or translational data which are

channeled to the joints of a virtual character. As the ac-

tor moves, the input devices send data to the computer

model. These data are used to control the movements

of the character in real time, and to generate the ani-

mation curves. Motion capture animation is often used

when the animation of the 3-D character needs to match

the performance of the actor very precisely.

Camera Setup

The point of view from which a scene is observed is

deﬁned by the CG camera. The point of view is deter-

mined by two components or nodes: the location of the

camera and the camera center of interest. The location

is a point in space, while the center of interest can be

speciﬁed as a location in space (i. e., a triplet of XYZ

coordinate values) or as camera direction (i. e., a triplet

of XYZ rotational values). In addition to location and

center of interest, important attributes of a CG camera

include the zoom parameter [whose value determines

the width of the ﬁeld of view (FOV) angle], depth of

ﬁeld,andnear and far clipping planes,usedtoclip the

viewable (and therefore renderable) 3-D space in the Z

direction.

Lighting

The process of lighting a CG scene involves selecting

the types of light to be used, deﬁning their attributes,

and placing them in the virtual environment. Common

types of light used in CG lighting include ambient

lights, spotlights, point lights,anddirectional lights.

An ambient light simulates the widely distributed, indi-

rect light that has bounced off objects in the 3-D scene

and provides a uniform level of illumination; a point

light emanates light in all directions from a speciﬁc

location in space (simulating a light bulb); a spotlight

is deﬁned by location and direction and emits light in

Part D 37.6

Computer-Aided Design, Computer-Aided Engineering, and Visualization 37.6 3-D Animation Production Process 649

a cone-shaped beam of variable width; and a directional

light (orinﬁnite light) is assumed tobe located inﬁnitely

far away and simulates the light coming from the sun.

Common parameters of CG lights are: intensity, color,

falloff (i.e., decrease of intensity with distance from

the light source), and shadow characteristics such as

shadow color, resolution, and density. In general, CG

shadows are calculated using two popular techniques:

ray-tracing and shadow depth-maps. Ray-tracing traces

the path of a ray of light from the light source and deter-

mines whether objects in the scenes would block the ray

to create a shadow; depth-mapped shadows useapre-

calculated depth map to determine the location of the

shadows in the scene. Each pixel in the depth map rep-

resents the distance from the light source to the nearest

shadow-casting surface in a speciﬁc direction. During

rendering the light is cut off at the distances speciﬁed by

the depth map with the result of making the light appear

to be blocked by the objects. The shadows in Fig.37.7

were generated using this technique.

Rendering

Rendering is the process of producing images from

3-D data. Most rendering algorithms included in 3-D

animation software packages use an approach called

scan-line rendering.Ascan line is a row of pixels

in a digital image; in scan-line rendering the program

calculates the color of each pixel one after the other,

scan line by scan line. The calculation of the pixels’

colors can be done using different algorithms such as

ray-casting, ray-tracing,andradiosity. The idea behind

Lights–91 spotlights

Intensity– 0.2

Falloff–52

Depth map resolution–600

Shadow density–1.35

All geometry resides within

the blue circle

Fig. 37.7 A 3-D image rendered using a lighting setup composed of a dome of spotlights with depth-mapped shadows

(courtesy of Purdue University, with permission from N. Adamo-Villani, C. Miller)

ray-casting is to cast rays from the camera location, one

per pixel, and ﬁnd the closest object blocking the path

of that ray. Using the material properties and the effect

of the lights on the object, the ray-casting algorithm can

determine the shading of this object. Ray-casting algo-

rithms do not render reﬂections and refractions because

they render the shading of each surface in the scene as

if it existed in isolation, in other words other objects

in the scene do not have any effect on the object be-

ing rendered. Ray-tracing addresses this limitation by

considering all surfaces in the scene simultaneously. As

each ray per pixel is cast from the camera location, it

is tested for intersection with objects within the scene.

In the event of a collision, the pixel’s color is updated,

and the ray is either recast or terminated based on ma-

terial properties (such as reﬂectivity and refraction) and

maximum recursion allowed. Although the ray-tracing

algorithm can represent optical effects in a fairly realis-

tic way, it does not render the diffuse reﬂection of light

from one surface to another. This effect happens, for in-

stance, when a blue object is close to a white wall. Even

if the wall has very low specularity and reﬂectivity it

will take a bluish hue because light bounces in a very

diffuse way from the surface of the object to the surface

of the wall. It is possible to render this phenomenon us-

ing a radiosity algorithm, which divides the surfaces in

the scene into smaller subsurfaces or patches. A form

factor is computed for each pair of subsurfaces; form

factors are a coefﬁcient describing how well the patches

can see each other. Patches that are far away from each

other, or oriented at oblique angles relative to one an-

Part D 37.6

650 Part D Automation Design: Theory and Methods for Integration

other, will have small form factors while patches that

are close to each other and facing each other 100%

will have a form factor close to 1. The form factors are

used as coefﬁcients in a linearized form of the render-

ing equation, which yields a linear system of equations.

Solving this system yields the radiosity, or brightness,

of each patch, taking into account diffuse interreﬂec-

tions and soft shadows [37.14].

Key activities of the preproduction and production

phases are illustrated in Fig.37.8.

37.6.3 Postproduction

The postproduction phase includes two main activities:

digital compositing and digital output.

Digital Compositing

In general an animated sequence includes images from

multiple sources that are integrated into a single,

seamless whole. Digital compositing is the process

of digitally manipulating and combining at least two

source images to produce an integrated result [37.15];

1. Conceptual design 2. Modeling 3. Texturing

4. Rigging 5. Lighting/camera setup (3-D layout) 6. Animation and final rendering

Fig. 37.8 Image showing key activities of the 3-D animation production process (courtesy of Purdue University and

Educate for Tomorrow, Inc., with permission from N. Adamo-Villani, R. Giasolli)

for example, Fig.37.9 is a composite created from

three different original images: the buildings, roads,

and grass areas are a CG rendering produced from

a three-dimensional model; the trees are also computer

generated 3-D imagery rendered in a different pass as

paint effects strokes; and the sky is a digital photograph

projected onto a 3-D dome and used as a backdrop. In

addition many of the elements in the image had some

additional processing performed on them as they were

added to the scene (for example, color and size adjust-

ments).

Digital Output

Computer animation sequences can be output in the

form of digital ﬁles, video or ﬁlm. Some of the most

popular digital ﬁle formats for saving animation se-

quences include: QuickTime, Motion Pictures Expert

Group (MPEG), audio video interleaved (AVI), and

Windows Media. The QuickTime format stores both

video and audio data. It is a cross-platform format

that supports different spatial and temporal resolutions

and provides a variety of compression options. MPEG

Part D 37.6

Computer-Aided Design, Computer-Aided Engineering, and Visualization References 651

(developed by the Motion Pictures Expert Group) is

a popular format for compressing animations; the data

compression is based on the removal of data that are

identical or similar not just within the frame, but also

between frames [37.12]. The AV I format (introduced

by Microsoft in 1992) is a generic Windows-only for-

mat for moving images; a more recent version of this

format is Windows Media, which offers more efﬁcient

compression for streaming high-resolution images.

Output on video can be done on a variety of video

formats (both analog and digital). Commonly used dig-

ital formats include high-deﬁnition formats such as D6,

HD-D5, HDCAM, DVCPRO and standard-deﬁnition

formats suchas D1, Digital Betacam, IMX, DV(NTSC),

D-VHS, and DVD.

Conclusions and Emerging Trends

This chapter provided an overview of computer-aided

design and computer-aided engineering and included

elements of computer graphics, animation, and visual-

ization. Today’s commercial brands of 3-D modeling

tools essentially contain many of the same functions,

irrespective of which software vendor is chosen. CAD

software programs are dimension driven, parametric,

feature based, and constraint based all at the same time,

and these terms have come to be synonymous when de-

scribing modern CAD systems. Computer animations

and simulations are commonly used in the engineering

design process to visualize movement of parts, deter-

mine possible interferences of parts, and to simulate

design analysis attributes such as ﬂuid and thermal dy-

namics. Today most CAD systems 3-D model ﬁles can

be converted into a format that can be used as in-

put into popular animation software programs, such as

Maya and 3ds Max. However, in the future it is antic-

ipated that there will be a tighter integration between

CAD and animation programs. CAD vendors will be

Fig. 37.9 An example composite image (courtesy of Purdue Uni-

versity, with permission from N. Adamo-Villani, G. Bertoline and

M. Sozen)

under greater pressure to partner large enterprise soft-

ware companies and become more product life cycle

management (PLM) centric. This will result in CAD

and animation becoming a part of a larger suite of soft-

ware products used in industry. The rapid development

of information technology and computer graphics tech-

nology will impact the hardware platforms and software

development related to CAD and animation. This will

result in even more feature-rich CAD software pro-

grams and capabilities. Faster screen refresh rates of

largeCAD models,the abilityto collaborateat greatdis-

tances in real time, shorter animation rendering times,

and higher-resolution images are a few improvements

that will result from the rapid development of infor-

mation technology and computer graphics technology.

Overall, there is an exciting future for CAD and ani-

mation in the future that will result in many positive

impacts and changes for industries and businesses that

depend on CAD and animation as a part of their day-to-

day business.

References

37.1 E.N. Wiebe: 3-D constraint-based modeling: ﬁnd-

ing common themes, Eng. Des. Graph. J. 63(3),

15–31 (1999)

37.2 P.J. Hanratty: Parametric/relational solid mod-

eling. In: Handbook of Solid Modeling,ed.

by D.E. Lacourse (McGraw-Hill, New York 1995)

pp. 8.1–8.25

37.3 G.R. Bertoline, E.N. Wiebe: Fundamentals of

Graphic Communications, 5th edn. (McGraw-Hill,

Boston 2006)

37.4 N.W. Hartman: Deﬁning expertise in the use of

constraint-based CAD tools by examining practic-

ing professional, Eng. Des. Graph. J. 69(1), 6–15

(2005)

37.5 N.W. Hartman: The development of expertise in

the use of constraint-based CAD tools: examining

practicing professionals, Eng. Des. Graph. J. 68(2),

14–25 (2004)

37.6 S.K. Bhavnani, B.E. John: Exploring the unrealized

potential of computer-aided drafting, Proc. CHI’96

(1996) pp. 332–339

37.7 S.K. Bhavnani, B.E. John: From sufﬁcient to efﬁ-

cient usage: An analysis of strategic knowledge,

Proc. CHI’97 (1997) pp. 91–98

Part D 37

652 Part D Automation Design: Theory and Methods for Integration

37.8 S.K. Bhavnani, B.E. John, U. Flemming: The

strategic use of CAD: An empirically inspired,

theory-based course, Proc. CHI’99 (1999) pp. 183–

190

37.9 M. O’Rourke: Principles of Three-Dimensional Com-

puter Animation, 3rd edn. (Norton, New York 2003)

37.10 J.A. Wright: Animation Writing and Development

(Focal, Oxford 2005)

37.11 C. Winder, Z. Dowlatabadi: Producing Animation

(Focal, Oxford 2001)

37.12 I.V. Kerlow: The Art of 3-D: Computer Animation

and Effects, 2nd edn. (Wiley, Indianapolis 2000)

37.13 S. Dyer, J. Martin, J. Zulauf: Motion capture white

paper (1995)

37.14 C. Goral, K.E. Torrance, D.P. Greenberg, B. Bat-

taile: Modeling the interaction of light between

diffuse surfaces, Comput. Graph. 18(3), 213–222

(1984)

37.15 R. Brinkmann: The Art and Science of Digital Com-

positing (Morgan Kaufmann, Sand Diego 1999 )

Part D 37

653

Design Autom

38. Design Automation for Microelectronics

Deming Chen

Design automation or computer-aided design

(CAD) for microelectronic circuits has emerged

since the creation of integrated circuits (IC). It

has played a crucial role to enable the rapid

development of hardware and software systems

in the past several decades. CAD techniques are

the key driving forces behind the reduction of

circuit design time and the optimization of circuit

quality. Meanwhile, the exponential growth of

circuit capacity driven by Moore’s law prompts new

and critical challenges for CAD techniques. Moore’s

law describes an important trend in the history of

the semiconductor industry: that the number of

transistors per unit chip area would be doubled

approximately every 2 years. This observation was

ﬁrst made by Intel co-founder Gordon E. Moore in

a paper in 1965. Moore’s law has held true for the

past four decades, and many people believe that it

will continue to apply for at least another decade

before reaching the fundamental physical limits of

device fabrication.

In this chapter we will introduce the funda-

mentals of design automation as an engineering

ﬁeld. We begin with several important processor

38.1 Overview.............................................. 653

38.1.1 Background

on Microelectronic Circuits............. 653

38.1.2 History

of Electronic Design Automation .... 656

38.2 Techniques

of Electronic Design Automation ............ 657

38.2.1 System-Level Design..................... 657

38.2.2 Typical Design Flow ...................... 658

38.2.3 Veriﬁcation and Testing ................ 662

38.2.4Technology CAD ............................ 663

38.2.5Design for Low Power ................... 664

38.3 New Trends and Conclusion ................... 665

References .................................................. 667

technologies and several existing IC technologies.

We then present a typical CAD ﬂow covering all

the major steps in the design cycle. We also

cover some important topics such as veriﬁcation

and technology computer-aided design (TCAD).

Finally, we introduce some new trends in design

automation.

38.1 Overview

Microelectronic circuits are ubiquitous nowadays. We

can ﬁnd them not only in desktop computers, lap-

tops, and workstations,butalso inconsumer electronics,

home and ofﬁce appliances, automobiles, military ap-

plications, telecommunication applications, etc. Due

to different requirements of these applications, cir-

cuits are designed differently to pursue unique fea-

tures suitable for the speciﬁc application. In general,

these devices are built with two fundamental and or-

thogonal technologies: processor technology and IC

technology.

38.1.1 Background

on Microelectronic Circuits

Processor technology refers to the architecture of the

computation engine used to implement the desired

functionality of an electronic circuit. It is categorized

into three main branches: general-purpose proces-

sors,application-speciﬁc instruction setprocessors,and

single-purpose processors [38.1]. A general-purpose

processor, or microprocessor, is a device that executes

software through instruction codes. Therefore, they are

Part D 38

654 Part D Automation Design: Theory and Methods for Integration

software-programmable. This processor has a program

memory that holds the instructions and a general data-

path that executes the instructions. The general datapath

consists of one or several general-purpose arithmetic

logic units (ALUs). An application-speciﬁc instruction

set processor (ASIP) is a software-programmable pro-

cessor optimized for a particular class or domain of

applications, such as signal processing, telecommuni-

cation, or gaming applications. To ﬁt the application,

the datapath and instructions of such a processor are

customized; for example, one type of ASIP, the digital

signal processor (DSP), may have special-purpose data-

path components such as a multiply–accumulate unit,

which can perform multiply-and-add operations using

only one instruction. Finally, a single-purpose proces-

sor is a digital circuit designed to serve a single purpose

– executing exactly one program. It represents an exact

ﬁt of the desired functionality and is not software-

programmable. Its datapath contains only the essential

components for this program and there is no program

memory. A general-purpose processor offers maximum

ﬂexibility in terms of the variety of applications it can

support but with the least efﬁciency in terms of per-

formance and power consumption. On the contrary,

single-purpose processors offer the maximum perfor-

mance and power efﬁciency but with the least ﬂexibility

in terms of the applications they can support. The ASIP

offers a compromise between these two extremes.

IC technology refers to the speciﬁc implementation

method or the design style of the processing engine

on an IC. It is categorized into three main branches:

full custom, semicustom,andprogrammable logic de-

vice (PLD) [38.2]. Full custom refers to the design style

where the functional and physical designs are hand-

Semicustom/ASIC

Array-based

Gate arrays

Sea of gates

Structured ASICs

Cell-based

Standard cell

a) b)

Platform-based

System-on-a-chip

IP-based

PLD

CPLD

EPROM-based

EEPROM-based

Flash-based

FPGA

SRAM-based

Anti-fuse-based

Flash-based

Fig. 38.1a,b Classiﬁcation of (a) semicustom design and (b) PLD design (SRAM: static random access memory;

EPROM: erasable programmable read-only memory; EEPROM: electrically erasable programmable read-only memory)

crafted. This would provide the best design quality but

also requires the extensive effort of a design team to

optimize each detailed feature of the circuit. Since the

design effort and cost are high, this design style is

usually used in high-volume (thus cost can be amor-

tized) or high-performance applications. Semicustom

design is also called application-speciﬁc integrated cir-

cuit (ASIC). It tries to reduce the design complexity by

restricting the circuit primitives to a limited number.

Such a restriction allows the designer to use well-

designed circuit primitives (gates or functional blocks)

and focus on their efﬁcient interconnection. Such a re-

striction also makes it easier to develop computer-aided

design tools for circuit design and optimization and re-

duce the design time and cost. Today the number of

semicustom designs outnumbers custom designs sig-

niﬁcantly, and some high-performance microprocessors

have been designed partially using semicustom style,

especially for the control logic (e.g., IBM’s POWER-

series processors and SUN Microsystems’ UltraSPARC

T1 processors).

Semicustom designs can be further partitioned into

several major classes. Figure 38.1a shows such a par-

tition. Cell-based design generally refers to standard

cell design, where the fundamental cells are stored in

a library. Cells often are simple gates and latches, but

can be complex gates, memories, and logic blocks as

well. These cells are pretested and precharacterized.

The maintenance of the library is not a trivial task

since each cell needs to be characterized in terms of

area, delay, and power over ranges of temperatures

and supply voltages [38.2]. Companies offer standard

cell libraries for use with their fabrication and design

technologies, and amortize the effort of designing the

Part D 38.1