Design Automation Technology Roadmap:Design Automation and Historical Perspective

Design Automation: Historical Perspective

The 1960s: The Beginnings of Design Automation

Early entries into design automation were made in the areas of design (description) records, PCB wiring, and manufacturing test generation. A commercial EDA industry did not exist and developments were made within companies with the need, such as IBM [1] and Bell Labs, on mainframe computers such as the IBM 7090. The 7090 had addressable 36-bit words and a limit of 32,000 words of main storage (magnetic cores). This was far less than the typical 256+ Mbytes of RAM on the average notebook PC, and certainly no match for a high-function workstation with gigabytes of main store. Though computer limitations continue to be an on-going challenge for EDA development, the limitation of the computers of the 1960s was particularly acute.

In retrospect, the limitation of computers in the 1960s was a blessing for the development of design automation. Because of these limitations, design automation developers were forced to invent highly creative algorithms that operated on very compact data structures. Many of the fundamental concepts developed during this period are still in use within commercial EDA systems today. Some notable advances during this period were:

• The fundamental “stuck-at” model for manufacturing test and a formal algebra for the generation of tests and diagnosis of faults.

• Parallel fault simulation, which provided simulation of many fault conditions in parallel with the good-machine (nonfaulty circuit) to reduce fault-simulation run times.

• A three-valued algebra for simulation which yields accurate results using simple delay models, even in the presence of race conditions within the design.

• Development of fundamental algorithms for the placement and wiring of components.

• Checking of designs against prescribed electrical design requirements (rules).

Also, there was development of fundamental heuristics for placement and wire routing, and for divide and conquer concepts supporting both. One such concept was the hierarchical division of a wiring image into cells, globally routing between cells, and then performing detailed routing within cells, possibly subdividing them further. Many of these fundamental concepts are still applied today, although the complexities of physical design (PD) of today’s large-scale integration (LSI) is vastly more complex.

The 1960s represented the awakening of design automation and provided the proof of its value and need for electronics design. It would not be until the end of this decade when the explosion of the number of circuits designed on a chip would occur and the term LSI would by coined. EDA development in the 1960s was primarily focused on printed circuit assemblies, but the fundamental concepts developed for design entry, test generation, and PD provided the basics for EDA in the LSI era.

Design Entry

Before the use of computers in electronics design, the design schematic was a hard-copy drawing. This drawing was a draftsman’s rendering of the notes and sketches provided by the circuit designer. The drawings provided the basis for manufacturing and repair operations in the field. As automation developed, it became desirable to store these drawings on storage media usable by computers so that the creation of input to the automated processes could, itself, be automated. So, the need to record the design of electronic products and assemblies in computers was recognized in the late 1950s and early 1960s. In the early days of design automation, the electronics designer would develop the design using paper and pencil and then transcribe it to a form suitable for keyed entry to a computer. Once keyed into the computer, the design could be rendered in a number of different formats to support the manufacturing and field operations. It was soon recognized that these computerized representations of the circuit design drawing could also drive design processes such as the routing of printed circuit traces or the generation of manufacturing test patterns. Finally, from there, it was but a short step to the use of computers to generate data in the form required to drive automated manufacturing and test equipment.

Early design entry methods involved the keying of the design description onto punch cards that were read into the computer and saved on a persistent storage device. This became known as the design’s database, and is the start of the design automation system. From the database, schematic diagrams and logic diagrams were rendered for use in engineering, manufacturing, and field support. This was typically a two-step process, whereby the designer drew the schematic by hand and then submitted it to another for conversion to the transcription records, keypunch, and entry to the computer. Once in the computer, the formal automated drawings were generated, printed, and returned to the designer. Although this process seems archaic by today’s standards, it did result in a permanent record of the design in computer readable format. This could be used for many forms of records management, engineering change history, and as input to design, analysis, and manufacturing automation that would soon follow.

With the introduction of the alphanumeric terminal, the keypunch was replaced as the window into the computer. With this, new design description languages were developed, and the role of the transcrip- tion operator began to move back to the designer. Although these description languages still represented the design at the device or gate level, they were free format and keyword oriented and design engineers were willing to use them. The design engineer now had the tools to enter design descriptions directly into the computer thus eliminating the inherent inefficiencies of the “middleman.” Thus, a paradigm shift began to evolve in the method by which design was entered to the EDA system. Introduction of the direct access storage devices (disks) in the late 1960s also improved the entry process as well as the entire design system by providing on-line, high-speed direct access to the entire design or any portion of it. This was also important to the acceptance of design entry by the designer as the task was still viewed as a necessary overhead rather than a natural part of the design task. It was necessary to get access to the other evolving design automation tools, but typically, the real design thought process took place with pencil and paper techniques. Therefore, any change that made the entry process faster and easier was eagerly accepted.

The next shift occurred in the later part of the 1970s with the introduction of graphics terminals. With these, the designer could enter design into the database in schematic form. This form of design entry was a novelty at first, but not a clear performance improvement. In fact, until the introduction of the workstation with dedicated graphics support, graphic entry was detrimental to design productivity in many cases. Negative effects such as less than effective transaction speed, time lost in making the schematic esthetically pleasing and the low level of detail all added to less than obvious advances. In contrast, use of the graphics display to view the design and make design changes proved extremely effective and was a great improvement over the red-lined hard-copy prints. For this reason, use of computer graphics represented a major advance and this style of design entry took off with the introduction of the workstation in the 1980s. In fact, the graphic editor’s glitz and capability was often a major selling point in the decision to use one EDA system over another. Further, to be considered a commercially viable system, graphics entry was required. Nevertheless, as the density of ICs grew, graphics entry of schematics would begin to yield to the productivity advantages of (alphanumeric) description languages. As EDA design and analysis tool technology advanced, entry of design at the register-transfer level (RTL) level would become commonplace and today the design engineer is able to represent his design ideas at many levels of abstraction and throughout the design process. System- level and RTL design languages have been introduced and the designer is able to verify design intent much earlier in the design cycle.

By the 1990s and after the introduction of synthesis automation, design entry using RTL descriptions was the generally accepted approach for entry of design, although schematic entry remained the accepted method for PCB design and many elements of custom ICs. There is no doubt that the introduction of graphics into the design automation system represents a major advance and a major paradigm shift. The use of graphics to visualize design details, wiring congestion, and timing diagrams is of major importance. The use of graphics to perform edit functions is standard operating procedure.

Large system design, often entails control circuitry, dataflow, and functional modules. Classically, these systems span across several chips and boards and employ several styles of entry for the different physical packages. These may include:

• Schematics—graphic

• RTL and behavorial level languages—alphanumeric

• Timing diagrams—graphic

• State diagrams—alphanumeric

• Flowcharts—graphic.

Today, these entry techniques can be found in different EDA tools and each is particularly effective for different types of design problems. Schematics are effective for the design of “glue” logic that interconnects functional design elements such as modules on a PCB and for custom IC circuitry. Behavorial languages are useful for system-level design, and particularly effective for dataflow behavior. Timing diagrams lend themselves well to describe the functional operations of “black-box” components at their I/Os without needing to describe their internal circuitry. State diagrams are a convenient way to express the logical operation of combinational circuits. Flowcharts are effective for describing the operations of control logic, much like use of flowcharts for specification of software program flow. With technology advances, the IC is engulfing more and more of the entire system and all of these forms of design description may be prevalent on a single chip. It is even expected that the design of “black-box” functions will be available from multiple sources to be embedded onto the chip similar to the use of modules on a PCB. Thus, it is likely that future EDA systems will support a mixture of design description forms to allow the designer to represent sections of the design in a manner most effective to each. After all, design is described in many forms by the designer outside the design system.