Pioneering Design-Class
Thermal Analysis for Electronics - II

<This article is the second in a series>

In the last article, we introduced an emerging class of analysis software called Design-Class Analysis (DCA) and compared it to traditional analysis software. We concluded that the main difference between the two is a matter of focus. Traditional analysis software is primarily focused on particular fields of analysis (i.e. FEA, CFD, FNM). DCA is primarily focused on applying analysis techniques to design problems within a specific field of engineering. For Flomerics, physical design of electronics is the engineering field of choice. The differences are summarized below:

In this article, we further examine the differences and explore the ways in which each impacts the first stage of a typical physical design process. In future articles, we examine later stages of design.

Opposing Forces: Performance Requirements and Design Constraints

Consumer demand and fierce competition drive the pace of innovation for the electronics industry. Severe performance requirements and stringent design constraints are the norm. The "first law of product design" says that performance requirements work against constraints and visa versa. In practice this means that if a product must run faster and longer, more time and money is required for design. Conversely, if a product must be designed quickly and cheaply, you can be sure it will break easily.

In recent years, consumers have increased their demand for high performance products resulting in faster and hotter running products. Just ten years ago, typical computer systems ran at 33 Mhz and 38 mW/sqin. Today, they run at 2 Ghz and 333 mW/sqin - a 60-fold increase in speed and 10-fold increase in thermal density. Over that same period, companies have strived to increase (or at least maintain) profits by reducing design time and manufacturing costs. These unrelenting business factors have placed tremendous pressure on design teams to move beyond traditional tools and processes and defy the first law of product design. What will take productivity and efficiency to the next level for physical design of electronics?

Why Traditional Design Processes Fall Short

At the highest level, a typical physical design process can be illustrated by the spring image introduced in the first article and shown in Figure 1 below. For many companies, stage one of the design process involves conceptualizing a baseline design in software. This is shown as "Model Build" in the figure. Stage two is the design optimization phase. This stage will be covered in the next article.

Figure 1: A Typical Design Process

How many software tools are used at your company to Model Build for design? Two, three or more? It is common for electronics companies to Model Build with three separate tools to complete a design. The first, and primary model is built with a solid modeling tool. Configuration design issues are worked out here. Additional models are built to address compliance issues and to increase performance. These may include Model Builds for thermal, structural, EMC and acoustics issues. How much duplication of effort can one company stand?

Over the past 10-15 years, traditional analysis companies have spent significant resources improving the Model Build process. CAD-like interfaces have been adopted and many complex tasks have been automated. As a result, we have seen steady, incremental improvements in ease-of-use and speed over the years. Despite this, a huge, unavoidable barrier remains for traditional analysis tools that prevent their wide spread use and force companies to spend more money on design than they would like. Model Building in traditional tools is a separate process by design.

The evidence of this is ubiquitous. For years, CAD companies have tried with little success to merge Model Building in traditional analysis with that of CAD. Just about every CAD company has an "integrated" analysis module but true, single model design remains a pipe dream. This is especially true in the electronics industry

To further complicate matters, the disconnected nature of traditional analysis creates inefficiencies in the design process. Design engineers find it tough to communicate with thermal engineers. Thermal engineers find it difficult to work with the EMC engineers. And so on. Ideally, performance, compliance and configuration issues drive the design early in the process so that trade-offs between competing design issues can be performed cost effectively. This will become practical when traditionally separate design processes become integrated through a single model. High performance heat sink design highlights the benefits of a single model design process. Significant time is spent designing heat sinks for high performance IC's only to find out later that they double as an efficient source of electromagnetic radiation. EMC and thermal design studies from a single model can save significant design effort and redesign costs.

Single model design is fundamental to DCA software. This concept is called Design Flow Integration and it defines a software design environment that works as a complete, single platform for physical design within a specific field of engineering.

FLOTHERM is being developed under Design Flow Integration principles and the results are starting to show. The example below is an ATX form factor PC chassis. The ATX form factor was introduced in the mid-1990's to improve expansion and packaging of PC's. Since then, system speed and thermal density have increased by 14 and six times respectively while the configuration has changed little. These factors have conspired to make significant the interplay between thermal and EMC performance issues. With traditional tools, multiple Model Builds are an undesirable but necessary reality for PC manufacturers.

Figure 2: Tower PC

The first Model Build is performed in a solid modeling tool. Ideally, this model is used directly to perform thermal and EMC design. Today, FLOTHERM is already well on they way to achieving the ideal because it automatically converts solid models to thermal and EMC models through sophisticated filters. The conversion process is extremely fast and requires very little user intervention. An example is shown below:

Figure 3: Automated conversion from solid to thermal analysis model with FLOTHERM

Once converted, thermal and EMC performance are checked and optimized within the FLOTHERM design environment. Competing design trade-offs can be explored easily and thermal and EMC engineers have the opportunity to drive the design early in the design process.

Figure 4: Airflow distribution	Figure 5: Surface currents

Future versions of the FLOTHERM design environment will perform multiple types of analyses on a single model, further simplifying the Model Build stage. The end goal for FLOTHERM is to become a true Design Flow Integration environment - a complete physical design of electronics environment that operates on a single model and covers configuration, performance and compliance design issues.

Design Flow Integration is an exciting concept for forward-looking manufacturers of electronic products. Streamlining design processes with DCA software is one way to bring high performance products to the market faster. Image the impact of being able to defy the first law of product design when your competition can't.

In the next article, we explore the design optimization (study) phase of a typical design process and compare the differences between DCA and traditional analysis software. Again, the differences are tremendous and the benefits significant. For the first time, design optimization may become a practical, routine part of product design.