The need for a change in thermal design philosophy

Figure 1: Standard situation

Figure 2: Required situation

Introduction

The future of many electronic companies will depend to a large extent ontheir ability to initiate techniques that bring schedules, performance, tests,support, production, life cycle costs, reliability prediction and qualitycontrol into the earliest stages of the product creation process. The essentialquestion for the thermal community is: where does temperature fit into thispicture? Temperature plays a role in the following fields:

While the first two areas do not pose any problems, the last one does. Mostdesigners have learned that according to popular protocols such as MIL Hdbk 217,temperature specifications are related to field failure rates in a well-definedmanner. Unfortunately, there are very serious doubts about this approach (seesurvey ref.1), and there is no doubt that the future will demand a drasticchange in the way the relationship between reliability and temperature (ortemperature gradients) should be handled. For the time being, we will take itfor granted that quality and reliability are significantly dependent ontemperature.

Why Thermal Management Becomes Increasingly Important

As a result of the widespread introduction of micro-electronics, togetherwith the increasing demands upon functionality and reliability, thermalmanagement is becoming a hot issue in almost every branch of industry.

This trend is not restricted to professional and consumer electronicssystems, but also applies to automobile electronics, electronic lamps anddomestic appliances.

The cooling of generally complex electronic systems has become a toughchallenge indeed, resulting from the combined effects of increasing heat fluxes,miniaturisation and the striving for zero defects. In addition, new designs mustbe completed in less time and at a lower cost than previous designs. Forexample, failures in the later stages of a product creation process areinvariably costly; a rule of thumb is that for every new link in the productdevelopment chain, a factor of 10 can be added to the costs. It is clear thatredesigns adversely affect the base price of the product, but even moreseriously, the timely introduction of the product to the market. The preventionof thermal redesigns by identification of possible problem areas is a strategythat is worth investment.

There are many reasons why thermal management is of ever increasingimportance. For the sake of clarity, a subdivision into three areas has beenmade.

1. Some Technological Reasons

At the system level, designers are confronted with technological demands formore plastic, dust sealed, no noise, more features, higher speeds,miniaturisation. All of these factors can be expected to raise system andcomponent temperatures.

At the component manufacturer’s level, the constant pressure on designers toreduce package dimensions while logic power requirements are at their highestlevels in history, makes the problem of minimizing the thermal resistance fromjunction to case heat removal a crucial part of the package design.

At the customer’s level, we see another remarkable change. In both theprofessional sector and in the consumer market, it is only fair to admit thatJapanese companies should be acknowledged as the first to realize that customersshould be taken seriously. Therefore, the industry is confronted with a plaincall for a better reliability, lower noise levels, a striving towards a zerodefect philosophy and ever-increasing demands for EMC (Electro MagneticCompatibility). EMC is defined as the ability of a device, of equipment or of asystem to function satisfactorily in its electromagnetic (EM) environmentwithout introducing intolerable EM disturbances to anything in that environment.Indeed, EMC could pose a far more serious problem than people realize. With theincreasing application of digital ICs in the consumer sector, it can be foreseenthat the measures to be taken to achieve an optimum EMC performance will haveserious consequences from a thermal management point of view.

2. Some Logistical Reasons

On account of historical reasons, electronic and mechanical design methodsare always part of every design consideration. This permits incorporation of thefunctional requirements from the very beginning. In many industries, thermaldesign is just an afterthought, to be addressed if measurements in a prototypereveals any temperature problems. The diagram at the start of this articledepicts the situation, which is standard in many enterprises.

In the past ten years or so, the electronics industry has put a considerableeffort into the development of sophisticated tools for schematic capture,component placement, electrical simulation, routing and CAD-M. Improvements ofthese individual methodologies will be limited unless product developmentfactors such as appropriate technology selection, adequate testing, productreliability evaluation and manufacturing, are assured. In many cases, the mostdifficult problem to overcome is the required change in culture, from ‘Throwingover the wall’ to ‘Removing the wall’, and in convincing the technical peoplethat the ultimate goal is not merely superior design, but superior design tocreate a superior product.

Let us define concurrent engineering as the systematic approach to theintegrated, concurrent design of products and their related processes. Itsobjective is to make all design criteria an integral and upfront part of thedesign process. Concurrent engineering is intended to cause designers toconsider all elements of the product life-cycle, right from the conceptionthrough to disposal. Managers are talking about continuous improvement,first-time-right, quality control, instead of quality check.

Unfortunately, the realization of these commendable ambitions requires thefulfilment of many conditions. For example, it should be possible to sharedesigns, libraries and databases, which would require company-widestandardisation on almost everything from software to hardware. Designers shouldbe encouraged to take part in design simulation and to attend major trainingprograms. Databases should be filled with reliable data, for which knowledgeablepersons are responsible. An essential part of the ultimate success of such aphilosophy is to treat thermal management in exactly the same way as all othertechnological disciplines.

In conclusion, thermal management should be an important and necessary partof any concurrent design environment, and must be recognised and accepted by allthe people involved. The past has shown that such is by no means a simple task.The introduction of thermal management is in fact much more time-consuming thanother technological disciplines. One of the major reasons is the lack of heattransfer curricula focusing on a pragmatic approach to deal with the complexphysical phenomena that rule electronic systems.

3. Some Physical Reasons

Many products rely on air cooling. The drive for increased performance at areduced cost for cooling, is expected to change the thermal design drastically.For example, the entire spectrum of devices, from heat sinks to fans, are beingpushed to their limits.

It is to be expected that these limits will be reached in the coming years.It is clearly important for Managers to know exactly what their limits are. Forexample, the introduction of controlled fans appears in many cases to be adecision that is a psychological rather than a rational barrier.

And such physical limits have other important consequences. For decades,established design rules have reigned in many design environments, often withconsiderable success, despite the fact that extrapolation towards situations forwhich they are not derived, is often very tricky. (Maybe the ultimate designrule is that the mere fact that design rules seem to be successful, is anindication that the system to which the design rules are applied is not criticalat all.)

In the case of natural-convection-driven systems, which encompass themajority of consumer products, there exists an upper limit to what can beachieved regarding the maximum temperature even if all variables that influencethat temperature are optimised.

Unfortunately, the designer with some background in heat transfer, cannotusually rely on handbooks or literature results. Despite the fact that hundredsof convective heat transfer correlations have been developed, their accuracy isstrongly dependent on the degree of similarity existing between laboratory andactual design geometry, fluid velocities, temperature gradients, etc.. Costlyredesign and/or over-design commonly results from the inability to accuratelydescribe the boundary conditions.

Design rules have a relatively wide margin, and added complexity tends towiden these margins even more. Their area of application therefore changesdrastically when the physically possible limits are reached. Hence, the old setof design rules should, in any case, be replaced by some temperature predictionmethod that narrows the design rule margins considerably.

Examples of current design rules that need careful attention are:

• The use of nomograms for heat sink design
• ‘Ambient’ temperature measurements combined with thermal resistances fromdata sheets to derive junction temperatures
• Handbook Nu correlations to calculate heat transfer correlations
• Cup-mixed-mean temperatures for channel flows

 

In all these cases, it can not be expected that the resulting data is usefulfor reliability prediction.

A Few Words on the Accuracy of Temperature Predictability

Imagine a scenario in which everything is settled and concurrent engineeringprinciples have been introduced everywhere. Suppose also that a system levelPackaging Engineer, who is responsible for an adequate thermal design of a newproduct, is asked by the Product Manager to come up with an estimation of theimportant temperatures, such as the die temperature of an active component, thecoil temperature of a transformer, or the peak temperature inside anelectrolytic capacitor. Then it is very important for both parties to agree uponthe underlying objectives of the request; whether, for example, the objectivesare to have an estimated worst case situation, to support a feasibility study,or to have some figure of merit to compare different designs where theestimation of critical temperatures for the purpose of a parametrical studymakes sense. However, if the objectives are to give an estimation of theexpected field failure rates, the Packaging Engineer is in deep trouble.

The strong dependency on temperature makes it crucial that temperaturepredictions are within narrow tolerances, i.e. within ±3°C.Unfortunately, such accuracy is difficult to achieve on a regular basis.

We have at least four large problems, involving lack of accurate values inthe areas of heat transfer coefficients, of thermal resistances of components,of input parameters (such as physical properties and air resistancecoefficients) and of temperature-reliability relationships. This missinginformation is an essential point which marks the difference between the use ofCAD tools in the early and in the final design stages. Early in the productcreation process, the focus should lie on proven design rules and rough systemanalyses to support concept decisions. In the final stages, the emphasis shouldbe on accurate system analysis supported by accurate component/PCB analyses, theoutput of which should serve as an input for reliability prediction tools.

Concluding Remarks

In recent years, the electronics-industry-specific maturation of CADsoftware, together with the availability of powerful low cost workstations, havemade possible the simultaneous solution of conductive, radiative and convectivephenomena, thereby greatly increasing the predictive capabilities of thedesigner.

Although the use of these tools is not the panacea which some people wouldlike to think, there is a growing acceptance that their use is indispensable forsolving what would otherwise be very costly or even untraceable problems.

In conclusion, thermal management should become an important designconsideration to ensure that thermal performance at all packaging levels ispredicted with sufficient accuracy during the early design phase so that theoperating and reliability constraints will be met in the final product.

Today’s CAD codes, tailored for use by designers, are very promising toolsto realize the required accuracies.

References

For further reading, the reader is referred to thefollowing survey paper:
Lasance C.J.M., Accurate Temperature Prediction in Consumer Electronics: aMust but still a Myth, in ‘Cooling of Electronic Systems’, eds. Kakac S. and YüncüH., Kluwer.Academic Publishers, 1994, pp. 859-898.