This article will discuss several aspects of front and back end thermal design. It will also expose some front end methods that have been practised by the author for more than 10 years and still are competitive.
Fig.1: Intensity of the design process.
The Design Process
There are many ways to represent the design process. In its most extreme form it is a complicated combination of checklists and time planning schedules. The representation shown in figure 1 is however sufficient for the purpose of this article. The period of low cost intensity in the beginning of the process will here be called the front end design phase.
This phase can coarsely be divided into of two parts, the conception phase and the pre-study phase. The purpose of the conception phase is to scan through all possible alternatives and surface the one that is believed to be the best. The purpose of the pre-study phase is to ensure that the proposed design is realistic and to predict the resources needed to implement it.
The major characteristics of the final product, including the cost, are decided in the front end design phase. It is therefore a very exciting time period and most engineers nourish a dream to once be a part of the small team that is involved in this activity.
The back end phase
To discuss the back end phase ahead of the front end phase may appear somewhat awkward. It is nevertheless a good refection of how the human mind works. Front end design must by necessity involve approximations and estimations and it is difficult to feel comfortable in such an environment unless one first knows how to be accurate. A natural reaction from someone who gets involved in thermal design is therefore to start by exploring the back end methods.
The purpose of the back end phase is to work out all the details and to ensure that the final product will have the properties desired. The market is full of thermal software for this purpose. Each particular tool has its pros and cons and miles of texts have been written about that. What not so often is written about, is how penalising it is to make major changes in this phase. The penalty sometimes appears as a cost increase, other times as a time plan violation and often as a combination of both. A project is therefore in big trouble if a thermal simulation in this phase indicates that major changes have to be made.
Although this is an article about technical issues, watts and degrees, it is nevertheless important to realise that the background is economical. Hardware projects are extremely expensive to run. Each hour, added or saved, often amounts to thousands of euros. In this environment it is obvious that the purpose of thermal design not only must be to predict temperatures and suggest changes but also to do it without time plan violations. The best way to accomplish this aspiration is to perform well in the front end phase.
The front end phase
The front end phase is radically different from the back end phase. There are few details available and important parameters such as the heat dissipation are often roller coasting.
There are much less market offers for front end thermal design tools than there are for back end tools. Considering the importance of these two phases, it should rather be the other way around. So, if somebody claims that thermal design is an awkward discipline, the author would not object.
Thermal engineers often complain that important decisions are taken without their knowledge and that their main activity therefore consists of fire fighting in the back end. A key matter for changing this state of things is to become an attractive discussion partner by learning how to handle front end thermal design issues.
The design team that works out a concept usually starts off by creating a large number of imaginable alternatives. Whether these are conventional, innovative or visionary does not matter at this stage. The next part of the job is to eliminate them one by one until only the best remains. Very little detailed information is available at this stage. The elimination process can therefore only advance on the bases of reasonable assumptions and approximations. A thermal designer that supports this process is always faced with a demand for estimations that cover a large range of possibilities. That is, overviews, directions and maps are much more important than precise answers to specific questions.
To create a thermal world in which a front end design team can navigate is a challenge. In addition it must be done fast, otherwise there will be no discussion. The bulk of the design tools used must therefore be fast, easy to handle and have good graphical representation possibilities. Back end tools are often useful in the less hectic moments of the process and should therefore also be a part of the tool arsenal.
Being involved in front end thermal design is also a matter of having the courage to do estimates, which always risky. The main risk is to be misunderstood. Hardware engineers are however a group of bright people that very well understands what estimates are. This risk is therefore quite small. It is nevertheless good practice to always let an estimation be accompanied by a reference scale and a clarification of which the main uncertainties are.
The questions below are examples of issues that can be brought up during the front end design process. They are not presented in any particular order and the sole purpose is to give the reader a taste of the atmosphere.
- Should the system be implemented on 3 or 5 PCBs?
- If natural convection is selected, what are the thermal consequences?
- Should this function be implemented with 3 off the self components or is it better to make 1 single custom circuit?
- Introducing an additional voltage level can save some 30% of the heat dissipation. Is it worth while?
- What would the thermal difference be between a 6 layer and an 8 layer PCB?
- Can the metal front serve as a heat sink for a component that needs to be located very close to a front connector?
- What space is required for a heat sink that can cool 10 W?
- Can sporadic 50 millisecond bursts of 50 W jeopardise the safe function case?
- Which is most advantageous, 1 or 2 daughter PCBs?
- What is the thermal difference between 15 mm and 20 mm PCB pitch?
What is striking about front end thermal design is how fast thermal problems can be identified. This is very important for pre-studies. The main problem in this phase is not lack of data but lack of time to collect existing data. Most component temperatures must therefore be estimates. Potential thermal problem must however imperatively be focused and solved. If this rule is respected, the rest of the thermal voyage will be a smooth.
There are nevertheless chronic trouble makers. One example is systems that involve custom design circuits. The heat losses in these circuits are always highly uncertain and can not be verified until they have been designed and manufactured. The only reasonable approach a thermal designer can take in this situation is to prepare both for the best and for the worst. More than one alternative must as a consequence be kept alive in the back end phase, which always is costly.
Front end thermal design offers great advantages and large efficiency gains. Yet it seems to be little practised. As always there are many possible reasons and quite a few of them belong to the psychological sphere.
When you respond to a fire alarm you enter into a situation where many persons are warred. If you can solve the problem you are the hero of the day and you will be called indispensable. If you are less successful, it is frustrating for everyone but who is to blame? You did not put that 6 W component on the board.
Back end thermal design also gives you the opportunity to produce impressive coloured temperature images and flow visualisations. Anyone likes to walk into a conference room with a couple of those in their hands.
The mental atmosphere in the front end phase is quite different. The people you work with get used to your performance very fast and after a short time they take it for granted. They want quick answers and can sometimes be irritated if you do not deliver fast enough. Your achievements will never give you any big headlines and if thermal problems appear in a later phase, you are partly to blame. Front end thermal design can therefore be quite frustrating, even if it is very exciting on a personal level.
The management attitude is also very important. One problem on this level is that of efficiency references. What do you compare the efficiency of front end thermal design with? If it is newly introduced it is always possible to make a before and after comparison but this option fades away after a couple of years. If you work hard and manage to keep all thermal problems away from the back end phase, there will be no fire alarms. An inexperienced management can as a result conclude that thermal design not is an important activity.
What sometimes also is difficult for the management to understand is that front end thermal design requires quite a lot of preparations. The hardware world changes fast. New thermal problems appear all the time and if you do not know how to deal with them in a smooth and fast manner you are just not attractive for the front end design team any more. If the management is not willing to give you the time you need to keep fit, things start to degrade very fast.
Another management problem is that internal book keeping routines can disfavour front end design. Thermal designers are sometimes paid by the hour they work in projects. Why bother about front end design under those circumstances when back end design will give you many more work hours and if something happens, you can always blame an electrical engineer.
The cooling efficiency method
Assume that you have designed a PCB that dissipates 10 W and works well. In the next project you have to do something similar but the PCB size will be 25% larger, the maximum board temperature will be 5 C lower and the air velocity will be 10% higher. How do you estimate the maximum heat dissipation for that PCB by reusing your former experience? This problem is not particularly difficult for a thermal engineer. It can be worked out on a spreadsheet. The draw back with this solution is that it requires quite a lot of input data for the reference PCB and that it does not produce a good overview.
The alternative is to use the cooling efficiency concept. The cooling efficiency can simply be said to be a measure on how well the PCB surface is used for cooling. An isothermal plate would have the value 100%. Real PCBs have lower values. Figure 2 shows an example of how experience values can be represented.
If you know the cooling efficiency it is a simple matter to calculate the maximum possible heat dissipation. All you have to do is make the calculation for a corresponding isothermal plate and then compensate the result with the cooling efficiency. The problem here of coarse, is that the cooling efficiency for a PCB not is known until a back end phase calculation has determined it. Front end calculations based on the cooling efficiency concept must therefore always cover a range of probable values.
A typical result from such a calculation is shown in figure 3. Although the result is quite coarse it is often a great help. Figure 3 would for example tell a designer that a PCB at the top left of the target window would need a cooling efficiency above 80 %. A quick look at figure 2 reveals that this hardly is possible.
The cooling efficiency method has been used for more than 10 years at Ericsson. The concept is also understood, used and appreciated by a large number of electrical designers in that company. One might therefore wonder why it not has penetrated outside its walls. There are many possible reasons. One is that the method requires easy access to experience values. The smoothest way to get these is to let a back end program publish them as a calculation result. Up to this year no thermal design tool that is available on the market has however done this.
Fig.4: A typical result from an air efficiency calculation.
The air efficiency method
The definition of the air efficiency is almost the same as the definition of the thermal efficiency for a heat sink. It is a measure on how well the air is used for cooling. 100 % would signify that the outlet air has the same temperature as the maximum board temperature. Most applications have an air efficiency well below 100 %.
There is nothing particularly inventive about the air efficiency concept itself. Thermal designers have used similar ideas for years. It is nevertheless often a very powerful help in the front end design process.
Figure 4 shows a typical result for a PCB without heat sinks. What is striking is that the air efficiency attains painfully low values when the air velocity increases. All this air must after all be transported and filtered and most of it does not seem to be of much use.
The air efficiency can however be significantly increased by surface enhancements. For air velocities around 1.5 m/s it is sometimes possible to approach values as high as 50 %. For the example in figure 4 this would increase the cooling capacity of the PCB with a factor 4!
The air efficiency concept can therefore be used to estimate how much heat that possibly can be dissipated from a PCB, all efforts made. Hardware designers have a great appreciation for such values because it gives them a landmark and it makes it easier for them exclude megawatt ideas.
The thermal territory method
One is inevitably confronted with component cooling questions in the front end design phase. These questions are often difficult to tackle because the temperature in a component does not only depends on itself but also on its adjacent components.
The thermal territory method is one way to deal with the problem. It is not a perfect method but it is very helpful as a first approach. A thermal territory is defined as the smallest rectangular surface of a PCB that a component needs for its cooling. What is a bit unconventional with this method is that it does not calculate the temperature of a component. It calculates the cooling surface need to attain a predefined temperature.
Thermal territories are ideal for making preliminary lay outs and they can also be used on an individual component bases, see figure 5. A simple rule of thumb is that a design will be very difficult to realise if the thermal territories cover more that 70 % of PCB surface.
A thermal territory is also associated with a thermal efficiency. It is defined as the cooling capacity of the territory compared with that of an isothermal plate of the same size. One of the characteristics of hot spot components is that they have thermal territories with low thermal efficiencies. The pain limit varies from case to case but if it is lower than 50 % it is definitely a strong candidate for the waste basket.
To avoid redesigns is a key matter for an efficient design process. High quality decision support in the early phases of the design process is therefore a necessity. The thermal design discipline must adjust itself to this situation and accept the fact that calculations based on preliminary and uncertain data also can be highly profitable. The days when thermal design could start with the first lay out attempt are over.
Front end thermal design is a concept with a positive flavour. Many software manufacturers want to be on the train with their specific product and are therefore pushing the concept. What many of them actually propose is to use back end tools as early as possible in the design process.
The author of this document advocates a different idea. In his mind front end thermal design is an activity that by its nature requires overviews, directions and maps rather than specific answers to precise questions.
The market offers few software tools for this kind of design, which suggests that it is not much practised. This is unfortunate since solving thermal problems in the early phases of the design process often results in large efficiency gains.
Article written by Ake Malhammar, Heat Transfer expert. Reproduced by kind permission in the Mechanical Design Forum. Link to original source material is here: http://frigprim.x10.mx/articels/front_end_design.html