As users grow ever more demanding, and manufacturers look to squeeze greater capabilities into smaller devices, design engineers must look for increasingly complex ways to dissipate heat and reduce the risk of their products overheating.
As such, it is often essential to consider how the design of the PCB will affect the thermal management of the device. The number of layers, their thickness and copper content, and the number and placement of vias can all play an important role in keeping the components cool. One of the most efficient ways to examine how these components impact the overall heat of a device is through thermal simulation.
Thermal simulation is widely used by designers to predict the temperature and airflow throughout a device to assist with the overall thermal design. This allows engineers to identify thermal design issues early, so they can optimise their designs and ultimately reduce costs.
To produce accurate simulation results, it is essential that the thermal properties of the PCB are accurately represented in the simulation. The average PCB is composed of many layers of conductors and dielectric material, with complex patterns of traces on each level. This can include thousands of extremely thin vias, making detailed PCB simulation a very computationally intensive process.
All of these complexities mean that even the simplest boards have traditionally required some degree of approximation in the thermal modelling process. However, as processing power and the capabilities of thermal modelling software have advanced, the ability to represent individual layers and complex structures has improved.
Back in the 1990s, simulation software could only represent PCBs as simple thin blocks, with varying degrees of roughness used to represent the components. Within the simulation these blocks would then simply ‘dump’ heat into the first adjacent air cell, providing a predictably inaccurate portrayal of heat transfer from the PCB surface.
As the technology evolved, it was eventually possible to model PCBs and components as thick solid blocks, taking account of their layered structure. This opened up a discussion of which thermal conductivity values should be used. The mechanical engineer Tony Kordyban once demonstrated that a single value (10W/mK) could produce good results for some situations. And for many years, the ‘Kordyban Conductivity’ was used in the absence of better data.
Today, modern modelling systems such as 6SigmaET are able to take advantage of the abundance of high speed computing and processing power. As a result, it is possible not only to represent each individual layer in a PCB design, but even to include complex details of layout traces, vias and on-board copper content.
Brute computational power, however, is not the be-all and end-all of thermal simulation and management. Modelling solutions must also consider how and when that power is best utilised. For example, instead of using the detailed layers directly in the simulation, it can be helpful to use the information to drive simplification and approximation. An image of the copper content of a layer can be analysed to give either the average copper content over the whole layer, or values on a grid cell-by-grid cell basis; it’s simply a case of working out the best balance between granularity and processing time. In this way, details such as individual traces and pads will no longer be modelled as solids. This providing a slightly less accurate but ultimately much less resource-intensive thermal simulation.
Regardless of the detail available, the best conclusion to take away from this debate is that designers need to treat the choice and layout of PCBs as a vital stage – if not the very first stage - in the thermal management process. Through intelligent PCB modelling, and a proactive consideration for thermal issues at this early stage, designers have the opportunity to improve the efficiency and usability of their devices. This means they can possibly save energy and cost in areas that they previously may not even have considered.
Author profile:
Chris Aldham is product manager at 6SigmaET
