Compact Cooling for High Power Density Electronics

Everything is getting more powerful. And smaller.

Electrical drives typify this. Small, compact and power dense drives are essential to meet the technical requirements in certain sectors. As an example, automotive EVs are driven by low cost and long drive ranges – requirements that feed up to designers to create smaller, lighter and more efficient products.

Effective cooling is essential for power electronics designers (from device level, to power modules, to full converter systems) to satisfy the power density requirements needed for these electrification applications. Just because you can fit all your components into a small and compact box, it doesn’t mean you can keep them cool!!

For products that are driven by power density, a good fundamental understanding of cooler design together with a holistic view of the electronics and the coolant system at a system level is essential. Without good thermal engineering knowledge and experience, the route to an effective and optimised cooling system is a long one.

Many simulation packages allow you to simulate cooling with relative ease. But do you really have confidence in those results (which turbulence model are you using, and is your near wall mesh size correct to model the boundary layer?) A strategic approach to optimisation is necessary to design efficiently, as opposed to a random ‘Monte Carlo’ style approach.

Discussed here are some of the considerations necessary for developing effective cooling for high power density electronics, whether it be individual models or fully packaged converter.

Turbulence

High heat transfer rates from hot surfaces to cooling fluids is achieved through the creation of turbulence. Boundary layers formed at surface interfaces are prohibitive to heat transfer, effectively insulating the wall and bulk cooling fluid. Disturbing this boundary layer is essential to achieve effective heat transfer to the cooling fluid. This is done by creating turbulent flow.

Where flow rates through a passage are not sufficient to create super high velocities and Reynolds numbers inside a passage, we must be more inventive.

Turbulence can be created by constantly forcing coolant to change direction, not allowing it to stabilise a flow pattern. Impingement is also used to create localised cooling, forcing the coolant through small apertures at high velocity, directing them onto surfaces that need cooling.

There are several options for creating high Heat Transfer Coefficients (HTCs) which are essential for removing heat from high power density components. Selecting the right method and optimising it for best performance with minimal trade-offs can be challenging.

A ‘serpentine’ approach in the cooling channels. Forcing the fluid to constantly change direction keeps turbulence and cooling high

Example of impingement used to direct high velocity flow onto a surface for increased heat transfer.

Pressure

All this additional cooling doesn’t come free. Every time we force a fluid to change its path, or contract through smaller flow area, a pressure drop is created. The relationship between pressure and cooling coefficient does not favour the cooling coefficient – pressure drop increases more speedily than the cooling benefits. This is demonstrated in the plot, which shows the pressure vs  HTC correlation for different cooling design options of an electronics package. Plotting Log Pressure vs HTC, each data point fits closely to a linear correlation.

Localised cooling

Because of this unfavourable relationship between heat transfer coefficient and pressure drop, as engineers we must be clever and selective about where to use our pressure ‘budget’. After all, it’s likely the pressure drop and flow rate are predetermined and our cooling design must satisfy it.

Understanding the component losses of the electronics package is essential. This will give us a good idea about which parts of the wetted surface are in need of enhanced cooling (those directly beneath high loss devices).

Quick thermal resistance calcs are essential to understand the necessary heat transfer coefficient required. Hopefully this was done by the package designers at an early stage to verify the cooling required is achievable!!  Summing the conductive resistances through the package layers to the cooled surface determines the fixed conduction resistance that, at this stage, we cannot change.

Then, knowing the device loss, maximum operating temperature and the maximum possible inlet coolant temperature, a calculation of the maximum permissible resistance can be made.

This should be higher than the conduction resistance we calculated, and the difference between the two provides the convection resistance we must work to.

A note about heat spreading…

In making the conduction resistance calculations, the Area used should be that of the device in contact with the substrate – a safe-side value. In reality as the heat conducts through the layers, it will spread laterally – heat spreading. Thick layers of highly conductive materials are excellent at doing this. Not only does this reduce the conductive resistance, but the area at the wetted surface for us to cool is larger, meaning a lower HTC is required!

Designing heat spreading into the package layers really helps with the coolant design. However, in ultra-compact designs these can often be omitted to save weight and space. A holistic view must be taken – a larger cooler design or bigger pump to deliver the higher cooling requirements may quickly remove those savings made in the package layers!

 

Parallel cooling paths

Electronic packages with multiple arrays of devices and components that require cooling face a further logistical challenge for the coolant. What is the best route to navigate the coolant to all these parts of the module or converter?

Knowledge of the incoming coolant flow rate and allowable pressure drop will help in determining this. Multiple parallel routes of coolant means less flow per parallel path, but if we can achieve the correct cooling coefficients in all areas, then it is likely the overall pressure drop will be lower. It also allows us to be more ‘aggressive’ with our turbulence creating cooling designs without impacting the overall pressure drop as much

A further consideration in this is the coolant temperature rise. A single coolant path serving multiple parts of the package will continually accrue heat, so the last components being cooled by the coolant loop will be done so by a fluid at a higher temperature than it arrived into the converter. Therefore, it may operate higher than an equivalent component at the beginning of the cooling circuit.

Electronic load balancing is dependant on having balanced temperatures. Hotter device temperatures create higher electrical resistances.  This leads to greater heat dissipation (it is a function of electrical resistance). This leads to higher temperatures… uh-oh, we have thermal runaway!!

A note about component losses…

The loss behaviour of the electronic components can be complex. Devices may be rated to handle very high power loads, but only be expected to see these loads occasionally and for very short periods of time. In these cases it is usually not appropriate to design the coolant to handle these powers. Short transients get absorbed by high capacitance materials, and the heat pulse may not even reach the coolant before the transient surge is over.

Understanding the duty cycles to determine the most suitable loss values to consider is essential to provide a cooler design without over-engineering.

Summary

Consideration of electronics cooling needs to be made at the beginning of the design process. Just because you can fit all your components into a small and compact box, it doesn’t mean you can keep them cool!! Some simple calcs can make sure your design is evolving towards something achievable.

High cooling rates are essential for power electronics designers (from device level, to power modules, to full converter systems) to satisfy the power density requirements needed for many electrification applications. There are many ways to create these high cooling rates, but they require a good understanding of the electronics arrangement and component loss profiles, as well as careful consideration of the impact on the whole system cooling path.

Localised cooling is highly effective and usually necessary to achieve the heat transfer rates without creating excessive pressure drops such that the required flow cannot be delivered.

ECS have tremendous insight into the complexities and common pitfalls in cooling electronics at device, power module and converter package levels. Talk to us about how we can help improve your electronics cooling solutions so you can acheive the best power density possible.