Managing heat through computational fluid dynamics is becoming more common and more complicated.
With thermal issues and constraints increasing becoming integral concerns of electronics design, computational fluid dynamics technology is gaining traction as a way to model, analyze, predict, and ideally prevent thermal problems from materializing.
From cooling a board to cooling a chip with a fan and heat sinks, all of this relies on air flow for the cooling, or the flow of liquid in some extreme cases. In all cases, fluids cool the chip.
“It’s one thing to say how much energy the chip will produce,” Marc Swinnen, product marketing director for the semiconductor business unit at Ansys noted, “But what temperature it will achieve depends on its cooling environment. To really get a handle on that, CFD is needed to model the fans, the heat sinks, and whether two hot components placed in proximity of each other will be different than if they’re far apart. And once you look a little bit above the chip to see its thermal environment, CFD is needed to model that.”
CFD is an important element in simulation on a much larger scale. “It’s really complicated to eyeball it. There are good mathematical models, so CFD is ideal for modeling,” Swinnen said. “Traditionally, airplanes have been designed in wind tunnels. You built something, tried it out, fiddled with it, then rebuilt it. Now, there’s a move toward simulation to catch problems earlier. In the case of a hypersonic missile, for instance, those are really hard to test, so they need to be simulated.”
Fig. 1: Modeling airflow with CFD. Source: Ansys
Lockheed Martin’s F-35 fighter jet is a case in point, because it has an engine that allows it to take off vertically. “It’s an engineering marvel,” said John Chawner, senior group director for CFD product management at Cadence. “When you see it, you think there is no way this thing can fly. It is such a complex system, but then you step back, and at the same time you say, ‘of course it can fly because it was designed to do exactly that,’ because somewhere behind a panel of that aircraft, underneath, there’s a box of electronics.”
There’s also a lot of simulation involving airflow. “CFD relies on Navier-Stokes equations to look at the flow of a fluid, or air in this case,” Chawner said. “EDA and CFD overlap in thermal. Electronics get hot. Your cell phone gets hot. Your laptop gets hot. Data centers get hot. To cool them, you have a couple of choices. You rely on natural convection of the air, the heat rises. Or you put in a fan somewhere and it blows cool air. But it has to blow in the right places, in the right way, with the right force. And so, from a simple standpoint, the cooling of the electronics once they’re in an enclosure is the intersection of those Venn diagrams. Then it becomes a very interesting discussion.”
In addition to aerospace, there is a significant use of CFD in automotive. In the case of electric vehicles, the electric motors are so much quieter than an internal combustion engine that one of the biggest sources of noise is the airflow over the side mirrors. Robert Schweiger, director of automotive solutions at Cadence, noted that CFD can be used to design the mirrors in a way that can minimize the airflow noise over those mirrors.
CFD also can be used to extend the range per charge for electric vehicles by designing vehicles with lower drag. “Aside from price, the top concern in buying an electric vehicle is the range,” Schweiger said. “The range needs to be beyond 400 kilometers or 300 miles, because the charging takes quite some time. It means an additional stop on your journey, and the range can be significantly improved by optimizing the aerodynamics. Here, CFD will play a major role in optimizing just the aerodynamics of a car, and this will have a positive impact on the range.”
CFD plays a role in cooling systems within a vehicle, as well. “For a combustion engine, the thermal problem is while you are stopped in a traffic jam, because you need to cool your engine, and you need to have all kinds of cooling systems,” Schweiger said. “The challenge for an EV is during charging. So CFD can be used for aerodynamics, and it also can be used for cooling systems to simulate the water flow and how it is able to cool down systems. It could be a converter of some kind. For batteries, there will be new concepts coming up where you have your battery pack and you spray water from the top onto the battery while you are using a supercharger. And then, if a tank is underneath, it pumps this liquid up again and sprays it down. This is one way to cool down batteries, which eventually will help to improve the lifecycle of a battery. Also, theoretically, if you take your electric vehicle on the racetrack, the battery would get very hot, so for performance cars, there will be a special thermal effort needed to cool down the battery and make sure it is not reaching a certain temperature range, which would eventually reduce the performance of the car. For all of these applications, from turbo machinery to cooling systems and fluid dynamics, CFD is used to simulate it and optimize it.”
Advanced packaging drives CFD use
Computational fluid dynamics technology also is getting increasing attention as Moore’s Law slows down. Rather than scaling, chipmakers are leveraging new and increasingly heterogeneous architectures, including multiple dies in an advanced package. Stacking die inside a 2.5D or 3D package greatly increases the challenge of thermal dissipation, which can cause problems ranging from accelerated and uneven aging to mechanical stress caused by thermal mismatch of different components and materials.
Thermally-driven chip-substrate-package interactions can be significant, and their successful management is key to ensuring acceptable reliability, noted John Parry, electronics and semiconductor director for Simcenter products group at Siemens Digital Industries Software. “Die can be made of different base materials, depending on the functionality — silicon, silicon germanium, gallium arsenide, etc., and the interposer made from silicon, glass, ceramic, or organic materials. When heated, these different materials expand at different rates, according to their differing coefficients of thermal expansion. Semiconductors are very hard materials that do not bend easily. While the structure does bend, the materials between the different layers experience shear stresses, which can damage the electrical interconnections. The overall package expansion and bending also affects the interconnect to the board.”
To ensure reliability, systems integrators and OSATs need to know the 3D temperature distribution throughout the package structure within their application environment. The temperature distribution through the package depends where and on the how heat is lost from the package into the board, and any attached cooling solution like a heatsink or chiller block. These heat flow paths interact, so any change to the design affects the heat flow everywhere. CFD is essential to predict the heat flow throughout the system, and hence throughout the package structure.
Technically, CFD is mainly an enabling technology for doing accurate thermal simulation, as it accurately predicts the heat transfer rates on surfaces due to convection and thermal radiation, Parry noted. “The focus over the years has moved down the system packaging hierarchy, from box, to PCB/board, IC package, and now down at the die level. The accuracy requirements have also gone up. Full simulation of conduction, convention and thermal radiation is needed, with accurate thermal models for both the components, PCB and any component-mounted heat spreaders or heatsinks.”
What’s changed?
CFD is being discussed more lately because of its earlier use in the electronics design process. It is now an important aspect of EDA design flows, such as closing power delivery nets (PDNs) in PCBs. “Temperatures from a CFD simulation of the board within its application environment can be fed back into power integrity tools. Updated power information is then sent back to the CFD software and the process repeated. Experience has shown that two or three repetitions of this process are all that is needed to achieve consistent power dissipations and temperatures, allowing the PDN design to be closed with confidence,” he explained.
Additionally, CFD is being applied in areas that might not have been previously thought of, according to Shawn Carpenter, program director for 5G and space at Ansys. “CFD is an integral part of the analysis chain for modeling thermal effects. How do you get the heat out? You need to be able to model fluid. Air is a fluid. You need to be able to model passive cooling approaches versus active cooling. You might have a forced-air fan for base station electronics — a baseband unit or a control plane unit. You’ve got to get the heat out, and how you do it probably needs to have margin that the unit can work in Siberia or in Death Valley. That heating unit needs to be set up so you don’t burn the electronics up when you’re running it at high capacity with a lot of users.”
CFDs also is being used for modeling how to get heat out of devices during heat transfer, and it’s being used in 5G base-station modeling, where antennas are attached to a tower.
“They put them on a tower or on the side of a building,” Carpenter said. “To do that, you have to understand the wind loading field that that device is going into. An interesting effect that happens with both 5G towers, as well as the backhaul units — which are wireless links with highly directional antenna relays — is that when they get exposed to strong wind, they rock in the wind, and the pole has a resonance. You need to couple the fluid dynamics of a wind field to a mechanical structural analysis of the tower they’re mounted on, to then understand what engineering is needed in the mounting structure. Can they be made lighter by applying a topology optimization to the mounting bracket? We helped a 5G equipment provider actually reduce weight in the mounting brackets for their antennas, which enabled them to be able to put more units on the same tower before they had to either reinforce the tower or build a new one. With CFD, we can look at things like this. It couples thermal analysis to optimizing what goes into the engineering of these systems.”
Fig. 2: Understanding heat in a system is critical. It can affect reliability, aging, and overall performance and power. Source: Ansys
Applying CFD
With 2.5D and 3D assembly techniques increasing in popularity and become pervasive, Parry anticipates thermal and thermomechanical issues will continue to grow and diversify into other industries like aerospace and automotive where safety and reliability are paramount. “The value of thermal design increases with earlier use in the design process. When integrating multiple die/chiplets together as a heterogeneously integrated package, there are multiple assembly options available which need to be explored/evaluated with non-viable configurations being discounted. This requires the adoption and use of predictive analysis to virtually prototype the performance at this early stage, when limited design data is available. The purpose is to guide the development effort and assure satisfactory performance for the selected assembly option before significant resources are committed to the detailed design effort. Done well, this ensures rework is minimized later when resource commitments and related costs are much higher.”
Another interesting trend at the system level is using reduced order thermal models, which are created using CFD toolsets to improve power estimation during schematic capture. “These models can support a wide range of thermal environments, are highly accurate, and very fast to solve,” he said. “They are available in a range of formats, including VHDL-AMS, so are ideally suited for electrothermal simulation within circuit simulators. This trend moves CFD-based technology even further up the electronics design flow, with leading component suppliers recognizing the merit in providing such models to assist their customers design activities.”
Others agree. “As circuits get smaller and denser, and devices get smaller and denser, it’s great because we deliver all sorts of value, but at the same time we’ve created thermal challenges and packaging challenges,” Chawner said. “So it turns into a materials challenge, because you end up having to deal with and better understand things like how materials may warp under heat load, how they’ll fatigue over time, like from a laptop on-off, on-off. This means as CFD and other technologies like digital twins are brought to bear, we could start looking at co-simulation, where we can do that upfront. If we look at the big idea of digital twin, we can carry forward all of these technologies together to ensure that over the lifespan of the object or the device, that we’re able to anticipate those problems and deal with them in the future as opposed to trying to solve them all immediately up front.”
In fact, he noted, the U.S. Air Force is doing that with some of its procurement. “They’re going into the design of this next-generation fighter knowing that about every five years they’re going to do an upgrade and a refresh, change the avionics, change other packages on it. If you build that in — which can be done through a digital twin process, whether it’s electronics, an airplane or a car — all these tools can be brought to bear on those problems as we move these designs into the future. As long as we build up these techniques, where they’re all layered and interconnected, and they can all talk and share data, then at any point going in the future it becomes easier to do those types of simulations and studies and anticipate the unanticipated. It allows you to monitor things that become damaged or worn. It allows you to anticipate when, for maintenance, you need to overhaul and replace, but also when it is time for a refresh. So, it goes into a bigger plan.”
CFD and EDA are just part of the picture.
“It’s solid mechanics,” said Chawner. “It’s aeroacoustics, i.e., the noise that bounces back because you’ve got small electronics. You put them in a box, and you’ve got to keep them cool. You’ve got to move a lot of air, which means you’ve got to put in a fan that’s turning at high RPMs. Guess what? It’s noisy. So all of these things have to be taken together holistically as part of the design, and that includes CFD. There are so many different applications, but the common aspect is energy generation or power generation, while another aspect where this comes into play is minimizing energy use, and harvesting energy where you can. There are applications where you want to minimize the amount of power you need, obviously for electric vehicles, and that’s really another place where a CFD comes into play. It’s an integrated way to look at the overall physics of a system instead of isolating it where you’re an expert. Look at the engineered device as a whole. And I say that as though it’s easy. It’s not.”
Leave a Reply