No Hot Products

While marketers strive to launch the next “hot” product, engineers struggle to prevent literally hot products! A recent breakthrough in thermal modeling comes just in time as electronic component manufacturers and their OEM customers increasingly battle thermal design issues.

Analog electronic component manufacturers have traditionally provided models in SPICE format so customers can simulate their application circuits and better understand the features, capabilities, and interactions of those parts in the system context. Over the past 50 years, an ecosystem and “supply chain” of SPICE models, which run in a variety of available SPICE circuit simulators, has evolved.

More recently, an IEEE and IEC standard model format has been developed, capable of representing not only analog electronics, but also digital, mixed-signal, and even non-electrical systems. It can be used together with traditional SPICE components to support a more comprehensive model-ecosystem.

A new thermal system modeling method has also been developed, BCI ROM, or Boundary Condition Independent Reduced Order Model. This modeling approach has been implemented in VHDL-AMS to enforce heat-flow conservation, like KCL for electronic circuit simulation. This allows design engineers to create thermal “schematics,” or networks, to represent multi-component thermal interactions. Component manufacturers can accurately capture the thermal characteristics of their component packages and provide them to the OEM circuit/system designers, who can then connect them together with BCI ROMs of their PCBs. These schematics can also include functional electronic circuits, so they can accurately simulate electro-thermal interactions in both steady-state and transient operation.

Fig. 1: Electro-thermal model supply-chain.

But the accuracy of these electro-thermal circuit simulations is dependent on having accurate models of the components.

Boundary condition independence

Why do we care about boundary condition independence for thermal models of electronic components? It’s because of the way circuit designs are created. Electronics engineers select and assemble component models into networks and simulate their interactions under the assumption that the simulator will enforce Kirchhoff’s Current Law, or KCL: “The sum of the current flows at each connection point in a network must equal zero.” This means that a model of a part only needs to define the internal relationships among its currents and voltages. For a resistor, that’s Ohm’s Law: v = i*R.

That relationship needs to be enforced regardless of the external voltages on its two terminals. There may be a large voltage difference in a circuit, and a small or even negative voltage difference at a different time or in another circuit that uses that model. The model just needs to report the appropriate current level for each situation. The simulator, by enforcing KCL, can solve for all the currents and voltages in the network at every time step. This means that SPICE electrical models are essentially “boundary condition independent!”

However, when component manufacturers make SPICE “equivalent circuit” models of their component thermals, they often base the behavior on a temperature transient response measured in the lab. Unfortunately, that transient is dependent on the external mounting and heat sinking of the test apparatus, which may be different than the end-user’s actual PCB implementation. This can lead to inaccurate predictions of junction temperatures during simulation.

BCI ROM solves this problem by isolating the “internal” thermal characteristics of the component from its external boundary conditions. BCI ROMs implemented in VHDL-AMS can be networked because the simulator enforces conservation of heat flow.

Switching circuits and the “time constant divergence” problem

A frequently occurring problem when simulating electro-thermal circuits is the result of widely differing time-constants that exist in electro-thermal circuits. This is not a problem when the electronic circuits have continuous current during normal operation. The problem is more significant in “switching” power electronics circuits, for which proper thermal design is essential.

The example shown in figure 2 is an on-board charger for an electric vehicle. The inverter is being used in reverse, to boost the available residential voltage to charge a 600 V battery. Because of the fast rising and falling edges of the inverter switches, the simulator must take small time-steps to resolve the transient voltages and currents during each switching cycle. This means the simulation will require a relatively long computation time, even for a relatively short simulated operating time.

Fig. 2: In situ power-loss calibration of three-phase boost converter.

A solution is to use an in situ calibration procedure to perform simulation-based measurements of the power losses under various operating conditions. While the inverter is using highly efficient Rohm SiC MOSFETs, there are still conduction and switching losses that can depend on the specifics of a circuit, such as gate driver timing and interactions with other components. That is why it is important to perform this calibration in situ, or in the context of the actual application.

In step 1 of this calibration process, multiple charging current levels and battery voltage levels were simulated over a 50 msec simulation run. At each operating condition, the average power loss of each component is recorded for use in step 2. For reference, this simulation took a total computation time of 56 minutes.

The second step, shown in figure 3, is to create a “state average” model of the boost converter, which eliminates the fast-switching features but keeps the important lower frequency power conversion and efficiency characteristics. The individual component power dissipation data from step 1 is entered into a look-up table model that injects heat flows into the thermal elements on the schematic. Those include the BCI ROMs for the six MOSFETs, which are then thermally connected to a BCI ROM of a cold plate.

Fig. 3: State-average boost converter with BCI ROM thermal network.

The look-up table model accounts for the current operating state of the system, as does the state-average electrical part of the system model. Thus, long term charging cycles and other operational sequences can be accurately simulated.

Note that this simulation end-time was 1200 seconds, or 20 minutes of charging operation. But the simulation was completed in four minutes of CPU time. This is a 300,000 times speed-up over direct simulation of the switching circuit!

Tools and flows for electro-thermal simulation

VHDL-AMS (BCI ROM) thermal models can be generated by Siemens thermal analysis tools, Simcenter Flotherm and Simcenter FLOEFD. Then, together with SPICE or VHDL-AMS electrical function models, complete electro-thermal simulations can be performed in either the PartQuest Explore (cloud-based) or the Xpedition AMS simulator. As they are both part of the Siemens Xcelerator business platform, Xpedition AMS is tightly integrated into the Siemens PCB design flow, so that a circuit’s layout-aware electrical performance and thermal integrity can be analyzed.

Leveraging IEEE standard VHDL-AMS modeling, a true electro-thermal model ecosystem can now develop, similar to and in parallel with the SPICE/electronics model supply chain that has served the electrical design community for many years. Electronics design engineers, who have recently been blowing on their thermally stressed products, can now breathe a sigh of relief!

For a deeper look into the motivations behind this breakthrough technology and the edge this solution gives to engineering teams, particularly in the accurate simulation of electro-thermal interactions in both steady-state and transient operations, please read the new whitepaper from Siemens, Electro-thermal design breakthrough.