Integration Of S-Parameters For Power Module Verification Into The Engineers’ Design Environment

By Wilfried Wessel (Siemens EDA), Simon Liebetegger (University of Applied Sciences Darmstadt), and Florian Bauer (Siemens EDA)

Developing a power module requires enhanced design and verification methods. Currently, multiple iterations are needed to get the design done. Today, design and manufacturing processes are heavily dependent on physical prototypes. The reason for this is the unique switching characteristics for voltage up to more than 1700 V and currents up to 225A in the same device. This article shows an automatic solution to modify S-Parameters in a way so that they can be back annotated directly. The S-Parameters are extracted by simulation tools from Siemens EDA HyperLynx Advanced Solver using the Method of Moments. After that, they are read into a mixed signal solver called Xpedition AMS. The article starts with an introduction of the original problem to solve. Then the automatic solution will be described. The article concludes with additional research tasks.

Power modules are high-power switching circuits that convert DC- in AC-currents in electric vehicles, renewable energy, photovoltaics, wind turbines, and many more applications. Due to the higher electrical power that must be switched, such power modules consist of more than one IGBT [1], MOSFET and diode [2] share a single package.

Fig. 1: Power Module PI2000 Siemens EDA.

Only if such a multi-device package (see figure 2) is designed correctly can higher voltages and currents be achieved at stable, uniformly distributed temperatures in the package and with acceptable low E- and H-field emissions. Those characteristics become a four-dimensional design challenge, as shown in figure 2. Due to such a polylemma, designers can no longer predict the potential contradictory effects of a single design change. This makes electromagnetic, thermal, and functional simulations essential.

Fig. 2: Power modules are a four-dimensional design challenge.

To get simulated results from an electromagnetic simulation, e.g., HyperLynx Advanced Solver, into a functional simulation for power modules works great for lumped elements. For S-Parameter, some manual rework is required. In the future, S-Parameters will become more critical. This change is driven by wide bandgap semiconductors such as Silicon Carbide (SiC) and Galium Nitride (GaN). These devices support much high switching frequencies. Compared to Silicon IGBT, they also have a reduced package size. This makes a frequency dependent analysis of parasitics effects more critical. This high demand for simulation does not allow time-consuming and error-prone manual rework. For this reason, an automated solution was developed.

Cir. 1: Standard S-Parameter wrapper file.

An example of such a manual rework is shown in Cir. 1. A subset of the transformed wrapper file is shown in Cir. 2. All required changes are marked in yellow. For a small example like this, this process is straightforward and can help to define all requirements for an automated version. For a complex power module topology, this is no longer feasible and requires automation.

Cir. 2: Enhanced S-Parameter wrapper file.

Fully automatic SPICE wrapper transformation

The single steps to manually transform an S-Parameter wrapper can now be implemented in code; several options are available. For a direct integration into the Siemens power modules design suite, Xpedition VB.Net is recommended. VB.Net is a flexible, object-based, powerful programming language well-suited for developing customizations based on COM APIs such as Xpedition.

Fig. 3: VB.Net power module S-Parameter merger.

Before implementing functionality, a new class with two items was defined. In figure 3, the class is called Port. This class has two elements: a signal and a reference port. This technique helps to use VB.Net embedded list functions. If the overhead for the user interface is ignored, the implementation has four functions: get ports, add sub-circuit pins, merge reference pins, and the function to write a new wrapper file. In the Get Ports from the S-Parameter function, the S-Parameter file handed over from the user interface is parsed, and all port pairs are stored in a list of the class Port. This sort of dictionary format is required since multiple nets and reference pins are possible. With the complete port information available, the function Add Sub-Circuit Pins adds the missing pins sub-circuit pins, as was shown in line 4 when transformed from Cir. 1 to Cir. 2. For this function VB.Net has a filter function available which is called Distinct. This function is applied to avoid duplicated pins in the sub-circuit header. The function Merge Reference Pins of figure 3 replaces all 0 in line 12 of Cir. 1 with the correct reference pin. A new wrapper file is generated at the end of the merge process, which can be used without manual intervention.

Fig. 4: User interface of the power module S-Parameter merger.

The easy-to-use user interface only has two buttons: one to point to the SPICE wrapper file and another button called Merge to start the fully automatic merge process. The application can be directly embedded into Xpedition, allowing the engineer to seamlessly integrate the whole power module verification process. Now we will discuss additional benefits and a function test.

Function test

The function test is done with a power module schematic shown figure 5. For this reason, a complete SPICE simulation in Xpedition AMS was configured. The circuit shows three parallel IGBTs and freewheeling diodes for the high-site path and three parallel IGBTs and free-wheeling diodes for the low-site path. This is a typical optimization use case where design engineers want to achieve equal static and dynamic switching losses for each device. This translates into equivalent resistance and inductance behavior for electrical parameters for each device. Optimizing these parameters helps to achieve equal device temperatures [3]. This is important for an extended power module life cycle and reliability.

Fig. 5: PI2000 schematic Siemens EDA.

The power module was designed with Si IGBTs of Infineon IGC193T120 [4] since they are still widely used and have a better price/performance ratio than SiC or GaN due to the established manufacturing process. The simulation is performed three times, first without any parasitic behavior, second using the SPICE parasitics composed by the RLCG matrix as described above, and the third simulation using S-Parameters. The simulation result is analyzed for static DC behavior, stray inductance [5] and simulation performance measured by time.

Figure 6 shows the emitter current through the IGBT U1, U3 and U5. The current through U1 with 76.46 A is approximately 14.5 A higher than the current through the two other devices. This leads to the static switching phase to a much higher self-heating. Referring to table 1, there are minor differences in the DC path. As described before, the S-Parameter requires wave signal propagation, so the DC point must be calculated by interpolation. This is not the case for the RLCG representation. The result shows that these differences are so small that they are still suitable for optimizing the DC path.

Fig. 6: Currents through U1 (red), U3 (blue), and U5 (green) using S-Parameter.

For the description of higher frequencies and switching behavior, S-Parameters are superior. Figure 7 shows the collector-emitter voltage and the derivate of the emitter current during the IGBT switch-off phase. These values can be used to calculate the stray inductance each device sees. It can be identified in figure 7 that the voltage overshoot is similar (25.6 V) for all IGBT devices. The derivate of the emitter current shows some differences. This leads to a stray inductance [5] of 18.62 nH for U1 and 15.22 nH for U5. For the S-Parameter representation, the difference is not as high as the difference using SPICE parasitics, with 35.34 nH for U1 and 21.14 nH for U5. There are two main reasons for the difference. The first reason is the better high-frequency accuracy of the S-Parameter. The second reason is due to the high node and device count of the SPICE parasitic. This causes excessive ringing effects, and the stray inductance results strongly depend on the simulation time step. The ringing also adds difficulties measuring at the correct time location. Minor differences in time have huge effects on the stray inductance.

Fig. 7: Top emitter-collector voltage with an overshoot of U1 (red), U3 (blue), and U5 (green). Bottom derivative of the emitter current of U1 (red), U3 (blue), and U5 (green) using S-Parameter.

The function test proved that the automatic S-Parameter wrapper generation works correctly. Table 1 shows a huge benefit in switching accuracy and a performance enhancement by approximately 20. The result is essential for holistic and fully automated power module optimization.

Table 1: Simulation summary.

Summary

Lumped parasitics as SPICE subcircuits are standard to perform time domain functional simulation for power modules. Table 1 showed the advantages of using an S-Parameter in the functional simulation: increased simulation performance by a factor of 20 and increased high-speed accuracy. The article showed how to automatically modify the S-Parameter wrapper. After the modification it can be back annotated in the mixed signal simulation using the standard functionality of Siemens EDA software. The work of this article helps to improve the power module design and verification process, especially for future usage of wide bandgap semiconductors.

Another advantage is that the simulation result no longer depends on the engineer’s experience. The work of this article explores the use of S-Parameter in the time domain analysis. It was demonstrated that future optimization methods in the electronic domain could be performed in two steps: in the first step of the process, the power module S-Parameters are compared to reference S-Parameters in a compliant check. In the second step, the same set of S-Parameters is used to perform and analyze the transient time domain behavior.

Currently, there is no industry standard to specify power modules with S-Parameter. This could mean that the solution must be enhanced to support Z-Parameters. The future work must also be verified by measurements. That comparison will show which solutions fit best.

For even more on the topics introduced in this article, please download the new whitepaper from Siemens DISW, Why are S-Parameters superior for power module optimization?

References:

1. D.R. Stauffer, J.T. Mechler, M.A. Sorna, “High Speed Serdes Devices and Applications,” Springer Vieweg, pp 208, 2008.
2. Peng Luo, “GaN HEMT Modeling Including Trapping Effects Based on Chalmers Model and Pulsed S-Parameter Measurements,” TU Cottbus, 2018.
3. S. Chowdhury, Z. Stum, Z. Li, K. Ueno, T.P. Chow, “Comparison of 600V Si, SiC and GaN power devices,” Trans Tech Publications Ltd, Switzerland, Vols. 778-780, pp 971-974, 2014.
4. Infineon, IGC193T120, https://www.infineon.com/dgdl/Infineon-IGC193T120T8RM-DS-v02_01-EN.pdf?fileId=db3a30433784a04001382e2b934d28b4, March 2023.
5. P. Fay, D. Jena, P, “High-Frequency GaN Electronic Devices,” Springer, Cham Switzerland, 2020.

Simon Liebetegger is of the University of Applied Sciences Darmstadt.

Florian Bauer is a technical marketing engineer at Siemens EDA.