Here’s our latest EV Tech interview with Christian Weber from Infineon:
Could you give us an introduction to the TC4 and TriCore?
Most engineers are familiar with the AURIX TC3 MCU family, which is more or less the industry standard for electric or combustion subsystems. The TC4 isn’t meant to be a replacement, but more of an extension/upgrade of the TC3 by adding advanced features to cover high-end applications and increase the accuracy for specific use cases.
Let’s focus on the new functionalities of the TC4. There are a lot of improvements, but I would like to focus on two specific ones.
First, there’s an internal AURIX Low Latency Interface (LLI). It connects peripherals like Analog-to-Digital Converter (ADC), enhanced timers and high-resolution Pulse Width Modulation (PWM) with minimal delay. This allows for exploiting the value that Wide Bandgap solutions (WBG; such as silicon carbide (SiC) and gallium nitride (GaN) bring to power conversion and inverter control.
A second new feature is the parallel processing unit (PPU), which is a separate core, independent to the TriCoresTM. Unlike the TriCores, the PPU executes software code in parallel, rather than step by step as in a pure scalar architecture. It’s a vectorized core and commonly used in Digital Signal Processors (DSP). For example, if you are running graphic code where you’re applying the same code to many instances (such as pixels on a screen), you can run that Single Instruction, Multiple Data, known as SIMD processing.
The PPU on the TC4 is particularly useful for AI-on-the-edge applications. It can efficiently execute complex algorithms, such as electrochemical battery models or neural network inference, with a high level of accuracy. It can perform these tasks more than 20 times faster than a TriCore.
In the context of battery management systems (BMS) for electric vehicles, there are a few key market concerns: fast charging without damaging the battery, accurate prediction of range and state of charge, and determining the second life value of the battery pack. These tasks require intelligence and real-time analysis using algorithms that are typically too complex to run on the edge in real-time, many OEMs aggregate the data, and send that to the cloud where a powerful machine can compute this data. This approach introduces several issues, including latency, cloud dependence, cloud costs, and data security concerns.
To solve this, you can always simplify the algorithm on the BMS, but unfortunately, you miss out on the complex granular data. Our solution is to use the TC4 with the PPU, which enables complex and high-performance algorithms to be run directly on the edge device, without relying on the cloud. This allows for real-time analysis, eliminating latency issues. Another advantage of going this route is if you have the same models in your R&D lab and on the edge, then updating the models becomes easy. This also saves R&D time because you may not necessarily need to run the model order reduction.
What sort of questions do they ask you? Those engineers working at tier one suppliers or OEMs, what are they commonly asking you about the features?
There are two common topics.
One, of course, is the price. Integrating a DSP or SIMD unit into the BMS to enable code vectorization gives a processing efficiency advantage, but the costs are high because you must purchase an additional part, perhaps an FPGA, which can cost $10 or more. The TC4 comes with this functionality on the chip, which makes it an economically highly attractive solution.
The second topic is adapting new hardware, which requires new code. However, with the TC4, we not only provide hardware but also the toolchain with it, which is one of our main value propositions. Customers are free to design their algorithms using model-based methods, like MATLAB/Simulink, without investing significant effort in hardware-oriented coding. In simple terms, you can stay focused on making your model work without fretting over the intricacies of the physical hardware it will run on. Our toolchain, Hardware Support Package (HSP), autonomously generates vectorized code from the model. This shift can help OEMs decrease dependence on hardware suppliers once they’ve invested time into their models.
Tying it all together, how does this translate into a better electric vehicle?
At a high level, although this advanced hardware may cost more upfront, the TC4 helps you improve overall energy and system cost efficiency. It lets you squeeze more range out of the battery pack, which allows you to decrease the number of cells in a pack and save money down the road. Also, the TC4 can help you avoid lithium plating, which helps preserve the pack life by decreasing the degradation.
So to tie it together, the benefits of the TC4 not only directly save costs on the amount of cells but also give you indirect benefits down the road. OEMs that are known to have highly advanced battery packs with long Remaining Useful Life (RUL) will be perceived very positively by the market, and the TC4 can help them achieve this.
Click here to learn more about Infineon’s TC4x.
Special thanks to Christian Weber for the interview!
