Thermal Management in Power Electronics Nanoscale Sensors Control Hotspots in GPUs And Batteries

Related Vendor

Taiwan Machine Tool & Accessory Builders' Association (TMBA)

Smaller sensors, faster data, easier integration: Digid brings nanoscale sensor technology to industrial mass production. By combining organic electronics with a unified signal platform, thermal and mechanical stresses can be measured in places where conventional sensors cannot fit.

Thermal management plays a central role in the massively parallel architectures of modern graphics processing units (GPUs) and in the highly compact battery packs used in electric vehicles. This is because waste heat is the limiting factor for performance. Existing solutions are mostly based on discrete NTC thermistors or integrated silicon sensors. These face two main challenges: thermal inertia and size.

The Problem of Thermal Blind Spots

Traditional sensors have a relatively high thermal mass. As a result, they can only respond to rapid temperature transients—such as those caused by load spikes in AI data centers—with a delay. Furthermore, they are often too large to be placed where the heat is generated. This is immediately between the battery cells or directly on the interposer of a high-performance chip. The result is what is known as “thermal blind flight,” which engineers have so far had to compensate for by using generous safety margins and early thermal throttling.

Nanotechnology at the Transistor Level

This is where Digid comes in. By utilizing organic thin-film transistor (OTFT) technology, developed in collaboration with partners such as Smartkem, the company has succeeded in developing sensors that are barely larger than a standard transistor. These tiny sensors can be integrated directly into circuit layouts.

The key physical advantage is that the extremely low thermal mass of the nanoscale sensors enables response times in the millisecond range. Whereas a conventional sensor still measures the ambient temperature of the housing, the nanosensor already detects the initial rise in temperature at the semiconductor die.

For developers, however, it is not only the sensor physics that matter, but also the ease of integration into existing systems. Due to their design, nanosensors produce extremely weak analog signals that are highly susceptible to electromagnetic interference (EMI). This is a critical issue in power electronics.

An End-to-End Approach

Digid's approach is as follows:

• Analog Front End (AFE): A dedicated interface conditions the weak signals immediately near the source.
• Digitalization: The system converts the data into a robust digital format (e.g., I²C or SPI).
• Firmware and API: Instead of laboriously calibrating calibration curves, developers can access pre-processed temperature data via an API.

This so-called "plug-and-sense" approach significantly shortens time-to-market, as the design team no longer has to deal with the complex analog conditioning of nanomaterials.

HPC And Battery Management

In practice, this opens up two strategic areas of application:

• High-Performance Computing (HPC): By integrating sensor arrays (thermal mapping), the power consumption of GPUs can be controlled more dynamically. Instead of throttling the entire chip, only the affected areas can be regulated, which increases the hardware’s overall throughput.
• Battery Management Systems (BMS): In EV batteries, so-called thermal runaways should be detected early on. Nanoscale sensors can be placed in situ, i.e., directly inside the cell or on the busbars. They detect local hotspots before they can spread to the entire module. This represents a massive improvement in safety compared to the selective monitoring commonly used today.

A Takeaway for Design Practice

The future of sensor technology is not only smaller, but above all more digital. For hardware developers, this means a shift in thinking. The challenge is shifting from circuit design (analog design) to intelligent data analysis (system design). Those who understand the thermal limits of their hardware more precisely can push them further, which also provides a competitive advantage that cannot be underestimated in the age of AI and e-mobility. (heh)

Battery Technology

The Race to Find the Battery of the Future