IGBT On-Chip Temperature Sensing: Achieving More Accurate Junction Temperature (Tj) Monitoring and Over-Temperature Protection

In the pursuit of higher power density and system reliability, accurately monitoring the Insulated Gate Bipolar Transistor’s (IGBT) junction temperature (Tj) has become a critical design challenge. While traditional NTC thermistors have served as a foundational thermal protection method, they present inherent limitations that can compromise performance and safety in demanding applications. The industry is increasingly shifting towards on-chip temperature sensing technologies that offer a direct, real-time measurement of the semiconductor’s actual operating temperature. This approach unlocks the potential for more precise control, enhanced protection, and optimized system performance, moving beyond simple thermal trip mechanisms to intelligent thermal management.

This article delves into the principles of IGBT chip-level temperature sensing, compares it with conventional NTC-based methods, and provides practical guidance for its implementation. We will explore how this technology enables engineers to push their designs closer to the thermal limits safely, maximizing the utility of the power module without risking catastrophic failure.

The Problem with Indirect Temperature Measurement: Why NTCs Fall Short

For years, the Negative Temperature Coefficient (NTC) thermistor, typically mounted on the module’s baseplate or Direct Bonded Copper (DBC) substrate, has been the standard for thermal monitoring. The concept is straightforward: as the module heats up, the NTC’s resistance decreases, providing a voltage signal that the system controller can interpret. While simple and cost-effective, this method measures the case temperature (Tc), not the junction temperature (Tj) where the heat is actually generated.

This indirect measurement introduces two significant problems:

• Thermal Gradient: There is a substantial temperature difference between the silicon chip (Tj) and the baseplate (Tc), governed by the thermal resistance of the intervening layers (solder, DBC, baseplate). During operation, Tj can be 20-50°C higher than Tc, and this gradient is not constant; it fluctuates with the load current and switching frequency. Relying on Tc means you are always working with an estimation, often forcing engineers to apply a conservative safety margin and underutilize the IGBT’s full capability.
• Thermal Lag: The thermal mass between the chip and the NTC creates a significant delay in temperature reporting. During rapid load changes or short-circuit events, the junction temperature can spike dangerously in milliseconds, long before the heat propagates to the NTC. This lag can result in a delayed or missed over-temperature trip, leading to accelerated device aging or immediate failure. You can find more details on this topic at Integrated NTC: The Key to IGBT Module Safety and Reliability.

Essentially, an NTC-based system is always looking in the rearview mirror. It reports on a thermal event that has already happened, not the one that is happening in real-time on the silicon.

The Principle of On-Chip Temperature Sensing: A Closer Look

On-chip temperature sensing overcomes the limitations of NTCs by integrating the sensing element directly onto the IGBT or freewheeling diode (FWD) silicon die. This provides a direct, low-latency measurement of the actual junction temperature. Two primary methods are prevalent in the industry.

Diode-Based Sensing: Leveraging Temperature-Sensitive Electrical Parameters (TSEPs)

The most common and accurate approach is to integrate a dedicated temperature-sensing diode onto the main IGBT die. The forward voltage (Vf) of a diode exhibits a highly linear and predictable negative temperature coefficient. The process works as follows:

1. A precise, constant sense current (typically in the range of 1-10 mA) is injected into the integrated thermal diode.
2. The resulting forward voltage drop (Vf) across the diode is measured by the gate driver or a dedicated monitoring circuit.
3. Since the relationship between Vf and temperature is linear (e.g., -2 mV/°C), the measured voltage can be directly translated into the junction temperature after an initial calibration.

Because the sensor is part of the chip itself, it has virtually no thermal lag, enabling it to track rapid Tj fluctuations during high-frequency switching or fault conditions. This provides the control system with immediate, actionable data.

Using Vce(sat) as a Temperature Sensor

An alternative method utilizes the IGBT’s own collector-emitter saturation voltage, Vce(sat), as a Temperature-Sensitive Electrical Parameter (TSEP). At a low collector sense current, Vce(sat) also demonstrates a relatively predictable negative temperature coefficient. This technique involves injecting a small sense current through the IGBT during its off-state and measuring the resulting voltage.

While this method avoids the need for a separate integrated diode, it can be less accurate and more complex to implement. The Vce(sat) characteristic can be influenced by factors other than temperature, such as collector current and gate voltage, and often requires more sophisticated calibration and signal processing to achieve reliable results.

Core Analysis: On-Chip Sensing vs. Traditional NTC Thermistors

The decision to use on-chip sensing versus a traditional NTC involves trade-offs in performance, complexity, and cost. For high-performance systems where maximizing power output and ensuring reliability are paramount, the benefits of on-chip sensing are clear.

Feature	On-Chip Temperature Sensing (Diode)	NTC Thermistor
Measurement Point	Directly on the silicon die (actual Tj)	Module baseplate or DBC substrate (indirect Tc)
Accuracy	High (typically ±2-5°C after calibration)	Lower (estimation of Tj, error varies with load)
Response Time	Extremely fast (microseconds)	Slow (seconds to minutes)
Thermal Lag	Minimal to none	Significant, especially during transients
Implementation Complexity	Requires signal conditioning, amplification, and calibration logic, often integrated into an advanced intelligent gate driver.	Simple voltage divider circuit connected to an ADC.
System Benefit	Enables precise thermal management, performance optimization, and predictive maintenance.	Provides basic, steady-state over-temperature protection.
Ideal Application	EV inverters, solar converters, high-frequency drives, and any system pushing power density limits.	General-purpose motor drives, power supplies, and applications with slow thermal cycles.

Practical Implementation and Design Considerations

Successfully integrating on-chip temperature sensing requires careful attention to the entire signal chain, from the sensor itself to the control system’s response.

Calibration and Signal Conditioning

The raw voltage signal from an on-chip sensor is small and susceptible to noise in a high-power switching environment. Therefore, robust signal conditioning is essential. This typically involves:

• Amplification: Using an operational amplifier to boost the signal to a level suitable for an Analog-to-Digital Converter (ADC).
• Filtering: Implementing low-pass filters to remove high-frequency noise induced by IGBT switching.
• Calibration: A one-time, two-point calibration (e.g., at 25°C and 125°C) is often required to establish a precise voltage-to-temperature conversion formula for each specific device, accounting for minor manufacturing variations. This data is then stored in the controller’s memory.

Integration with Gate Driver and Control Logic

Modern intelligent gate drivers from manufacturers like Infineon are often equipped with dedicated inputs and processing logic for on-chip thermal sensors. These drivers can provide the constant current source, amplify the return signal, and even trigger a fault signal or initiate a soft shutdown autonomously. This tight integration between the sensor and the gate drive is crucial for achieving the fast response times needed for effective protection.

Fault Response Strategies: From Derating to Shutdown

With real-time Tj data, the system can deploy a much more sophisticated protection strategy than a simple on/off trip. A multi-stage response can be programmed into the system controller:

1. Level 1: Warning & Derating. When Tj exceeds a first threshold (e.g., 150°C), the controller can issue a warning and start actively derating the system’s output power. This can be achieved by reducing the switching frequency or modulating the pulse-width modulation (PWM) scheme to lower thermal losses and stabilize the temperature without interrupting operation.
2. Level 2: Soft Shutdown. If the temperature continues to rise past a second, more critical threshold (e.g., 165°C), a controlled, soft shutdown can be initiated. This prevents the abrupt current changes that can cause voltage overshoots and further stress on the system.
3. Level 3: Hard Fault Trip. If Tj spikes suddenly or exceeds the absolute maximum rating (e.g., 175°C), the gate driver initiates an immediate shutdown to prevent catastrophic failure. This real-time data is invaluable for post-failure analysis.

The Future of IGBT Thermal Management and Market Trends

The adoption of on-chip temperature sensing is a key enabler for next-generation power electronics. In applications like electric vehicle inverters and wind turbine converters, where maximizing efficiency and lifetime is critical, direct Tj monitoring is becoming a standard requirement. Power module manufacturers such as Mitsubishi and Semikron are increasingly offering modules with integrated sensors.

This technology also paves the way for advanced predictive maintenance. By logging Tj data over the device’s lifetime, it’s possible to analyze thermal cycling patterns and use models like the Coffin-Manson equation to estimate the remaining useful life of the IGBT module. This allows for proactive servicing and replacement, dramatically improving overall system uptime and reliability. The principles of advanced thermal management will continue to be a cornerstone of power system design.

Key Takeaways for Engineers

• NTCs Measure the Past: NTC thermistors provide a delayed and indirect measurement of temperature, making them unsuitable for protecting against fast thermal transients in high-performance systems.
• On-Chip Sensing is Real-Time: Integrated sensors on the IGBT die measure the actual junction temperature with minimal lag, providing accurate, real-time data for control and protection.
• Implementation Requires a System Approach: Effective use of on-chip sensors demands careful signal conditioning, calibration, and tight integration with an intelligent gate driver and system controller.
• Unlock Higher Performance Safely: By eliminating the guesswork associated with Tj estimation, on-chip sensing allows engineers to operate systems closer to their true thermal limits, increasing power density and efficiency without compromising reliability.
• It’s a Foundation for the Future: Direct Tj monitoring is becoming essential for intelligent power systems, enabling sophisticated thermal management, lifetime prediction, and enhanced safety features in critical applications.