Thermal-Resistance of DC/DC Converters with Power MOSFETs

| Autor / Redakteur: Peter Miller, Rich Nowakowski * / Thomas Kuther

Figure 1. TPS543C20 40-A thermal image at 0.9-V output with 12-V input, at 25°C ambient
Figure 1. TPS543C20 40-A thermal image at 0.9-V output with 12-V input, at 25°C ambient (Source: Texas Instruments)

Since operating temperature contributes to the performance and reliable life span of semiconductor devices, conducting thermal analysis on power devices is an important part of the design process.

Understanding thermal information from a DC/DC converter data sheet is a critical component of thermal analysis. The data sheets for most modern power devices provide thermal-resistance tables and thermal images (Figure 1) to help designers with their thermal analysis. However, understanding the information in the data sheet is not always direct or easy.

This article presents assumptions that analog designers may make for thermal analysis. The analysis for each assumption is followed with insights to decipher the actual thermal information in the data sheet. The device used for analysis is a high-current DC/DC converter capable of delivering 20 A to 30 A of output current at a low output voltage from a 12-V input.

Power Architecture for Point-of-Load Consumer Applications

The thermal information section of a data sheet typically appears before the electrical characterization section. Table 1 shows the thermal information table from the data sheet of the TPS543C20 in the 5- by 7-mm, quad-flat, no-lead (QFN) package.[1] For more information about the thermal metrics shown, please read the application report, “Semiconductor and IC Package Thermal Metrics.”[2]

Analog designers may look at Table 1 and see the junction-to-ambient thermal resistance is 28.9° Celsius per watt (°C/W). For example, when the maximum ambient temperature inside the system is 65° and the maximum recommended operating temperature is 125°, the estimated allowable power dissipation of the DC/DC converter is 2.07 W (125 – 65/28.9) according to the RθJA specification rating.

For this simple analysis to be correct, assume that the system uses a Joint Electron Device Engineering Council (JEDEC)-standard, high-K board. Essentially, a high-K board is a 1.6-mm thick, four-layer board with 2-ounce copper on the top and bottom layers, 1-ounce copper for the internal planes, and a copper area of 74.2 mm by 74.2 mm. The package device pins are connected to the top layer and routed with 0.25-mm wide traces, 25 mm to the perimeter, and equally spaced no more than 2.54 mm apart. But questions remain as to what the rest of the thermal specification numbers in Table 1 mean and how they pertain to a high-K board.

Components of Thermal Resistance

The junction-to-ambient thermal resistance, RθJA, is the most commonly reported thermal metric and the most often misused. Note the difference between the first value in Table 1, junction-to-ambient thermal resistance, and the last value, junction-to-case (bottom) thermal resistance.

Of the 28.9°C/W thermal resistance, only 1°C/W is inside the package. Therefore, only about 3% of the total thermal resistance is moving the heat from the power metal-oxide semiconductor field-effect transistors (MOSFETs) into the ground pad of the circuit board and to the surface of the integrated circuit (IC). The remaining 97%, or 27.9°C/W, is moving the heat from the IC through the circuit board, thus spreading the heat through the ground plane and surface traces and into the ambient area.

The 27.9°C/W thermal resistance is made up of several parts. The first part—over which designers have very little control—is the thermal resistance from the circuit board into the surrounding air. Without forced airflow, each square millimeter of solder-masked FR4 has a thermal resistance to air of about 65,000°C/W, or about 12°C/W for the 74.2- by 74.2-mm JEDEC board. Although larger boards will offer improved performance, other heat sources in close proximity will begin to contribute to the total power dissipated by the circuit board local area.

Lateral heat flow through the circuit board’s copper will reduce the heat dissipated beyond 35 mm from the IC. This second part of thermal resistance, which is the majority of the remaining 15.9°C/W, comes from this lateral heat flow, and is where designers can make the most thermal improvements.

Reconsidering JEDEC Board Design

The internal ground plane of the high-K board is only a single layer of 1-ounce copper, which is the primary conduction path for heat drawn from the thermal pad under the DC/DC converter. A second 1-ounce copper ground plane can easily reduce the lateral thermal resis- tance by half, while additional ground layers will continue to reduce the lateral resistance.

The surface layers that connect to the converter’s pins of the JEDEC board are not copper pours but 0.25-mm- wide traces, even when adjacent package pins are intended for connection to a common copper pour. The traces are fanned out from a 0.3-mm-wide space at the IC to 2.54 mm apart at the perimeter of the board. Compared to a copper pour, this fanned pattern has two to four times as much lateral thermal resistance as a solid copper pour. In power devices such as TI’s PowerStack clip-QFN package of the TPS543C20 in Figure 2, the PVIN, GND and SW pins are connected to the power MOSFETs through wide, thick copper clips. Copper pours at these pins can significantly improve the lateral heat flow from the FETs across the circuit board and eventually into the ambient air.

Using 2-ounce copper for the top layer with copper pours at PVIN, GND and SW, and at least two 1-ounce copper ground planes connected to the package thermal pad reduces the 15.9°C/W thermal spreading resistance by 50 to 70%, thus reducing the practical RθJA from the JEDEC reference value of 28.9°C/W to 20°C/W the case, and how it’s now possible to measure the die temperature effectively through the top of the package. ψJT is very low because almost no heat is actually flowing from the top of the case into the ambient area.

Remember that the thermal resistance of the circuit board is 65,000°C/W for 1 mm2 of circuit-board area. The top side of the package has about the same thermal resistance rating, but it’s size is much smaller. The 30-mm2 package area has a thermal resistance of about 2,000°C/W, or about 1,000°C/W with forced air cooling, so very little heat will flow through the top of the package. This high top-side case-to-ambient thermal resistance enables the top-side package temperature to serve as a very good proxy for the die temperature.

The Common Question of Top-side Cooling Assumptions

A common question that arises is top-side cooling. While some benefit can be had through top-side cooling, the small thermal surface area of the converter severely limits its effectiveness.

With heat generated over an area of approximately 10 mm2 on one side of the package, a 0.1-mm-thick layer of thermal grease introduces 5 to 10°C/W of thermal resistance. A 0.2-mm-thick “super-thermal” gap filler intro- duces approximately 20°C/W of thermal resistance, and conventional gap fillers add 4- to 10-times higher resistance. Compared to the 20°C/W thermal path through the circuit board, very little heat can be drawn through the package unless a thin layer of thermal grease conducts the heat directly to a heat sink.

It is interesting to note the low value of ψJT despite the high value of the actual thermal resistance to the top of package are applicable to other packages to similarly reduce their thermal resistance. The answer is that these techniques can be tried, but their effectiveness depends on how the JEDEC-standard board limits the junction-to- ambient thermal path of these other packages.

Not all packages can improve their circuit board’s thermal spreading resistance as effectively in a practical design as much as the PowerStack clip-QFN package. Consider whether the JEDEC board moves most of the heat generated by the DC/DC converter through the top layer traces or through the ground plane. Also, consider whether the DC/DC converter package allows connection of the pins to the wide planes, or if the pins are limited to the narrow traces of the converter.

Designers often Misapply the Thermal Metrics

Many thermal metrics exist for semiconductor and IC packages, which range from RθJA to ψJT. Often, designers misapply these thermal metrics when trying to estimate the junction temperatures in their systems. Ultimately, thermal performance depends on the circuit-board layout and using standard JEDEC-referenced thermal numbers.

How to Power Smart Multivariable Sensor Transmitters

How to Power Smart Multivariable Sensor Transmitters

10.11.17 - Power supplies for smart transducers in industrial applications must meet the highest requirements. This article presents a transmitter design with BLE connectivity for a multivariable sensor. lesen

Considering EMC Emissions Early in the Design Process

Considering EMC Emissions Early in the Design Process

09.11.17 - If developers of power supplies do not want to experience an unpleasant surprise during the prescribed EMC conformity test, it is a good idea to consider the EMC as early as possible. lesen

Highly Integrated Chips:
Super-Brains for the IoT Age

Highly Integrated Chips: Super-Brains for the IoT Age

13.11.17 - Electronic devices are becoming continually smaller, more flat and multifunctional and their “inner life”, the microelectronic systems, along with them – thanks to 3D integration. lesen

References:

[1] “TPS543C20 4-VIN to 16VIN, 40-A Stackable Synchronous Step-Down SWIFT Converter with Adaptive Internal Compensation.” Texas Instruments data sheet (SLUSCD4A), September 2017.[2] Darvin Edwards and Hiep Nguyen, “Semiconductor and IC Package Thermal Metrics,” Texas Instruments Application Report (SPRA953C), April 2016.

* Peter Miller is a Systems Engineer with Texas Instruments in Manchester, NH, USA.

* Rich Nowakowski is a Product Marketing with Texas Instruments in Dallas, TX, USA.

Kommentar zu diesem Artikel abgeben

Schreiben Sie uns hier Ihre Meinung ...
(nicht registrierter User)

Zur Wahrung unserer Interessen speichern wir zusätzlich zu den o.g. Informationen die IP-Adresse. Dies dient ausschließlich dem Zweck, dass Sie als Urheber des Kommentars identifiziert werden können. Rechtliche Grundlage ist die Wahrung berechtigter Interessen gem. Art 6 Abs 1 lit. f) DSGVO.
Kommentar abschicken
copyright

Dieser Beitrag ist urheberrechtlich geschützt. Sie wollen ihn für Ihre Zwecke verwenden? Infos finden Sie unter www.mycontentfactory.de (ID: 45525111 / English)