Electronics 101

 

5. Power Losses and Thermal Considerations

Power semiconductors are thermally limited and a good thermal design is the key to their cost effective utilization. Thermal design is important in all applications but is particularly important for power semiconductors for two reasons:

1. They operate with very large current densities and with a steep temperature gradient between junction and ambient.
2. The thermal mass of a semiconductor is very small and it can go into a thermal runaway in a fraction of a millisecond.

It is the task of the designer to select the heatsink or other cooling method, i.e. to do the “thermal design”.

In general, the objective of the thermal design is the selection of the best device-heatsink combination that has the following characteristics:

  • It is capable of delivering the required amount of power to the load;
  • It is designed to dissipate the semiconductor losses into the ambient;
  • It keeps the junction temperature within safe limits.

This process may require a number of iterations but can be speeded-up significantly with the help of simple device models plugged into an application-specific spreadsheet.

 

5.1. The first step: power dissipation in the device

The starting point of a thermal design is the calculation of power dissipation in the semiconductors. We distinguish between conduction and switching losses:

  • Switching losses occur when the device is transitioning from the blocking state to the conducting state and vice-versa. This interval is characterized by a significant voltage across its terminals and a significant current through it. The energy dissipated in each transition needs to be multiplied by the frequency to obtain the switching losses;
  • Conduction losses occur when the device is in full conduction. The current in the device is whatever is required by the circuit and the voltage at its terminals is the voltage drop due to the device itself. These losses are in direct relationship with the duty cycle.

The first step is to calculate these losses. The calculation of conduction losses is relatively simple but the calculation of switching is dependent on a number of factors that make it complicated:

  • The impedance of the gate drive circuit has a significant impact on the switching performance of MOSFETs and a lesser impact on that of IGBTs;
  • Temperature increases switching losses in all minority carrier devices, specifically diodes and IGBTs;
  • The reverse recovery transient causes losses in the diode that are difficult to quantify. It also has a significant impact on the turn-on losses of the transistor that carries the reverse recovery current, in addition to the load current;
  • All transistors are different: some have been designed to achieve superior switching performance, some with better conduction characteristics.
  • The parasitics Shown in Figure 1 can have a significant impact on switching losses.

Tools to calculate losses are available from a variety of sources. They fall into two classes:

  • Those that rely on device models, with SPICE being a typical example;
  • Those that rely on mathematical models generated from test results.

More sophisticated tools go beyond the simple calculation of losses and calculate junction temperature in a specific application, like a motor drive or a buck converter.

Many designers use experimental methods to calculate losses. The waveform labeled B in Figure 10 is the integral of the voltage times the current and represents the energy dissipated inside the device during this specific switching transition. Switching losses in the application can be calculated from a limited number of intelligently chosen scope pictures. This method has the advantage of being specific to the application in its own environment, as opposed to models and simulations.

It should be kept in mind that the losses thus calculated are an average for a load condition that is in steady state. The temperatures calculated in these conditions are also steady state temperatures.

Most power circuits are rated for peak loads and the power dissipation under peak load conditions can be much higher, even if their duration is short. Thus, the assumption of a steady state operation does not apply any longer. The losses during peak loads can be calculated in ways that are similar to the ones we have seen, but the calculation of junction temperature is much more complicated, as we will see in the next section.

For a simple square wave operation the following tool calculates IGBT losses and steady state temperature: http://igbttool.irf.com/

A description of the underlying methods can be found here: http://www.irf.com/product-info/igbt/hardswitchinglosses.pdf

 

5.2. The second step: the calculation of the junction temperature

Simple thermal equations are frequently used to calculate junction temperature in steady state in a known thermal environment. This is explained in detail in a number of publications, including AN-1057: www.irf.com/technical-info/appnotes/an-1057.pdf

The method therein described is commonly used by power engineers and it is frequently made part of the circuit that is being modeled with SPICE. It should be made clear, however, that it’s a gross oversimplification of the thermal phenomena. The term “heatsink temperature” is very misleading because there is no such thing: different places in a heatsink have different temperatures and they can be 20°C apart.

It is, however, a simple and convenient way of getting a first approximation of temperatures in a power system.

When it comes to transients and peak loads, these methods are hopelessly inadequate as they can’t account for the fact that heat spreads into a system like a wave, with a finite, variable and unknown speed. Attempts have been made to extend these traditional methods to transient conditions with the “transient thermal impedance curve” being the most notable example (see Section 5 of AN 949 http://www.irf.com/technical-info/appnotes/an-949.pdf). These methods assume a constant case temperature, a condition that is somewhat valid if the thermal time constant of the heatsink is large with respect to the duration of the transient. This is certainly not the case when devices are mounted on a PCB.
The only proper way to calculate temperatures in a power system during a transient overload is by means of Finite Element Analysis simulation. Several simulation packages are widely advertised. They require a CAD drawing as an input and a detailed description of the material in the system. They provide a 3-D output with temperature for every coordinate. They can also provide a movie clip of temperature evolution, as shown in the enclosed video file.