COOLING OF POWER SWITCHING DEVICES
Thermal resistance
We have seen that by adopting a switching strategy the power loss in the switching devices is small in comparison with the power throughput, so the converter has a high eYciency. Nevertheless almost all the heat which is produced in the switching devices is released in the active region of the semiconductor, which is itself very small and will overheat and breakdown unless it is adequately cooled. It is therefore essential to ensure that even under the most onerous operating conditions, the temperature of the active junction inside the device does not exceed the safe value.
Consider what happens to the temperature of the junction region of the device when we start from cold (i.e. ambient) temperature and operate the device so that its average power dissipation remains constant. At Wrst, the junction temperature begins to rise, so some of the heat generated is conducted to the metal case, which stores some heat as its temperature rises. Heat then Xows into the heatsink (if Wtted), which begins to warm up, and heat begins to Xow to the surrounding air, at ambient temperature. The temperatures of the junction, case and heat- sink continue to rise until eventually an equilibrium is reached when the total rate of loss of heat to ambient temperature is equal to the power dissipation inside the device.
The Wnal steady-state junction temperature thus depends on how diYcult it is for the power loss to escape down the temperature gradient to ambient, or in other words on the total ‘thermal resistance’ between the junction inside the device and the surrounding medium (usually air). Thermal resistance is usually expressed in 8C/W, which directly indicates how much temperature rise will occur in the steady state for every watt of dissipated power. It follows that for a given power dissipation, the higher the thermal resistance, the higher the temperature rise, so in order to minimise the temperature rise of the device, the total thermal resis- tance between it and the surrounding air must be made as small as possible.
The device designer aims to minimise the thermal resistance between the semiconductor junction and the case of the device, and provides a large and Xat metal mounting surface to minimise the thermal resist- ance between the case and the heatsink. The converter designer must ensure good thermal contact between the device and the heatsink, usually by a bolted joint liberally smeared with heat-conducting com- pound to Wll any microscopic voids, and must design the heatsink to minimise the thermal resistance to air (or in some cases oil or water). Heatsink design oVers the only real scope for appreciably reducing the total resistance, and involves careful selection of the material, size, shape and orientation of the heatsink, and the associated air-moving system (see below).
One drawback of the good thermal path between the junction and case of the device is that the metal mounting surface (or surfaces in the case of the popular hockeypuck package) can be electrically ‘live’. This poses a diYculty for the converter designer, because mounting the device directly on the heatsink causes the latter to be dangerous. In addition, several separate isolated heatsinks may be required in order to avoid short-circuits. The alternative is for the devices to be electrically isolated from the heatsink using thin mica spacers, but then the thermal resis- tance is appreciably increased.
Increasingly devices come in packaged ‘modules’ with an electrically isolated metal base to get round the ‘live’ problem. The packages con- tain combinations of transistors, diodes or thyristors, from which vari- ous converter circuits can be built up. Several modules can be mounted on a single heatsink, which does not have to be isolated from the enclosure or cabinet. They are available in ratings suitable for converters up to hundreds of kilowatts, and the range is expanding. This develop- ment, coupled with a move to fan-assisted cooling of heatsinks has resulted in a dramatic reduction in the overall size of complete con- verters, so that a modern 20 kW drive converter is perhaps only the size of a small briefcase.
Arrangement of heatsinks and forced air cooling
The principal factors which govern the thermal resistance of a heatsink are the total surface area, the condition of the surface and the air Xow. Most converters use extruded aluminium heatsinks, with multiple Wns to increase the eVective cooling surface area and lower the resistance, and with a machined face or faces for mounting the devices. Heatsinks are usually mounted vertically to improve natural air convection. Surface Wnish is important, with black anodised aluminium being typically 30% better than bright.
A typical layout for a medium-power (say 200 kW) converter is shown in Figure 2.19.
The fans are positioned either at the top or bottom of the heatsink, and draw external air upwards, assisting natural convection. The value of even a modest air-Xow is shown by the sketch in Figure 2.20. With an air velocity of only 2 m/s, for example, the thermal resistance is halved as compared with the naturally cooled setup, which means that for a given temperature rise the heatsink can be half the size of the naturally cooled one. Only a little of this saving in space is taken up by the fans, as shown in Figure 2.19. Large increases in the air velocity bring diminish- ing returns, as shown in Figure 2.20, and also introduce additional noise which is generally undesirable.
Cooling fans
Cooling fans have integral hub-mounted inside-out motors, i.e. the rotor is outside the stator and carries the blades of the fan. The rotor diam- eter/length ratio is much higher than for most conventional motors in order to give a slimline shape to the fan assembly, which is well-suited for mounting at the end of an extruded heatsink (Figure 2.19). The rotor inertia is thus relatively high, but this is unimportant because the total inertia is dominated by the impeller, and there is no need for high accelerations.
Mains voltage 50 or 60 Hz fans have external rotor single-phase shaded-pole motors, which normally run at a Wxed speed of around 2700 rev/min, and have input powers typically between 10 and 50 W (see Chapter 6). The torque required in a fan is roughly proportional to the cube of the speed, so the starting torque requirement is low and the motor can be designed to have a high running eYciency (see Chapter 6). Slower-speed (but less eYcient) versions are used where acoustic noise is a problem.
Low-voltage (5, 12 or 24 V) d.c. fans employ brushless motors with Hall eVect rotor position detection (see Chapter 10). The absence of sparking from a conventional commutator is important to limit inter- ference with adjacent sensitive equipment. These fans are generally of lower power than their a.c. counterparts, typically from as little as 1 W up to about 10 W, and with running speeds of typically between 3000 and 5000 rev/min. They are mainly used for cooling circuit boards directly, and have the advantage that the speed can be controlled by voltage variation, thereby permitting a trade-oV between noise and volume Xow.