Thanks to growing component counts, falling board space, and increasing current requirements, effective thermal management has become one of the key challenges facing designers of distributed power architectures in applications ranging from telecommunications base stations to industrial control systems. In particular, a key goal for the designer is to ensure high-efficiency power conversion while achieving a balanced temperature across the PCB.
The advent of intermediate bus architectures has fueled the proliferation of dc bus converters and non-isolated synchronous buck point-of load (PoL) converters. Using dc bus converters, which convert a 48V input to an intermediate bus voltage that is then fed into a POL converter, will help to reduce board size and deliver more output power. But this size reduction places greater pressure on the power MOSFETs used. In particular, good dc bus converter design demands optimised discrete power MOSFETs on both the primary and secondary side of the converter. PoL converters deliver power directly to the board-mounted components and, while they offer small size and high efficiency, rapid increases in boardlevelcurrents means that good thermal design is becoming more critical for these conversion stages.
To help designers address the thermal-management challenges posed by modern distributed power architectures, International Rectifier (IR) conducted a number of tests to compare the various MOSFET package options that can be deployed in DC bus converters and synchronous buck converters. These tests demonstrate that, as system power continues to increase, selecting the appropriate device package style plays a vital role in meeting thermal performance requirements.
DC BUS CONVERTER TESTS
To evaluate the efficiency and performance of the MOSFET package options that can be deployed in dc bus converters, IR created a 48V dc bus converter demonstration board (Fig. 1). This board, which was smaller than a conventional 1/8th brick dc-dc converter format, was designed to deliver 27.5A at an 8V output voltage (i.e. 220W) with a typical efficiency of around 96%.
The dc bus converter circuit was constructed using the IR2085S primary side controller and driver IC. This is a selfoscillating half-bridge driver IC with 50% duty cycle ideally suited to half-bridge bus converters in the 36 to 75V range. In the test, primary-side 100V n-channel power MOSFETs, which satisfy applications in this power range, were compared. On the secondary side, the comparison involved two 30V n-channel power MOSFETs. These were configured using a self-driven synchronous rectification topology to provide an 8V output. It should be noted that designing for a 12V output could be achieved using 40V n-channel power MOSFETs. For both the primary and secondary side, comparisons were made between conventional SO-8 MOSFETs (paralleled and unparalleled) and MOSFETs packaged using DirectFET technology
The traditional approach to addressing dc bus converter thermal management, including the avoidance of hot spots, has been to parallel multiple SO-8 devices. However, IR's experiments show that paralleling two devices actually increases board temperature thanks to the increase in switching losses. Furthermore, tests show that the primary side transfers heat to the secondary, increasing the overall board temperature of both stages.
Table 1 shows the comparison of board temperature when using two versus four IRF7495 100V SO-8 single n-channel HEXFET power MOSFETs. In each case, the secondary is based on two IRF6618 30V DirectFET devices. Observations while paralleling FETs on the primary side showed an increase in secondary-side board temperature as a result of the higher losses. In summary, even though paralleling FETs increases efficiency (Fig. 2) and makes the two stages more balanced in terms of temperature, the board temperature remains relatively high.
The third column in Table 1 shows the same situation but with the SO-8 devices replaced with DirectFET equivalents. DirectFET has been designed to offer improved efficiency, power-handling capabilities, and thermal performance. Using this technology in the primary of the dc bus converter shows a marked reduction in the temperature of both the primary and the secondary stages. Furthermore, the primary side MOSFETs now allow for balanced temperatures across both of the stages. With SO-8s on the primary side, there is a temperature imbalance in the region of 20°C, which limits the power capability of the board. Specifically, when using the SO-8 alternatives, it was observed that the primary side heats up significantly when pushing higher power densities, such that the SO-8-based topology was typically limited to a power of 150W. In this situation, the only way to increase power output is to parallel the primary-side FETs so as to remove the hot spot. With the DirectFET alternative, the temperatures between primary and secondary are within as little as 3°C of each other.
In terms of efficiency, if we again refer to Figure 2, we see the efficiency curve showing that the DirectFET MOSFET outperforms the equivalent SO-8 solution by around 1%. This may not sound like a lot, but taken in the context of a circuit-that is already operating at 95 to 96%, this represents a significant incremental improvement. Finally, the results from the test also validate the fact that single MOSFETs based on the enhanced DirectFET package technology can successfully be used to replace two paralleled SO-8 parts.
POINT-OF-LOAD CONVERTER TESTS
To conclude the MOSFET package comparisons, IR also evaluated MOSFET solutions for PoL converters. To do this, the company created a two-phase synchronous buck converter demonstration board that allowed MOSFET comparison in a system where input voltage was 19V, output voltage was 1.35V, and switching frequency was 300kHz. MOSFETs from a variety of suppliers were tested, and package options comprised LFPAK, PowerPAK, SO-8, and DirectFET. In all cases the devices operated without airflow across the PCB, and case temperatures were tested at 35A and 40A.
Results showed that in the transition from 35A to 40A, the case temperature of the SO-8 devices was around 10°C higher than that of the other package types. This remains the case even when two SO-8s are paralleled together. The reason for this temperature differential is due to the higher resistance of the SO-8s, which means that as currents rise, power dissipation increases much faster than other advanced packages. Due to the higher thermal impedance of SO-8s, these devices are unable to dissipate heat as well as other advanced packages in which metal or die is exposed to the pc board and/or the ambient.
Finally, these tests demonstrated that in most cases the temperature on the synchronous FET is higher than that on the control FET. At 35A and 40A, the two FET temperatures are best balanced with pairs of devices in the same package, from the same supplier.
From the tests performed on both dc bus and synchronous buck converters, it is clear that MOSFET package thermal performance plays a major role in achieving thermal-management targets in the face of increasing system power levels. Using package technologies such as DirectFET, performance can be improved with a simpler design rather than through the added complexity of paralleled devices. Furthermore, improved thermal performance means that the PCB size and the amount of copper used can be maintained as current densities rise.