Small Is Cool

In times past big was always beautiful when it came to heatsink technology. Less is more when it comes to the dual-sided cooling of DirectFET power MOSFET devices using miniature heatsinks.

Significant increases in the power density of multiphase voltage regulator module (VRM) circuits have been enabled by the use of source-mounted power MOSFET devices. Increases in current density have been enabled by two factors.

Firstly, package parasitic impedances have been reduced, and secondly, the thermal resistance between the power device junction and the external ambient has been dramatically cut.

Exploiting the advantages of source-mounted packages depends very much upon the thermal management strategy adopted by the system designer. Here we look at the thermal management strategies available for source-mounted packages and, using empirical data from International Rectifier's own tests, assess the advantages of using small heatsinks to improve heat dissipation in these devices. The impact of direct heatsink attachment on device reliability under power cycling conditions is also investigated.

Figure 1a is a schematic showing the construction of International Rectifier's DirectFET source mounted power MOSFET technology and Figure 1b is an illustration of this device assembled onto a PCB. As the diagram in Figure 1a shows, the device is based on a power MOSFET die housed in a copper clip assembly. This MOSFET die has solderable area array gate and source electrode connections on its front surface, while the rear surface of the die, which forms the MOSFET drain electrode, is connected to the clip using an electrically conductive adhesive. The clip extends over the edge of the MOSFET die to enable drain connections to be made to the circuit board.

Use of the copper clip to house the power MOSFET die delivers significant thermal advantages when compared to conventional package technologies such as the SO-8. The main thermal advantage is that the top of the package (i.e. the metal clip) offers a low thermal resistance connection to the power MOSFET die. Exploitation of this low resistance thermal path to remove heat from the power die can result in less heat being transferred to the circuit board. Alternatively, heat may be transferred from the power die for dissipation through both the circuit board and the top of the package. Known as 'dual sided cooling', this last technique offers the circuit designer significant flexibility in implementing effective thermal management schemes.

One method of implementing a thermal management scheme in a DirectFET-based design is through the use of small heatsinks that are mounted directly to the surface of the device. There are a number of commercially available heatsinks from which to choose, with options including different finishes and differing fin heights. To help engineers assess the best solution for a particular design, International Rectifier has performed a number of tests using the heatsinks shown in Figure 2, all of which have been prepared by machining commercially available heatsinks to fit the source-mounted package. Heatsink A is a multiple plate array type with a silvered finish, and heatsink B is identical in dimensions to heatsink A but has a blackened finish. Heatsink C is a pin array type with taller dimensions than heatsinks A and B. Each heatsink was assembled onto two of the source-mounted packages using a thermal interface compound, and thermocouples were attached to the device to measure the temperature of the device casing.

In order to fully assess real-life conditions, IR performed tests using an experimental wind tunnel that simulates the effects of various forced air cooling conditions. Figure 3 shows the experimental wind tunnel set-up used to characterise the source-mounted devices. This tunnel consists of a clear insulating plastic cylinder with a variable speed AC fan positioned at the cylinder inlet. The air ejected from the fan runs through an extended mesh grid to provide laminar airflow into the system. Air velocity is measured using a pitot tube close to the exit of the fan.

During the experiments, power dissipation in the MOSFETs was controlled using a voltage follower op amp circuit connected between the source and gate of each device. This enabled each MOSFET to operate in a linear mode, dissipating constant power without drift in the gate-to-source voltage applied to the device. For each gate-to-source voltage setting the device power dissipation was calculated using the voltage and current measurements.

The time taken for the heatsink to reach stable temperature was established by monitoring the temperature of the device case whilst applying constant power to the MOSFET junction. Constant case temperatures were obtained within approximately 300 seconds of applying the power.

Table 1 shows the case-to-ambient thermal resistance of the MOSFET devices with and without additional heat sinking, and under forced air and natural convection conditions. With no heatsink, natural convection heat flow between the DirectFET package and ambient results in an Rthc-a of 50°C/W. The addition of a heatsink in natural convection conditions reduced the Rthc-a to between 41.8 and 45°C/W depending on heatsink type used. A key conclusion from this data is that the drop in Rthc-a produced by the addition of a heatsink under natural convection is unlikely to be of a large enough benefit to a design's thermal management scheme to justify the additional cost of components, materials and assembly. However, under forced convection the effect of adding heatsinks to the source-mounted package become significant. As the table shows, with a forced convection of 2.3m/s, the Rthc-a drops from over 21°C/W to between 12.3 and 12.5°C/W upon addition of the heatsink.

The effects of varying the air speed upon the cooling characteristics of a heatsink mounted to the DirectFET package are shown in Figures 4(a) and (b). Figure 4(a) shows case temperature versus device power dissipation over a range of linear air speeds. With increasing air speed the slope of the temperature versus device power dissipation plot decreases. This is indicative of a decrease in case-to-ambient thermal resistance with increasing air velocity. A graph of Rthc-a versus air speed for the tall black heatsink is shown in Figure 4(b).

The experiment described above clearly demonstrates that the application of miniature heatsinks to the DirectFET package enables significant reductions in the thermal resistance between case and ambient. However, heatsinking can also have an impact on reliability. There are three cases to consider:

  1. The effect a heatsink enabled reduction in junction operating temperature can have on MTTF.
  2. The effect a heatsink enabled reduction in operating temperature range can have on reduction in board attach solder fatigue.
  3. The effect that the additional mechanical stresses potentially caused by adhering a heatsink to the top of the DirectFET may have on thermal cycling performance.

Case 1: Reducing the junction operating temperature in application is widely known to increase the lifetime of a MOSFET. If we examine the results from the forced air cooling experiments they show that the addition of the miniature heatsink reduced the Rth(j-a) from 21°C/W (no heatsink) to 12.5°C/W (with heatsink). In an application dissipating 4 Watts this would mean the addition of a heatsink would reduce the junction temperature by 34°C. Using an Arrhenius based calculation it is possible to determine that a reduction of 34°C would equate to greater than a ten times increase in the MTTF for the device.

Case 2: With the reduction in operating temperature then there will also be a reduction in the temperature range over which the device/board system is thermally cycled. This will reduce the rate of fatigue in the board attach solder. It is possible to use Coffin-Mason to determine that a reduction of the operating range of 34°C will equate to a four times increase in the MTTF through a reduction in solder fatiguing.

Case 3: In order to determine the effect of the additional mechanical stresses caused by adhering a heatsink to top of the DirectFET it was necessary to run a comparative experiment between devices with and without heatsinks. To perform this experiment, samples of DirectFET devices were assembled onto test cards. Machined aluminium blocks, used to represent heatsinks, were mounted upon half of the samples using a thermal interface material, while the remaining samples without a heatsink were used as a control.

The power cycling tests were based on dissipating a known power through the MOSFET junction for two minutes before turning the device off and allowing it to cool for a further two minutes — this equated to a single cycle. During each cycle the junction temperature was therefore raised from 35°C to 105°C and allowed to cool back down to 35°C again. This cycle was completed 16,000 times and the on resistance (RDS (ON)) of each device was monitored before, during, and after the tests.

Figure 5 shows the effects of power cycling on DirectFET devices with and without heatsinks mounted to the 'can'. With no heatsink attached the average RDS (ON) shift after 16,000 power cycles is 0.15m_. However, with a heatsink in place, the average RDS (ON) shift is significantly less at just 0.05m_. This reduced level of RDS (ON) shifting when the heatsink is attached is a consequence of the heatsink thermal capacity. During periods of power dissipation and MOSFET cooling, the heatsink reduces the rate of temperature change experienced by the device interconnects. Consequently this reduces the solder joint fatigue of the DirectFET interconnects. This phenomenon is likely to extend rather than reduce the long-term device reliability although the exact results will be a function of heatsink design, the method used to attach the heatsink to the DirectFET and the interface material used.

While offering little benefit with natural convection, the use of small heatsinks to cool source-mounted power MOSFET devices such as IR's DirectFET products results in a significant reduction in the junction-to-ambient thermal resistance under conditions of forced convection cooling. The experimental data shows that the difference between silver and blackened finished heat sinks is minor, confirming that radiation plays little part in the removal of heat. Rthc-a was found to decrease significantly with increasing air velocity in the range 0.5 to 5m/s. Under natural convection the addition of heatsinks did slightly reduce Rthc-a but the decrease is unlikely to be sufficient to justify the additional cost of the heatsinks.

In addition, it has been shown that when employing heatsinking in an application to enable greater power dissipation for the same rise in junction temperature the reliability of the DirectFET product is not compromised. Power cycling results with heatsinks applied show that the MOSFETs experienced lower shifts in RDS (ON) with temperature than those devices to which no heat sinking was applied.

By applying Arrhenius and Coffin Manson equations to a simple application example it is possible to show that heat sinks similar in size and performance to those tested can give substantial increases in the MTTF for a device and further increased heatsinking would result in even larger improvements.

Heatsinking should therefore be considered where increases in power density or system reliability are required. Whilst heatsinking of through hole packages has been possible for many years, DirectFET technology has enabled, for the first time, surface mount power MOSFETs to employ a heatsink and take advantage of the gains described in this paper.

Hide comments


  • Allowed HTML tags: <em> <strong> <blockquote> <br> <p>

Plain text

  • No HTML tags allowed.
  • Web page addresses and e-mail addresses turn into links automatically.
  • Lines and paragraphs break automatically.