For the PDF version of this article, click here.
The bulleted items on the front page of a dc-dc power module's data sheet often highlight electrical performance that the product cannot actually deliver in system. This presents a challenge to system designers, who must compare the electrical/thermal performance of different manufacturers' modules. System designers must ensure the dc-dc power modules chosen for their end equipment offer the requisite electrical/thermal performance across the application's full temperature range. Estimating the minimum and maximum load currents a module can supply in an actual system environment can be the most important factor for determining the cost and reliability of a power supply. This helps designers choose a module that is capable of the required output current at the most economical cost.
The electrical/thermal performance of a power module is characterized by its thermal-derating curves. The curves are the best and most commonly used metric for judging the overall performance of a module. Power-module manufacturers conduct extensive thermal testing to generate the curves, which are published in data sheets. The thermal-derating curves in Fig. 1 show the maximum current a module can deliver under various airflow velocities and ambient temperatures. This defines the device's safe operating area (SOA) — the operating condition where the maximum electrical output can be achieved without exceeding the recommended thermal design limits.
Each point on a thermal-derating curve represents a combination of output current and an environmental condition that causes the temperature of some component within the module to reach a predetermined limit. In the example above, an application may require 30-A load current. Environmental conditions include 50°C ambient temperature with air velocities as low as 1 m/sec (200 lfm). Upon consulting the module's data sheet, the SOA curves (Fig. 1) may reveal that a module with a 30-A maximum-output current rating can reliably deliver only 23 A continuously under these conditions.
Thermal-derating curves tell the designer if a chosen module will deliver the desired current at the desired ambient temperature, if additional airflow is needed, and how much margin or reserve is available in case of clogged filters or cooling-fan failure in an enclosure. Moreover, thermal data tells the designer if he must either derate (operate the module below its maximum-output rating), or supply increased amounts of cooling air or, in some instances, attach a heatsink.
In actual applications, many dc-dc power modules do not achieve the output current rating on the front of their data sheets. One reason for this is power-module manufacturers' specsmanship. Another is the fact that the power-module industry has no standard thermal-derating characterization process for isolated and nonisolated dc-dc power modules.
System designers face the challenge of selecting modules from a large number of suppliers. As a result, the dc-dc power-module business is highly competitive. An aspect of the intense competition is specsmanship, which has led power-supply manufacturers to describe product performance in increasingly creative ways.
Unfortunately, when it comes to interpretation of manufacturers' thermal-derating data, a true comparison is not so simple. The system designer should take into consideration differences in derating test details. Differences such as airflow and ambient-temperature measurement method and location, the maximum component temperature allowed, board pitch and test fixtures have significant influence on the derating curves. Because of these differences, the derating curves published by different manufacturers cannot be easily compared without first understanding their measurement method.
There is no industry standard for measuring thermal performance. Two of the traditional approaches use an air velocity measurement inside a wind tunnel. This setup replicates the typical thermal environments in most modern electronic systems with distributed power architectures. The electronic equipment in networking, telecom, wireless and advanced computer systems operates in similar environments and uses vertically mounted pc boards or circuit cards in cabinet racks.
Fig. 2 shows a typical restricted SOA test setup. The power module is mounted on a test board and is vertically oriented within the wind tunnel. An adjacent board is used to simulate a card-rack environment. The adjacent board forces air over the power module. The spacing between the two boards is typically twice the height of the module. The wind-tunnel setup uses a probe to measure airflow and ambient temperature at a single point.
Power modules are typically tested at their nominal input voltage. Thermocouples or a thermal-imaging camera is used to measure the temperature of key components, while load current is varied from no load to maximum load. Data is collected at several representative airflows, typically from 0 m/sec to 2.5 m/sec.
Smoke is sometimes used in wind tunnels to qualitatively describe airflow. Fig. 3 is a representation of a wind-tunnel test for the restricted test setup. This figure reveals how the restricted test setup reduces the spacing of smoke filaments across the power module. This indicates that airflow across the module has increased from the airflow measured in front of the module. The parallel-facing pc board can increase airflow from 1 m/sec to 2 m/sec. Manufacturers that use this method claim it simulates a card-rack environment.
Fig. 4 shows an unrestricted SOA test setup, in which the power module is soldered to a test board within the wind tunnel. The setup has no parallel-facing pc board. The unrestricted test setup allows air to move over the module without restricting airflow. This does not decrease the cross-sectional flow area (which would increase the air velocity) as in the restricted test setup.
As shown in Fig. 5, there is relatively constant spacing of the smoke filaments in front of and across the surface of the module in the unrestricted test setup. This indicates that airflow across the module is the same as airflow measured in front of the module. In the restricted test setup, airflow across the module is much greater, which will result in a more aggressive SOA curve. (The module will deliver more current at a given airflow.)
Where the temperature measurements are taken is critical to the accuracy of the SOA curves. Some manufacturers recommend measuring temperature at a spot on the pc board. Rarely is this the hottest point in the circuit. For accuracy, the hottest components (typically the FET, control IC or magnetics) should be measured directly. FET temperatures must be monitored at the component's case or tab. Most manufacturers use automatic measurement processes to determine thermal performance. This is done by using a thermocouple on all of the power components, such as FETs, magnetics or a thermal camera that can monitor many components under program control.
Thermocouples can affect the measurement of low-mass components. Because of its metal construction, a thermocouple transfers generated heat away from the part it touches, making it difficult to obtain an exact thermal profile. Thermocouples use a single point to measure temperature.
Since heat patterns are difficult to predict, it is not always possible to know where to attach the thermocouples necessary to make measurements. For this reason, power-supply manufacturers attach thermocouples at multiple points. Wires connecting a thermocouple to various points on a power module can block airflow across the part, causing it to run at higher temperatures.
Many manufacturers now use thermal (infrared) imaging to help design and characterize their products. The thermal-imaging camera provides an alternative to thermocouples for measuring the temperature of key components. Thermal imaging uses multiple points for measuring thermal performance. It can be used with either the restricted or unrestricted test setup. As shown in Fig. 4, thermal images of a power module are taken through a window on the side of the wind tunnel.
Thermal imaging is often used where the power components are visible so that the surface temperature of the individual components can be measured. The images provide a good overall thermal profile of the module and can identify layout problems or overstressed components. What's more, thermal images allow power-supply manufacturers to evaluate the effectiveness of cooling and “shadowing” from adjacent heatsinks and components.
The measurement of a component's surface temperature provides a direct indication of its internal core temperature. Of significance is the junction temperature of the semiconductors and the winding temperature of the magnetic parts. By varying the temperature limits placed on these components, the module's derating curves — and its output rating at a specific ambient temperature and airflow — can be manipulated.
Some manufacturers push their module's rating by setting higher-than-normal internal component temperature limits. This contributes to an improved thermal rating. As an example, one manufacturer may operate FETs at junction temperatures close to the component's absolute maximum rating, while another may limit it to a lower, more conservative value. These opposing design conditions can have a major effect on overall power-module performance and reliability. For instance, if the operating temperature of a FET is increased from 115°C to 125°C and all other operating conditions remain the same, the module's reliability changes from a MTBF rating of 929,368 hours (1076 FIT) to 822,368 hours (1216 FIT).
Manufacturers use these higher ratings to claim superior thermal performance on their data sheets. These performance claims and the SOA graphs on the data sheet's inside pages lead designers to believe they can reliably operate a particular module in their systems at higher temperatures. Designers don't realize that the life of the power module will be reduced if it is used consistently in these operating conditions.
There is no right or wrong way to measure thermal performance. Each approach has unique advantages. SOA curves obtained from restricted test setups are valid only in an environment similar to the test setup. SOA curves obtained from unrestricted test setups can be used in a wider variety of environments. Because many applications do not use parallel boards with restricted airflow, the unrestricted test setup yields the most conservative approach.
In addition to the SOA test setup, there are several other factors that can affect test results. Is the airflow measured by anemometer or by volumetric calculation? Hot-wire anemometers used to measure airflow directly in front of the module ensure the highest airflow accuracy. Is the airflow type turbulent or laminar? Laminar airflow is the more conservative approach.
Some of today's dc-dc power modules are available in both horizontal- and vertical-package styles. Some of the orientations yield better thermal performance, which is typically highlighted in the module's data sheet. The designer must question the performance at other orientations and whether or not the derating curves are based on best-case or worst-case orientation.
Thermal Test Results
Even though most thermal performance is calculated by using data from thermal-imaging cameras, the actual test setup and method of measurement will have a significant impact on the results.
Fig. 6. shows two sets of thermal-derating curves for an isolated quarter-brick module rated to deliver a 3.3-V output at 30 A. The unrestricted thermal-measurement method was used to generate the thermal-derating curves shown on the left in Fig. 6. The restricted measurement method was used for the curves shown on the right in Fig. 6. The maximum component temperature, mounting orientation and airflow direction were the same in both tests.
At 70°C and 1 m/sec airflow, the derating curve derived in an unrestricted setup indicates that the module should be operated at a maximum of 18 A, as shown on the left in Fig. 6. When the same module is measured in a restricted airflow setup, the derating curves indicate it can be operated up to a maximum current of 23 A, as shown on the right in Fig. 6. If the system designer's product configuration is not identical to the restricted setup, there is significant risk that the module's internal components will run at temperatures much higher than the manufacturer's recommendation. This could lead to future reliability problems.
System designers often find that the output current rating on a data sheet's front pages do not match the actual output current shown on the thermal derating graph. This situation can turn product comparisons into quite a challenge. The bulleted items on the front page often fail to mention the test conditions under which the derating curves were measured. This is why, before comparing thermal performance, the designer must look beyond the data sheet's front page, searching its inside pages. In many instances, the actual output current that a power module can deliver is usually less than what is claimed on the front of the manufacturer's data sheet. This is often due to differences in the test setups and operating conditions.
To understand a module's thermal performance, the system designer must determine if temperature measurements were taken using a thermal-imaging camera or thermocouple. The system designer also must understand if the temperature was measured at a single point on the pc board or, for greater accuracy, directly at multiple components such as the FET, control IC or magnetics.
Another consideration is the thermal test setup. Some manufacturers use the unrestricted setup, while others use the restricted setup, which results in a more aggressive SOA curve. Lastly, system designers must understand if a manufacturer has allowed internal component temperature to approach or reach maximum limits when evaluating thermal performance.
To eliminate confusion in the dc-dc power-module selection process, system designers must beware of creative marketing. Be sure to compare thermal performance by carefully scrutinizing thermal data and derating curves or, better yet, by evaluating the module's thermal performance in the actual application.
Strassberg, D. “Tiny Titans: Choose ‘Em And Use ‘Em With Care,” EDN, p. 41, May 2, 2002.
“Discover a New World without Baseplates,” SynQor Inc., p. 7, 2001.
“Thermal Derating Curves for Logic-Products Packages,” Texas Instruments Inc., p. 4, 1999.
“Reliability Prediction of Electronic Equipment,” Military Handbook, MIL-HDBK-217F, Dec. 2, 1991.
Narveson, B., and Jones, G. “Understanding Thermal Performance of DC-DC Power Modules,” Applied Power Electronics Conference (APEC) 2006.