Thursday, February 20, 2020

How do I evaluate a convection cooled power supply’s performance?

I discussed convection cooling in a previous article, giving guidance on how to position (mount) a power supply in a system.   It did not cover how to evaluate a product’s performance though, a subject worthy of mention as testing methods across the industry differ significantly.
First, a quick recap on convection cooling as it is widely assumed a convection cooled power supply does not need any airflow to operate.  One definition of convection is “The transfer of heat by the circulation or movement of the heated parts of a liquid or gas”.  In our case – the circulation or movement of hot air.

An open frame power supply is typically mounted on a flat surface upon standoffs.  Figure 1 shows how the air behaves.  As the hot air rises, cooler air is drawn in from the sides.  Although the airspeed is quite low, just 0.3m/s, it is sufficient to reduce internal temperatures.



Figure 1: Natural airflow around a convection cooled power supply

Before making the choice of vendor and power supply it is advisable to download not only the datasheet, but also the manufacturer’s evaluation report, reliability data, application information and safety files for the “Conditions of Acceptability”.  If it is not available from the website request it.  This “homework” is paramount to ensuring you are not going to have program delays due to non-performing power supplies or worse, excessive field failures.

The report listing the thermal measurements is a key section and may raise some suspicions.

Most power supplies are capable of operating over a wide range input.  From a thermal aspect 115Vac is tougher than 230Vac, due to the higher currents (I2 x R=losses) in the input filtering and rectification circuitry.  If the report only states results measured at 230V, this may be an issue.

Convection cooled power supplies often have a higher forced air cooling rating stated on the datasheet.  In this case the report should confirm if the test results were recorded with convection cooling.  If that statement has not been made it should be noted for your own follow up testing.  Check that the output loading is at 100%.

The mounting orientation affects the thermal performance of the power supply.  This too should be identified in the report.

Where was the ambient temperature measured?  Some reports show the temperature recorded above the power supply.  This falsely overstates the true ambient temperature. Ambient should be recorded on both sides of the power supply, or underneath for a vertically mounted or DIN rail mounted supply.

At what ambient temperature were the measurements recorded at and how long was the power supply running before the tests took place?

A convection cooled power supply will take between two and three hours for the temperatures to stabilize.  The larger, hotter components like the isolation transformer and power semiconductor heatsinks have a greater thermal mass than a small capacitor.  If the report states that the power supply was run for one hour at 25oC and for one hour at 50oC then there should be at approximately a 20 oC delta between the two sets of results.  Power semiconductors do operate more efficiently at higher case temperatures, and the natural convection airflow will be higher, so there will be some discrepancies.  This is why extrapolating results can give errors due to non-linearity.

Having received your sample power supplies, you can start testing.  If a thermal chamber is not readily available, it is advisable to bench test with the unit covered with a large enclosure.  This avoids external airflow from air-conditioning systems affecting the results.  Even in a thermal chamber, there may be an internal fan to circulate air.  TDK-Lambda recommends placing the power supply in a sealed enclosure to shield that air from interfering with the results.  Remember to test at all the input voltages your system will be operating from.

Ideally the temperature of each electrolytic capacitor should be measured using thermocouples attached to the exposed metal case at the top of the component.  This data can be used to determine each capacitor’s life.  TDK-Lambda can assist, supplying a sample with thermocouples already attached*.

In the end system it is very important to ensure that there is adequate space for the air to be drawn in from the sides and allowed to exit above the power supply.  A distance of 50mm is considered safe, less than this will cause interference with the natural convection airflow.  Power supply thermal testing has to be repeated again in the end system to verify the naturally circulating airflow has not been restricted.

The cooler a power supply operates, the longer it will last.  Care taken during the early stages of product development can avoid last minute launch delays.

*A follow-up article will be available shortly on assessing capacitor life.

Power Guy

Popular Posts