Rather be cool than dead

"RATHER BE COOL THAN DEAD." in white capital letters on a blue background
Temperature is the number one enemy for the life of power supplies. As a rule of thumb, every 10°C increase in temperature decreases the lifetime of a power supply by half. Similarly, every 10°C decrease in temperature doubles their lifetime. Hence, the title of this piece “Rather be cool than dead”. [1]

The reason why temperature plays such an important role in the lifetime of power supplies can be found in the Arrhenius equation. The Swedish chemist, Svante Arrhenius, developed this equation back in 1889 to calculate the relationship between the rate of chemical reactions and temperature. As many electronic component failures are chemical in origin (corrosion, leakage currents, diffusion, migration effects, etc.), high ambient or operating temperature is usually the most aggressive stress factor leading to early failure of power supplies.

Arrhenius equation
Fig. 1: Arrhenius Equation

All power supplies generate heat. This is caused by their internal power dissipation. This power dissipation is dependent on efficiency and load. The higher the efficiency, the less the internal power dissipation, and similarly, the lower the output power, the lower the internal heat dissipation. Consequently, the amount of power dissipated inside a power supply can be found from the following simple equation:

Fig. 2: Internal power dissipation, Pdiss, is dependent on the efficiency, η and the output power (load)

If the internal heat cannot be dissipated (lost to the surroundings) then thermal runaway can occur, leading to the destruction of the power supply. To avoid power supply overheating, the output power must be derated (or reduced) at high ambient temperature. For example, the following derating diagram shows that this power supply can deliver full power up to an ambient temperature of 68°C, but must be limited to 55% of full load for operation at +85°C ambient temperature:

A derating diagram is valid only for a unique operating condition, typically nominal VIN and free air convection cooling. The graph also assumes that the natural convection thermal resistance between the power supply case and the surrounding air (θ_CA) is constant, so that the internal temperature rise, Trise, is directly proportional to the power dissipation – once the maximum component temperature has been reached, further increases in ambient temperature must be balanced out by corresponding reductions in the output power.

If more output power is needed at higher ambient temperatures, then the thermal impedance needs to be reduced. With forced air cooling, more heat can be convected away from the power supply, leading to a family of derating graphs at different flow rates:
Graphic shows the percentage performance of a device as a function of the ambient temperature in degrees Celsius

Fig. 3: Calculated derating diagram and temperature rise relationship

Graph shows output power versus ambient temperature
Fig. 4: Calculated derating diagram with different air flow rates and the heat transfer relationship
The derivative Q-dot is the heat transfer rate per unit of time, h is the heat transfer coefficient, A the surface area and Trise is the temperature increase due to the internal power dissipation.

For the same temperature rise and surface area, increasing the air flow improves the heat transfer coefficient, h, leading to an increase in the heat transfer rate. In the example above, derating starts at +68°C for free air natural convection (0m/s air flow), but the same power supply could deliver full power at +85°C with 2m/s forced air flow and above +90°C with 3m/s forced air flow.

It must be mentioned that the derating diagram is a calculated result. The graph has straight lines because the calculations assume that the power dissipation remains constant over output load and input voltage (not true) and that the thermal impedance remains constant for a given air flow rate (also not true).

If the derating is measured rather than calculated, the result is a derating curve.

Graph shows the voltage as a function of temperature

Fig. 5: Example of a measured derating graph

Measuring the derating curve requires a calibrated wind tunnel and an automated control system with real-time temperature monitoring. Figure 6 shows the RECOM setup:

The airflow inside the wind tunnel is regulated and laminar down to 0.05m/s. The fully automated control system monitors the component temperatures contactlessly using an infra-red (IR) camera and adjusts the output load for real-time maximum power dissipation monitoring (Figure 7). The IR camera has an ethernet link data feed attached to the control computer so that the system can monitor the temperature of multiple components simultaneously, ensuring that none of them exceed their limits.
Laboratory setup with equipment and a computer

Fig. 6: RECOM’s in-house wind tunnel setup

Circuit diagram with hub, IR cam, DUT and computer

Fig. 7: Schematic representation of the automated derating control system and the live IR camera feed.

The air flow is increased in steps and the component temperatures are monitored. The system automatically controls the output power to keep all hot-spots below their critical temperature.

Graphic shows the relationship between output power and air flow over time
Control system interface with displays, graphics and settings

Fig 8: Stepped airflow (blue trace) and output power (red trace). Operator’s GUI.

The computer logs the data from each test and then calculates the thermal transfer coefficient (temperature rise per watt of output power in °C/W) at each airflow step to automatically plot accurate derating graphs based on the thermal coefficient measurements.

Thermal derating curve: power vs. temperature

Fig. 9: Measured derating curves based on the wind tunnel measurements

Conclusion

Derating calculations are only as reliable as the data they are derived from. Additionally, they are based on several assumptions. Consequently, real-life testing is needed to generate precise parameter values and accurate derating curves.

Automating this process guarantees repeatability and can speed up the testing significantly. For instance, in order to properly characterize a power supply, measuring the maximum temperature of multiple critical components manually at different airflow rates used to take several days. Now, accurate derating curves can be generated in a matter of hours, thanks to automated processes.

When a user installs a power supply in harsh operating conditions with high ambient temperatures, they have to understand that the lifetime of the device will be significantly reduced. Nevertheless, there may be situations where the power supply needs to be pushed to the extremes of its operating envelope, but this must be done without risking catastrophic failure such as thermal runaway or burnt-out components. However, knowing that the derating curve is measured rather than calculated renders a degree of confidence by implying that the power supply will survive such extreme events without failing in the process.

Our tests also revealed that different components reach their maximum temperature limits under different operating conditions. For example, at low air flow, the temperature of the input bulk capacitor could be the “weakest link” that limits the output power. However, at higher air flows, vortex shedding around the cylindrical shape of the capacitor increases the effective heat transfer rate and a different component (typically, the switching transistor) reaches its critical maximum temperature before the capacitor. Such real-life variabilities are difficult, if not nearly impossible, to simulate accurately using thermal flow modelling, but by using multiple hot-spot IR thermal imaging, the RECOM system can automatically selects the most critical component and uses that to control the load.

Thus, derating curves are not always similar for the same power supply under different operating conditions. For RECOM customers who need reassurance that our power supplies will not be pushed beyond their limits under extreme conditions, we can program our wind tunnel set-up to replicate those operating states and give a clear pass/fail reply.

[1] If you think the title sounds familiar, Kurt Cobain sang the line "I’d rather be dead than cool" in the track Stay Away in his album Nevermind by Nirvana. He achieved both, of course.
Applications