Extending the operating temperature of DC-DC converter modules by understanding their components
As power density increases, the internal operating temperature of key DC-DC converter components potentially becomes a major barrier to extended temperature operation. At the same time, users are increasingly keen to dispense with cooling fans that require regular maintenance. Understanding the mechanisms behind component derating allows judicious selection that ensures reliable DC-DC converter operation in high ambient temperature environments such as telecoms racks and server farms.
In recent years, with geometric growth in logic density, the general trend for integrated circuits has been towards ever-lower operating voltages that speed switching while limiting on-chip power dissipation to acceptable levels. While much peripheral logic operates at 3.3 V – itself a step down from the traditional 5 V level that many industrial users still prefer – many processor and FPGA cores operate at much lower levels, with 1.8 V being common and lower levels becoming more usual. The combination of denser logic, faster switching speeds, and lower operating voltages creates the need for DC-DC converters that handle increasing load current levels, in turn implying more heat generation from resistive and semiconductor volt-drops.
In distributed-power architectures, the converter must also accommodate increasingly large step-down ratios that can pose challenges to conversion efficiency. In large systems, designers are adopting system-level modelling techniques to establish the optimal physical points for down-converting 24 VDC and 48 VDC distribution rails to the target load voltages and to help manage thermal loads. While variable-speed fans are an extremely efficient cooling tool, avoiding their use cuts noise and simplifies equipment maintenance schedules that range from cleaning filters to replacing units before the onset of bearing failure. In many instances, specifying a DC-DC converter that can operate reliably and with a long useful life in elevated temperatures eases thermal management concerns and may dispense with the need for fans.
Appreciate component derating characteristics
The question then becomes, how can an equipment designer confidently specify a DC-DC converter for extended temperature operation, knowing that the underlying converter design issues have been addressed? Most engineers will know a rule-of-thumb that states that for each 10°C decrease in operating temperature from a component’s maximum rating, there is an approximate halving of its failure rate during its useful life. The origins of this guidance lie within the Arrhenius equation that classically quantifies chemical reaction times at varying temperatures. This model has been empirically shown to apply equally well to the diffusion and migration processes that apply to various classes of electronic components, providing a solid basis for predicting mean-time-to-failure due to temperature rises.
Many components require derating above about 75°C, but the key to design success is to appreciate the mechanisms that affect different component classes and to specify parts according to the environmental stresses that they will endure. For instance – and possibly the best example for visualising chemistry under electrical stress – the lifetime of electrolytic capacitors shows a clear correlation between operating temperature, electrical stress, and the rate of electrolyte diffusion that this generic lifetime prediction formula demonstrates:
Here, L is the predicted lifetime in hours; Lr is the manufacturer’s rated endurance at maximum temperature Tmax, in hours; T is the operating temperature that we expect for the capacitor; Vmax is the capacitor’s rated maximum operating voltage; and V is the circuit’s operating voltage. If we assume that the designer operates a 25 VDC rated part at 70% of its maximum voltage rating, a normal commercial-grade component that is rated for 2,000 hours at 85°C might be expected to have a lifetime of around 50,000 hours at 50°C; simply substituting a 105°C rated part would extend this period to almost 80,000 hours.
This baseline model is far from perfect, but it serves to illustrate the influence that judicious component selection can have upon service lifetime. In practice – and because wear out of aluminium electrolytics is a prime cause of premature power supply failure – Murata Power Solutions for example, avoids their use, preferring multilayer ceramic parts wherever possible.
While the prime cause of ceramic capacitor failure is mishandling, it is again essential to understand the effects that temperature and voltage stress create over time. These effects depend upon the dielectric material and are far more pronounced as capacitance values approach the technology’s limits although the parts lose effective capacitance rather than fail. For instance, the X7R material that is often the dielectric of choice has a basic capacitance tolerance of ±15% from -55 to +125°C while the Y5V dielectric can drop its capacitance by more than 80% at +85°C. Also compounding this is the effect whereby a ceramic capacitor’s effective value will decrease significantly under dc bias due to a phenomenon called “polarisation reversal” which is an intrinsic characteristic of the underlying BaTiO3 ceramic material.
Similar component-specific considerations apply to power inductors whose basic performance depends upon the core material they embody, with different materials exhibiting varying degrees of losses at different temperatures depending on circuit conditions. Happily, power inductors rarely fail unless they are operated way out of their specification, such as a gross overload that should never occur given appropriate circuit design. Engineers who wish to explore the effects of ac and dc bias levels, frequency, and temperature on a wide range of capacitors and inductors can take advantage of online simulation tools such as Murata’s SimSurfing suite that runs in a browser. The results may be surprising—for instance, the dc bias pane in Figure 1 shows that a 22 uF/25 VDC X7R part has an effective capacitance of just 7.75 µF when subjected to 15 VDC bias. It is worth remembering that internal temperature rises occur in capacitors that handle ripple currents, as the Temp Rise pane in Figure 1 shows.
Figure 1. The SimSurfing online simulation tool reveals the behaviour of ceramic capacitors and inductors under varying environmental conditions. Characteristics of a 22 uF 25 V capacitor are shown here.
The opto-isolators that often appear in feedback circuits can exhibit issues with variations in current transfer ratio with time that continual exposure to high temperatures exacerbates, reducing useful system lifetime as the opto changes, potentially causing instability and premature converter failure.
Substituting MOSFETs for diodes in synchronous-rectifier configurations can help mitigate issues with Schottky rectifiers, as well as improving efficiency. Nevertheless, there are circumstances where using a Schottky is difficult to avoid – such as the freewheeling diode across the synchronous switch of a buck converter. With Schottky diodes and opto-isolators now available that withstand junction temperatures of 150°C, careful component selection and design that avoids hot-spots again allows circuits to operate reliably at high temperatures. Yet key restrictions remain that limit high-temperature operation, such as UL listing requirements dictating a maximum of 130°C for typical circuit boards.
User selection criteria
From the DC-DC converter designer’s viewpoint, understanding the characteristics of each component is essential for reliable long-term circuit operation – and especially when extended temperature operation is necessary. It is then critical to design for an efficiency “sweet spot” within the temperature band that users wish to apply to the product.
As a result, DC-DC converter users wishing to specify modules for extended temperature operation must look beyond the banner electrical specifications that many manufacturers publish for 25°C and the typically blanket statement regarding the part’s operating temperature range. To know the actual temperature stress on a converter, users must know where to measure the converter’s temperature, which is often stated at a reference point on the device’s surface. Such measurements must be made within a representative operating environment – that is, with air temperature and flow that reflects the target application. When air flow and its temperature is not specified, only ‘operating temperature’, the user should ensure that they understand where that temperature is to be measured. It could be on the body of the converter, in the air local to the converter or even in an ideal, isolated ‘stirred air’ environment.