Though advances will continue, you probably won’t see the same degree of efficiency improvements that characterized the previous two decades.
Prof. Dr. Werner Wöfle, Traco Power Solutions
Over the past 40 years, the switching power supply has evolved from a 60-Hz transformer to a sophisticated high-frequency device. The evolution was primarily driven by the development of ever-faster semiconductor switches.Switch technology advanced from relatively slow-switching thyristors to bipolar transistors, first with a small and later with a high blocking voltage. Improving bipolar technology allowed the realization of switching frequencies up to the 60-kHz range. Field-effect transistors were technically sophisticated in the 1980s and had a decent price/performance ratio. Their availability made it possible to raise switching frequencies again, this time to several hundred kilohertz.
Of course, the continual rise in power supply switching frequency does not serve as an end in itself; the physical properties of magnetism lead to use of ever-smaller power transformation devices and correspondingly smaller switching power supply designs. On the one hand, power transformation devices must be insulated from the dangerous mains voltage; on the other hand, they adapt the output voltage to a level safe for the consumer. But a higher switching frequency leads to higher switching losses which may force the use of additional cooling methods, thus working against smaller designs.
For this reason, today’s switched supplies use more complex switching topologies where the switching elements are switched on and off either in a voltage- or current-free state. Where the switching technology makes zero-voltage or zero-current switching impossible, extremely fast-switching gallium-arsenide (GaAs) or silicon-carbide (SiC) switching elements are also used. Compared to MOSFET technology, these components are still rather expensive; however, their prices are slowly trending downward and are thus increasingly suitable for industrial applications.
The shrinking size of power transformation devices
In the 1970s, switching power supplies with large and heavy 60-Hz transformers were still in use. A 250-W power supply weighed over 20 lb and was larger than a shoe box. The power transformation device continues to be a significant component in every power supply, greatly affecting switching overall supply size. The transferable energy in a power transformation device primarily depends on the cooling, the transformer core volume, the winding and rate-of-change of the magnetic field, and the transfer frequency. Thus, the transfer frequency must rise to increase a transformer’s transferable power or reduce its size while maintaining the same power level.
Disregarding the insulation requirements, a power transformation device can transfer power, in first approximation, at a level inversely proportional to the transfer frequency’s square root. This is the reason modern switching supplies first rectify the 60-Hz mains voltage and then generate a higher frequency alternating voltage by means of electronic switches. For example, if the frequency of this alternating voltage is 60 kHz, the required power transformation device is about 30 times smaller than at 60 Hz, which naturally also affects the volume and weight of the switching supply. At frequencies of 600 kHz, the size of a power transformation device can further drop to a third. This means any additional rise in frequency can only lead to a moderate reduction in the size of the power transformation devices.
Capacitors are used in switching power supplies to buffer voltages during breaks in the current flow, to smooth the residual ripple of currents and voltages, or to filter out high-frequency interference. The size of these capacitors can also drop linearly with frequency which, in turn, leads to smaller switching supplies. However, the buffer capacitors at a switching power supply’s input generally won’t shrink because they operate at 120 Hz. This is also the reason why the size of switching power supplies cannot be reduced arbitrarily, unless buffer times are omitted.
One factor frequently ignored is that the maximum transferable power of a switching power supply frequently depends on the maximum permissible operating temperatures and the component cooling. Manufacturers often make ambitious statements that can lead to problems for the user if sufficient cooling is unavailable. So when selecting a supply, it is best to consult the efficiency specs or the power loss indicated by the manufacturer.
One manufacturer may list a significantly higher nominal power than another, simply by allowing higher operating temperatures. However, the higher operating temperatures may reduce operating reliability. In general, it can be said that switching power supplies today are optimized to such a degree that any further volume reduction can only come by adding cooling via heatsinks or forced air. But additional cooling adds costs, and the use of forced air, in particular, is problematic because of noise and potential contamination. So the more promising approach for reducing supply size is to make the supply more efficient at converting power.
In the 1980s — the early days of switching power supplies — industrial switching supplies were about 70% efficient. By the 1990s they had improved to more than 80%. In the past ten years, switching power supplies in the 90% range have become the technical standard.
Today, power transformation in switching power supplies mainly comes via resonantly switched FETs. These semiconductors are inexpensive and have a low rate of loss because they switch on or off at the point of zero voltage or zero current. They are well-suited for power supplies handling about 800 W.
Boost converter topologies are regularly used on the input side of switching power supplies over 100 W, providing a significantly higher power factor (over 95%) than is possible when only using a rectifier. This topology requires an additional inductance. To keep it as small as possible, supply designers may employ fast-switching semiconductor switches based on GaAs or SiC. The transit frequency and switching processes of these switching elements is about ten times higher than that of traditional silicon semiconductors. GaAs and SiC switching elements still cost more than silicon MOSFETs; however, their price is dropping.
Modern switching power supplies over 100 W usually have a two-stage design. A converter generates a pre-regulated direct current while maintaining a high power factor such that the converter input current is nearly sinusoidal. A second stage, usually designed as a resonance converter, transforms the voltage to a lower level and separates the input voltage from the output.
Switching power supplies will continue shrinking to a moderate degree, and the power density will continue to rise, albeit not to the same degree seen during the past 10 to 20 years. More than in the past, the limiting factor will be the amount of power lost as thermal energy. Continuing reductions in size will make it increasingly hard to dissipate wasted heat.
As a final bit of advice, users are well advised to consider the performance data of switching power supplies, in particular the power loss information, in light of the physical volume of the supply. Potential difficulties should always be clarified and questioned in the interest of reliable application. The phrase “Small is Beautiful” only applies when the resulting power loss during operation is correspondingly small as well! DW