The most common type of power supply today is the switching supply. These units use pulse-width modulation (PWM) to regulate output. Supplies today employ several different PWM circuit configurations. In all cases, the PWM logic signal drives the switching power transistor, and the power transistor drives the load.

The switching transistor turns on and off rapidly, producing a chopped dc voltage. The chopped dc voltage is fed to a transformer which converts the pulsating dc to high-frequency ac. This ac is then fed to a second bridge rectifier which produces the final dc output. A sensing circuit continuously monitors the output voltage, adjusting the switching duty cycle to maintain a constant voltage output.

Switching power supplies are more efficient than series-regulated ones because little power is dissipated in a switching transistor. Switching supplies are physically smaller than series-regulated types because components operating at the switching frequency (typically 20 kHz) are much smaller than those used in a nonswitching supply operating at 50 to 60 Hz. These power supplies are well suited where compactness, efficiency, and moderately accurate regulation is required. But switching-type power supplies are electrically and sometimes audibly noisy. Thus, they are unsuitable for powering circuits that are sensitive to electrical noise unless those circuits are filtered and shielded. Finally, switching power supplies are generally more costly than other power supplies.

Switching frequencies are continually rising. The advantages of higher frequencies include reduced component size, lower ripple voltage, higher power per unit volume, and quiet operation. While 20 to 30 kHz seems to be the most widely used frequency today, 100 to 500 kHz are also being used. And some PWM-type integrated circuits are capable of handling switching frequencies to 1 MHz and more.

Circuitry that produces the PWM drive signal is now available on numerous standard ICs. These chips sport a variety of features. Many of the features protect the chip and power supply from start-up current surges, overvoltage, and short-circuit faults. Others let the supply designer build in more flexibility, as in remote on/off control, remote error sensing, and proportional load-current sharing. Custom ICs and microprocessors are now being built into more complex power supplies, especially those that interface with host computers through standard buses.

Supply makers say that options quickly become standard features as users demand better products. And as systems become more complex, standard features become basic necessities. Overvoltage protection, adjustable voltage, and active soft-start are among the most common capabilities found on today's power supplies which were once options. Additional options that are candidates for standardization include specialized EMI filters, power fail and power valid indicators, and current balance circuits for proportional load sharing.

Switching is typically implemented in one of three ways. The first is a flyback circuit configuration. It is suitable up to 100 W and is the most economical of the three types because it contains the fewest number of parts. A second is called a forward converter. It is most cost effective between 80 and 200 W. The third is a more complex type that comes either as a center-tapped push-pull circuit or a half-bridge push-pull circuit. These two are widely used in the 150 to 600-W range.

An off-line switcher rectifies the incoming ac main voltage and is considered a dc-to-dc converter. Rectified and filtered 115 Vac produces about 145 Vdc; therefore, some converter designs work from 145 Vdc input as well as 115 Vac. The input rectifiers become steering diodes that allow either input lead to be positive or negative. And off-line models with a 115/230-Vac selectable input handle up to 290-Vdc input.

Switchers that operate directly off the mains require an input current surge-limiting circuit. Since there is no transformer impedance to help limit current charging the filter capacitors, peak currents can be high enough to destroy the rectifiers.

In its basic configuration, the flyback switcher contains one transformer, one pulse-width-modulator circuit, one power transistor, and one output diode. The transformer steps down voltage, provides line isolation, and acts as an inductor. When the power transistor switches on, current in the primary stores energy in the transformer core. The polarity is such that the diodes do not conduct. When the transistor switches off, the voltage polarity reverses and flies back, passing current through the output diode to the output capacitor and load. The amount of energy stored in the core is varied by the on time of the PWM and transistor.

As power increases over 100 W, the flyback transformer size increases rapidly because of increased current requirements. Also, the saw-tooth waveform produced by the flyback circuit needs twice the peak current for a given power level compared to a forward converter. Over 100 W, the maximum allowable peak current for the flyback transistor occurs quickly.

A forward converter uses an additional flywheel diode and a filter choke in its output compared to the flyback circuit. Also, the transformer steps the voltage up or down and provides line isolation.

During the switching transistor on-time, current flows through the output inductor to the filter capacitor, so the inductor stores energy. When the transistor turns off, the stored energy continues to flow through the flywheel diode, causing less output ripple voltage than the flyback design. The peak current is only half that of the flyback, but the forward converter has two magnetic components which increase size and cost.

As power demands further increase, push-pull circuits are widely used up to about 600 W. Two versions are available. One is a center-tapped push-pull circuit, and the other is a half bridge. The center-tapped circuit looks like two forward converters with alternate on periods. Both converters share a single output inductor. Depending on the pulse width, the inductor supplies current to the capacitor while both switches are open. Both push-pull circuits produce the lowest ripple voltage of all switchers.

While switch-mode power supplies have many advantages over linear types, they also have several drawbacks. Among them is the production of noise during switching which requires special care in circuit design and printed-circuit-board layout to filter. A well-designed linear power supply has an output noise level of less than 1 mVpp, compared with 10 mVpp for the same capacity switcher. Both conducted and radiated noise and switching frequency harmonics extend into the radio frequency spectrum. Designers must keep these noise levels within specifications that are set and controlled by regulatory agencies around the world.

Another drawback concerns limited response to dynamic load changes. Unlike linear supplies with very low output impedance, load voltage correction in a switcher takes place only after a full cycle of the oscillator. In addition, the control-loop time constant is set to integrate the output voltage change over several cycles to prevent continuous ringing.

Typically, line and load regulation for linears are about ten times better than switchers for the same voltage and current ratings. But this quality comes at the expense of power dissipation. For example, a 2:1 ratio in efficiency of a switcher over a linear can account for as much as a 6:1 advantage in power dissipation at an 800-W level.