Power Converter Topologies
How to Choose the Right One for Your Design
by Dennis L. Feucht
Innovatia Laboratories

With so many different power converter circuits to choose from for a given design application, how does a non-power engineer determine (and quickly) which one is best? This article presents an overview of the decision process.

The first consideration is whether the output needs to be isolated from the input. If not, then consider first the three basic first-order converter configurations: buck (or common-passive, CP), boost (common-active, CA), or buck-boost (common-inductor, CI.) These three topologies are supported extensively by commercial power ICs. For output voltage, V_o, less than source voltage, V_g, use the CP (buck) converter. For V_o > V_g, use the CA (boost) configuration. The CI is versatile in that it has a D/(1 - D) transfer function (where D is duty ratio) and can have an output voltage that is either less or greater than V_g.

For isolation, the isolated topologies corresponding to the basic configurations are the forward (CP) and flyback (CI) converters. Using a transformer of arbitrary turns ratio, the forward converter can supply V_o that is less or greater than V_g. So can the flyback converter, which uses the magnetic component as a coupled inductor instead of a transformer. The isolated CA does not exist because it transfers a dc component of current between input and output while the active switch (on the primary side) is off.

Coupled Inductors Versus Transformers

Flyback converters use magnetic devices with at least two windings in a different way than do forward converters. In a flyback, the primary winding supplies energy to the magnetic field during the on-time of the switch. Then during off-time, the field delivers energy to the secondary circuit. All of the transferred energy is stored in the magnetic circuit (the core, mainly), as an inductor. But instead of having only one winding, the second winding allows multiple electrical access to the magnetic circuit.

A transformer transfers energy directly from the primary to the secondary winding without storing it in the magnetic circuit. To do this current flows simultaneously in both windings, and their fluxes are of opposite polarity and cancel. In other words, essentially no energy is stored in the magnetic circuit. For forward converters, the primary and secondary circuits conduct at the same time.

Another way to describe the difference is that the mutual coupling to the magnetic circuit in a coupled inductor is positive in that the fluxes generated by the windings are of the same polarity and add, causing the flux to have a non-negligible value, and magnetic energy to be stored in the core. For a transformer, coupling is negative in that the fluxes oppose and cancel, leaving a negligible net flux in the magnetic circuit.

Yet another way to distinguish - from an electric circuit standpoint - is that transformers have no net dc (unipolar) currents in their windings while inductors usually have a large dc component. Transformers are driven bipolar while inductors have unipolar currents. All inductive devices require bipolar voltage drive for flux balance, just as capacitors require bipolar current drive for charge balance.

How magnetic devices are designed depends on whether they are inductors with large dc and small ac current components, or transformers, with small dc and large ac currents. High-frequency switching losses from ac currents require low-loss cores at switching frequencies, and these are generally ferrite cores. But if the ac component is small powdered-iron cores (or variations on them) are more optimal. The ac losses in them are greater (requiring small ac currents), but their flux saturation values are 3 to five times higher than ferrite, and can maintain their inductance values at higher currents.

Flyback coupled-inductors are, nevertheless, usually designed with ferrite cores because their ac currents are large, though they are unipolar. (This is not necessary true for continuous-current mode, CCM, flybacks however.)

Cuk-derived converters use the magnetic component at the boundary between inductor and transformer. The dc component of currents in the windings aid, like an inductor. But the ac components oppose, like a transformer. What is needed is a general word for magnetic devices with multiple turns that includes both transformers and coupled inductors. How about transductor?

The second consideration is the amount of power involved. If you need a few mA at under 100 V, non-isolated, then switched-capacitor converters are probably optimal. They use no magnetic components, and are easy to design, whether you use an IC designed for the purpose or a low-cost 555 timer to drive a capacitor-diode network. V_o can exceed V_g by using voltage multiplier circuits. For instance a 12 V output at 5 mA can be derived from a 5 V source using a voltage tripler. (If you need somewhat more power, a boost converter, using one small inductor and a converter IC is optimal.)

For isolation, between 0.25 W and 50 W, the discontinuous-current-mode (DCM) flyback converter is optimal only because it is simple; it uses only one transductor. Thus it is low-cost. It has no performance advantages over other converters, and has the notable disadvantage of high current shape factors (peak/avg.). The high peak input and output currents produce noise while the desired current (the average, dc component) is relatively low. This is caused by the alternating input-output conduction, causing the ac/dc ratio to be large. For converters (not inverters), dc, not ac, is desired. DCM flybacks have the advantage of being relatively easy to stabilize in a feedback control loop.

The CCM flyback is a different story. Not all input energy each switching cycle is delivered to the output. Instead, the core always has a dc flux value that is supplemented by the primary each cycle. When the primary switch turns on the initial current is not zero but some value relatively close to the peak value at turn-off. During off-time, the current does not decay to zero but only by some Delta_is that is a fraction of the peak current. CCM flyback converters can be the optimal choice for up to 100 W or more. Their main drawback is that they are harder to stabilize in a feedback loop. (CCM introduces a RHP zero.) However, they are not so hard that they should be rejected as a design alternative; for this power range, they are generally optimal.

For conversion of power over 100 W, isolated, some variation on the CP configuration is optimal because use of the magnetic range (over ± saturation) doubles due to bipolar currents in the transformer. A separate output inductor is required to store energy, and the two magnetic components make this an inherently more expensive alternative. The primary circuit basically chops the voltage to drive the transformer, and it is then rectified by the secondary circuit. The secondary inductor sets the current for both sides of the transformer, and if current ripple is kept small, the ac magnetic losses are small.

If you go this route don't underestimate the design problems in the secondary circuit with the rectifier. Though diode switches are passive, they are demanding. Not only are their switching characteristics important, but transformer-related resonance can also overvoltage them.

For 500 W to 2 kW, the transformer driver is more optimally a bridge circuit than a single switch. A half-bridge uses two series switches and two series capacitors for the other branch. The switches are rated at the full source voltage and twice the current of full-bridge switches. The primary winding has half the source voltage applied to it. Full-bridge switches carry half the current, but are double in number.

High-side switches require high-side drivers, which add cost. But that cost is in silicon, not ferrite and copper, and is often a better tradeoff in overall cost than a larger, center-tapped transformer and ground-based switches.

Another design consideration is power-line effects of noise and impedance. A power factor corrector, which is usually a boost converter, minimizes ripple and presents a resistive input to the line. This adds cost and inefficiency as a second converter stage. It is better to start with a converter with near-constant input current characteristics, such as a Cuk converter (described below), and control it to track the line for resistive input. The inner current loop is made to track the phase of the varying line voltage while the amplitude of the current is controlled by an outer voltage loop that regulates output voltage. Some additional control circuitry is required to build this variable-gain current source, and can be a challenge for the non-power designer.

Multiphase converters are another category of control, in that they introduce no novel converter circuit but use multiple circuits, phased to transfer energy at different times. They reduce the ac component in amplitude by pushing it to a higher frequency. The phasing of the interleaved switchers allow near-constant input and output currents, but at the expense of multiple converter circuits and a more complicated controller, usually an IC.

Cuk-Derived Configurations

Challenging the three first-order configurations, in all their variations, are the Cuk-derived topologies: Cuk, SEPIC, zeta, and inverse zeta. They too have isolated versions. These converters are definitely harder to understand (and thus design), but they have better performance features than the basic configurations. The Cuk topology is essentially ideal, though the isolated version requires both a coupled inductor and a transformer. Its input and output ac currents can be reduced to very small values, thereby eliminating noise problems in both input and output circuits. The SEPIC configuration of the basic four-terminal Cuk inductive switch cell has dc (continuous) current input but ac output. The zeta is the reverse. The non-isolated versions are dc-isolated in that no dc paths exist from input to output. This is not sufficient for meeting line-operated safety requirements, but is otherwise a benefit. Shorting the output does not short the input source, except transiently.

The Cuk category of converters also have few parts and multiple outputs are possible. But even for the non-isolated versions the required coupling capacitor between input and output can be a significant cost item, making these converters somewhat more expensive than a CCM flyback.

Multiple Outputs

The isolated converter circuits can easily accommodate multiple output windings, extending their usefulness. The main problem with the multi-secondary approach is that not all the outputs can be regulated. The feedback to the controller will come from either the most demanding output or else a weighted sum of some or all outputs. Also, a transient load on one output can couple into the others, and some knowledge of secondary decoupling techniques is required, though they are imperfect. For more demanding multi-output applications, separate post-regulators are the most common solution. In a big system, a semi-regulated power distribution bus sources individual post-regulators at the loads. (These post-regulators are sometimes called "voltage regulation modules," or VRMs.) But for the smaller, and more cost-sensitive the system, this two-stage approach becomes less optimal.

Closure

The overall design situation is that the more sophisticated converter types provide greater optimality in operation, but are more demanding of the designer. If you are not a power-electronics expert, and the converter is not a critical subsystem for the performance of the application, then you can use a simpler-to-design topology without incurring system-level sub-optimality. But if the distinctive features of the system require optimal converter performance, then find a power electroniker to design it for you.

The above claims of optimality for various converter types can be found to have exceptions. But if you are in a hurry and do not intend to master power electronics before designing a converter as part of your system, the above maxims should guide you reasonably well for most applications.

In summary, for most applications, for up to about 100 W, go with the CCM flyback. Over 100 W, line-powered and isolated, use either the isolated Cuk topology or a half-bridge, controlled as a PFC.

Contact the author

• Dennis Feucht