There are many possible topologies for a Switch-Mode Power Supply (SMPS), but the most popular for circuits supplying a load less than 150W is the flyback converter. Some estimates go as far as to suggest that up to 75% of offline power supplies use the flyback topology.
Many power-system designers are therefore faced with the challenge of developing a flyback converter circuit. To achieve the best performance, satisfy electrical specifications and stay within cost and space limits, the designer will need to implement some form of customisation in the design; and the most important custom element of a flyback converter is the transformer.
In the design engineering community, transformer design and prototyping is generally regarded as a black art. To the uninitiated, the wide range of parameters affecting transformer performance – from selection of core material and size to the arrangement of the windings around the core – can appear confusing. In fact, the process of transformer design can be worked through in an orderly way by applying a small number of important equations, combined with a certain degree of trial and error, perhaps better described as ‘experienced guesswork’.
The team of designers at the Future Electronics EMEA System Design Centre (SDC) in London, UK, has gained much practical experience from its work on the development of custom power supplies for OEM customers. The purpose of this article is to share some of this experience and reveal effective ways to optimise transformer design in flyback converter circuits.
Flyback topology: theory of operation
The flyback converter is an isolated form of buck-boost converter, as shown in Figure 1. It consists of:
• a primary-side switch, typically a MOSFET
• two inductors in the form of a primary and a secondary winding around a magnetic core, as shown in Figure 2. The windings are turned around a plastic bobbin which provides mechanical support and a set of pins for the wire connections and through-hole mounting on a PCB. In its operation the arrangement of the two inductors is more correctly called a ‘magnetically-coupled inductor’. But because of the two separate windings, it is commonly referred to by designers as a ‘flyback transformer’. Strictly this is a misnomer, but for convenience this article will refer to it this way.
• a secondary-side switch, typically a diode
• an output capacitor
Feedback for control purposes across the isolation barrier is generally implemented with an optocoupler and compensation circuitry.
When the primary switch is turned on, current is drawn through the primary winding, generating a magnetic field which is readily transferred through the low-reluctance core to a small air gap in the centre of the core, where stored magnetic energy accumulates. When the primary-side switch is turned off, the stored magnetic energy induces current to flow through the secondary winding and the output diode to the load.
The various advantages of this converter topology explain its widespread adoption:
• Isolation is readily achieved via the flyback transformer and optocoupler feedback compensation.
• Component count and cost are low.
• The turns ratio of the flyback transformer allows for a high ratio between input and output voltages, such as a 3.3V output directly from an AC mains voltage input.
• A single power stage can provide multiple output-voltage rails, of both positive and negative polarity.
• The flyback topology supports both step-up and step-down operation: it is a buck-boost topology.
But there are drawbacks to the flyback converter. The most important are:
• Voltage stress on the MOSFET and output diode are high, and widely variable from design to design.
• Relatively high noise, due to high peak currents and high voltage peaks at both switch elements during switch transitions. The flyback transformer may also contribute noise via coupling across and radiation from the windings.
How converter specification affects transformer design
Optimisation of the flyback transformer is determined by the key parameters specified by the designer, which are:
• Output power
• Switching frequency
• Peak and average current values in the primary and secondary windings (taking account of the worst case of maximum load at minimum input voltage)
• Primary inductance
• Maximum flux density
• Turns ratio
Before the designer can begin the process of designing the flyback transformer, however, the conduction mode of operation needs to be chosen: Continuous Conduction Mode (CCM), Discontinuous Conduction Mode (DCM) or Critical Conduction Mode (CRM). The process of transformer design is the same for all three conduction modes, but in any power converter the operation is different, and fundamentally so in the case of the flyback converter, because the transfer function of the converter is different in each case, affecting feedback compensation.
There is substantial literature available to guide the designer’s choice, so this article does not deal with conduction modes in detail. Practical experience at the Future Electronics SDC suggests that the choice is most often determined by:
• Size and cost pressures, in which case DCM has the advantage because of its lower inductance requirement
• Requirement for low conduction losses and high efficiency at higher output power levels, in which case CCM is preferred because peak and Root Mean Square (RMS) output current are lower for any given output-power value.
A further decision to be taken early in the development process is the choice of core material. The main parameters affecting the choice of core material are maximum flux density, reluctance and cost. For flyback transformers the magnetic material most commonly used is ferrite. This is a cheap material which suffers from low losses at switching frequencies up to around 500kHz. Ferrite cores become saturated at a relatively low flux density, typically around 0.4T. This means that, in designs using a conventional ferrite core, flux density should be kept to a value no higher than 0.3T at the peak primary-side current to avoid saturation.
The causes of losses, and how to manage them
It is a rare power-converter design project in which the engineer’s attention is not firmly focussed on power efficiency, and the minimisation of power losses. In general, loss reduction helps to reduce thermal stress and the need for cooling devices, improves system reliability, and enables the creation of a smaller, lighter and cheaper end-product.
In a flyback converter, there are many sources of loss, including MOSFET and diode conduction and switching losses, output capacitor ripple-current loss, snubber losses, and input and output filter losses. But in most cases by far the greatest proportion of total losses is attributable to the flyback transformer. There is therefore considerable benefit to be gained from efforts to reduce transformer losses.
It is helpful to start with an understanding of the various sources of loss within a flyback transformer. These are:
• Copper losses due to the DC and AC resistance of the copper wire used for the primary and secondary windings.
• Proximity losses due to the effect of closely-coupled currents within a strong magnetic field, concentrating current flow in a portion of the copper wire’s cross-section.
• Leakage inductance: magnetic field leakage results in electrical power loss. This must also be taken into account in the circuit design, since the level of leakage inductance directly affects so-called ‘snubber losses’. A basic requirement for avoiding magnetic field leakage is to locate the air gap inside the winding.
• Loss in the magnetic core material due to the switching action and the inherently hysteretic behaviour of core materials.
The amount of loss in a winding’s copper wire is influenced by:
• The current waveform, and the relative sizes of the DC and the AC components
• The overall DC and AC resistance of the windings
• Switching frequency
• Proximity loss
In particular, a high switching frequency and a relatively high AC component in the current waveform will increase resistance due to the so- called ‘skin effect’. The skin effect causes high-frequency AC components to be conducted towards the outer surface of the wire, effectively reducing the cross-sectional area of the conductor, and therefore increasing its resistance. Future Electronics’ practical evaluation of real-world transformer designs operating at switching frequencies below 100kHz has shown that the skin effect – and copper losses – can be minimised by using single-strand copper wire with a diameter of ≤0.5mm.
Proximity loss also adds to the losses in copper wiring: in essence, a conductor which carries a high-frequency current induces copper loss in an adjacent conductor by a phenomenon known as the proximity effect. This effect causes copper losses to compound with each additional layer in a multiple-layer winding.
To minimise the effect of proximity losses, therefore, the designer must keep the number of winding layers to a minimum: ideally no more than two or three for the primary and secondary windings, particularly when the current waveforms have a high proportion of AC components, which is the case in DCM operation.
Leakage inductance is a function of the number of turns squared (N2) and the winding geometry. To minimise leakage inductance for a given core and bobbin, the designer should choose a core that provides an appropriate cross-sectional area, thus minimising the number of turns required to reach the target inductance.
Another important step is to provide the best possible coupling between the primary and secondary windings. The best results are achieved when the winding widths of the primary and secondary layers are matched, and kept on adjacent layers, or when the secondary layer is sandwiched between two primary windings, as shown in Figure 3.
Core losses: energy is required to effect a change in the magnetisation of the core. Not all of this energy is recoverable in electrical form; a fraction is lost as heat. This power loss can be observed electrically as the hysteresis of the B-H loop. The losses are generally proportional with the change in flux density (ΔB) and the square of the switching frequency (Fsw2).
For magnetic components in general, there is a trade-off between saturation flux density and core loss.
The use of materials with a high operating flux density offers benefits in the form of reduced size, weight and cost. For example, silicon steel cores typically have saturation flux densities of 1.5-2T. Unfortunately, such core materials also suffer from high core loss.
In contrast, ferrite cores are ceramic materials which have low saturation flux densities in the range 0.25-0.5T. But because their electrical resistivity is high, their core losses are low. Ferrite core materials commonly used in flyback transformers include 3C90 from Ferroxcube and the Magnetics® ‘R’ material, as shown in Figure 4.
Curves showing core losses at various switching frequencies, typically plotted as core loss in kW/m3 over ΔB measured in Teslas, are provided in material datasheets and can be used to estimate the core loss in any given application.
All of the considerations of loss above also have an effect on the calculation of core size. Readily available technical papers explain various methods for determining core size. In Future Electronics’ experience, it is often better to start with a slightly larger core size than strictly necessary, if space and cost constraints allow, since this will reduce the number of turns, core losses and leakage inductance.
In addition, it is best to choose a bobbin for the core of choice which provides the best winding length-to-height ratio: this will minimise the number of winding layers required.
The next step: hands-on prototyping
This article has outlined the important theoretical factors and design decisions that have to be taken into account in developing a design for a flyback transformer on paper. It also provides some guidance drawn from Future Electronics’ practical experience of transformer design, affecting factors such as core sizing and winding arrangement.
With this information to hand, the designer is ready to embark on the practical process of building a transformer prototype in the laboratory, a subject which will be addressed in a second article from Future Electronics’ EMEA SDC.