Fullscreen HTML5Med Res (??MB)HTML4

By Andrew Levido Power Electronics Part 3: Isolated DC-DC Converters Isolated DC-DC converters offer both improved safety and flexibility compared to the non-isolated kind. To design an isolated converter, first we need to understand how transformers work, plus their potential roles in DC/DC and AC/DC converters. This article will start with transformer/inductor fundamentals, then move on to using them. T Fig.1: this shows the relationships between the key magnetic quantities: field intensity, flux and flux linkage. Together, these allow us to determine the inductance of the coil. The left side of Fig.1 shows a winding of several turns of wire around a magnetically permeable core (we’ll consider what that means soon). The N-turn winding carries a current i, which induces some kind of magnetic flux (or field in some textbooks) denoted by the Greek capital phi (Ø), shown in red. The core itself has a cross sectional area A and an average length l. By convention, the direction of the flux follows the ‘right-hand rule’ illustrated at the bottom left of the figure. If you curl the fingers of your right hand in the direction of current flow, the resulting flux points in the direction of your thumb. Gauss’ law for magnetism tells us that the net flux entering and leaving any closed region is zero, which means that flux lines have no beginning or end – they must always be closed loops. For us, this just means that all of the flux leaving the coil at the top enters it again at the bottom. For now, we will assume that all of the flux is confined to the core. You can think of this arrangement in two ways at the same time; as loops of current enclosing a flux, or as loops of flux enclosing a current. Flux is driven around the core by a ‘driving force’ known as the magnetomotive force (mmf) that describes the amount of current linking with the flux. This mmf is N times the current i, since the current passes through the flux loop once for each turn of the winding. We can also describe this magnetising force as a magnetic field intensity, which we denote with the letter H. This is defined by Ampère’s law to be equal to Ni ÷ l, or the mmf per unit length of core. The right-hand side of the figure shows a section of the core with the Australia's electronics magazine siliconchip.com.au he DC-DC converters we have looked at so far have been non-isolated types. That means there is a direct electrical connection between the input and output. In many cases, we want the output to be isolated from the input; for safety reasons, if the input is connected to the mains, or because we need the output to be referenced to a different potential than the input. Isolation is usually achieved using a transformer. Adding a transformer to a switching converter can provide a host of other benefits – for example, it can reduce the range of duty cycle (and consequent component stress) required to achieve high step-up or step-down ratios. It can also allow us to get multiple outputs from a single converter. As we have seen in this series so far, we cannot get very far in the field of power electronics before coming up against magnetic components such as inductors and transformers. Like everything else, it is the second-order non-ideal behaviour of these components that will catch us out if we are not careful. Transformers, in particular, are highly specific to the application, so 26 Silicon Chip sooner or later we will have to roll our own. This means that a solid understanding of magnetics theory is necessary before we get to isolated DC-DC converters, so let’s dive in. Back to basics One of the reasons magnetics can be confusing is the number of terms that sound similar but have different meanings. This comes about because magnetics was one of the earliest areas of electrical engineering to be studied by the likes of Gauss, Ampère and Faraday, who each invented their own terms, which we live with to this day. With terms like magnetic field intensity, magnetic flux, magnetic flux density, magnetic flux linkage, magnetic permeability and magnetic permeance, it is no wonder many of us get confused. Don’t even get me started on Maxwell’s equations! (We covered those in the November 2024 issue – siliconchip.au/Article/17029). I promise that it is not as hard to wrap your head around as you might think. As usual, I will not cover this topic in a rigorous academic fashion, but from the perspective of a working engineer. We will employ just a little basic algebra. flux lines passing through it. As we have seen, the magnetic field intensity H drives a total flux Ø through the core, equal to µHA. This results in a flux density, B, which is simply the amount of flux divided by the cross-sectional area of the core. B is also related to H by the magnetic permeability, or µ, of the core according to the relationship B = µH. Permeability is a measure of how ‘easily’ a given field intensity can create a flux density in the core. Materials with a higher permeability will develop a higher flux density for a given field intensity. The permeability of a material is usually expressed as the permeability of free space, µ0 (which is equal to 4π × 10-7H/m) multiplied by a unitless relative permeability, µr. The permeability of free space is the basic measure of how much flux a given current will produce in a vacuum or in air. The relative permeability is a measure of how many times more permeable a material is than this baseline. Thus, you often see permeability expressed as µ0 × µr. Relative permeabilities for magnetic materials range from a few hundred for mild steel, to 5000-10,000 for transformer steel, up to 40,000 or more for some ferrites. In addition to the flux and the flux density, we need to introduce the concept of flux linkage, which describes how a flux links with a winding. From Fig.1, you should be able to see that the red flux lines pass through each turn of the winding, giving it a flux linkage λ = NØ (λ is the Greek letter lambda). So, to sum up, the current flowing in a winding produces a magnetic field intensity, which drives a flux around the core. This produces a flux density in the core that is related to the field intensity by the permeability, and to the total flux by the cross-sectional area of the core. The resulting flux passes through the winding, resulting in a flux linkage, λ. Fig.2: reducing magnetic geometries to equivalent circuits makes analysis much easier, since you can use all the usual circuit theory tricks. magnetic field. In our example, the magnetic flux will change proportionally to changes in the current, so it follows that a voltage is produced across the terminals of our winding as the current through the winding changes. The changing coil current effectively induces a voltage in itself. This is known formally as self-inductance, but we usually just refer to it by the shorter name “inductance”. Inductance kind of wraps up all of the N, H, B, Ø, and µ malarky into a relationship between the current and the flux linkage. In fact inductance, is defined as the flux linkage per unit of current, L = λ ÷ i. For the mathematically inclined, you can see how this works by working out an expression for the inductance of the arrangement in Fig.1. Combining the equations for field intensity (H = Ni ÷ l) and flux density (B = µ0 µr H) we get B = µ0 µr Ni ÷ l. Multiplying by the area gives us the total flux Ø = µ0 µr NiA ÷ l. We can then use the formula for flux linkage (λ = NØ) to get λ = µ0 µr N2iA ÷ l, and finally use the formula for inductance to arrive at L = N2 × µ0 µr A ÷ l. The inductance is therefore the product of the square of the number of turns multiplied by a term related to permeability of the core and its dimensions. This latter term is referred to as the permeance of the core (not to be confused with the permeability). If you know the permeance of a core, you can easily calculate the number of turns required to obtain a given inductance. Manufacturers of cores usually provide the permeance in their data sheets as an “Al” value, in units of nanohenries per turn squared or similar. Electric circuit model The inverse of permeance is reluctance, which is an extremely useful quantity we can use to build a model of magnetic systems that is analogous to electric circuits. In these circuits, reluctance (denoted R) is equivalent to resistance and the mmf (F), equal to Ni, is equivalent to voltage. The resulting flux (Ø) is analogous to current. Fig.2 shows the electrical circuit equivalent of the magnetic circuit in Fig.1. The circuit obeys the magnetic version of Ohm’s law, so F = ØR. We can use all of our usual circuit analysis techniques, so this is really helpful to calculate inductances and the like when faced with more complex core geometries such as that in Fig.3. At the top, we have a classic E-I core with a winding on the centre leg that is wider than the outer legs. Fig.3: analysis of an E-I core and a gapped core using the magnetic equivalent circuit shows how easy it can be to calculate the inductance of complex geometries. Inductors You might have noticed that this is a bit circular – the current in the winding produces a flux that links with the winding. By introducing the third member of the magnetic holy trinity, Mr Faraday, we can use this to understand how inductors work. Faraday’s law states that a voltage is induced in a winding by a changing siliconchip.com.au Australia's electronics magazine January 2026  27 The magnetic equivalent circuit is shown to the right. It is easy to calculate the reluctance of the centre leg, R1, and that of the two outer legs, R2, based on the dimensions of the core and its permeability. You can then use what you know about resistors in series and parallel to calculate an equivalent reluctance and therefore the inductance as shown in the figure. Fig.3 also shows another common configuration, where a very narrow air gap is included in the magnetic circuit. This is done to increase the stability of the inductance and the amount of energy that can be stored in the core. The reluctance of the air gap is much higher than that of the core, because the relative permeability of air is one, compared to many thousands for the core. This means the inductance is dictated by the air gap, and is largely independent of the core material. This can be a good thing for the stability of the inductance, since the relative permeability of most core materials changes with temperature and flux density, as we will see below. I won’t cover the maths, but for the same reason, for a given flux density, the amount of energy that can be stored per unit volume is much higher in the air gap than in the core. In fact, it is common to assume that all of the energy is stored in the gap, and it’s not unusual to start the design process for inductors by selecting a core with sufficient gap volume (the area of the core times the gap length) to store the required energy each switching cycle. Leakage As usual, I have made a few simplifications in the above discussion, and some other factors come into play when we get into the nitty-gritty of magnetics design. One of these is leakage flux. In the above examples, we assumed that all of the flux was constrained to the core. Fig.4 shows that this may not be the case – some flux may leak away from the core and pass through the air where there is a lower reluctance path; however, it will always return, due to Gauss’ law. The effect of this on the magnetic circuit is shown in the right-hand side of the figure. The leakage paths form a leakage (generally high) reluctance, which appears in parallel with the core reluctance. You can see from the formula that this leakage will result in a slightly increased inductance over the ideal case. It will be useful later to think of the leakage producing an extra ‘leakage inductance’ in series with the main core inductance, as shown in the lower equation. Saturation, hysteresis and residual flux We have also assumed up until now that the relationship between field intensity and flux density is linear, described by a simple proportional relationship of permeability. The reality (as always) is a little more complex. A typical magnetic material has a B-H characteristic like that in Fig.5, although it is shown a bit exaggerated for clarity. Remember that H is the magnetising ‘drive’, proportional to the winding current, and B is the resulting flux density. There are three important things to note. First, the path that B follows when H is increasing is not the same as it does when H is decreasing; there is hysteresis. Second, neither path passes through the origin. When H is zero, there will be some residual flux density Bres (either positive or negative) present in the material when it is not excited. Thirdly, the slope of the characteristic (the permeability) saturates at some flux density, Bsat – it deviates from the ideal value of µ shown in red as the field intensity increases. We normally want to avoid saturation (although there are some notable exceptions where this characteristic is used to advantage), so we take some care to ensure the maximum flux density stays well below the saturation level. Losses We have also so far assumed inductors are lossless, but of course, we know this cannot be the case. Losses in magnetic components fall into two categories: copper losses (sometimes called winding losses) and core losses. Copper losses include the familiar ohmic loss determined by the resistivity of the winding material, its cross sectional area and its length. Be aware that the resistivity of copper, the most common winding material, is temperature-dependent and increases by approximately 40% for every 100°C temperature rise. Make sure to calculate losses using the resistivity at the highest operating temperature you will see in the winding. Resistive losses are exacerbated by a phenomenon known as ‘skin effect’. As the frequency of a current passing through a conductor increases, eddy currents create magnetic fields inside the conductor that force the current outward, so it flows only in the outside ‘skin’ of the conductor. The higher the frequency, the more the current is forced to the outside of the conductor. The ‘skin depth’ is a measure of how much of the conductor is effectively useful. For copper, the skin depth is about 9.2mm at 50Hz, which explains why high-current AC busbars tend to be broad but are rarely more than 10mm thick. At 100kHz, the skin depth is about 0.2mm, and at 1MHz, it is just Fig.4: flux leakage in an inductor produces an extra leakage reluctance in parallel with the core’s reluctance, and increases the total inductance slightly. Fig.5: magnetic materials have a less-than-ideal B-H characteristic that includes hysteresis, saturation and residual flux density. 28 Silicon Chip Australia's electronics magazine siliconchip.com.au 65µm. There is no point in using a cylindrical conductor with a diameter greater than twice the skin depth, since there will be no corresponding reduction in resistance. At 100kHz, for example, any conductor with a diameter larger than 0.4mm will be a waste of copper. For this reason, the magnetics in high-power switching converters use multiple thin conductors in parallel, or are wound with copper foil (or even multiple parallel layers of copper foil). You can also use Litz wire, which is an intricate arrangement of fine insulated wires twisted together into bundles, which are themselves twisted together. It is lovely stuff, but expensive. Core losses also come from two sources: eddy currents and hysteresis. Eddy current losses are resistive losses caused by currents circulating within the core material. In conductive core materials like steel, these losses are mitigated by making the core from thin laminations that are insulated from each other by an oxide layer. You will have seen these laminations in the cores of E-I type mains transformers. Toroidal mains transformer cores are wound from a long thin strip of steel (like a roll of sticky tape) to achieve the same end. At higher frequencies, we tend to use cores made of materials that have poor electrical conductivity, such as ferrite or sintered metal oxides, to avoid eddy current losses. Hysteresis loss is caused by the shape of the B-H curve. When the material is magnetised in one direction, it takes some magnetic force in the other direction to overcome the residual flux and bring the flux density back to zero. This takes energy, which becomes heat in the core. The amount of loss is proportional to the area within the hysteresis loop, so choosing a material with a narrower Fig.7: a realistic transformer has leakage inductances due to imperfect coupling of the flux, and a magnetising inductance due to the finite permeability of the core. hysteresis curve will help minimise these losses, as will limiting the maximum flux density excursions. Manufacturers usually provide a measure of core losses for their materials in kW per cubic metre for a (usually pretty small) range of frequencies and flux densities in the data sheets. You have to multiply these by the core volume to get an estimate of core loss in your application. Transformers Adding a second winding to our inductor, as shown in Fig.6, produces a transformer. If the flux is perfectly linked by both windings, as shown here, the transformer is said to be perfectly coupled. While we are at it, let us also assume that the core is so permeable that the reluctance is zero. This is, after all, an ideal transformer. Since the flux linked by each turn on both windings is identical, so is the voltage produced across each turn. Each winding voltage is therefore proportional to the number of turns in that winding. The ratio of voltages v1 : v2 is equal to the turns ratio, N1:N2. Due to the sense of the windings in the diagram and the right-hand rule, the total mmf is the sum of the Ni values of each winding. If current flows into the dotted ends of either winding, it produces a clockwise flux as shown. I have indicated the direction of flux for a current into the dotted terminal by a red arrow on the electrical equivalent circuit on the right. Fig.6: adding a second winding to a core produces a transformer. The currents in the windings oppose each other, reducing the flux to almost zero. siliconchip.com.au Australia's electronics magazine The magnetic circuit shows that even though the two mmfs are pushing flux around the core in the same direction, when one is excited, the other will see an mmf of the opposite polarity. The mmf seen at F2 due to a positive F1 will be negative and vice versa. This means that a current entering the dotted terminal of one winding will force a current out of the dotted terminal on the other winding. If the reluctance is zero, the mmfs will be equal in magnitude as well as opposite in sign. In other words, the ratio of transformer currents i1 : i2 is equal to -N2:N1. Don’t get too worried about the negative sign here; it is just there because convention says positive current flows into the dotted terminal. In an ideal transformer, then, perfect flux linkage means the voltages are related by the turns ratio (v1 : v2 = N1:N2), and zero reluctance means the currents are related by the inverse of the turns ratio (i1 : i2 = N2:N1). The net flux in the core must be zero since the input and output mmfs cancel out as they are identical but opposite in sign. The impedance looking into one winding with the other open will be infinite (it will look like an open circuit), and when the other is short-­ circuited, it will be zero. A transformer model While ideal transformers are handy for circuit analysis, they are not realistic. We can summarise these non-­ idealities in the equivalent circuit of Fig.7. There is an ideal transformer in the centre of the diagram. The inductances in series with each side are leakage inductances caused by incomplete coupling of the flux, as we saw in Fig.4. I have shown a leakage inductance on either side of the transformer, but it is also sometimes handy to have it all ‘lumped’ onto one side or the other. For example, if we wanted to show Ll 2 on the same side as Ll 1, we would shift it over but multiply its value by (N1÷N2)2. We could equally move Ll 1 January 2026  29 to the same side as Ll 2 by multiplying its value by (N2÷N1)2. Since any real transformer core has a finite reluctance, the opposing mmfs of each winding will not completely cancel out, and there will be some level of residual flux in the core. This is represented by the parallel inductance known as the magnetising inductance. This is responsible for the small current that will flow in a transformer’s primary winding when its secondary is open circuit. There are also copper losses and core losses in transformers, driven by the same mechanisms as discussed above for inductors. The copper losses can be represented by appropriate resistances in series with the leakage inductances, and the core losses by a resistor in parallel with the magnetising inductor. I have not bothered to show them here to keep things simple. The forward converter Knowing what we do about magnetics, we can begin to understand isolated DC-DC converters. In the upperleft corner of Fig.8 is a non-isolated buck converter that we are by now very familiar with. To its right is an isolated version of the same topology, known as a single-ended forward converter. An ideal (for now) transformer with a turns ratio of N:1 has been inserted pretty much in the middle of the buck converter’s switch Q1, which now consists of a Mosfet (Q1) on the primary side and diode (D2) on the secondary side. When Q1 is on, current flows into the dotted primary terminal of the transformer. A current, scaled by a factor of N, emerges from the dotted secondary terminal and passes through D2, forming the second half of the switch feeding the filter inductor. When Q1 and D2 are off, the filter inductor current flows via diode D1, just as it does in the non-isolated converter. The transfer function of the forward converter is the same as the buck converter, but scaled by the transformer turns ratio, N. I have drawn the forward converter with the Mosfet in the positive input line to match the buck converter, but in reality, it is typically moved to the ‘ground’ side of the transformer primary to make its gate drive simpler, as shown in the circuit at lower left in Fig.8. In this circuit, I have also added the magnetising inductance to the transformer, and a clamp circuit consisting of diode D3 and zener diode ZD1. You can probably already see why these are necessary. We don’t need to worry about the leakage inductances, as they are in series with the ideal transformer, so there is always a path for their current to flow. They will affect voltage regulation a bit, but we won’t worry about that now. This is a ‘single-ended’ converter because power flows through the transformer only during the part of the cycle when Q1 is conducting. This means that when the Mosfet switches off, the magnetising current needs a path to flow or else the Mosfet’s drain terminal voltage will spike and it will be toast. The clamp circuit limits the Mosfet drain voltage to the sum of Vin plus the zener voltage. The energy stored in the magnetising inductance is dissipated in the clamp every cycle. The price we pay for the convenience of the transformer is the additional complexity and power dissipation of a clamping circuit, and the additional voltage stress on the switch. There are several other ways to implement the clamping circuit, two of which are shown at lower right in Fig.8. The first, a resistor-capacitor-diode (RCD) clamp, relies on a capacitor to absorb the energy, which is then dissipated in the parallel resistor. This is probably the cheapest option and is often seen in low-cost designs. A more efficient option is to add an extra winding to the transformer and a diode, as shown in the energy recovery clamp partial circuit. When the Mosfet switches off, the magnetising current causes the transformer terminal voltage to rise until the clamp diode conducts. If the clamp winding had the same number of turns as the primary, the Mosfet drain voltage would be Fig.8: the single-ended forward converter is essentially a buck converter with Q1 replaced by a Mosfet, transformer and diode. The transformer’s magnetising inductance requires the addition of a clamp circuit, as shown along the bottom of the figure. 30 Silicon Chip Australia's electronics magazine siliconchip.com.au clamped to twice Vin. You can choose the number of turns on the clamp winding to limit the drain voltage even further if necessary. As the name implies, the advantage here is that the magnetising energy is returned to the supply. Another similar variant of the single-­ ended forward converter, the isolated hybrid bridge converter, is shown in Fig.9. It is still a single-ended converter, because the Mosfets can only drive flux through the transformer in one direction, but it solves the magnetisation current problem by clamping both ends of the windings to the supply rails when the Mosfets are off. The Mosfets are never subject to more voltage stress than the supply rails, but the gate drive for the upper Mosfet is more complex in this configuration. Double-ended forward converters An obvious(?) next step would be to replace the diodes in the hybrid bridge converter with Mosfets to produce a full-bridge converter. This has the huge advantage of driving flux in both directions in the transformer and allowing us to use a full-bridge rectifier on the secondary side. The single-ended converter can only drive the magnetising flux around the transformer core in one direction (remember, the magnetising flux is what’s left in the core after most of the Fig.9: the isolated hybrid bridge converter solves the problem of magnetising inductance at the cost of circuit complexity. flux is cancelled out). In a double-ended converter such as this one, the magnetising current can change sign. We can therefore utilise the full range of flux density in the core material, making for a more efficient and smaller transformer. Being able to use a full-bridge rectifier means we can use a smaller filter inductor, since the frequency at the output of the rectifier is twice the switching frequency. This type of converter does require that we take care to limit the duty cycle of each phase to less than 50%, or we run the risk of switching on the upper and lower Mosfets at the same time, with catastrophic results. The price to pay for such advantages is complexity. There are now two highside and two low-side Mosfets to drive, and four output diodes. Moreover, the arrangement means that two of these Mosfets and two diodes are in series each cycle, so the efficiency is less than ideal. If only we could get the advantages of the double-ended converter without these disadvantages! Well, you can, by using a more complex transformer, as shown at the bottom of Fig.10. This is the transformer-coupled half-bridge or push-pull topology, and it has all the advantages of the full bridge, but is considerably simpler. There are only two switches, and both are ground-referenced. Only two diodes are required, and the output current only ever passes through one of them. Nice. The flyback converter Next, I want to cover the flyback topology, which is an isolated topology derived from the boost converter (Fig.11). This time, we split the circuit in the middle of the inductor, creating two coupled sections. It looks a lot like a transformer, but strictly speaking, does not behave that way. Fig.10: double-ended converters can drive flux through the transformer core in both directions, increasing the efficiency of the magnetics. siliconchip.com.au Australia's electronics magazine January 2026  31 Fig.11: the flyback converter is derived from the boost topology. The ‘transformer’ is actually two coupled inductors – the windings never conduct at the same time. When Q1 is on, current flows into the dotted primary terminal of the ‘transformer’. True transformer action would require it to emerge from the secondary’s dotted terminal, but it cannot, because it is blocked by the diode (D1). The secondary winding is effectively open circuit, so the ‘transformer’ acts like an inductor, building up flux and storing up energy in the core. When the Mosfet switches off, the primary winding is open-circuited and the magnetic field begins to collapse, reversing the voltage on both windings and allowing D1 to conduct. Now that the primary winding is open-circuit, the flyback transformer secondary acts like an inductor and the current ramps down, just as it does in the boost converter. The only difference is that the turns ratio means the secondary current is scaled by a factor N. The voltage transfer function for the flyback converter is therefore the same as for the boost converter, but scaled by the turns ratio. While the flyback circuit looks a lot 32 Silicon Chip like the forward converter, the crucial difference is in the operation of the transformer. A forward converter has a true transformer in that the net flux largely cancels (except for the magnetising flux), and no appreciable energy is stored. The output filter inductor remains the primary energy storage element. In flyback converters, the ‘transformer’ is also the energy storage element. Since the two windings never conduct simultaneously, the flux increases significantly. Flyback transformers are really two-winding inductors and usually have gapped cores. If you are ever unsure about what topology you have, take a look at the dots on the transformer, and work out if both windings can conduct at the same time or not. The fact that flyback converters combine the energy storage element and the isolation element into one piece of magnetics (and because they use The Mornsun LM25-23B12 25W 12V isolated power supply. one less diode) is one of the reasons why they are the most common topology for small mains converters. That includes many phone chargers, plugpacks and other low-power DC-DC converter modules below about 50W. Flyback converter transformers do not have the magnetising inductance concern that single-ended forward converters do (because they are not really a transformer), but they do have a problem with leakage inductance, as shown at the bottom of Fig.11. When the Mosfet switches off, the energy stored in the core is delivered to the secondary, but any that is stored in the primary side leakage inductance has no place to go since it is, by definition, not linked by the secondary winding. So weirdly, it turns out that a practical flyback converter needs a similar type of clamp as a single-ended forward converter, but for a different reason altogether. A professional design To add a practical twist, I want to take a close look at the design of a commercial flyback converter, because you can learn a lot by looking at designs by experts. The converter I chose is a Mornsun LM25-23B12, an offline isolated 25W switcher with a 12V DC output. It can accept input voltages in the range of 100V to 277V AC and can deliver 2.1A at 50°C. You might think this is an AC-DC converter and not a DC-DC converter, and by some definitions, you would be right. Still, I will argue that, like many mains-powered supplies, it is an AC-DC converter followed by a DC-DC converter. The AC-DC side of this converter is a simple bridge rectifier, so all the interesting stuff is happening in the DC-DC part. These converters are built to a price, but they do claim to meet a bunch of international specifications for safety & EMC (electromagnetic emissions compliance). Looking at the construction & component choice, I don’t siliconchip.com.au Fig.12: the reverse-engineered circuit of a commercial 25W switching converter (the Mornsun LM2523B12). The text describes some of the interesting design features. doubt that this is a well-designed unit. The accompanying photos show the power supply and both sides of the PCB. It is a single-sided board with through-hole components on the top side and SMT parts on the bottom. The slots milled into the board are to provide creepage isolation between primary and secondary and between high voltages. The circuit, as best as I could reverse engineer it, is shown in Fig.12. Starting on the left is the mains input, with a fuse and an inrush-limiting NTC resistor. An X2 capacitor and common-­ mode inductor provide some filtering to minimise the amount of EMI (electromagnetic interference) conducted back onto the mains. This filter is followed by a full-wave bridge rectifier and three 15µF 400V DC capacitors in parallel, to smooth the input to the flyback converter. The negative side of the high-voltage DC supply is tied to mains Earth via a 2.2nF X1 capacitor, and to the output negative rail by two 1nF X1 capacitors in series. These capacitors provide a path to shunt high-frequency noise to Earth without compromising the safety or isolation. The power circuit looks like any flyback circuit, with one side of the transformer primary connected to the positive supply and the other to the drain of the Mosfet switch, which is incorporated into the SDH8666Q control IC. The Mosfet’s source is connected siliconchip.com.au to the current sense (CS) pin, which is connected to the negative supply via a 0.5W shunt resistor (three parallel 1.5W resistors). This chip uses current-mode control, and this is the current-­sense resistor. An RCD clamp with a 120kW resistor (two parallel 240kW resistors) and an unmarked capacitor protects the Mosfet from spikes caused by the transformer leakage inductance I described above. On the secondary side, the rectifier consists of two SK3150AS schottky diodes connected in parallel. These are 150V 3A diodes, and I guess two are used in parallel since the peak current could easily be twice the maximum output current of 2.1A. The diodes are followed by two parallel 470µF electrolytic filter caps. I am a bit surprised not to see a ceramic cap in parallel with these, given the switching frequency is in the 65kHz range. These caps must be working hard from a ripple current perspective (but presumably within their specifications). A secondary filter comprising a small inductor and a 47µF capacitor helps eliminate a lot of the switching noise on the output. Isolated voltage feedback and control loop compensation is provided via the circuit at lower right. This uses a TL431 shunt reference and opto-­coupler in a clever (but common) arrangement. I think this circuit is worth a bit of Australia's electronics magazine a closer look, so I have redrawn it in Fig.13. The converter’s output voltage is divided down and compared to the 2.5V reference internal to the TL431. The resulting error voltage at the TL431’s anode is converted to a current by Rb to drive the opto-­ coupler’s LED. A current proportional to this will flow into the coupler phototransistor’s collector and be converted back to an error voltage by the pullup resistor internal to the control chip. Fig.13: the Mornsun voltage feedback, error amplifier, isolation and loop compensation circuit uses a TL431, and opto-coupler and a handful of passives. The compensator is a Type II circuit, suitable for current-mode controllers, as described last month. January 2026  33 I calculated the circuit’s small-­signal transfer function using the complex impedance method we covered last month. As we would expect with a current-mode controller, the result is characteristic of a Type II compensator. I will spare you the maths. The constant terms at the front of the transfer function relate to the opto-coupler’s current transfer ratio (CTR), and the resistors on each side that convert voltage to current to voltage. The interesting part is inside the brackets. There is a pole at the origin and another formed by the capacitor Cb and the pullup resistor Rpu inside the controller. There is also a zero formed by Ra and Ca. This zero cancels the output capacitance/load resistance zero, and the Cb/Rpu pole cancels the zero formed by the output capacitor and its ESR. The purpose of the capacitor directly under the opto-coupler in Fig.12 is a bit of a mystery, but as it measures about 22pF and is connected at the output of the TL431’s internal op amp, I suspect it is there for stability and plays no meaningful part in the control loop. The next interesting part of the flyback converter is the power supply for the control chip. This is derived from an auxiliary winding on the flyback transformer via a diode and a couple of capacitors to supply the Vcc pin of the controller. You may ask yourself how this circuit can possibly start, given that the chip is powered by its own output. The chip has a clever trick up its sleeve in that there is an internal high-voltage depletion-mode (normally on) Mosfet connected to the DRAIN pin (pin 6) that initially provides a small trickle current to charge the 22µF capacitor on the Vcc pin. To keep the power dissipation in the depletion Mosfet to a minimum, this current is very small – nowhere near enough to power the chip – so the chip is not enabled until the Vcc voltage reaches some fairly high predetermined level. At this point, the chip switches on and uses the charge stored in the 22µF cap to run for long enough for the external supply to take over. Once everything is up and running, the depletion-mode Mosfet is switched off to save energy. This leaves only the DEM pin, which is fed from the auxiliary secondary winding prior to the rectifier diode. This pin is used to (roughly) sense the output voltage to provide output overvoltage protection, and something called “valley lockout”, which appears to be a mechanism to prevent the chip restarting too quickly due to small dips in the input voltage. The DEM pin can indirectly sense the converter’s output voltage because the voltage on the auxiliary winding is proportional to the output voltage (less one diode drop) according to the turns ratio. I suspect that the valley lockout works by ignoring short dips in the input voltage (sensed at the DRAIN pin) if the output voltage (sensed at the DEM pin) does not also drop. This would prevent the start-up sequence described above from happening unnecessarily for very short mains interruptions that don’t impact the output. That’s it for this month. Next month, we will have a look at AC-to-DC converters, and we will go through the design process for a simple DC power source in detail. We will build and test the circuit to see how well the theory SC and practice align. Silicon Chip as PDFs on USB ¯ A treasure trove of Silicon Chip magazines on a 32GB custommade USB. ¯ Each USB is filled with a set of issues as PDFs – fully searchable and with a separate index – you just need a PDF viewer. ¯ 10% off your order (not including postage cost) if you are currently subscribed to the magazine. ¯ Receive an extra discount If you already own digital copies of the magazine (in the block you are ordering). The USB also comes with its own case THE FIRST SIX BLOCKS COST $100 OR PAY $650 FOR ALL SEVEN (+POSTAGE) NOVEMBER 1987 – DECEMBER 1994 JANUARY 1995 – DECEMBER 1999 JANUARY 2000 – DECEMBER 2004 JANUARY 2005 – DECEMBER 2009 JANUARY 2010 – DECEMBER 2014 JANUARY 2015 – DECEMBER 2019 OUR NEWEST BLOCK OF ISSUES COSTS $150 → JANUARY 2020 – DECEMBER 2024 WWW.SILICONCHIP.COM.AU/SHOP/DIGITAL_PDFS Ordering the USB also provides you with download access for the relevant PDFs, once your order has been processed 34 Silicon Chip Australia's electronics magazine siliconchip.com.au