Introducing Cutting Edge Technology into Low Power DC/DC Converters

DC/DCs are just a collection of active and passive components, so why have they not become just ‘another’ integrated circuit and shrunk in the same way? One reason is that they often run with significant power dissipation, needing some surface area to dissipate from, but this is becoming less of an issue as efficiencies rise with new conversion techniques. The main reason is the magnetic components required in most converters, which have stubbornly stayed with the same fabrication technology and around the same size over decades. As a comparison, in 1988, Taiwan Semiconductor Manufacturing Company (TSMC) was providing ICs with a 3µm geometry, and today, the figure is a thousand times smaller at 3nm [1].

In the same period, typical discrete, surface-mount passive components, that are practical to machine-place, have also reduced from 1206 to as low as 01005 sizes, which is a reduction in footprint by more than 50 times. In contrast, the sizes of magnetic cores in DC/DC converter transformers and chokes have barely changed since the 80s, set by the inherent maximum flux density of the material and operating frequency, which then set the minimum number of winding turns.

To be fair to the generations of power engineers, power density has improved, with lower losses from new conversion topologies, better components, and advanced thermal design. This has allowed more output power from a given DC/DC module size, by perhaps a factor of just 3x in the case of the SIP7 format, for an unregulated type (Figure 1).

There have always been options to reduce the size of power conversion magnetics by increasing switching frequency, which generally reduces core size, winding turns, or both, in some combination. However, with higher switching rates, semiconductor efficiency drops and core losses increase, so overall case size does not necessarily reduce without higher internal temperatures. The solution is a more complex converter designed for high efficiency, but this has been seen as prohibitively expensive.

Magnetic parts are also relatively expensive to make and fit in a typical converter; there has been little change to assembly techniques that Faraday would have found familiar—winding insulated wire around a core and then soldering ‘flying’ wires to a substrate (Figure 2). The typical wire size is 0.18mm,and the core diameter is 6mm outside and 3mm inside. Bobbins generally take up too much space and techniques, using printed windings, have not been practical due to the number of turns and windings needed and the unacceptable cost of multi-layer substrates, at least for low-power products.

The route taken by most low-power DC/DC converter manufacturers has been to make the circuit as simple and low-cost as possible, for example, by using the traditional ‘Royer’ circuit (Figure 3). The savings achieved then offset the high labor cost of winding simple toroids and hand-soldering wires to a double-sided PCB, with encapsulation or over-molding, to protect the fragile terminations. The circuits and assembly techniques have been refined, over the years, so that a simple unregulated converter might only employ around ten discrete components, and a regulated version uses fifteen.

With transformer manufacture and module assembly in a low-cost location, the product is reasonably efficient, provides isolation, a wide operating temperature range, and quite accurate voltage conversion between fixed levels. An upside of the manual assembly method is that variants of the products, for different input/output voltages and power ratings, are relatively easy to achieve in the manufacturing process with a simple operator instruction to wind more or fewer turns.

There are inevitable downsides to this approach. However, manual assembly produces variation between samples, and it is difficult to provide comprehensive fault protection in simple circuits, and isolation to a safety-certified level is not practical without more complexity, cost, and larger case sizes. A basic Royer converter has no line or load regulation, and the output voltage can significantly rise at very light or no load. Additionally, labor costs only increase over time, while end-customers expect price reductions, and the labor element does not even decrease with production volume. At the same time, there is market pressure to increase functionality and efficiency and reduce the size of power converters to suit modern space-constrained applications.