Most systems are now built around a microprocessor, digital signal processor (DSP) or other complex system-on-chip devices which have specific power requirements. High on that list is the trend to ever-lower voltage rails that carry correspondingly high currents.
Traditional power distribution architectures, in which centrally-regulated bulk supplies were bused around a system at the same voltage as they were consumed, were abandoned generations ago. Sharing currents of tens, and then hundreds, of Amps from a single source quickly became impractical. Not least among the problems was that, with the smallest IR (current x resistance) drop in the wiring being a high proportion of the load voltage, precision regulation became all but impossible.
At the same time as sub-micron semiconductor process technology was dictating the shift from 5V to 3.3V, then 1.8V and so on down, power transistor progress (among numerous other trends) enabled the construction of compact and efficient switch-mode DC-DC converters (“bricks”). This in turn launched a progression in power distribution architectures that routed power around the system at (initially) a relatively high DC level of 48V and converted it locally to the load voltage.
Voltage converters are available for distributed bus, intermediate bus and factorised power architectures (FPAs): in an FPA there is a fundamental reassignment of the locus of several key functions, compared to earlier schemes, that is essential to understanding the role of one the most efficient DC-DC conversion schemes yet devised: sine amplitude conversion, or SAC.
In an FPA power distribution scheme, the function of regulation – maintaining a stable voltage in the present of load and supply variations – takes place at the intermediate voltage level. The functions of isolation, and that of conversion to the final desired load voltage, are combined in a separate circuit block, a voltage transformation module.
This function can also be described as a current multiplier, as its main feature is that it acts as a ‘DC-DC transformer’. That is, it converts from DC to DC with a fixed ratio. In parallel with the evolution of power distribution schemes – and enabling that development – DC-DC converters also followed a path of increasing efficiency and power density.
Early designs “hard-switched” DC into an alternating square wave that was applied to the primary of a transformer. The resulting secondary voltage was rectified to yield the required output. Unsurprisingly, the presence of fast, high-current switched waveform edges gave rise to significant harmonic components that had to be filtered; and there were considerable losses, both conduction and switching, in the power devices.
These limitations began to be addressed by the use of a conversion circuit arrangement that contrived to turn power switches off and on at the point in the waveform where both current and voltage was at a zero-crossing (Figure 1).
A resonant tank circuit (comprising the added capacitor and the transformer winding) on the primary provides the context for the quasi-sinusoidal waveform. Switching losses were reduced; contemporaneously, improved transistors cut conduction losses. The primary waveform approximated to sinusoidal, and harmonic generation diminished.
An important point in this circuit’s operation is that one cycle of the primary provides one “unit” or “quantum” of energy to couple to the secondary. If this converter is to respond to an increase in secondary load, the control loop must react by increasing the switching rate at the primary; the principal regulatory mechanism is via switching frequency.
The next architecture in the progression is the SAC. It is important to note that it also employs ZVS-ZCS – in that its power switches turn off and on in near-ideal conditions – but its operation is fundamentally different.
The conceptual circuit arrangement is shown in Figure 2. At the heart of the power path is the power transformer T1. A sinusoidal waveform is excited in the primary, the resonant elements being the leakage inductance of the transformer, and the inter-winding capacitor. The multi-winding transformer T2 is a control/drive component that provides signals to the primary power switches and synchronous rectifiers on the secondary side, to switch at the zero-crossing points of the waveform.
That waveform is fully-sinusoidal, continuously oscillating, and fixed frequency.
Response to load changes is by a different mechanism to the previous case. A load increase prompts an increase in amplitude in the primary waveform, driving more energy to the secondary, and maintaining the fixed conversion ratio from input to output. The analogy with an AC transformer is useful (as long as the limitations of the analogy in understanding circuit operation are kept in mind); the VxA product at input and output is identical, allowing for losses. Losses are very small – use of the latest semiconductors keeps conduction losses to a minimum and all switching is at zero-crossing points. Conversion efficiency of between 98% and 99% is achieved, contributing to very high power density and reduced thermal management requirements. With sinusoidal waveforms and negligible switching transients, harmonic content, both radiated and conducted, is also minimal.
Naturally, there are many subtleties to the detail design that are not apparent from the outline circuit. The designer must, for example, generate correctly-shaped drive waveforms for each of the power switches from the waveforms coupled by the drive transformer; the designer must also ensure (again, taking just one example) that the converter starts under all load conditions.
In common with other devices in the most advanced power-conversion families, the properties of every component – both primary and parasitic effects – are modelled, understood and fully exploited as an integral part of the design.