In a relatively short time, semiconductor-based electronic systems have developed at an exponential rate. It started with the first integrated circuit, developed in the 1950s, comprising just two transistors on a substrate that measured almost 13 square centimetres.
We now have devices that integrate millions of transistors in just a few square millimetres. But the penalty for this achievement is the need for ever more complex power systems to support these highly integrated devices.
The advent of the switching power transistor in the 1960s had a huge impact on power supply design, as it enabled the DC-DC converter. This served the early development of the semiconductor industry well, but demand was already building for more. In the early 80s, Vicor introduced a new approach to DC-DC conversion which ushered in a new era in power supply technology. The concept of quasi-resonant zero-current switching (ZCS/ZVS) was a departure from existing pulse-width modulation (PWM) based switched mode DC-DC converters, allowing them to reach new levels of efficiency and power density.
Power goes modular
Thanks to this increased power density provided through quasi-resonant zero-current switching, Vicor was able to develop what is now widely accepted as the industry’s first modular based power solution: the brick. This format afforded engineers much more flexibility when developing systems, giving them the engineering and economic freedom to create, much more quickly, systems that were smaller and more cost effective.
Being so radical in design, the first generation of bricks were poorly supported by component manufactures, and so power supply companies were forced to use components that were intended for signal processing applications. This had implications in terms of size: the surface mounted discrete components available consumed around 70% of the available board space, while the pot core transformers introduced noise and heat dissipation issues.
However, the evolution of the brick – and subsequently half-bricks and quarter-bricks – continued unabated, making full use of developments in semiconductor technology. As integration technology developed, manufacturers were able to combine more and more functions in a single device. Through the integration of control circuitry into one device, and active circuitry into another discrete component, the volume of second-generation bricks was drastically reduced. With just two discrete devices needed for control and active circuits, the volume they consumed dropped to 1,6 cm³, with a subsequent reduction in the component count from 113 to just 35 and a commensurate reduction in the cost per Watt of 50%.
By the late 1990s there was little doubt that a modular approach to power design had been adopted by most engineering companies. Its benefits were far reaching, thanks to its constant evolution. However, it was also becoming more apparent that the applications were also evolving. Power density, which for so long had been crucial in power supply design was, it transpired, a double-edged sword. Integrated circuits were becoming more densely populated with transistors, switching at ever higher rates. This had the effect of increasing the junction temperature of integrated transistors, which had no effective way of dissipating that heat.
The semiconductor industry was in danger of going into meltdown – quite literally. The solution was to prove revolutionary; by lowering the supply voltage of certain integrated circuits used around a system, developers were able to reduce the active power of end applications without sacrificing performance. This was to have an impact on the power supply industry, however, which now needed to provide much greater flexibility in terms of how power was supplied. With lower supply voltages – in some cases sub-1 V – delivering power in the right way became more challenging.
While the vast range of brick-based products could accommodate most requirements – by the end of the ’90s Vicor was offering more than 8000 standard modules – it was still desirous to develop an optimised system that could meet the increasingly diverse power requirements of modern electronic systems. Optimisation and flexibility are normally contradictory parameters, but through another innovative development Vicor changed the direction of the power supply industry by introducing the factorised power architecture.
Factorised power architecture
With the new millennium came a new approach to power distribution, created by the need for greater power density and flexibility. In 2003, Vicor introduced the factorised power architecture (FPA) which took the concept of distributed power and extended it beyond the limitations of distributed power architecture (DPA). Based on ZCS/ZVS as used in the early bricks, FPA benefits from developments in integrated circuit design and is empowered by three components, collectively referred to as VI chips. The development of FPA and VI chips was not only significant; it was described as revolutionary in power supply design.
VI Chips are the key strength of FPA; each of the three components work independently to enable optimally designed power supplies, without sacrificing any of the flexibility needed to meet the design requirements of modern systems. In essence, power systems built using FPA and a combination of VI chips deliver multi-MHz fixed frequency switching with a power density measured in kW/in³. This is achieved through combinations of the intermediate bus converter, known as the bus converter module (BCM), a pre-regulation module (PRM) and a voltage transformation module (VTM). Together these elements overcome the limitations of non-isolated point-of-load (niPOL) regulators when used with an intermediate bus voltage of around 48 V, in systems that require VCC as low as 0,8 V.
Using FPA, a 48 V intermediate bus voltage can be transformed at the processor to 1,2 V or below, something niPOL regulators struggle to achieve due to the duty cycle of the FETs used in their construction. This means that modern systems which require multiple VCC, voltages can be constructed using the VI chips and FPA approach much more efficiently than systems using a contemporary intermediate bus voltage approach. Because FPA gives the engineer much more scope to design systems that use only the power required, at the point of load, it offers a much more efficient power solution.
Thanks to the FPA’s topology, only the VTM needs to be located at the point of load, which reduces the losses inherent with high-current PCB traces. The VTM effectively provides 48 V to 1,2 V conversion at the processor, saving significant board space in point-of-load conversion. A single VTM can deliver between 0,8 V and 55 V at the point of load – at up to 100 A – and accepts a regulated input from a PRM of between 26 V and 55 V. That equates to a power density of more than 60 W/cm³.
Depending on the diversity of output voltages required and power to be delivered, a power system could be constructed from a single PRM and VTM. For systems where more voltage levels or power are needed, VTMs and PRMs can be connected in parallel, to provide greater power output or a level of redundancy. In fact, VTMs have been designed to inherently operate in parallel, which avoids the need for a power sharing protocol or interface signals, as well as remote sense connections.