A new type of power-conversion topologies known as "sine amplitude converters" is the driving force behind the high performance achieved by Vicor's line of voltage transformation modules (VTMs). The BGA-packaged VTMs are integral dc-dc conversion blocks within the Factorized Power Architecture (FPA), the company's new approach to distributed power. Though the capabilities of these components and architecture were described at length earlier this year *("Distributed Power: Novel Architecture Yields New DC-DC Building Blocks," *Electronic Design*, April 28, p. 40-42)*, the details of the unique underlying technology are only now being made public.

Sine amplitude converters (SACs) enable VTMs to attain very high power density of about 1 kW/in.^{3 }at the point of load, high efficiency, and fast transient response. (Existing pulse-width modulation (PWM) converters and zero-current/zero-voltage (ZC/ZV) resonant converters achieve power densities of around 100 W/in.^{3}) This is all possible due to a unique design that combines the best attributes of ZC/ZV switching resonant converters and hard-switching PWM converters. SACs owe much of their performance to their high switching frequency, which lets them use very small magnetics.

SACs switch at frequencies in the megahertz range. In the current VTMs, the switching frequency is 3.5 MHz. This is well above the operating frequency of classic ZC/ZV switching resonant converters and PWM converters, as deployed in isolated dc-dc converter bricks.

While some nonisolated buck converters interleave multiple phases to achieve high effective switching frequencies, SACs are single phase. So, the switching frequency isn't an *effective* frequency, it's *real*.

Comprehending how SACs achieve their high-frequency performance requires an understanding of the limitations of existing designs. ZC/ZV switching resonant converters like the forward-type dc-dc converters in Vicor's second-generation product series can switch at frequencies up to 1 MHz. These converters work by storing energy in the power transformer's leakage inductance.

In classic resonant topologies, energy is intentionally stored in the transformer of the resonant converter, and the Q of the energy storage elementsâ€”specifically the transformer's leakage inductanceâ€”must be made high. However, the transformer's high Q and leakage inductance are obtained at the expense of resistive losses in the transformer's windings.

In PWM topologies, like those typically applied in non-resonant isolated dc-dc converters, the leakage inductance in the power transformer is minimized. Therefore, resistive losses in the transformer aren't the limiting factor for operating frequency. But in this case, the switching losses associated with PWM hard switching generally limit switching frequencies to values in the 100- to 400-kHz range.

SACs overcome the resistive losses of resonant converters and the switching losses and switching frequency limitations of PWM converters. Like the classic resonant converters, SACs use ZC/ZV switching, minimizing frequency-dependent switching losses. Yet like the PWM converters, SACs minimize the transformer's leakage inductance. That allows them to use a more-efficient power transformer capable of operating at a multi-megahertz frequency.

The shift from maximizing to minimizing leakage inductance is a fundamental change. Instead of maximizing energy storage in the transformer, the SACs minimize energy storage. With this change, the principle for processing energy changes.

The classic ZC/ZV resonant converter works on the principle of "quantum energy transfer," whereby a fixed amount of energy is transferred to the load in each switching cycle. The output power is proportional to the switching frequency times the energy stored in the converter's tank circuit. To a first-order approximation, P_{OUT} is proportional to f times Li^{2}, where P_{OUT} is the power output of the converter, f is the switching frequency, L is the transformer's leakage inductance, and i is the current through the transformer.

To increase the power delivered to the load, the resonant converter raises its switching frequency. In contrast, the SAC works by "charge transfer." The amount of charge processed by the converter per switching cycle grows as a function of output power. The SAC creates a low-voltage sinusoidal oscillation by resonating the small amount of leakage inductance present in the power transformer with a primary-side capacitor *(Fig. 1)*. The amplitude of this oscillation varies as a function of the current drawn by the load.

In other words, as the load current rises, the amplitude of the oscillation (as measured across the resonant capacitor) increases, while its frequency remains fixed. When no current is drawn by the load, the amplitude drops to zero.

A VTM with 12-V output provides an example of the SACs' operation *(Fig. 2)*. Here, the VTM is operated open loop with no output-voltage sensing and regulation, which a preregulator module (Vicor's PRM or other dc-dc converter) would otherwise provide. In the example, the VTM exhibits 1% load regulation when exposed to load step changes from 100 to 200 W. But thanks to the low Q, the output waveform and the resonant capacitor's internal waveform settle quickly with no ringing. If the VTM were operated closed loop with a PRM, load regulation would improve to 0.1%.

In some VTMs, the SAC is implemented with full-bridge configurations on both the primary- and secondary-side circuits *(Fig. 1, again)*. But it also is possible to build SACs using half-bridge and push-pull configurations, depending on application requirements.

The SAC's fixed-frequency operation distinguishes it from the classic ZC/ZV resonant converter. Another distinguishing characteristic is its elimination of serial (intermediate) storage elements, which slow the transfer of energy (or charge) to the load. In Vicor's forward converter with ZC/ZV switching, energy stored in the transformer's leakage inductance is transferred to a secondary-side resonant capacitor, then fed through a smoothing inductor to the load.

The capacitor and inductor are serial energy storage elements that take up space within a brick and slow down the brick's transient response. In the SAC, the serial energy storage components aren't required because the converter works by transferring charge, rather than by storing and transferring packets of energy.

Serial energy storage also applies to PWM converters. Even in a forward-type PWM converter without an output inductor, the leakage inductance of the power transformer appears as a reactive term in the output impedance of the converter. This reactance coupled with the dc resistance in the circuit produces voltage droop on the output when the load changes.

With the SAC, the reactance associated with the power transformer's leakage inductance is cancelled out by locking the switching frequency to the resonant frequency set by the primary-side resonant capacitor. So, whatever small amount of serial output reactance would be contributed by the transformer's leakage inductance is eliminated. In a 1.5-V output VTM, the open-loop output impedance is just 1.3 m(omega), flat from 0 to 1 MHz.

This impedance characteristic leads to a fast transient response of about 200 ns. That's better than what's achieved by multiphase buck converters seen in the voltage-regulator modules that generate CPU core voltages. But the SAC holds another unique advantage: A bidirectional power train lets it handle load dumps and surges without requiring voltage clamps of any kind. If a CPU produces a load dump, the SAC will recycle that energy from the output back to its input voltage source. At the same time, the SAC will provide effective bypass-capacitance multiplication in proportion to the square of the input-to-output voltage ratio.