The most obvious is to connect batteries as a constantly-connected back-up power source, permitting the battery to be charged from a power bus when offline power is available, and to feed power from the battery to the same bus when it is not. Conventionally, in the traditional uninterruptible power supply, stored energy is used to generate an AC waveform to replace the failed line supply. There are attractions to configuring an equivalent function within the DC distribution level of a power network.
As with the design of all power conversion systems, the architecture of such designs is defined by the location of major functions such as voltage level changing, isolation and regulation. In most of the portable devices we carry, the power bus is effectively the battery potential and is allowed to occupy a range of voltage levels, between that necessary to fully charge the battery, and the minimum the battery can deliver without internal damage. The charging supply and the functional part of the product have a common connection – with refinements – to the battery. Regulation becomes, in effect, part of the application’s function, enabling it to run from any voltage in the permitted range.
For higher power systems, such as larger-scale battery back-up installations or uninterruptible power supplies implemented in DC, this configuration is not satisfactory. There, the requirement is for a constant voltage power bus that is stable at the same level when fed from an offline power supply – powering both client systems and charging the battery – and when the battery comes on-line. Between power bus and battery, therefore, there must be both level-changing and regulation, and, preferably, isolation.
Regulation – maintaining a constant bus voltage over the range of a fully-charged to exhausted battery: or holding the correct level to accurately charge a battery – can be handled by a buck-boost DC/DC converter, and with today’s best zero-voltage-switching architectures, such devices can perform that function with high efficiency. Higher-power battery backup installations will typically operate with a much higher voltage at the systems bus than at the battery, so a large step-up/step down function is indicated. This provides a role for a recently-introduced (by Vicor) functional block termed the BCM or bus converter module.
The BCM was launched largely in the context of power-distribution architectures for, typically, telecommunications or IT server facilities. It provides an isolated DC/DC conversion function that provides a fixed input-to-output voltage ratio. Making a comparison to the role of the transformer in handling AC is irresistible; in almost every respect, the device can be regarded as a “DC transformer”. The BCM, among its other attributes, is bi-directional: power can flow from input to output or vice-versa, with equal efficiencies – depending on the circuit configuration, the terms ‘input’ and ‘output’ begin to change their definitions.
In a telecommunications or server rack, power flow through such a module is likely to be resolutely uni-directional; the reversible capability is relegated to interesting footnote. (A typical role might be to step down from a 380V DC bus to the classic telecoms 48V distribution bus level.) However, in a battery backed power system, that feature comes into its own.
The BCM employs the circuit configuration that Vicor calls Sine Amplitude Conversion. A simplified circuit is shown in Figure 1 – this depicts the power paths through the circuit: there will also be control and gate-drive, and other supervisory, functions that are not shown.
Figure 1. Power paths through Vicor’s BCM fixed-ratio conversion design.
The topology is transformer-based series-resonant, and operation depends on maintaining oscillation at the resonant frequency of the tank on the primary side of the power transformer windings, which is labelled T1 in the figure. The control architecture locks the operating frequency to the tank resonant frequency and the four MOSFETs in the H-bridge are synchronously switched so as to sustain a sinusoidal oscillation. The FETs are switched at the zero-crossing points of the sinusoidal waveform, virtually eliminating switching transients, the losses that inevitably accompany a “hard” switching cycle and reducing the generation of high order
noise harmonics. The current in the primary resonant tank has a sinusoidal shape, and this also contributes in reducing spectral content and provides a cleaner noise signature.
The output load, when translated to the primary side, appears in series to the LC tank (series resonant). An increase in output load therefore causes an increase in amplitude of the sinusoidal current in the tank, hence the name Sine Amplitude Converter. The increase in current amplitude increases the amount of energy coupled into the secondary, countering the load demand. When the load is reduced, the amplitude of the sinusoid decreases, approaching zero under ‘no load’ conditions.
As the amount of energy is only associated with the current amplitude within a small switching period, large, reactive storage elements are not required for this type of topology. The transformer characteristics can be optimised, allowing for a size reduction of the transformer itself, and for high frequency operation, which is also made possible due to the reduced switching power losses. This translates overall into both high power density and efficiency.
The secondary circuit, which could be based on a centre-tapped configuration, as in Figure 1, or a full-bridge of four FETs and a single transformer winding mirroring the primary side, operates as a synchronous rectifier, converting the sinusoidal current to a rectified-AC waveform at twice the switching frequency; the output capacitance carries the vast majority of this high frequency current, which is used to maintain an output voltage strictly proportional to the input voltage.
All of which leads to the question; how does the circuit operate when power flows in the reverse direction?
next; reverse power flow...