New power architecture addresses PFC load efficiency issues

Thomas Lawson, Cognipower


Compound Power Converter architecture promises flexible, efficient power factor correction

Click image to enlarge

Figure 1: Block Diagram of an Isolated AC/DC Compound Converter

As Power Factor Correction (PFC) migrates to lower power AC/DC converters, maintaining efficiency at low loads becomes an increasing challenge. The approach usually taken involves adding an extra boost conversion stage feeding a second buck conversion stage. The boost stage performs PFC at the input while the buck stage performs regulation at the output. An active bridge front end, sometimes called a "Bridgeless" system, is a more efficient variant on this approach, because one series diode is eliminated from the input path. The PFC boost stage produces a relatively high DC voltage, which is stored in a bulk capacitor. The stored voltage serves as the input for a subsequent down-regulation stage. All the power moves through both stages of power conversion in series. Therefore, the efficiencies of the two stages multiply, i.e. 90% * 90% = 81%. The two-stage approach does not take full advantage of the flexibility implicit in a bridgeless input stage. Compound Converter architecture PFC and regulation functions are blended together for better efficiency in CogniPower's approach to address the issue, Compound Converter architecture. The key difference in a Compound Converter is the optimization of the placement and use of the storage element. The storage could be moved to the output, by providing extra capacitance at the output filter, but the result would be to maximize the amount of capacitance required for good regulation. That would also maximize the size and cost of the filtration. A more efficient approach is to provide separate storage to support the output during the zero crossing periods of the AC input, but not to route all the power that is converted into, and out of, the storage element. The two stages in a Compound Converter operate in parallel, not in series. The result is that the majority of power moves through a single stage of power conversion. That difference enables a substantial efficiency gain. When a PFC boost conversion stage is running with a near constant ON time, the Power Factor will be near ideal. At a particular fixed ON time, the correct amount of power will be transferred from input to output during each single AC cycle. However, during parts of the cycle, there will be excess energy available, and during other parts, there will be a deficit of available energy. In a compound converter, the storage capacitor both accepts the surplus and provides the shortfall, as appropriate, but is not involved with the transfer of the majority of energy from input to output. During periods of sufficient AC energy availability, the load is serviced directly, with any excess energy going into storage. At times in the AC cycle when not enough energy is available to supply the load, supplemental energy is provided from storage. In that fashion, most of the energy transferred can move from input to output through only a single stage of power conversion. Efficiency rises. The storage voltage varies at twice the frequency of the AC line, and is regulated only on the average. The loop that controls the storage voltage should be slow compared to the line frequency to obtain the best PFC behavior. Because the storage voltage can vary over a wide range, a much smaller capacitance is required to support a given load than when the capacitance is placed at the output. Also, the storage voltage can be lower, eliminating the need for an expensive, high-voltage capacitor for storage. In a Compound Converter, 2/3s of the power might move through a single stage at 90% efficiency, and 1/3 might move through both stages at 81% efficiency, for an overall efficiency increase from 81% to 87%. When the output voltage is allowed to droop a little more during supplemental converter operation, even higher efficiencies gains can be achieved. In addition, all the usual steps known to further increase the basic efficiency of either the main or supplemental stage can be applied here, as well. Because a larger percentage of the power moves through only a single stage of power conversion at lighter loads, Compound Converter architecture makes it easier to meet efficiency standards at light loads. There are advantages when heavily loaded, also. In a conventional two-stage converter, both power stages must be built to withstand the maximum power levels. In a compound converter, the main and supplemental stages share the maximum load. The main power path can be optimized for efficiency at average power levels, while the supplemental path can supply surge currents to satisfy transient loads, improving both efficiency and regulation. The hardware and control requirements to create such a system are surprisingly simple. The component count can actually drop in comparison to the normal two-stage approach. Regulation in a compound converter is better because of the presence of a local, supplemental supply of energy at a convenient voltage. The supplemental conversion stage can run at a higher frequency than the main converter for even better regulation and lower output ripple. Wide application area Compound architecture can be applied to a wide variety of converter topologies including forward converters, zero voltage switching converters, DC to DC, AC to AC, multiple output converters, and even power amplifiers. Figure 1 shows a Compound Converter in block form. The left-hand, AC-side of the figure is a bridgeless boost stage acting on the dual primaries of the transformer. Alternatively, a single-primary transformer could do the same job if connected in a full bridge. The topology shown in Figure 1 minimizes the number of series diodes or switches at the expense of an additional transformer winding. The AC-side switch intelligence is located on the right-hand, isolated side. A single bit of data representing ON or OFF is sent across a simple digital isolation barrier. Switch 1 or 2 energizes one of the transformer windings based on the polarity of the AC input voltage. Switch 3, operating as a synchronous rectifier, completes the main flyback function. With the addition of a control input to disable Switch 3, that switch does more than just synchronously rectify the main flyback stage. It also provides local regulation. Switch 3 is opened when the load is satisfied, allowing surplus inductive energy to be transferred to the storage capacitor through the auxiliary flyback diode. The remaining circuitry is an auxiliary buck converter to move energy from the storage capacitor to the output. The auxiliary control circuitry will cause the supplemental converter to support the load during zero crossing, or during interruptions in the AC input. If the supplemental converter regulates to a slightly lower voltage, it will only act after the main regulator begins to drop out. That effect increases efficiency. The lower the load presented, the greater the improvement in efficiency. The supplemental regulator can operate out of phase with the main regulator to reduce ripple. If at any time a transient load should disturb the output, the supplementary converter can act immediately to help restore regulation, reducing the peak requirements for the main stage.

Control is simpler than one might think The main converter can operate in discontinuous or continuous mode. The basic controls involve three simple loops. Loop 1, Synchronous rectifier: Switch 3 closed if the load requires energy when inductive current would flow in the desired direction. Loop 2, Supplemental buck: Close Switch 4 whenever the output is below the supplemental reference voltage Loop 3, Constant ON time for transformer primary switches: Heavily filtered slow loop to keep the storage voltage within broad limits. The constant ON time will not change significantly during a single AC cycle. A maximum ON time prevents excess current from flowing during transients or during start-up. Figure 2 shows a simulation of a 100 Watt Compound Converter running from 120 VAC at 60 Hertz. The output is here 20 VDC, though a wide range of output voltages is practical. The storage voltage varies at twice the line frequency. In this case, the nominal storage voltage is 40 volts. Higher storage voltages increase the holdover time for a given value storage capacitance, but require higher voltage components. The minimum storage capacity should support the maximum load during one AC zero crossing. More storage will support a missing AC cycle, or longer periods of AC drop-out. With substantial storage available, the Compound Converter becomes a Uninterruptible Power Supply.

Figure 3 shows simulations of the converter running at 50% load. The top trace is a single cycle of the AC line voltage. The second trace shows the current supplied to the load through the main flyback path. The third trace shows the current supplied to the load through the supplemental buck path. The bottom trace shows current flowing into the storage capacitor during the period of excess energy availability. Note that on average, the power going into storage will nearly equal the power being supplied by the supplemental buck converter. Other Considerations Because control of the switches on the AC side is provided by circuitry on the DC side, a chicken-and-egg situation must be avoided at start-up. There are several ways to handle that issue. One is to add a separate, bootstrap, self-oscillating supply that has two secondary windings. One winding provides power for efficient switch drive on the AC side, the other powers the Control block on the DC side. CogniPower has a simple, efficient design for such a bootstrap supply. Another possibility is to add start-up circuitry on the AC side which operates only briefly, and without feedback when power comes up, to move enough energy through the transformer to allow the Control block to begin to function. As with other types of power converters, further light load efficiency gains can be achieved by reducing the underlying frequency of operation. When synchronous behavior is not required, regulation and good PFC can also be achieved with a completely fixed ON time and a slowly changing frequency of operation. Cognipower