Choosing the Right Topology for Energy-Efficient Power Supplies

Various initiatives around the world to reduce the energy consumption of electronic systems are driving single-phase AC input power supply designers to adopt more advanced power supply technologies. To achieve higher power levels, these initiatives require efficiencies of 87% and above. Since traditional power supply topologies such as standard flyback and two-switch forward do not support these high efficiency levels, they are gradually being replaced by soft-switching resonant and quasi-resonant topologies.

How it works

Figure 1 shows the voltage and current waveforms of switches using three different topologies: a quasi-resonant flyback, an LLC resonant, and an asymmetric half-bridge using soft-switching techniques.

Figure 1. Comparison of quasi-resonant, LLC, and asymmetric half-bridge topologies.

The output diode current drops to zero

Ramp changes when the primary is coupled back to the secondary

The body diode conducts until the MOSFET turns on

These three topologies use different techniques to reduce the turn-on loss of MOSFET. The calculation formula for the conduction loss is as follows:

In this formula, ID is the leakage current just after turn-on, VDS is the voltage on the switch, COSSeff is the equivalent output capacitance value (including stray capacitance effects), tON is the turn-on time, and fSW is the switching frequency.

As shown in Figure 1, the drain current of the MOSFET in the quasi-resonant topology is zero when it is first turned on. Because this converter operates in discontinuous conduction mode, the switching losses are determined by the voltage at turn-on and the switching frequency. The quasi-resonant converter turns on when the drain voltage is minimum, thereby reducing switching losses. This means that the switching frequency is not constant: at light loads, the first minimum drain voltage comes earlier. Previous designs always turn on at the first minimum, and the efficiency at light loads decreases as the switching frequency increases, offsetting the advantage of the lower turn-on voltage. In Fairchild Semiconductor's e-Series quasi-resonant power switches, the controller only needs to wait for a minimum time (thus setting an upper frequency limit) and then turn on the MOSFET at the next minimum.

Other topologies use zero voltage switching technology. In this case, the voltage VDS in the above formula will be reduced from the bus voltage of about 400V to about 1V, which effectively eliminates the turn-on switching loss. Zero voltage switching can be achieved by allowing the current to flow through the MOSFET in the reverse direction through the body diode and then turning on the MOSFET. The voltage drop of the diode is generally about 1V.

The resonant converter achieves zero voltage switching by generating a sinusoidal current waveform that lags behind the phase of the voltage waveform, which requires a square wave voltage to be loaded on the resonant network. The fundamental frequency component of the voltage causes the sinusoidal current to flow (higher order components are generally negligible). Through resonance, the current lags behind the voltage, thus achieving zero voltage switching. The output of the resonant network is rectified to provide a DC output voltage. The most common resonant network consists of a transformer with a special magnetizing inductance, an additional inductor and a capacitor, hence the name LLC.

The asymmetric half-bridge converter achieves zero voltage switching through soft switching technology. Here, the voltage generated by the bridge is a rectangular wave with a duty cycle far below 50%. Before this voltage is applied to the transformer, a coupling capacitor is required to eliminate the DC component, which also serves as an additional energy storage unit. When both MOSFETs are turned off, the energy in the leakage inductance of the transformer causes the voltage polarity of the half-bridge to reverse. This voltage swing is ultimately clamped by the body diode of the associated MOSFET, which suddenly appears in the primary current.

Selection criteria

These energy optimization efforts result in excellent efficiency. For a 75W/24V supply, the quasi-resonant converter design can achieve efficiencies of over 88%. With synchronous rectification (plus additional analog controllers and a PFC front end), it is possible to increase efficiency to over 90% at 90W/19V. At this power level, while LLC resonant and asymmetric half-bridge converters can achieve higher efficiencies, the high implementation costs of both solutions have led to the widespread use of quasi-resonant converters in this power range. The e-Series family of integrated power switches is effective for applications ranging from 1W auxiliary power supplies to 30W set-top box power supplies and even 50W industrial power supplies. Above this power level, the FAN6300 quasi-resonant controller with external MOSFETs is recommended, which provides additional flexibility in handling very high system input voltages and helps optimize cost-performance due to the wide range of external MOSFET options.

The quasi-resonant flyback topology uses one low-side MOSFET; the other two topologies require two MOSFETs in a half-bridge configuration. Therefore, at lower power levels, the quasi-resonant flyback is the most cost-effective topology. At higher power levels, the size of the transformer increases, and efficiency and power density decrease, and two zero-voltage switching topologies are often considered.

System design is affected by four factors: input voltage range, output voltage, ease of synchronous rectification, and leakage inductance implementation.

Figure 2 compares the gain curves of the two topologies. For illustration purposes, we assume that the input voltages to be supported are 110V and 220V. For the asymmetric half-bridge topology, this is not a problem. Under the operating conditions we assume, the gain is 0.2 at 220V and 0.4 at 110V. At 220V, the efficiency is lower because the magnetizing DC current increases as the duty cycle decreases. For the LLC resonant converter, the maximum gain is 1.2, and it is important to note that the full-load curve is very close to resonance. A gain of 0.6 would result in very high frequencies and poor system performance. In summary, LLC converters are not suitable for a wide operating range. By externally adjusting the leakage inductance, LLC converters can be used for the European input range, but at the expense of a higher magnetizing current; they work best if a PFC front end is used. The asymmetric half-bridge configuration has a PFC stage at the input, so the circuit can operate over a wide input voltage range.

Figure 2. Gain curves for asymmetric half-bridge and LLC converters

For output voltages above 24V, we recommend an LLC resonant converter. The high output diode voltage will reduce the efficiency of the asymmetric half-bridge converter because the forward voltage drop of the higher rated voltage diode is also higher. Below 24V, the asymmetric half-bridge converter is a good choice because the output capacitor ripple current of the LLC converter is much larger, which increases as the output voltage decreases, thereby increasing the cost and size of the solution.

Both of the above topologies can use synchronous rectification. For the asymmetric half-bridge topology, this is very simple to implement (see Fairchild Semiconductor Application Note AN-4153). For the LLC controller, a special analog circuit is required to sense the current flowing into the MOSFET, but the technique is relatively simple if the switching frequency is limited to the second resonant frequency (100kHz in Figure 2).

Finally, both designs rely on the leakage inductance of the transformer: in the LLC converter to control the gain curve (Figure 2), and in the asymmetric half-bridge converter to ensure soft switching at light loads. For most applications, we recommend using two separate inductors to achieve this. Leakage inductance is a parameter that is not easily controlled in the transformer. In addition, to achieve an unusual leakage inductance, a non-standard bobbin is required, which increases the cost. For the asymmetric half-bridge structure, the resonant switching speed is at least 10 times the switching frequency if a standard transformer is used, resulting in greater losses. In summary, for the LLC converter, it is recommended to use an additional ordinary ferrite inductor, while for the asymmetric half-bridge converter, it is recommended to use only a high-frequency ferrite inductor.

The circuit diagram of an asymmetric half-bridge converter is shown in Figure 3. The diagram is very similar to the LLC resonant converter, with one difference: the LLC resonant converter does not require an output inductor, and the asymmetric half-bridge controller requires frequency setting instead of PWM control.

Figure 3. Asymmetric half-bridge converter based on FSFA2100

The 192W/24V asymmetric half-bridge converter has an efficiency of around 93%. The AN-4153 360W/12V current doubler also has a full load efficiency of over 93% at 20%-100% of rated load.

Under the condition of 200W/48V power supply including PFC front end, the efficiency of LLC resonant converter is about 93%. Through synchronous rectification, the efficiency can be improved to 95%-96% at this power level.