Switched-mode power supply (SMPS) topologies have led to advances in efficiency, thanks to continual tweaks and optimisations to circuit design, says Martin Brabham, Stadium Stontronics
The first step in increasing efficiency in power supply design was the adoption of the pulse width modulation (PWM) techniques.
In a half-bridge arrangement with high and low-side MOSFET switches, the PWM converter supplies a regulated output voltage by controlling the on-time of the two transistors. It turns on the high-side switch. During this on-time, energy is transferred to the output via a transformer, and the output voltage swings upward and current starts to flow in the output inductor. When sufficient energy has passed to the inductor, the control logic turns off the high-side transistor, and the inductor current begins to fall. After a delay (which also prevents shoot-through between the two transistors) the low-side switch half-bridge turns on and current in the output inductor starts to increase once more, and so on.
The PWM controller samples the output voltage and compares this to a reference voltage to generate an error signal. An error to the low side indicates that a longer on-time, and therefore a higher duty cycle, is needed, and vice versa. The signal is usually compared with the output of a ramp oscillator, and the result is used to control the duty cycle for the two phases. The continuous adjustment that PWM makes compensates for swings on the input rail and variation in demand from the load. Thereby providing a stable supply of energy to the output load.
The key issue with basic PWM is hard switching – the tactic of turning off the devices while current and voltage are above zero. It can lead to large losses while devices are in a transitional state. The high dV/dt and dI/dt levels encountered during hard switching also lead to considerable EMI. One of the biggest contributions to the surge in efficiency has been the adoption of architectures that make use of resonance. Reducing current or voltage to zero, or as close as possible, before activating or deactivating the power path, reduces the losses of hard switching. Since its development in the 1980s, the main way of implementing this is to move to a resonant architecture.
Schematic of ZCS topology
This architecture takes advantage of reactive passive components in the switching path. Combinations of capacitors and inductors result in frequency resonances that cause the voltage and current to oscillate around a zero point. The simple forms of these are the quasi-resonant types: in this strategy the voltage across the primary winding is monitored and the switching of the MOSFET is synchronised to a point where current and voltage is at its minimum. This reduces the turn-on switching loss but has little effect on the turn-off.
One strategy was to combine PWM switching with resonant tank circuits, or soft-switching converters. The strategy is dominated by the PWM switching cycle but the voltage and current waveforms are controlled in such a way either is pushed towards zero just before the power transistor switches state. One form is zero-voltage switching (ZVS); the other is zero-current switching (ZCS). Typical ZCS strategies send the current to zero at turn-on and reduce the current before turning the transistor back off. The design still incurs losses at turn-on, particularly in MOSFETs because of the capacitive effects in the transistor’s body diode.
Another problem associated with ZCS switching is when the transistor starts to turn off or on. At this point, a high rate of change of voltage can be coupled to the gate drive circuit through the devices Miller capacitance, which can result in the device turning on or off at a slower rate, or even bouncing. This is not the case when ZCS is used in larger power supply designs based on IGBTs. In this application, the power transistors tend to suffer from a high tail current when switched off. This is generally due to the high levels of doping needed to build high-power IGBTs. Doping introduces a large number of minority carriers that need to be removed when the device is switched off, which tends to increase the switching time if high current levels are involved. The same effect also tends to limit the maximum achievable frequency when using IGBTs, so the switching frequency remains relatively low, which also limits any reduction in size that can be achieved.
In fully resonant architectures the power supply works at close to the circuit resonant point and most controllers run with set duty but have variable a switching frequency as the method of controlling the throughput of energy.
Resonant ZVS eliminates the capacitive turn-on losses encountered with MOSFETs. The ZVS architecture is similar to a conventional PWM converter, adding an inductor effectively in parallel with the switch path. The energy stored in the inductor is used to swing the main switching voltage to zero just before turn-on. At turn-off, the energy is transferred to the inductor slowing the initial rate of voltage change so the transistor can be fully switched off before any voltage develops across it. One drawback is that the energy stored in the inductor is not transferred to the output. Instead it circulates from the resonant components back to the input tank.
The reduction in switching losses makes it possible to increase the switching frequency of the conversion circuit, moving from 10s of kHz to half a MHz or more. This can lead to a reduction in the size of transformer, inductors and capacitors, supporting a push towards more compact power supplies. However, pure resonant architectures tend to push high peak power levels though the output path and can have high circulating currents in the resonant components, which places a limit on size reduction, as well as leading to higher conduction losses through the active and passive devices.
Synchronous outputs are one of the other areas that have helped achieve better overall efficiencies. The output rectification is handled by MOSFETs synchronised to the PWM switching and turned on at the same point a diode would be conducting current. A small amount of dead-time is included, as any misalignment will lead to shoot-through and significant losses. The advantage of synchronous rectification is that the voltage drop across the devices when passing current is less than the voltage that would be present across a diode when conducting.
The reduction of losses in both the turn-on and turn-off stages has allowed switching frequency to increase. This, combined with reduction in overall losses, has resulted in reductions in power supply size for a given wattage. Since the development of the original quasi-resonant and ZVS, ZCS strategies, variants have been developed, for example, the phase-shifted topology in a full-bridge AC/DC converter yields supplies able to deliver 10kW of power at an efficiency of more than 95%.
Some of the trade-offs between topologies are easy to see, while others are more complex. Such subtleties can lead to large differences in efficiency, reliability and cost and make it worth consulting power-design experts on the best way forward.