Jon Harper looks at how equipment makers can bring the efficiency advantages and lower EMI of quasi-resonant power conversion to lower power systems.
Well-established quasi-resonant conversion technology has been extensively used in power supplies in the consumer arena, with flyback and buck converters improving power supply efficiency.
The principle of quasi-resonant conversion is to reduce the turn on losses of the power switch in a topology. A resonant converter(1) minimises the turn on losses and works in a very different way. One way of explaining quasi-resonant operation is to consider it as an extension of discontinuous conduction mode operation.
Fig. 1 shows the drain waveforms in a current mode flyback converter operating in discontinuous conduction mode. Only one gate pulse has been applied. During the first time interval, the drain current ramps up until the desired current level is reached. The power switch is then turned off. The leakage inductance in the flyback transformer rings with the node capacitance. This causes the leakage inductance spike which is limited by a clamp circuit. After the inductive spike has diminished, the drain voltage returns to the input voltage plus the reflected output voltage. When the current in the output diode drops to zero, the drain voltage would immediately drop to the bus voltage, if the effect of the primary inductance and the node capacitance were ignored. However, the drain voltage rings down to this level as shown in the figure.
The primary inductance and node capacitance form a resonant circuit. Taking a value of 1.4mH for the inductance and 73pF for the node capacitance gives a resonant frequency of 500kHz using the equation 4[g16]2f2LC = 1. The resonant circuit is lightly damped. We note that the resonant frequency using this approximation is independent of the input voltage and load currents.
In the case of a discontinuous conduction mode flyback converter, the MOSFET is turned on at a fixed frequency (ignoring the effect of any frequency jitter). The device is turned on, turned off when the set current level is reached, and then turned on again at a fixed time after the previous device turn on. The device turn on time is not synchronised with the drain resonance. In some cases the device may turn on when the drain voltage is lower than the bus voltage plus the reflected output voltage, and in some cases the device will turn on when the drain voltage is higher. This characteristic is often seen on the efficiency curves of discontinuous flyback converters: when driving a constant load, the efficiency will vary with the input voltage as the device turn on time moves up and down the valleys and troughs of the resonance curve.
For quasi-resonant switching, the device does not have a fixed switching frequency. Instead, the controller waits for one of the troughs in the drain voltage and then switches on. Older quasi-resonant devices designed for the colour television market always switched on the first trough. This was a good solution for colour televisions where the load is always high. However, for loads with a wide dynamic range, this presents a problem.
The time between device turn off and the first trough is fixed by the resonant frequency. The time between device turn on and turn off is set by the controller. For lighter loads, the time is smaller as less energy in required in the inductor, resulting in a shorter on time, and also a shorter output diode conduction time. So for lighter loads the frequency increases, resulting in much higher switching losses.
On the FSQ series of Fairchild Power Switches this problem is avoided by using a frequency clamp circuit. This circuit ensures that a maximum frequency is not exceeded while at the same time switching on one of the troughs.
In comparison with both discontinuous mode and continuous mode operation of a flyback converter, quasi-resonant switching offers reduced turn on losses, resulting in increased efficiency and lower device temperature. The disadvantage of higher losses at light loads for a simple quasi-resonant circuit is removed by the frequency clamp circuit used in modern controllers or integrated power switches.
The EMI generated by the turn on process is reduced if this happens at low current and lower voltage, which is the case in quasi-resonant applications. This reduces EMI in the 1MHz to 50MHz range.
Further, there is an intrinsic frequency jitter in the quasi resonant process which spreads out the EMI noise, which further reduces filter costs. This is caused by the input voltage ripple on the bulk capacitor. Both the on time and the output diode conduction time are less at the maximum ripple voltage than at the minimum ripple voltage, for a constant load. This results in a linearly changing switching frequency with a frequency sweep equal to the ripple frequency (eg 100Hz for a full bridge rectified circuit operating from 50Hz AC). This reduces EMI in the switching frequency range 150kHz to 1MHz.
The range of quasi-resonant converters has recently been extended down to cover lower power levels (2W-50W) in response to the market demands for higher efficiency of low wattage power supplies and lower standby power.
Fig. 2 shows an example application designed in Fairchild's Global Power Resource Centre in Germany. R103-5, D104 and C103 form the extra components needed for the detection of the minimum voltage levels on the drain. It uses an error amplifier that combines the functions of a standard optocoupler and an industry standard '431 reference in one package. The remaining components are standard for a flyback converter.
The no-load standby power was measured to be less than 130mW in the 175VAC to 265VAC range used for the design. At lower input voltages the standby power is even lower for similar designs. The full load efficiency was greater than 86 per cent for the whole voltage range, which is very high for a multiple output flyback power supply at this power level. Line regulation was excellent - the measured voltages did not really change when the input voltage was modified. Load regulation for the regulated output was well within 5 per cent.
The availability of low power quasi-resonant devices has opened up interesting new possibilities. The Global Power Resource Centre has developed a quasi-resonant buck circuit for a 20V/100mA output. Here the standby power with a 10mA load was measured. The total power consumption was less than 400mW (including the 200mW base load) over the full 85VAC to 265VAC range, and less than 350mW for the range up to 180VAC. The better performance at low input voltage is due to the troughs of the drain voltage being far closer to zero than at higher input voltage.
For 85VAC to 160VAC the full load efficiency was greater than 80 per cent, dropping to 73 per cent for the range up to 265VAC. This is excellent performance for such a small power supply, entirely attributable to the use of quasi-resonant techniques. Both line and load regulation were well within 1 per cent over the full operating range. The temperature rise of the device without heatsink was measured to be 15 degrees at room temperature.
Jon Harper is Market Development Manager, Industrial & White Goods Systems for Fairchild Semiconductor Europe. www.fairchildsemi.com
1. R. W. Erickson and D. Maksimovic, Fundamentals of Power Electronics, Second Edition, Springer, 2001, Chapter 19, ISBN 0-7923-7270-0.