In this section, a brief review of existing efficiency enhancement methods for DC-to-DC con- verters is discussed. The knowledge obtained from this section can serve as a basis for control design of the flyback converter in Chapter 8.
2.6.1
Variable switching frequency
An approach to improve the light load efficiency using a variable switching frequency was first proposed by Arbetter et al. [43, 50]. By investigating different loss types of a fixed frequency converter as shown in Fig. 2.18(a), Arbetter et al. [43, 50] believed that the switching losses were a main reason in causing a decline in the converter efficiency at light load, and that it was possible to reduce the switching losses by allowing a variable frequency operation as demonstrated in Fig. 2.18(b). Based on this observation, Arbetter et al. designed variable-frequency constant-peak- current control for a buck converter in DCM [43], and subsequently extended the control method to both CCM and DCM [50]. Since the switching frequency fpwm is considered as a manipulated
variable to regulate the output voltage, the technique proposed in [43, 50] is commonly known as pulse frequency modulation (PFM) control. The main feature of PFM is to reduce the operational frequency under light load conditions and consequently lower the switching losses. Although the variable-frequency method can greatly improve the converter efficiency, it also poses other major problems, such as poor output voltage regulation, EMI issue due to the dependence of fpwm on
the load, and many others.
Power loss Load current 0 Total Switching Conduction Fixed Power loss Load current 0 Total Switching Conduction Fixed
(a) Fixed switching frequency (b) Variable switching frequency
Figure 2.18: Converter loss vs. load current for different switching frequency schemes [43]
Another form of PFM control, known as constant on-time control, is commonly applied to buck converters [45, 46, 51, 52]. In such applications, the transistor on-time Ton is selected to maximize
the overall converter efficiency, while the switching frequency is linearly modulated by the output current iout(t) in order to keep a constant output voltage. Though the authors in [51] claim that
constant on-time control can avoid the instability and large current ripple issues of constant peak- current control, both the control approaches turn in quite similar performances. As an optimal value of Ton is only valid for a given input voltage [45, 53], the system efficiency could be further
2.6.2
Pulse-skipping and burst-mode control
Pulse-skipping and burst mode control can be considered as a special case of PFM control during ultra light load and no-load conditions. Instead of applying a wide pulse with PFM, a burst of smaller pulses is typically employed to compensate for the energy wasted in the snubber circuit and possibly the output load [54–56]. The comparison between burst mode control and PFM control is best explained in Fig. 2.19. As a consequence of short pulses, burst mode control typically results in a smaller peak magnetizing current when compared with that of PFM control for the same values of iout(t) and fpwm. The benefit of burst mode control is twofold. It not only
simplifies EMI filter design but also brings a great benefit to acoustic noise cancellation.
PWM Tpwm im(t) vout(t) Ton t t t 0 0 0 PWM Tpwm im(t) vout(t) Tburst t t t 0 0 0
(a) PFM (b) Burst mode
Figure 2.19: Comparison waveforms of the control signal, inductor current and output voltage under (a) PFM and (b) burst mode control
Though burst mode control can easily be achieved using a hysteresis approach [54], it does not lend itself to the flyback converter with magnetic sensing (MS) regulation. The main obstacle is the indirect measurement of the output voltage.
2.6.3
Quasi-resonant operation
Since the voltage across the drain and source of the MOSFET vds(t) is high at the end of the
switching cycle, the switch-node capacitance losses, according to Eq. (2.68), become substantial under light load situations. One way to minimize such losses is to turn on the MOSFET at the point of the minimum value of vds(t). Such a technique has been successfully applied to the flyback
converter operating in DCM [57], where the MOSFET is turned on at the first valley of vds(t),
as exemplified in Fig. 2.20. In practice, the second or higher valley number could be selected, depending on the design specifications.
Since the choice of first or second or higher valley number switching directly decides the amount of power dissipation, a proper assignment is necessary in order to maximize the system efficiency over the entire operating range. The control strategy using valley switching is usually referred to as quasi-resonant control [57]. Like burst mode control, quasi-resonant control is also helpful in
PWM Tpwm im(t) vds(t) Ton t t t 0 0 0 Vin Vin+Vz Ton 2Tpwm First valley
Figure 2.20: Example of quasi-resonant control with valley switching, where the transistor is turned on at the first valley of vds(t).
reducing the size and cost of the EMI filter.
2.6.4
Multi-mode (hybrid) operation
All efficiency improvement methods in Sections 2.6.1, 2.6.2 and 2.6.3 have shown good perfor- mance under certain loading conditions, but none can maintain a high efficiency over the entire working range. For example, a fixed switching frequency converter can easily offer a high efficiency at heavy load, but not at light load. On the other hand, the variable frequency approach in [43] can achieve a constant efficiency over a relatively wide load range, except at heavy load and very light load. Under these two extremes, the variable-frequency controller needs too small and too large switching frequencies which make the design of the EMI filter and the magnetic components more difficult. Burst mode control is applicable in ultra light load and no-load scenario, while quasi-resonant control is only useful in DCM.
The idea of multi-mode (hybrid) control is to select certain existing power-saving techniques and combine them together. Ideally, multi-mode control should inherit all the advantages of each individual method and hence can guarantee a high efficiency operation under any loading condition. Various schemes of multi-mode control have been reported in [46, 47, 51, 58, 59]. Though most of these studies focus on a synchronous buck converter only, they can be a useful reference for flyback converter design.
2.6.5
Offline efficiency optimization
Table-based efficiency optimization control, which was proposed recently by Kang et al. [8, 44, 60], is the only study considering efficiency optimization for the flyback converter. In fact, the approach in [8,44,60] can be considered as multi-mode control, consisting of fixed-frequency control for CCM and variable frequency, with quasi-resonant control, for DCM. Unlike the formulae-based approaches in Section 2.6.1, the optimum switching frequency fpwm in [44] is computed offline
and looked up from a table, rather than calculated directly from iout(t) and vin(t). This offline
optimization method has several advantages, such as requiring less computation power, offering higher efficiency due to more accurate calculation of fpwm, and much simpler to implement. One
main drawback of the offline-efficiency optimisation is that the approach usually requires a-priori knowledge of the converter parameters, which is not always available in practice.
Notice that the use of a look-up table has been exploited in various commercial products [58]. However, the switching frequency is specified to ease EMI filter implementation and controller design, rather than efficiency maximization.
Chapter 3
A review of power transformer modelling
and simulation of DC-to-DC converters
3.1
Introduction
In this chapter, a review of the published literature related to modelling and simulation of DC-to-DC converters and power transformers, which is relevant to Chapters 5 and 6, is presented. Particularly, Section 3.2 begins with a brief review of approaches to the simulation of DC-to- DC converters in digital computers. Along with the simulation implementation, the selection of modelling techniques which maximizes the effectiveness of the simulation, i.e. minimize the com- putation time while preserving all the necessary information, is also discussed. We then proceed in Section 3.3 to summarize the existing studies into modelling of power transformers and identifica- tion of transformer models from experimental measurements. Finally, the discussion and rationale which leads to the research in Chapters 5 and 6, is presented in Section 3.4.