Tim Williams Seven Transistor abs, C http://www.seventransistorlabs.com/ June 19, 2016 Development Kit Phase Shift PWM Generator 1 Summary These plans describe a phase shift PWM generator, suitable for high power switching supply designs, induction heating control, research and development, and teaching power supply design and control concepts. Features include: true continuous phase shift control 0-180 ; wide frequency range; programmable dead time; multiple operating modes; two error amplifiers; ready integration into bridge type converters, frequencycontrolled resonant circuits, and demonstration and research applications. 1
Contents 1 Summary 1 2 Introduction 3 2.1 Pulse Wih Modulation...................... 4 2.2 Control Theory........................... 6 2.3 Current Mode Control....................... 8 2.4 But I Need Regulated ltage..................... 14 2.5 ltage Mode Control....................... 14 2.6 Bridge Converters.......................... 16 3 Design 17 4 Specifications 18 5 Examples 19 6 Assembly 20 2
2 Introduction Phase shift PWM (PSPWM) is a useful approach to power control and conversion. Although it is a more complicated method (requiring four independent drive signals, and more generator circuitry), it brings important benefits in control linearity, more consistent switching (allowing reduced EMI), and even sharing of load current and switching loss within the H- bridge. This paper is organized in three sections: The Introduction provides background on pulse wih modulation (PWM). If you re a beginner, take the time to understand what PWM is, why it works, and how it s done in a real circuit. Following closely, an introduction to control methods: PWM isn t very useful if it s just sitting there, not moving! Finally, an explanation of PSPWM, how it s used and its advantages over related modulation methods. Next, the PSPWM circuit is explained, discussing its Design and behavior, and presenting the Specifications which the circuit achieves. Application Examples are provided, covering a few typical applications, and including a complete 500 W power supply design. Finally, for kit-builders, a complete Assembly instruction is provided. Testing methods are included as well, allowing the builder to verify their specifications. 3
2.1 Pulse Wih Modulation If we want to control power efficiently, the easiest method is to switch a source on and off rapidly. As long as the switch has a low on resistance and a high off resistance (compared to the load it s switching), and spends a short fraction of the cycle going between off and on, then the efficiency will be high: the average input current will be less than the load current, and the average output voltage will be lower than the input voltage. Okay, so we re making an awful racket in the process, but in principle we can filter that out, leaving clean DC. 1 A 1 2 Iavg + Ipk Iavg Iavg - Ipk Figure 1: The most basic switching converter, an inductor switched between V in and 0 V. efficiency can t be better than the ratio of voltages V o Power can also be controlled in a linear (continuous) Current Sensor fashion, but the (for the most common, pass regulator topology, and similar rules for other types). inear amplifiers are still best at fast, precise and low-noise applications (even very powerful ones, like 10 kw MRI gradient coil drivers). Where efficiency orb low loss is paramount, power switching is the absolute winner. Raw switched power isn t very useful, so let s consider ways of filtering it. V_Isw V_ The most common type of filter has an inductive input (see Figure 1). (There are only two components to choose from: inductors or capacitors; nothing else stores energy, reversibly and quickly. But if we used a capac- D V in V CK 1 It doesn t have to be DC. The input or output side can have AC superimposed, as long as it s at a low enough frequency to look like DC to the filters. The polarity can even reverse, as long as we arrange the switches correctly. This is usually more bother than R it s worth, because semiconductor Vpk switches transistors only work inq one direction, but that s not a big deal. S C 4 CK Cp
2 3 Rev 1 Initial Release Iavg + Ipk Iavg Iavg - Ipk Current Sensor each cycle, which is no help here 2, so inductor it is.) V_ Figure 2: Waveforms for the buck converter. itor here, it would simply dump a huge pulse of current from the switch, et s take a close look at Figure 1. This is a study in dynamics: how Co Rz2 Clead a system evolves over time. This requires calculus for a thorough under- ESR R1 standing, but in fact, the only calculus you ll need to know is the fundamental equation of the inductor: Vm Clag Rz1 w D V = di V is the voltage applied to the inductor, is the value of the inductor, and di CK is the rate of change in current flowing through the inductor. R2 CK R S Note that this equation does not tell you what the current is, at any Transconductance Amplifier given point in time. Only how it changes over time (a velocity, not a position). The absolute value can be anything! Note also that, if you apply a Q constant voltage, the current rate is also constant, which means the current can rise forever! 3 Cp Incidentally, doesn t need to be constant either. Indeed, often it s a Rz Co R1 W function of current too. Inductors are usually chosen so as to avoid this effect (magnetic saturation), but it can be used to advantage in some cases 2 While I won t discuss them here, charge pumps, such as the Vref flying capacitor converter, V_ or diode-capacitor voltage multipliers, are useful at times. 3 Real circuits, with real inductors and ltage transistors, Error Amplifier have resistance that eventually limits current to I = V R. But if we get into the time and current scales where resistance takes Vmover, we ve already failed, because so much current is flowing that our transistors have already exploded... R2 CK 5 7 Seve abo
(such as swinging chokes and saturable reactors). For this article, I ll assume remains constant. Select your inductors accordingly! Anyway, if we apply a general voltage waveform, we have to do calculus with this equation. But we know a few things about our waveform. As long as V o is held constant (the dashed line in Figure 2), the voltage on the inductor is either V in V o while on, or V o while off. During each state, the voltage remains constant, and so di is constant too. We don t need to laboriously calculate anything over the whole waveform, we only need to know the applied voltage, and the duration. Change d s to s, and add up the t durations of each state, and their current changes I. If we end up in the same place where we started, then the average current didn t change; or if it s higher or lower, then given enough time, we can steer it anywhere we please! In this way, we can generate almost any current waveform we like. The exception is that: 1. it s always going to have that triangular shaped ripple on it, and 2. it can t be varying any faster than the di limits us to. But within that scope, there are many applications: audio amplification, motor control, AC and DC power control, etc. Also, note that, while the voltage across the inductor is a square wave ( high or low voltage, no inbetween), the current waveform has a triangle shape. The inductor is acting as an integrator (after all, it s solving calculus for us!). An integrator is a simple kind of filter the first stage towards filtering out all that noise! 6