Complexities of Switching Power Design
Even the simple buck converter can present very complex and changeable control characteristics when interacting with surrounding components and operating in different modes.
This column is going to be devoted to the complex and intriguing issues that continue to make switching power supply design a challenging and time-consuming task. I hope to show you the need for proper diligence and engineering in designing the power supply, a crucial and often-misunderstood part of electronic systems.
Outside of the power supply design community, there is also an ever-present feeling that power supply design is easy, and not worthy of time and attention. Nothing could be further from the truth, and numerous product recalls in recent year related to power and heat issues are a testament to this fact.
There is no doubt that power supply design is a mature industry. Standard circuits are used in around the world, with flyback circuits, buck converters, and forward converters dominating the field. This has held stable for many years now, and as a result, there are strong efforts from semiconductor companies to integrate functions into advanced chips with more and more capability inside the chip, and fewer parts on the.
As the switching power supply functions become incorporated inside the chip, we often lose design flexibility, and access to crucial functions. I’ll talk in a future column about the parts that are being integrated, and why I personally prefer access to many of them with discrete designs. In this column, we’ll focus on just one of the functions that can be overlooked, the feedback control loop.
It has always been a specialty skill within power supply design to work on control. Too often, in the latest round of IC designs, the pretense is made that this is no longer a function that needs attention, and integration of the feedback loop eliminates the chance for proper system design.
Integration of discrete power circuits has happened before in this industry. Thirty years ago, when we didn’t yet have switchers to deal with, the industry was dominated by linear regulators. Sophisticated designs were generated by experienced engineers to optimize parameters such as the minimum dropout voltage, transient response time, thermal characteristics, efficiency, etc.
Designers were highly knowledgeable in transistor characteristics, thermal design, and feedback analysis. Since all regulators use error amplifiers to precisely set the output voltage, feedback analysis and measurement was part of the design procedure for an optimized system.
Later, standard solutions arrived in the industry, leading to integration of the linear regulator. Today, few of us ever consider building their own linear regulator since it has all been effectively integrated. In the process, access to the feedback loop has been lost, but no-one seems to be concerned about this. Why not? Well, the integration of the linear regulator went fairly smoothly, and there are three reasons for this – 1) predictability, 2) consistency with line and load variations, and 3) low noise.
Feedback for the linear regulator is quite straightforward. The small-signal model is just that of a current source feeding a capacitor and load resistor. The only variation in the design of a control loop for the system is in the impedance of the output capacitor. Apart from this, the system is predictable, and can easily be simulated, and modeled, with modeling results agreeing closely with measuremkkents.
Figure 1: Once designed with discrete components, the three-terminal linear regulator is now almost always fully integrated. There is no longer any access to the control loop.
Once the linear regulator was integrated, there was no opportunity left to change the controller, except for changing the output capacitor. And the output capacitor was really the only uncontrolled component. However, there is little stress placed upon this part – it is there just to stabilize the system, and provide energy during load transient.
Once the linear regulator design is placed in the system, we can look at the characteristics of the regulator as the simple model shown in Figure 2. The input impedance of the linear regulator is a current source – regardless of changes in the input voltage, the current draw is fixed, equal to the output voltage divided by the load resistance. The output of the regulator model is just a voltage source.
Figure 2: As long as the linear regulator control loop is stable, the circuit model looks like a current source on the input (infinite impedance) and a voltage source on the output (very low impedance). Very little noise is introduced into the system.
Since the linear regulator does not generate any significant noise, we have no need to introduce any complex filtering on the system board. As a result, placement and integration of linear regulators on the board is a job that is almost trivial, and it does not require a power electronics designer to be involved (thermal issues notwithstanding – that’s often our responsibility, too.)
Does the future hold the same fate for the switching power supply? Will integration of the controller functions with power devices and auxiliary circuits render the switching power supply design a simple process?
Well, not quite so fast. As we’ll see below, even the simplest switching power supply has tremendous complexity of characteristics.
Continue reading this article?