Introduction to Multiphase DC-DC Converters
This article explores step-down switching power supplies that divide output current among multiple regulation subcircuits operating in parallel.
Multiphase DC-DC conversion can significantly improve the performance of a step-down switching regulator in high-current applications. In this article, I’ll explain the structure and functionality of multiphase buck converters, and in a future article, I’ll present the pros and cons to help you decide which design projects might benefit from a multiphase rather than a single-phase regulation scheme.
First, let’s briefly review DC-DC conversion basics.
Switch-mode Voltage Regulation Using a Buck Converter
The following circuit (Figure 1) represents a rudimentary step-down switching regulator (also called a buck converter):
Figure 1. This circuit is an asynchronous buck converter. In the synchronous buck topology, a low-side transistor replaces the diode. Image used courtesy of Texas Instruments
Unlike linear regulators, DC-DC converters can achieve high efficiency by exploiting the benefits of “switch mode”—i.e., on vs. off—current flow. Instead of dissipating power across a transistor functioning as a variable resistor, as is the case in linear regulation, a DC-DC converter’s transistor is switched fully on or fully off and consequently avoids operation in the low-efficiency intermediate region.
The switched voltage is filtered into a steady, reduced voltage by the inductor-capacitor circuit on the output side of the transistor. When the transistor is conducting, current flows to the load through the inductor. On the other hand, when the transistor is switched off, the inductor maintains current flow (recall that its current cannot change instantaneously). In this case, the output capacitor provides a charge reservoir for the required load current. Regulation is accomplished via a feedback loop that adjusts output voltage by pulse-width modulating the control signal applied to the transistor’s gate, thereby varying the ratio of on-state duration to off-state duration.
Example Multiphase Conversion Architecture
Next, let's look at the diagram below in Figure 2, which is taken from the datasheet for the DA9213/14/15 multiphase buck converters from Renesas.
Figure 2. This is the system diagram for DA9213. Image used courtesy of Renesas [click to enlarge]
These devices can supply up to 20 A and are intended for low-voltage, high-current applications such as generating power rails for microprocessors in smartphones and tablets. I like this diagram because it shows the structure of a multiphase buck converter without conveying an oversimplified idea of what it takes to implement multiphase conversion in a real-life application.
On the right, you see can four pairs of field effect transistors (FETs) and four inductors. One pair of FETs functions as a half-bridge driver that controls the current through one inductor, and each half-bridge-driver-plus-inductor subcircuit is a phase (i.e., the core of a separate buck converter). Phases operate in parallel and cooperate to supply current to the load (load current in the diagram is represented by the current source to the right of the output caps).
Though the diagram shows four separate output capacitors, all of these caps are connected in parallel; in other words, the output capacitance is physically divided but electrically united. This is true of the input capacitance as well. Thus, the phases do not share inductance, but they do share input and output capacitance.
Optimized multiphase conversion is a complex procedure, and you can see in the diagram that the DA9213 includes quite a bit of control circuitry. The serial interface allows a microcontroller to read and write data related to:
- Temperature faults
- Current limits
- Output-voltage target
- Output-voltage status
- Voltage ramp rate
- Phase shedding and many other operational details
Multiphase Conversion—Phase Timing
An important aspect of multiphase conversion is the interleaved timing applied to the phases, and actually, multiphase converters are also called interleaved converters. Interleaving activates phases in a cyclical fashion by applying a sequence of control pulses to the phase transistors.
The following schematic in Figure 3, from a research paper written by Reyes-Portillo et al. and published in World Electric Vehicle Journal, depicts an asynchronous multiphase buck topology designed for EV battery charging.
Figure 3. Example synchronous multiphase buck topology for EV charging. Image used courtesy of Reyes-Portillo et al
Additionally, the authors provide the following timing diagram (Figure 4) for the four phases.
Figure 4. Timing diagram covering four phases of that same example that was shown in Figure 3. Image used courtesy of Reyes-Portillo et al
The control signals for the transistors depicted as switches Q1 to Q4 in the schematic and implemented as metal-oxide-semiconductor field effect transistors (MOSFETs) create a cycle where the phases “take turns” entering the on-state. This is what is meant by interleaving. The particular scheme shown above includes phase-to-phase overlap in the control signals, but the overlap isn’t necessary.
One thing to note is that the authors of this study state that control-signal overlap is advantageous, at least in their usage scenario, because it eliminates discontinuities in the input current drawn from the power source.
Phase Current vs. Output Current
Before going much further, it’s important to recognize that though the phases enter the on state sequentially, they don’t “take turns” supplying all the load current. Just as the current supplied by a standalone buck regulator doesn’t drop to zero when the control signal turns off the transistor, the interleaved phases supply current during the off state, and the sum of these currents is available to the load. The following diagram (Figure 5) from a Texas Instruments app note will help to clarify this concept.
Figure 5. Example block diagram from a TI app note. Image used courtesy of Texas Instruments
First, notice how the phase control signals in this scheme do not overlap.
The phase current begins to decrease as soon as the control signal goes low and turns off the transistor, but this results only in the current ripple, not the loss of the phase current. The two rippling currents add together into a (rippling) sum current, and consequently, each phase in a two-phase system is responsible for only half of the maximum load current. Likewise, each phase in a four-phase system is responsible for one-fourth of the maximum load current.
The following diagram, shown in Figure 6, which was taken from a different TI app note on the benefits of multiphase conversion, more clearly shows the details of the phase currents and their relationship to the output current.
Figure 6. Example showing the phase currents and their relationship to the output current. Image used courtesy of Texas Instruments
The two phases have about 5 A of the inductor current, with a peak-to-peak ripple of about 2 A, and the total current delivered to the regulator’s output capacitance is the sum of the two 5 A phase currents. In a follow-up article, we’ll see that this technique of using multiple interleaved regulator subcircuits to supply a larger total power-supply current is key to the benefits of multiphase DC-DC conversion.
Overall, I hope that this article has given you some insights into a power-supply technique that is quite advantageous in certain applications and perhaps not as widely known as it should be. If you’ve had the opportunity to incorporate multiphase DC-DC conversion into any of your designs, feel free to leave a comment and share your experiences.