Design Feature: March 14, 1996

Step-down-converter choices include a variety of linear and switching regulators. Linear regulators have excellent regulation and dynamic-load response but are less effective than switch-mode supplies in meeting the efficiency and size requirements of today's power systems.

Buck converters are a logical choice for switch-mode down-conversion, but, as operating voltages drop, the switching elements, feedback methods, and silicon partitioning become critical. Synchronous-rectification techniques help increase efficiency, but you must also analyze the trade-offs between using voltage- and current-mode feedback methods and continuous and discontinuous conduction. Also, you must choose between using discrete MOSFETs or more integrated ICs. For high-current, low-voltage applications, integrating the power switches with some control and drive circuitry enables you to design a high-performance synchronous-rectified buck converter based on voltage-mode operation.

Conventional vs synchronous buck regulators

A standard low-voltage buck regulator uses a MOSFET and a Schottky diode as the two main switching devices (Figure 1). Turning the MOSFET on delivers energy to the load and to the inductor. When the MOSFET turns off, the energy in the inductor forces current to circulate through the load and Schottky diode. In this manner, the MOSFET switch pulse-width-modulates the energy. The inductor and capacitor act as a lowpass filter to restore a nearly constant output voltage. The idealized output voltage is equal to the MOSFET duty cycle (t_on/(t_on+t_off)) multiplied by the input voltage.

When designing a conventional buck regulator, you select the MOSFET based on the regulator's intended switching speed, efficiency goals, and thermal constraints. You can translate these requirements into MOSFET characteristics, such as R_DS(ON) and gate charge. As the duty cycle of the high-side switch increases, R_DS(ON) has the most influence on the converter's efficiency. In these types of applications, you typically use n-channel MOS (NMOS) FETs to achieve the lowest possible R_DS(ON). For a given die area and breakdown voltage, an NMOS FET's superior carrier mobility translates to about one-half the on-resistance of that of a p-channel device. The use of NMOS FETs in a high-side configuration complicates the design, however. Either an auxiliary supply, a bootstrap circuit, or a charge pump must bring the gate voltage above the source node (input voltage).

You select the buck regulator's Schottky diode on the basis of its forward-drop and reverse-leakage characteristics. As the output voltage drops, the Schottky diode's forward voltage becomes a limiting factor in improving the converter's efficiency. This limitation forces engineers to use synchronous rectification in applications in which size, efficiency, and thermal considerations dominate, such as laptop computers and mobile communications. Even designers of desktop PCs and workstations are turning to synchronous rectification as their power requirements increase and new ICs ease their implementation.

Synchronous rectification raises efficiency

Synchronous rectification increases the efficiency of a buck converter by replacing the Schottky diode with a low-side NMOS FET (Figure 2). The resultant voltage drop across the MOSFET is smaller than the forward voltage drop of the Schottky diode. Figure 3 compares a Schottky diode's forward drop to the third quadrant R_DS(ON) of a MOSFET. A more comprehensive comparison includes the switching losses for both the MOSFET and the Schottky diode. However, at typical operating frequencies and voltages, a buck regulator's switching losses are usually small in comparison to the conduction losses. The low-side MOSFET conducts current in its third quadrant during the off times of the high-side MOSFET. This synchronous switch operates in the third quadrant, because the current flows from the source to the drain, which results in a negative bias across the switch. A positive voltage at the gate of the device still enhances the channel.

As Figure 2 shows, conventional synchronous-rectified buck converters partition the PWM-control and synchronous-drive functions into a single IC that drives discrete MOSFETs. The control and driver circuits synchronize the timing of both MOSFETs with the switching frequency. The upper MOSFET conducts to transfer energy from the input, and the lower MOSFET conducts to circulate inductor current. The synchronous PWM control block regulates the output voltage by modulating the conduction intervals of the upper and lower MOSFETs. Under light loads, the control block usually turns the lower MOSFET off to emulate a diode.

Synchronous rectification with discrete MOSFETs causes variable switching delays because of the variations in gate charge and threshold voltage from one MOSFET to another. Standard control circuits compensate for these variations by delaying the turn-on drive of the lower MOSFET until after the gate voltage of the upper MOSFET falls below a threshold. This delay creates a dead time in which neither MOSFET conducts. The dead time eliminates the possibility of a destructive shoot-through condition in which both MOSFETs conduct simultaneously. Standard designs use the same method to delay the turn-on of the upper device. A typical design delays discrete MOSFET conduction with a 60-nsec dead time and limits converter switching frequency to 300 kHz.

During the dead times, the inductor current flows through the lower MOSFET's body diode and develops stored charge in the depletion region. This stored charge must sweep out to allow the body diode to recover its forward-blocking characteristic. The body diode in a discrete MOSFET has a slow reverse recovery that adversely affects the converter's efficiency. You can minimize or eliminate the stored charge by placing a Schottky diode in parallel with the lower MOSFET. This addition improves the converter's efficiency approximately 1%. The Schottky diode can have a lower current rating than the one that the standard buck regulator uses, because the diode conducts only during the dead times, which lowers the rms current.

Regulator feedback methods The synchronous-rectified buck converter of Figure 2 uses current-mode control to regulate the output voltage. This control mode allows the converter to respond to changes in line voltage without delay. Also, you can reduce the output inductance to increase the converter's response to dynamic-load conditions.

Although these features would appear to favor current-mode control in applications that require a fast dynamic response, this control method has some disadvantages. For example, it tends to be sensitive to noise in the control loop. Also, current-mode control method requires two feedback loops: a current inner loop and a voltage outer loop, thus complicating the design. Finally, the controller uses a current-sensing resistor in series with the output inductor. This current-sensing resistance typically dissipates as much power as do the MOSFETs, further reducing the current-mode converter's efficiency.

Voltage-mode control is attractive for low-voltage buck converters, because it involves a single control loop, exhibits good noise immunity, and allows a wide range for the PWM duty-cycle ratio. Also, voltage-mode converters do not require a resistor for sensing current. However, the transfer function of standard voltage-mode buck converters that use Schottky diodes changes from no load to full load, making it difficult to achieve fast response to large dynamic loads.

Figure 4 compares the power-stage transfer function of a standard voltage-mode buck converter at light loads with that of the same converter at full loads. The figure shows that the light-load transfer function exhibits no double pole at the LC filter-frequency that is characteristic of the full-load transfer function. This difference occurs because the Schottky diode in the standard buck configuration allows inductor current to flow only in one direction. This unidirectional current flow results in discontinuous operation at light loads in which the inductor runs "dry" during a portion of each cycle, resulting in a single low-frequency pole. The load-current boundary between continuous and discontinuous conduction occurs at a load current of one-half the peak-to-peak ripple current.

Note that current-mode converters do not exhibit this behavior. The transfer function of a current-mode converter changes only slightly from discontinuous operation to continuous operation. The current-mode controller has two loops. The purpose of the inner, or current, loop is to separate the high-Q double pole of the LC filter into two single, well-separated, low-Q poles. Discontinuous operation exhibits a single low-frequency pole.

Continuous conduction has advantages

Figure 5a illustrates the inductor current for a standard buck converter at light load. When the current drops to zero, the rectifier turns off. The voltage across the rectifier rings at a high frequency, due to the circuit's parasitic capacitance. Most converters use a snubber network to suppress this high-frequency noise, or EMI, and prevent it from interfering with other critical circuits.

Whether or not you design a current- or voltage-mode synchronous-rectified buck converter, you face this ringing issue with discontinuous conduction. Nevertheless, synchronous converters for portable or handheld applications often deliberately employ discontinuous conduction to save power at light loads. These power savings become important in applications in which the µP spends a lot of time in sleep mode or in suspended operation and the power source is a battery or is otherwise limited.

It can be difficult to assure stability in voltage-mode converters in either continuous or discontinuous operation with a single compensation network in the PWM controller. Many designers increase the output inductance so that discontinuous conduction occurs only below the minimum expected load current. Increasing the inductance eases the compensation-network design, increases the size and cost of the inductor, and moves the double pole to a lower frequency. Also, the large inductor limits the converter's ability to respond to large dynamic loads.

You can avoid the problems of the large inductor by designing a continuous-mode, complementary-switching synchronous rectifier—in which the lower MOSFET conducts under conditions that would reverse-bias a diode—because this circuit operates with continuous conduction even during no-load conditions. Therefore, the transfer-function issues of discontinuous current and minimum-load requirements don't constrain your inductor selection. Under light loads, the inductor current continues past zero (Figure 5b). The lower MOSFET conducts in its traditional direction; that is, the current flows from drain to source, and the bias across the switch is positive, pulling energy from the output capacitor. When the lower MOSFET turns off, the upper MOSFET turns on, returning energy to the input capacitor until the inductor current reverses.

During the negative-inductor-current interval, the circuit temporarily acts like a boost converter. Energy transfers from the output to the input with ohmic losses. Although these losses reduce light-load efficiency compared to a standard buck converter, only the output-ripple specification and the output capacitor's equivalent series resistance (ESR) now determine the minimum output-inductor value. You can use a small output inductance for a high-bandwidth dc/dc converter that can drive the fast dynamic loads that are characteristic of modern µPs. Thus, at the expense of light-load efficiency, you can eliminate one of the major disadvantages of voltage-mode control.

Figure 6 illustrates the operation of a voltage-mode, complementary-switching converter. The inductor current reaches 10A in less than 5 µsec (Figure 6a) vs 13 µsec for the standard buck converter (Figure 6b). Also, the output voltage of the standard buck converter sags to 2.75V and slowly recovers to its final value. In contrast, the output voltage of the complementary-switching regulator falls only to 2.8V and recovers quickly.

The conventional synchronous rectified buck converter in Figure 2 uses a single IC for control and synchronous drive of two external MOSFETs. Dead times limit the converter's switching frequency, and the current-sensing resistor reduces its efficiency. However, new power-process technologies enable you to repartition this circuit. Figure 7 shows one approach. In this circuit, a SynchroFET IC integrates the upper and lower MOSFETs, their drive circuitry, and the synchronous-control logic. You can use such a device along with a simple PWM controller to implement voltage-mode buck converters that have several advantages compared to converters that use discrete MOSFETs. These advantages include improved efficiency, higher switching frequency, reduced EMI, and simplified thermal design. The advantages of the conventional approach over this integrated approach are the wider choice of discrete MOSFETs with a wide range of available R_DS(ON)s from which to choose. Also, you can tailor discrete MOSFETs to meet an application's efficiency and output-current requirements.

However, in contrast to a converter with discrete MOSFETs, an integrated design takes advantage of matched-silicon parameters. Worst-case analysis is less severe, because parameters such as gate charge and threshold tend to track with process variations and operating conditions. Additionally, the body diodes of the integrated MOSFETs exhibit low stored charge and short reverse-recovery times. Integrated power devices also reduce parasitic inductances from the critical high-speed connections.

These performance improvements let you build a converter that reduces dead times to less than 20 nsec, that switches with rise and fall times lower than 10 nsec, and that operates at frequencies higher than 1 MHz. For Figure 7's circuit, voltage-mode operation eliminates the current-sensing resistor. Also, complementary switching eliminates the high-frequency ringing and the need for a snubber circuit. Finally, the fast switching, low stored charge of the body diodes, and short dead times eliminate the need for a parallel Schottky diode.

When you use discrete MOSFETs, which vendors fabricate using vertical technologies, the substrate is at drain potential. Thus, conductive cooling requires large pc traces. In synchronous buck converters, these traces are the input voltage for the upper MOSFET and the switching node for the lower MOSFET. Large traces increase pc-board area and parasitics that can increase EMI. In contrast, the substrate and the tab of the SynchroFET package in Figure 7 are at ground potential. Therefore, heat can transfer directly from the power switches, through the tab, and then to the ground plane.

Why not integrate all the required silicon for the synchronous buck regulator into a single IC? This level of integration is achievable but involves trade-offs. An IC that integrates the PWM controller, power switches, and drive and synchronous control has greater die size and pinout. Power-IC packages with the required pin count and thermal capabilities are expensive. Instead, the silicon partitioning of Figure 7's circuit allows you to pair a seven-pin IC with many different PWM controllers to make the final implementation trade-offs, such as current- vs voltage-mode operation.