Linear vs. switcher: It's a debate many face when deciding how to provide regulated voltage. From a pure engineering standpoint, the argument comes down to power dissipation and noise. But throw in the four-letter c-word, cost, and it gets a bit more complicated.
Essentially, the basic linear regulator consists of a pass transistor and analog control circuit. The control circuit uses an error amplifier to measure the differences between the sampled voltage (or current) and its intended value, and opens or closes the pass transistor to bring back the intended voltage level. The output and input may have additional capacitors for filtering and load transients. Control may be enhanced by an external divider resistor arrangement and ac compensation components.
Linear regulators have some drawbacks. They can only step down or produce an output voltage that is lower than the input voltage, and the pass element will dissipate a large amount of power under large load currents or large input-to-output voltage differentials.
But linear regulators do have several advantages. They are easier to design, with low parts count and low noise generation. Recent developments in linear regulators include highly integrated solutions and low dropout (LDO) regulators. Integrated designs are a function of package and silicon.
Power dissipation in a linear regulator is simply the voltage across the pass device or input voltage minus the output voltage, multiplied by the output current. If the package can handle the power, the pass device is integrated onto the silicon of the control circuit. When the power gets high, the pass device requires too much silicon area when compared to the supervisory silicon. The solution is to separate the control silicon into its own IC and couple it with a discrete pass device, such as a power MOSFET. Often, the control IC will be in a small package and the pass element in a larger package that is capable of dissipating the lost power.
Regulation with a linear regulator is simply a matter of biasing the pass transistor in a manner that provides a regulated voltage. For a bipolar junction transistor (BJT), the device is operated in the linear region and regulation "drops out" when the input voltage nears the typical saturation voltage of 0.2 V. For a MOSFET, the device is also biased in the linear region and regulation ceases when the input-to-output differential approaches the on resistance of the device multiplied by the output current.
Regulation is closely related to dropout. The dropout voltage is the voltage at which the input voltage is low enough to cause the output to go out of regulation. With the reduction of logic voltages, the dropout voltage becomes more critical. A case in point is when you want a 1.5-V alkaline cell to power a 1-V DSP. The alkaline cell can degrade to 1 V and the regulator can still provide power to within a few millivolts of 1 V. Dropout requirements dictate the type of pass element used and favor CMOS for very low-dropout regulators. Some regulators use a pass element and a low-loss switch that directly couples VIN to VOUT with a small voltage drop across the switch.
The typical linear regulator has about 70 percent efficiency, though some are run as low as 50 percent. Of course, this will increase the power dissipated in the pass element and, thus, the pass element's operating temperature. Higher operating temperatures mean decreased reliability and shorter battery life.
In the quest to decrease power loss, transistors are no longer used as pass elements in the linear region. They are now used as switches. If the forward drop across the switch is low enough, the conduction loss decreases the power dissipation in the switch. A new element of switching loss is introduced. Switching loss is typically 10 percent of the total loss in the switch for moderate power switching regulators at frequencies of 100 kHz and below for MOSFETs.
The key to a switching regulator is an energy storage element. For example, switched-capacitor-type designs work on a charge-transfer principal. The charge on the input capacitor is loaded by turning on switches S1 and S3 to charge capacitor C1. Switches S2 and S4 are then turned on and C2 charges.
While switched-capacitor designs offer low parts count and ease of design, they do not favor large load swings. Additionally, such designs are most efficient when producing voltages of 2x VIN or when producing -VIN, a negative rail. At voltages other than these, efficiency suffers and the switches dissipate most of the power causing many of the same problems that bedevil the linear regulator.
Switching converters use both an inductor and a capacitor to regulate. The basic regulator for stepping down voltages is the buck regulator. The principle is to build up current through an inductor by turning on the transistor or main switch. The transistor is then turned off and the inductor forces current to free-wheel through the rectifier diode. The size of the inductor coils can be minimized by increasing the frequency at which the transistor switch is turned on and off. Switching regulators operate from 30 kHz up through several megahertz, though most designs keep to 100 or 200 kHz.
By rearranging the inductor, transistor and rectifier in a switching regulator into different topologies, various regulation schemes result. Voltage can drop with the buck topology, or step up by the boost. But when comparing switching regulators with linear regulators it is best to stick with the buck, or step-down, topology.
The buck has a unique regulation feature where VOUT= VIN x D, where D = duty cycle or the on time of the main switch. Because D is always less than 100 percent, the buck produces an output voltage lower than the input voltage.
When moving to the switching regulators, we unfairly introduced that pesky inductor. In reality inductors are a combination of wire and some sort of magnetic path called the core. Along with this comes a mess of unit conversions such as gauss and oersteds. Also, inductors can saturate or cause excessive ripple and require calculations like window area, wire size, skin effect depth and gap length.
Today, switching regulators are achieving efficiencies up to 98 percent. But the switching introduces noise and additional components, and dropout can be a problem as duty cycles approach 90 percent. Still, these regulators do provide regulated voltages at high power.
The choice of a linear or a switching regulator requires in-depth engineering to meet the spec at a respectable cost.
See related chart