Typically, efficiency considerations drive a designer's final decision when selecting an application's battery-power-management strategy. The choices for this strategy are switched-power converters (SPCs), charge pumps, and low-dropout regulators (LDOs). As far as efficiency is concerned, the SPC is the overall best choice for almost all applications. If the application can tolerate the switching electromagnetic-interference (EMI) noise, an SPC's efficiency is relatively independent of line voltage and output current.
In terms of line voltage variations, the charge pump's power efficiency is nearly comparable to the SPC. It's only more efficient for a small range of load currents. Also, charge pumps generally have unregulated outputs. If a charge pump's output must be regulated, it needs to be connected to a linear LDO. So the efficiencies of the charge pump and LDO are multiplied together, resulting in a system with lower efficiency than either the LDO or charge pump alone.
The linear regulator's efficiency can be approximated by the VOUT/VIN ratio. Given this scenario, the efficiency is dynamic, decreasing linearly with increasing line voltage. If the source voltage varies widely, the linear regulator is the poorer choice for a power-supply design. But the good news is that a linear regulator's efficiency is relatively independent of output current load—if the application can tolerate the power dissipation.
With all of these options, it might seem that choosing the right power system for a given application would be difficult. But taking time to carefully select the power-supply strategy can give one a competitive edge by providing the most efficient, compact, low-cost solution for the intended application.
The power-supply management circuit discussed here can be implemented with discrete components, or as a combination of ICs and discrete components (Fig. 1). Its purpose is to match source-voltage changes to the desired output voltage, while complementing the output drive requirements. In the best cases, this task is achieved with optimum efficiency. As mentioned earlier, the ICs typically used in this type of application are SPCs, charge pumps, or LDOs. In all cases, the IC conditions the source voltage to a different output voltage.
Buck-SPC circuit efficiency: Figure 2 shows a simplified example of a buck SPC circuit. This style of converter uses a simple chopping network in combination with a low-pass LC filter. As described here, the buck converter is operating in a continuous inductor-current mode. The input is "chopped" using a pulse-width modulator (PWM) signal to the switch. Then, the resulting pulses are averaged to create a dc output voltage. This converter can only step down the input voltage (from a high value to a lower value).
In evaluating the circuit of Figure 2, the following assumptions are made: (1) the input voltage is always greater than the output voltage, as required for a buck SPC, (2) the output voltage is essentially dc, which implies that the output filter is large enough to completely average this voltage (typically less than 1% variation on VOUT), and (3) there's always a current through the inductor, which is required when keeping the converter in a continuous-conduction mode.
With these assumptions, the buck SPC stage has two states per switching cycle. The ON state is when Q1 is ON and D1 is OFF. The OFF state is when Q1 is OFF and D1 is ON. The duration of the ON state is equal to D × tS = tON, and the OFF state is equal to (1 − D) × tS, where D is the duty cycle set by the PWM control circuitry and tS is the switching period or 1/FSW.
In this circuit:
VOUT = VSOURCE × D
IOUT = IL(AVG)
Several issues affect the buck SPC circuit's efficiency. Four main sources cause most losses, three of which are associated with the MOSFET. The first is the MOSFET's gate-charge current, resulting from the PWM switching action. This loss is almost independent of load. The second is the power dissipation of VSW × ISW while the MOSFET is in the linear region during transition. Making the edges faster controls this, but introduces more conducted and radiated emissions noise.
A third source of loss is the RDS(ON) of the MOSFET and other resistances in the circuit. In a good design, the efficiency curve will peak at full load, or just before then. At that point, the switching losses about equal the conduction, or I2R losses. When the efficiency curve starts rolling off, the ON resistance of the MOSFET will begin to dominate. Finally, the output diode is a major source of power loss, particularly at higher currents.
Charge-pump efficiency: Figure 3 illustrates a simplified circuit example of a charge-pump power-management system. This inverting charge pump inverts the input voltage, VIN, to the VOUT node. For example, if +5 V is applied to VIN, −5 V will appear at VOUT. The ideal switched-capacitor charge-pump converter has a two-phase oscillator, four switches (S1 through S4) that are synchronized by the two-phase oscillator, and two capacitors (C1 and C2). Often, C1 is called the "pump" or "flying" capacitor, and C2 is called the "reservoir" or "output" capacitor.
The voltage inversion occurs due to the two-phase operation. During the first phase, switches S2 and S4 are open while S1 and S3 are closed. Also during this phase, C1 charges to VIN, and C2 supplies the load current. In the second phase, switches S2 and S4 are closed while S1 and S3 are open. In this configuration, C1 and C2 are parallel, and the top side of C1, which was connected to VIN, is now connected to ground. The bottom side of C1, which was connected to ground, is now connected to the bottom side of C2. With this ac-tion, the charge is transferred from C1 to C2. These connections result in a negative VIN value at the VOUT node.
The losses in this network occur be-cause of the effective switching resistance (as it relates to the switching frequency), actual switch resistances, and effective series resistance of C1 and C2. Usually, the actual switch resistances dominate the performance.
LDO efficiency: Figure 4 is a simplified circuit example of an LDO. In it, the input voltage directly supplies the power for the circuits and the output load. Usually, the quiescent current of LDO devices is relatively lower than the output current. Consequently, it can be ignored in a first order calculation.
For proper LDO operation, the input voltage should always be higher in magnitude than the output voltage. When an input voltage is applied, the low-drift, bandgap reference voltage is established. The bandgap voltage and an op amp are used to sense the resistor divider voltage at the output. This divider establishes the output voltage magnitude. Regardless of input voltage, as long as it remains higher than VDROPOUT + VOUT, the output voltage will stay constant.
The output current to the load is supplied from the input source through the p-channel MOSFET, Q1. For linear regulators, the input current is equal to the output current plus the internal current required for bandgap generation, op-amp bias, and p-channel MOSFET turn on. If the LDO is fabricated on a CMOS process, the current needed for band-gap generation, amplifier bias, and p-channel MOSFET drive is very low. Therefore, it can be ignored when calculating efficiency for appreciable load currents. If the LDO quiescent current is ignored, the LDO's efficiency is approximately VOUT/VIN. Naturally, the LDO is very efficient for input voltages that are close to the output voltage.
The efficiency of these types of circuits can be defined as the ratio between output power and source power, or:
% Efficiency = 100 × \[(VOUT × IOUT )/
(VSOURCE × ISOURCE)\]
With this formula, a snapshot can be taken of a power supply's effectiveness under specified conditions. While a study of this type of snapshot is interesting, changes in efficiency as it relates to changes in the source voltage and output current give the guidance needed to ensure that a circuit design is optimized.
A number of efficiency experiments were conducted with these three types of devices. In both figures, data was taken via the TC105 buck SPC, TC1185 LDO, and TC7662A charge pump from Microchip Technology. The TC105 is a step-down buck SPC that supplies up to 1-A output currents (maximum). This device normally operates in a pulse-width modulation mode, but automatically switches to a pulse-frequency modulation mode at low output loads for greater efficiency.
The TC7662A charge- pump voltage inverter converts voltages of 3 to 18 V to between −3 and −18 V. It has an on-board oscillator and can source output currents of as high as 40 mA. The TC1185 LDO is designed with CMOS construction to eliminate wasted ground current, in-creasing the efficiency. This device is stable with an output capacitor of 1 µF, and capable of 150-mA (maximum) output current.
Figure 5 plots the efficiency of these three devices with respect to the source voltage. As shown, the TC105 buck SPC is the best in its class, with the TC7662A charge pump a close second. It's important to note that the charge pump is unregulated, so its efficiency curves can be misleading. If a regulated charge pump is evaluated with these same conditions, the efficiency-versus-source voltage curve would degrade with increased voltage. In contrast, the TC1185 LDO's efficiency degrades linearly with increased input voltage, as expected.
The same devices also are evaluated in Figure 6, where the results of their efficiency are plotted with respect to output current. These plots show that the SPC is very efficient over a wide range of load currents and input voltages. For applications that have a wide input voltage swing, and over a 100-mA load-current requirement, the SPC clearly wins in efficiency.
Battery-operated systems: A typical challenge for power-management systems that use these types of devices is the battery-powered application. For example, a large number of battery applications use one lithium-ion (Li-ion), rechargeable battery cell and require an output of 1.8 V at a current from 0 to 300 mA. The nominal voltage of a Li-ion cell is 3.6 V, with an output voltage range between 2.8 and 4.2 V. For this application, the feasibility of SPC, charge-pump, and LDO solutions can be evaluated in terms of efficiency.
The charge pump is probably the most unlikely device to fit in this application, primarily because it's an inverting device. Consequently, it couldn't easily provide a 1.8-V output. Although this is obviously a very critical specification, the charge-pump circuit has other problems in this application. Its efficiency-versus-current performance is only optimum for output currents of 1 to 10 mA (Fig. 6, again). Although the charge pump is an inexpensive solution (because the external components are capacitors only), the device's efficiency doesn't suit this application well.
The LDO is a possible solution to this application problem. Implementing the TC1185 in a circuit is very simple due to the low external device count. Though the device can be designed into this application, the efficiency-versus-source voltage isn't as good as the other solutions. Because the battery has a large output voltage range, the frequency of battery charging is higher with the LDO, compared to the SPC solution. In fact, the total LDO power dissipation is 240 mW when the input voltage is 4.2 V and the output is loaded to 100 mA.
The SPC device is the best choice for this type of application. The TC105 can easily power a 1.8-V output with good regulation. Its efficiency-versus-source voltage performance is comparatively very high—approximately 90% over the entire source-voltage range (Fig. 5, again). Although the efficiency versus output current isn't superior at every current, it remains the best choice across the entire output-current range (Fig. 6, again).
The bottom line is that in battery-powered applications, efficiency considerations are at the top of the list when considering a new design's power management. Available choices include SPCs, charge pumps, and LDOs. For systems that have wide source-voltage variations, the SPC is by far the best choice. If the system operates over a small range of low output currents, the charge pump can yield the best efficiency in a system. Finally, if the system requires a good low-noise regulated output, and the power dissipation is manageable, the LDO will provide satisfactory results.