Consumer demand has stretched the capabilities of portable devices in terms of power and functionality while pressuring for smaller size and significantly longer battery life. Now consumers want it all: months-long battery life, 2-watt RF transmission and Web-surfing capability from a product small enough to comfortably fit in a shirt pocket.
The incremental evolution of semiconductor manufacturing technology has reduced IC size and supply voltage requirements while increasing integration and power requirements. Core logic voltage requirements have dropped from 5 V to 3.3 V, down to 1.8 V, with rumors of 1.1 V and 0.9 V on the horizon.
In the 1990s simultaneous developments in fine-geometry, low-voltage semiconductor technology; high-performance, dynamically clocked CPUs; gallium arsenide and silicon RF ICs; digital signal processors and advanced liquid-crystal displays fueled an explosion of portable products. Portable cellular phones emerged, first as products the size and weight of a cement brick, and then as the shirt-pocket-size products we know today. Portable computers entered the market, first with limited capabilities and monochrome LCD screens. Now they have fast, high-resolution color LCDs and computational capabilities rivaling top-end desktop machines. A myriad of other devices have entered the market too-personal digital assistants, talking toys and personal medical diagnostic instruments, such as glucose monitors.
The flood of portable devices, and diverse technologies enabling them, have complex power requirements. For example, plus/minus 20 V to 40 V for LCD bias, 40 V to 60 V for PIN diode bias, -4.5 V for GaAs RF device bias, 400 V ac for CCFT backlight tubes, 12 V for flash memory programming and so on. In particular, core logic voltage requirements began at 5 V for the venerable LS, HC and HCT logic families. In the early '90s, 3.3-V products emerged. Then subsequent reductions in semiconductor manufacturing geometries allowed 3-, 2.5-, 2- and now 1.8-V operation as digital designers struggled to integrate more functionality into their products, with higher performance without the products turning into chunks of molten slag from overheating, or sucking the batteries dry in a few minutes.
Combined with consumer expectations for battery life extending to hours, days or months, the complex voltage requirements of modern portable devices necessitated a fundamental rethinking of internal power strategies. Now products require complex multimode power-management systems. Product operation must be segregated into active, standby, sleep and shutdown modes for extending battery life while maintaining full functionality. Provision must be made for power sequencing during transition between operational modes. Thought must be given to both conducted and radiated electromagnetic interference (EMI) for preventing disruption of sensitive RF, analog and audio circuit blocks by noise from digital logic and switching dc/dc converters. These requirements must be implemented through a multipart power-management system encompassing battery chargers, switching regulators, charge pumps and LDOs (low-dropout linear regulators).
Power-management systems for modern portable devices are typically constructed using a modular, distributed architecture that places each power-conversion block near the subsystem it serves. This results in better voltage regulation through reduction of voltage drops across circuit-board traces. It also improves transient response by reducing delays from trace inductance. This architecture typically consists of an assortment of battery chargers, switching dc/dc converters, charge pumps, LDOs and supervisory ICs, depending on the specific needs of each subsystem.
In noise-sensitive devices, such as cell phones, the modular blocks of the power-management system are usually distributed across the board in a strategic manner for reducing interference. The board is divided into electrically quiet and noisy sectors. These are placed distant from each other. Since EMI field strength ebbs with distance, this strategy reduces interference from radiated noise. First, noisy switching dc/dc converters used to convert or stabilize the battery voltage are placed on a corner or edge of the circuit board, far from sensitive circuitry. Second, noisy CMOS digital processors and logic are placed around this. Next, relatively low-noise, noise-immune digital and analog circuitry is placed in sectors further down the board. Finally, noise-sensitive RF, audio and sensor circuitry is placed on a corner or edge of the board far from the dc/dc converters and digital circuitry. A similar strategy can be followed for heat management. Hot devices can be placed away from heat-sensitive sensors to reduce thermal interference.
Power to quiet zones on the circuit board, such as those used for RF, audio, sensor and other analog circuitry, is often post-regulated from a slightly higher system voltage using LDOs. This reduces conducted noise from switching-regulator ripple and transients from other circuit blocks, such as the PA (RF power amplifier). In extreme cases, such as for cell phones, the quiet circuitry may be protected using a metal EMI shield. This is typically constructed using a metal ring and sheet metal cap. In some cases, the LDOs may be distributed near the EMI shield junction to prevent conducted noise from generating radiated noise inside the shielded zone.
Modern portable devices achieve both high functionality and long battery life through careful planning of the system's operational profile and associated power needs. As continuous full-power operation and long battery life are mutually exclusive, this tension is resolved through multiple operating modes. These modes are typically shutdown, standby and full-power. Some devices, such as cell phones and smoke detectors, have intermediate modes entered periodically to perform a system function, such as checking for incoming calls.
Typically, portable devices spend most of their time off resting in someone's pocket or on a table top. Thus, the device must have a shutdown mode that draws low current, compared to the self- discharge rate of the batteries used. This is on the order of a few microamperes. Dc/dc converters and other power-management devices typically support this with a shutdown mode, controlled by a digital input. This input may be operated by an on-off switch or micropower CMOS logic.
The bulk of operational time is spent in some kind of a sleep or standby mode. In this mode, the device waits for some event, in a low-current state with minimal circuitry on. It may be waiting for an incoming phone call or RF transmission. It could also be waiting for some kind of sensor to activate, such as an oxygen sensor for a portable industrial safety device. In this mode, some circuit blocks may be biased up, but in a static or low-power state for a fast, smooth transition to full power operation, when needed. System current consumption in sleep or standby may be a few milliamperes or less.
Many modern dc/dc converters for portable devices support sleep or standby through some kind of low quiescent current state that allows precise regulation with minimal battery consumption. For switching dc-dc converters, this is typically done through some form of pulse frequency modulation (PFM). In PFM mode, the power converter only switches as needed to serve the load. This reduces switching losses at light loads and eliminates current consumption by the internal oscillator in the dc-dc converter, resulting in reduced quiescent current and improved light-load efficiency. This is achieved at the expense of variable frequency switching noise with potential for noise or harmonics crossing a sensitive IF, audio or data acquisition sampling frequency. Regulation is achieved by switching with a discreet energy/cycle and then modulating cycle rate to control the power sent to the load. In PFM mode, quiescent current is typically in the 20 to 200-microamps range, depending on performance trade-offs. High-efficiency load range typically extends over three decades vs. one decade for a comparable constant-frequency PWM (pulse width modulation) regulator.
A minority of operational time is spent in full-power operation. Full-power operation may be alarm activation in a safety or security device, or talk time for a cell phone. In this mode, the portable device may be drawing anywhere from a few hundred milliamperes to several amperes. Since the high load currents associated with full power can discharge the battery very quickly, it is important that this operation be short in duration and high in efficiency. Modern switching dc/dc converters for portable devices typically handle full-power operation through a pin-selectable PWM mode. The device may have a SYNC input to allow synchronization to a system clock to reduce interference by eliminating intermodulation (beating) with system frequencies. In PWM mode, devices for low-voltage operation often have synchronous rectification to reduce efficiency losses associated with the output diode. Efficiency with moderate to full loads is typically on the order of 80 percent to 97 percent while in PWM mode, depending on the input and output voltages, and other design parameters. In contrast with PFM operation, PWM mode quiescent current is typically in the 500-microamps to 5-mA range.
In some portable devices, the transition between shutdown or sleep/standby and full-power operation is done by incrementally powering up different circuit blocks. This is done to allow circuit parameters to stabilize before use. It prevents glitches and weird problems that can develop from an uncontrolled startup. Power-up sequencing is often performed under the control of some kind of low-power microprocessor or supervisory circuit. The modular and distributed design used for portable power-management systems allows separate activation of operating modes to achieve this purpose.
While in standby, many products intermittently switch to a medium-power task to perform key functions. For example, smoke detectors wake up every few seconds to sense the voltage generated by the ionization chamber. CDMA cell phones wake up for a multiple of approximately 1.28 seconds to check a paging channel for incoming calls. This can be done in either PFM mode or PWM mode, depending on the noise sensitivity of the application circuit.
Many portable system functions, such as the LCD display, PIN diodes, sensors and low-voltage logic (typically the CPU or DSP) often have special voltage requirements. These requirements necessitate the use of dedicated voltage conversion circuitry to support each of these functions. The correct type of voltage converter depends on the specific voltage required, maximum load, loading profile and battery configuration used. Additional considerations are performance, noise, cost and size.
The first consideration is dc/dc converter type. There are three basic options:
- Low dropout linear regulators (LDOs)
- Charge-pump voltage converters
- Inductor- or transformer-based switching dc/dc converters.
LDOs typically offer the simplest, smallest and cheapest solution, and generate no EMI. In portable devices, they are typically used to post-regulate supply voltage to sensitive IF, audio and analog blocks. For this use, their functions are to stabilize supply voltage, attenuate ripple from a switching dc-dc converter or smooth over power spikes and transients from other circuit blocks. Surface-mount devices designed for portable applications are readily available with output current capabilities of 50 mA, 100 mA, 600 mA or more. Quiescent current is low, typically on the order of 10 to 100 microamps, depending on performance and precision trade-offs of the LDO. However, efficiency is VOUT/VIN, with a loss of IOUT(VIN to VOUT) being dissipated across the device and generating heat. For portable devices, packages are available dissipating up to around 1 W. Above this, through-hole packages and bulky heat sinks are usmicroally required. Furthermore, the heat generated may be excessive for the application. Providing the input voltage range is close to the desired output voltage, LDOs provide good efficiency. For example, efficiency for post-regulation from a 3.6 V lithium-ion battery to a 3.3 V IF block is a healthy 91.7 percent.
High quiescent current
Charge-pump voltage converters offer a simple, compact solution for load currents up to around 100 mA. They offer high efficiency in the 70 percent to 98 percent range for cases where the output voltage is an integer of fractional multiple of the input voltage. Modern charge-pump devices can be used to step-up, step-down, divide or invert the supply voltage. They are available in both regulated and unregulated models. Regulated models use either pulse skipping, fractional voltage conversion or linear post regulation. In contrast with LDOs, charge pumps generate moderate voltage ripple and EMI. Because they typically switch at a constant frequency, quiescent current is relatively high, typically on the order of 200 microamps to 500 microamps. In portable devices, charge pumps are commonly used for doubling or inverting a supply or battery voltage for an analog function, such as for generating a negative bias voltage for GaAs RF circuitry. They can also be used to divide a battery or supply voltage for low-voltage logic.
Inductor- or transformer-based switching converters are the optimal solution where either the input-to-output voltage difference is relatively high, load currents exceed around 100 mA, the output voltage is not an integer or fractional multiple of the input, or step-up or voltage polarity inversion is needed. In these cases, switching dc/dc converters offer a performance advantage. Due to the inductor and relatively large input and output capacitors required for switching dc/dc converters, LDOs and charge pumps can have a cost, EMI and size advantage where applicable. Quiescent current for switching dc/dc converters is typically in the 50-microamps to 500-microamps range in PFM operation and 500-microamps to 5-mA range for PWM operation.
Depending on the input voltage range and output voltage, one of several switching dc/dc converter topologies can be chosen. Inductor-based buck (step-down), boost (step-up) and buck-boost (inverting) topologies offer the lowest size and cost. Efficiency is typically in the 70 percent to 97 percent range, depending on the circuit and application parameters. Similar to what happens with LDOs, efficiency improves for buck and boost topologies as the input voltage approaches the output. However, for most devices, there is a dropout voltage associated with maximum-duty cycle limits and this voltage is typically higher than for comparable LDOs. Transformer-based single-ended primary inductance converter (Sepic) and flyback topologies are useful in cases where the input voltage range crosses the output, but achieve step-up/ step-down operation at the expense of a 2 percent to 10 percent reduction in efficiency.
Care must be taken in selecting switching dc/dc converters for portables. This is because the specifications given in the data sheets' tables of electrical characteristics typically reflect out-of-circuit test routines used for screening parts with automated test equipment during manufacturing. Often these specifications differ from real-world performance. For example, some PFM device data sheets boast of quiescent currents down to 20 microamps. However, this is based on a deceptive specification made with the voltage feedback pin pulled high and the device not switching. In real-world applications, practical quiescent current will be significantly higher. This is not a case of dishonesty. It's a case of marketing going berserk over the specification for a test routine used to screen for defective units during production. Because some vendors have done this in the past, all must now do this to compete.
Portable power systems
Meanwhile, in portable power systems, dc/dc selection decisions are complicated by the discharge profile of the batteries used, output voltage requirements and current profile of the system. Basically, battery voltage declines as the battery discharges, and this can have a large effect on efficiency and available battery life. Reusing the previous example, if an LDO is used to power a 3.3-V IF block from a 3.6-V lithium-ion battery, initial efficiency will be (3.3 V/3.6 V)*100 percent = 91.7 percent. As the battery discharges, efficiency will rise eventually to 97 percent, assuming a 100-mV dropout voltage for the LDO. This is excellent efficiency. However, if precise regulation or noise attenuation is required, a considerable portion of the battery's voltage may go unused, especially at low ambient temperatures where the battery's nominal output voltage may be lower than normal. This is because the battery can only discharge to (VOUT + VDROPOUT) = 3.4 V before the LDO enters dropout, the output voltage slumps and noise rejection is compromised. Here, a step-up/step-down switching regulator may actually yield longer usable battery life, since it can operate until the battery is discharged (typically to 2.8 V). On the other hand, low-voltage digital logic typically has a 2.3-V to 3.6-V supply tolerance, allowing the LDO to run in dropout, with very high efficiency until the battery is discharged. In that case, the cost size and complexity of a switching dc/dc converter is unjustified.
See related chart