When comparing a typical cell phone today with one from a few years ago, the biggest surprise is not the smaller size or even the battery life: It is actually the complexity and feature-rich aspect of the handset. Everyone had predicted that phones would drop in size to fit nicely in a jacket pocket or purse and would contain a dial pad with approximately 1-cm pitch. But no one predicted that they would be dual-mode or tri-mode-with multiple complete handsets inside-able to receive text e-mail messages and to interface with a computer to store phone numbers and other information.
Today, those features are an absolute necessity. In the short term, it is clear that cellular phones are not going to get much smaller, and for the most part battery life is at an acceptable level. Nevertheless, the whole industry revolves around consumers' upgrading their phones at least every 18 months and carriers' offering new-generation phones as incentives to switch carriers. Even the PC industry cannot match this breakneck pace of increasing functionality. Those high volumes affect the whole phone design philosophy, especially power management.
In the typical cell phones manufactured today, power management incorporates several blocks, including lithium ion cell protection; battery cell charge control; external power control; dc-to-dc conversion, to produce multiple regulated rails; load management; and power sequencing and control for multiple power amplifier (PA) stages. All of those blocks have the same ultimate goal: to get power in and out of the cells and to provide the correct regulated power levels to the ICs or subcircuits of the handset as efficiently as possible.
Let's start with the battery pack. Present designs typically use an IC and two MOSFETs to protect the cell or cells. A fully integrated protection circuit on a chip has been discussed ever since lithium-ion cells were introduced. The answer as to why such ICs are not the norm today may lie in what is a common theme in the integration trade-off: voltage ratings, electrical ruggedness and cost. To minimize chip size and hence cost, application-specific ICs typically have the lowest voltage rating possible for the application. In the case of battery protection, the circuit must be able to withstand voltage spikes on the power rail being controlled. A 30-V input controller IC with integrated 30-V MOSFETs would be prohibitively costly, whereas a controller with a lower voltage rating driving discrete 30-V MOSFETs is a cost-effective solution.
In addition, MOSFET performance has observed its own version of Moore's Law, improving every two years, on average, by a factor of two. Devices that were available only in an SO-8 package a few years ago are now available in an SOT-23 or even an SC-70. That improvement comes only from companies that specialize in MOSFETs. If 20-V or 30-V MOSFETs were integrated onto an IC, not only would the initial cost be higher, because of the complexity, but the cost/performance ratio would have difficulty keeping up with the improvements in MOSFET technology. However, in the long run, it is likely that the initial hurdle of cost/performance will be overcome. And once fully integrated solutions are in volume production, the performance will keep up with discrete devices because of the R&D stream available.
Charging of the battery pack from the ac adapter or cigarette lighter cable adds a new variable: heat. For example, to charge a battery in two hours requires currents of over 1 A and regulation of the voltage and current. A MOSFET is typically used to regulate the charging voltage and current. Since the MOSFET usually performs as a linear regulator, the power dissipation easily exceeds 1 W, making integration difficult, since most IC packages are not designed to dissipate heat effectively
Inside the set
Nevertheless, a trend is emerging wherein the battery-charging IC, which may feature integrated MOSFETs for fast charge and trickle charge, is being placed inside the cellular handset. That level of integration is possible with submicron CMOS processes, where digital control functions require only a minimal die area and pulsing of the charge current rather than controlling it in a linear fashion. Even with the addition of two MOSFETs on chip, the finished die cost can be kept to a reasonable level, so that a cost-effective solution can be brought to the application.
That type of integration into the handset itself allows users to employ an inexpensive plug-in wall cube for battery charging and to utilize the intelligence of the microprocessor already inside the phone to provide the correct charging algorithms. That reduces overall cost and is very convenient for the user.
For the time being, some charge ICs will use external MOSFETs, and some will contain integrated MOSFETs, depending on the power dissipation and voltage ratings of the pass elements.
For the next generation of cellular phones, known as 3G (for third generation) or wideband code-division multiple access (W-CDMA), the output power will be controlled to use the minimum necessary power to conserve battery life. The output stage PA will therefore require a variable voltage that is tightly regulated and has very fast line and load transient response. That will preclude the use of a linear regulator, so a synchronous buck regulator must be used. That method of powering the PA will also lead to increased levels of battery life, since the efficiencies of conversion will be in the mid-90-percent range with a synchronous regulator, compared with the mid-70-percent range when a low-dropout regulator is used.
Since the volumetric constraints are very tight and the overall footprint of a dc/dc conversion solution is a function of frequency, a dc/dc converter in a cell phone typically operates at 1 MHz or higher. The higher the operating frequency, the smaller the inductor and filter capacitors. At those frequencies other factors come into effect, such as parasitic inductance and board layout. If the controller circuitry is included in an ASIC, a 1-cm wire bond, as found in high-pin-count ASICs or the trace of the printed-circuit board, becomes a significant inductor. Placing the components for a dc-to-dc converter around the ASIC is just not possible without excessive stray inductance.
It also results in too many other components competing for the space around the multipurpose IC. Another difficulty that can occur with an integrated dc-to-dc converter is that the high currents, with fast switching speeds, could cause noise problems elsewhere in the IC. That requires special up-front design consideration, since it is a problem that is hard to simulate in advance.
These factors support using a dedicated controller IC. If the power levels are low enough, integrating MOSFETs onto a dedicated pulse-width-modulation controller IC to make a one-chip dc-to-dc converter can provide a cost- and space-effective solution. This form of integration can actually reduce parasitic values and layout problems, and the IC designers are already experts at avoiding noise problems from high-speed switching. The concerns here are cost, power level and flexibility. MOSFET manufacturers keep making better and better MOSFETs at lower and lower prices for these applications. So some circuits are best suited for integrated MOSFETs and some for discrete ones in SOT-23, SC-70 or similar packages.
As cellular phones continue to become more complex and functional, power management will become even more important and more widespread in the handset. Once wireless Web browsing becomes even more widely used with the introduction of W-CDMA, and personal digital assistants (PDA) and phones finally combine, there will be even more pressure on longer battery life. For example, the notebook computer and PDA markets have shown us that color displays are very desirable and became a standard very quickly. This impending leap in complexity for mobile handsets will mean that a fully integrated power-management IC is just not practical, because of the number of different tightly regulated output voltages required and the necessity of making every single Joule of energy in the battery count. Even if a multioutput dc-to-dc converter could be integrated into the analog glue ASIC, it could never meet the performance and efficiencies of a dedicated PWM controller IC.
Total power management on an analog ASIC might be a good approach if we were to design yesterday's phones with today's technology. However, the need for added functionality means that the choices must be examined more carefully.
ICs with integrated MOSFETs have size and performance advantages over ASICs that require external MOSFETs. In addition, discrete MOSFETs, with some simple added functions such as level shift will always have a place because they have ideal specifications for switching power and higher voltage ratings and they can dissipate heat without affecting other parts of the circuit.