April 10, 1997

Mobile phones put the squeeze on battery power

Brian Kerridge, Editor

Powering mobile phones from fewer cells focuses attention on every aspect of battery discharge. Today’s designs call for a concoction of low-dropout regulators, inductive switching regulators, and charge pumps.

Ask any mobile-phone user to suggest three design improvements, and, without doubt, the answer will be battery power, battery power, and more battery power. Although we designers talk a lot about battery management, our focus too often dwells on controlling the charging side of the charge-discharge equation. Ensuring that batteries are fully charged, fast-charged, and never overcharged is certainly important, but battery management and a product’s runtimes have as much to do with the way your design discharges—and charges—its batteries.

With these issues at hand, today’s phone designers have placed a firm and miserly grip on every bit of battery power. Their main attention is on replacing earlier low-efficiency regulators with versions that promise efficiencies ranging from 80 to 95%. Other work includes designing for a minimum battery voltage of 3V, which is the recharge threshold of three nickel-cadmium (NiCd) or nickel-metal-hydride (NiMH) cells. Using this number of cells paves the way for mobile phones that will eventually run from a single lithium-ion (Li-ion) cell.

For most mobile phones now on the market, the battery consists of five NiCd or NiMH cells in series. This battery provides a 6 to 5V source for a fully charged to fully discharged condition. These phones currently develop a variety of supply rails from the single battery supply using three to seven linear voltage regulators and perhaps two charge pumps for negative rails. (Much of this voltage regulation is not especially efficient; for example, a 3V linear voltage regulator operating from 6V is 50% efficient.)

However, this overall distributed power scheme reflects the various supply requirements of different device technologies in any one product (Figure 1). For example, a typical subsystem includes an RF power amplifier (PA), a display, analogue and digital circuits, a subscriber-identity-memory card, and a voltage-controlled oscillator/phase-locked loop. Each subsystem sets its own, often different, power requirements—and these requirements present a moving design target as individual subsections undergo a technology hop.

Five- to three-cell hop

Five cells’ providing 6V gives mobile designers a degree of freedom in tracking subsystem evolution. In contrast, next-generation phones, and those currently under design, will severely restrict that freedom by using four (or even three) cells to power a similar combination of hardware. A three-cell source means the charge-to-discharge voltage becomes 3.6 to 3V, and this battery voltage range is the major focus of today’s battery-management designs. The great attraction of using this number of cells is the prospect of swapping the three NiCd or NiMH cells with one Li-ion 4.2V cell.

Despite their high cost and requiring close control of charge and discharge (Reference 1), Li-ion cells offer the biggest single advance in battery power capacity per unit volume and weight for the near future. Some top-end phones on the market already use Li-ion, but as a two-cell battery to produce approximately 8 to 4V. This range is neither convenient nor efficient, because the large regulator I/O volt-drop throws away some of Li-ion’s power-capacity advantage. One-cell Li-ion is the principal market driver, although a three- to four-times cost penalty (including the charger) precludes widespread adoption for mid-range models.

Li-ion cells also have drawbacks associated with high internal resistance. This inherent feature poses particular problems for some mobile-phone systems, such as GSM, that draw large pulse power. One solution is to use Li-ion cells in parallel, but then you’ll need cells with a matched internal resistance—and again you’ll face a cost penalty. So, phone designers’ strategies include working hard on the three-cell NiCd/NiMH design, knowing that Li-ion cell vendors are furiously developing their chemistry. Li-ion costs will fall, as will internal resistance, and, when the time is right, a single cell will neatly replace the three-cell design to provide an immediate battery power gain of approximately 250%.

Apart from short-term Li-ion concerns, moving from a 6 to 3.6V battery seems straightforward enough, especially when you consider that mobile phones increasingly use low-power 1.8 or 3V devices. But that simple five- to three-cell transition throws battery-power-management designs back into a melting pot. And designers are left to ponder the following:

• With a 3V battery, how do you power an existing 5V RF power amplifier?
• Can linear regulators operate reliably with battery-voltage inputs down to 3V?
• How can you maintain circuit block isolation and easily switch off idling circuits without using linear regulators?
• Using those nasty dc/dc switchers begins to look inevitable, so how about the effects of switching noise?

Thus, battery management now looks more complex—and in opposition to your overall requirement of reducing cost, size, and weight. In the end, you’re left with more questions than answers.

What is clear is that three-cell phone designs require a mixture of regulator types. Low-dropout regulators and charge pumps (in some form or another) remain essential requirements. But, whereas earlier designs could make do using only these two types of regulators, now designers have to accept that there is no alternative to introducing at least one inductive dc/dc switching regulator. Table 1 shows a selection of regulator ICs that includes all three types of regulator you’ll need for a complete mobile-phone battery-management design.

PAs pose power-pulse problem

Inductive dc/dc switching regulators frighten mobile-phone designers. The fear is that noise, both conducted and radiated, will modulate the phone’s baseband or RF signals. But, despite these fears, a step-up switcher is a must when your battery supply is at 3V and your PA requires 5V. Today’s 5V PAs use bipolar devices, and this technology will remain part of the next generation of 900-MHz mobile phones, despite threats from 3V gallium-arsenide (GaAs) PAs. Earlier five-cell battery designs took the easy option of powering the PA directly from the battery. (PA control circuits maintained power output levels as the battery voltage changed.)

Thus, you must not only come to terms with using a fearsome switcher, but also find board space for the switcher IC and its external parts. If you were planning to utilise the space vacated by the two missing cells, then forget it. Life is never that easy, particularly when you’re designing a GSM or DCS1800 mobile. The main problem is that the transmission mode is time-division-multiple-access (TDMA), which means the transmitter works on an 8:1 duty cycle. Essentially, the PA delivers a 577-µsec, 2W-pk antenna power burst every 4.6 msec. This RF power burst refers as a 1.5A surge on the 5V supply, and your step-up switcher must be ready to take the strain.

Switcher vendors have divergent solutions to this current burst-demand problem. Maxim suggests a circuit that delivers the burst from a massive 2000-µF reservoir capacitor on the 5V line (Figure 2). The capacitor is smaller and costs less than the two NiCd cells it replaces, and the 500-kHz switching frequency allows the switcher to use a small, low-cost inductor. Also, the switcher’s internal MOSFET negates the need for an external part. The phone design must allow the PA to switch off, and a switcher inherently has that feature, whereas earlier designs required additional parts.

Maxim’s switcher is approximately 80% efficient for output currents between 300 mA and 1.5A pk, which puts it in a similar efficiency class to low-dropout regulators operating close to their dropout threshold. By comparison, the efficiency of Philips Semiconductors’ switcher (Figure 3) ranges from 85 to 96% for power outputs of 3W continuous or 8W pk. By virtue of various switcher modes (Reference 2), this design can do some rapid internal gear-changing in response to an external trigger input.

For example, by thrusting the switcher into top gear in synch with the PA’s power burst, the circuit can momentarily deliver the 1.5A current surge. This high-surge capability in the switcher itself allows the design to use a 330-µF reservoir capacitor on the 5V output line. Because the circuit stores less burst power in the reservoir capacitor, more burst power comes directly from the battery. In general, current surges reduce the life of rechargeable cells, so this circuit adds another 330 µF on the input side to alleviate the problem. In addition to exceptionally high efficiency, the circuit usefully operates on battery inputs down to 1.6V, which prepares the way for replacement with a single-cell Li-ion.

With regard to the RFI fears associated with using an inductive switcher, designers do have some options. Naturally, you must take the usual layout, suppression, and screening precautions for minimising RFI (Reference 3). If you use a toroid in place of an open-frame bobbin inductor, then most RFI problems will be conducted rather than radiated. Philips provides a reference pc-board layout (30×20 mm) for its TEA1204 switcher that phone vendors are using without ill-effect. In general, other options include using a switcher’s fast on/off control to suspend switching at critical times, or to synchronise switching frequency with the phone’s baseband frequency, so that switching harmonics fall outside the phone’s IF passband.

Low dropout is efficient

Once you’ve developed a 5 to 5.5V rail from essentially a 3V source, you can turn your attention to supplies for other sections of the mobile. Although sections such as display, subscriber-identity-memory card, and PA all operate from 5V, these sections generally need isolation, and a linear regulator is a natural choice. Displays and subscriber-identity-memory cards that operate nominally at 5V also guarantee operation down to 4.7V, so you can use low-dropout regulators from the main switching-regulator output.

Similarly, low-dropout regulators are ideal for powering the general analogue and logic sections of the phone that now work at 3V or lower. Even nominally 3V devices operate at 2.7V, so, providing you select genuinely low-dropout regulators, you can drive the regulators directly from the three-cell battery. This arrangement provides an efficient design because you can allow the cell’s voltage to fall all the way to its 1V recharge threshold. Also, whereas earlier so-called "low-dropout" regulators weren’t quite low dropout, today’s devices achieve 120 mV at 100 mA, and that level of dropout is important in providing high efficiency.

For example, if you assume your three-cell source voltage declines linearly from 3.6 to 3.0V, and you also assume regulator output voltage is 2.7V minimum across the operating temperature range, then efficiency ranges from around 78 to 93%, yielding a mean value of 85%. Regulator quiescent current reduces these efficiency levels, but only <0.5% at full current load and generally no more than a few percent at 10% of full load (where power saving is less significant anyway).

Low-dropout regulators are simple to design in and require few external parts, but, nonetheless, pc-board layout is important. Low-dropout regulators employ very high ß transistors; if you do not follow a vendor’s layout rules, your design can easily oscillate. For the same reason, most regulator designs specify a stabilising output capacitor with a low ESR.

Analog Devices’ anyCAP regulators are the exception, and you can use any capacitor, even a low-cost 0.47-µF ceramic type. Also, anyCAP regulators adopt a "thermal coastline" chip layout that improves SO-8-package power efficiency by approximately 30%. Using SO-8 packages in mobiles is a prerequisite because size and weight constraints are major design drivers. Typical SO-8 packages dissipate 550 mW at 25°C, and often you need to add large copper pads to your pc-board layout to sink that power at high ambient temperatures. The anyCAP regulator design dissipates 1.15W at 25°C or 650 mW at 70°C without any external heatsinking.

Another important aspect of low-dropout regulators in this application is the accuracy of the output voltage. Working with such a small voltage overhead between input to output, you can soon lose the advantage of a 120-mV low dropout. For example, if regulator output tolerance is 5% as opposed to 1%, then you can’t allow your battery to discharge to the same level. This means that, in a group of regulators, the first regulator that hits its dropout voltage will directly pull the battery source and may immediately cause other regulators to hit their end-stops. Although individual circuit blocks of the phone’s subsystem may continue to function even under these circumstances, mutual isolation between circuit elements ceases to exist. Resulting interactions can cause a system failure, particularly if supply-voltage variations reach the phone’s voltage-controlled oscillator.

References

1. Schindler, Matt, "Proper handling helps make the most of Li-ion batteries," EDN, Dec 5, 1996, pg 179.
2. "Portable Power Battery Management Cookbook," October 1996, Philips Semiconductors.
3. Ott, Henry, Noise Reduction Techniques in Electronic Systems, John Wiley & Sons, 1988.

| EDN Access | Feedback | Table of Contents |

Table 1—Representative regulator ICs for mobile phones