Smart drivers reduce energy use and PCB clutter in portable electronic systems

The push for low-power full-featured products

With the growing feature lists and performance levels that these systems deliver, there is also a growing need to manage power dissipation for at least three reasons:

1. As function and feature counts increase, functional density and, if unchecked, dissipation density increase as well. Improvements in semiconductor integration moderate this tendency but are not sufficient to offset the trend.

2. Despite feature expansion, pressures continue to push portable electronic products into smaller case sizes, again increasing functional and dissipation densities. Here the limits appear to be set by the mechanical requirements that the MMI determines and by company standards for mechanical and electrical robustness.

3. As case sizes shrink, so do available spaces for energy sources. Though improvements in battery technologies, particularly in Lithium-Ion energy sources, balance this trend somewhat with increasing storage density, space constraints have prevented OEMs from capturing significant increases in per-charge energy storage.

The strategies OEMs have used to address power dissipation issues have also evolved. The first-level strategy focused on the efficiency of the energy-management subsystem including minimizing losses through DC-DC converter, LDO, battery-management, and battery-protection circuits.

This power-subsystem-centric approach depended largely on the semiconductor vendors' ability to produce components and integrated devices that are less dissipative than similar structures on the market. This left the OEM engineer with what was largely a component-selection task, balancing energy efficiency against other concerns such as component cost and package size.

Although this strategy has been quite effective, the component market, for the most part, has already realized its benefits. As has been the case with most analog and analog-dominant mixed-signal ICs, components supporting this strategy have not benefitted significantly from the ongoing process shrinkages that digital and digital-dominant mixed-signal ICs have been driving.

The second-level strategy cycles off the power to sections of the system or even sections of large ASICs that are not in use at any particular time. This strategy has been particularly effective when applied to large energy users such as radio-link hardware and display backlights but also extends per-charge-cycle operating times by powering off even moderate loads such as audio subsystems, I/O ports, or nonvolatile configuration memory. Current production mobile phones, for example, operate with 20 or more power domains.

In addition to saving the power dissipated through idle current in largely dissipative circuits such as radios and display backlights, this strategy helps whittle away at quiescent dissipation whenever the system can shut down a clocked-circuit section. As IC fabrication processes have aggressively pursued previously unimaginable dimensions (the not-long-ago-impossible 90 nm node is now commonplace and semiconductor-fabrication-equipment makers are already working toward the third generation after it) this strategy has effectively replaced clock gating to reduce idle currents.

This dissipation-reducing strategy depends on engineering contributions from system architects, software and hardware implementers, and ASIC vendors. Though successful, this approach has been somewhat limited by the amount that additional features have added to the application processor's load, pushing designers to draw on larger and more power-hungry computational resources. For example, mobile telephone handsets have moved from ARM-7 to ARM-9 and ARM-11 processors as the baseband and ancillary processing resource of choice. Other portable electronic products have exhibited similar trends though to a lesser extent.

The third-level strategy focuses on reducing the energy that various features and functions use without sacrificing performance. One technique is to take advantage of distributed intelligence to manage functions and features that do not require the enormous processing power and speed of the baseband or application processor.

This strategy allows the processor to turn over entire functions to semi-autonomous peripheral controllers. The results are operating modes during which the processor can enter a sleep state between tasks that occur at human-activity rates, not the data-processing or communication rates for which the processor's full capability is necessary. A smart display-backlight driver serves as an example.