Digital power

Many devices already manage power digitally, such as cell phones, laptops, and other advanced handheld electronic products. They keep track of how much power the system is using, and by managing device performance and power supply operation, maximize efficiency and extend battery life. Soon anything with a battery in it will have digital power management by necessity, as the consumer will not accept anything but maximum battery life and performance.

Digital power is more important now than it was when it was first introduced years ago because the concept is now beginning to be applied to larger systems that draw their power from the mains.

Electronic device density has gotten so high that important subsystems such as server and telecomm racks are handling so much more power than when they were first implemented that designers are looking for any way to manage the system for maximum efficiency.

Unlike power management in a single device, a system-wide control protocol must be able to accept commands remotely (even via the internet) yet be able to control power use all the way down to the individual point-of-load converter on the circuit board. Considering that a modern board can have over 10 separate voltages, with different switch-on times and operating characteristics, multiply that by the number of units in a rack, multiply that by the number of racks in a facility, and you get the magnitude of the situation.

The "killer app" of digital power will be system-level power management where a data- or telecomm facility's power requirements will be dynamically scaled up and down to correspond to the operational load of the computing systems in it. If a data center is only using 30% of its capacity, it can be idled down to the board level of every sub-system, maximizing power savings and ensuring that when the power has to be restored it ramps up properly and smoothly. In addition, since the communications bus is two-way, each power converter in the system can report on its operating status and temperature to provide accurate fault prediction and recovery.

The concept is being explored at every level by different players, and there is currently a debate on the software and hardware protocols the industry will need to implement the technology. The problem is that, as in every other area of competition, companies want to field proprietary systems to "lock-in" customers to their particular solution, which is bad for the industry in that such an approach will make system integration between the various levels more difficult and slow adoption of the technology.