The mobile-phone phenomenon continues to morph. Just look at the latest technologies insinuating their way into this arena— High-Speed Downlink Packet Access (HSDPA), Global Positioning Systems (GPS), and TV-on-mobile—all of which have increased consumers' expectations of mobile phones.
As a result, operators have begun marketing mobiles as multimedia devices, rather than simply a means of communication. Accordingly, recent predictions have TV-on-mobile becoming one of the next "killer applications" for handsets.
TV-on-mobile has progressed rapidly from the initial videodelivery services that were envisioned on the first 3G networks. Low bit-rate and resolution standards have advanced to the high-quality services available today. In early 2006, a U.K. trial revealed that, on average, users spent 23 minutes per session and one to two sessions per day watching television on their mobile phones.
In addition to TV-on-mobile being used by consumers on the go, the service is really set free when connected to larger displays. For example, ARM sees it being used in vehicles where the TV-on-mobile service is enabled via a head-rest display.
Whatever the specific use of TV-on-mobile, it's clear that the user must have a sustainable experience, whilst still being able to use the phone for its primary purpose of voice calls. Therefore, mobile-phone and IC designers have had to address the issue of power consumption.
Low power consumption is now one of the most important characteristics of embedded processors. In mobile phones, the applications processor plays a central role in managing the efficiency of the most power-hungry features. However, in highend mobile phones, such as the Nokia's N93 handset, the display may not be most powerhungry component in the system. During a phone call on a mobile or Wi-Fi network, the RF frontend and baseband will likely be the most power-hungry component, followed by the processor (typically 300MHz and above) and then the display.
In many use cases, particularly when it comes to video and audio, the applications processor itself can be the primary consumer of power. That's because the application processor is involved in almost every interaction of system components. For example, all Bluetooth and Wi-Fi data transfers will eventually go through the applications processor. Therefore, it's essential that the application processor's power be managed in order to cut power consumption.
Two crucial issues crop up when addressing power. Power is made up of two parts (Fig. 1). The first is active power, also referred to as "dynamic" power dissipation, which is consumed during active processing time. The second is "static" power dissipation, also referred to as leakage, and it's the power consumed when the processor is in idle or standby mode.
Active power directly affects the phone's talk-time, audio, video, and TV-on-mobile playback time. Static power directly affects standby time. A typical desktop processor wastes approximately 25% of its power due to leakage.
Dynamic power can be managed by a technique called Dynamic Voltage and Frequency Scaling (DVFS). DVFS lowers the frequency and voltage of the processor in real-time (dynamically) to match the workload on the processor. When used with a control mechanism, such as ARM Intelligent Energy Manager (IEM) technology, DVFS can save 20-25% of the system power.
Leakage has become a primary issue with large designs at 90nm and with nearly all designs at 65 nm and beyond. At these geometries, thinner oxides and shorter effective channel lengths result in power escaping, or "leaking" off the transistor. It then starts to act more like a sieve than a switch—the current paths continue to conduct, even when the device is turned off.
Now, particular attention is being paid to two sources of leakage. One is sub-threshold leakage (ISUB) between the source and drain (a result of the inability to close the channel completely); the other is tunneling leakage from the gate through the dielectric to the channel, known as gate leakage (IGATE) (Fig. 2).
Designers are investigating various solutions to overcome these problems. Sub-threshold leakage can be reduced by lengthening the channel, by biasing a well under the channel or by fabricating multiple-gate devices. Gate leakage can be reduced by switching to thicker, high-k dielectric materials and by using new gate materials, in bids to reduce the tunneling current.
Sub-threshold leakage dominates the overall leakage numbers. There are several ways to reduce sub-threshold leakage: Reduce the supply voltage (VDD), or use dual VT flows, power gating, or variable threshold CMOS (VTCMOS). The most effective methods of reducing sub-threshold leakage are power gating and VTCMOS.
Power gating principally isolates any inactive transistor by switching off its supply during periods of inactivity, therefore stopping leakage current from occurring in the switch. Power gating comes in fine-grain and coarse-grain formats. Fine-grain power gating places the switch in a transistor cell. Coarse grain, on the other hand, is shared amongst a number of cells. There's less overhead size per cell with coarse grain compared to fine grain. Nonetheless, coarse grain is more difficult to implement. Power gating is common in many SoC designs today. Texas Instruments' OMAP2420 processor has power gating for several of on-chip components, including the CPU, DSP, and hardware accelerators. Freescale's iMX31 Multimedia Applications Processor features similar techniques.
VTCMOS is another effective form of sub-threshold leakage. VTCMOS back-biases the substrate to reduce leakage. This is more complex to implement, but can reduce leakage by an order of two or three.
Simply put, gate leakage cannot be ignored. Over the past 40 years, the materials used to manufacture the transistor have remained relatively unchanged. However, new materials and architectures improve the amount of current that can be carried across the transistor while minimising the power needed to operate the device.
One of the first technologies to combat leakage, developed by IBM, used silicon-on-insulator (SOI) technology. An insulating layer sits below the transistors inside a chip and decreases power consumption by preventing current from leaking out. An SOI microchip's processing speed can be 30% faster than today's complementary metaloxide semiconductor (CMOS)- based chips, while power consumption can be reduced by 80%, which makes them ideal for mobile devices.
Furthermore, SOI is considered one of the better available technologies for making chips that offer high performance without the heat and power penalties that occur when using a standard substrate.
On this front, ARM recently acquired SOISIC. SOISIC is a leading company in Physical IP based on SOI technology, and plans are to integrate the company into its Physical IP Division.
FUTURE POWER TECHNOLOGY
In fact, ARM is working on several solutions to address both active power and leakage power consumption. The company expanded its on-chip component product line, which addresses low power. For instance, ARM AudioDE technology is a configurable audio engine with low power requirements when in operation.
ARM Artisan physical IP forms the library or cell building blocks for many SoCs. Artisan physical IP can help manage power in two areas. The first is through cell libraries that support DVFS for dynamic power (where the cell libraries are characterized for multiple voltage points). The second area is through the static power, using cell libraries that are able to help in reducing leakage. Artisan Metro, Advantage, and Sage libraries support both dynamic and static power-management techniques.
The company's IEM technology, a combination of hardware and software, can reduce power and energy consumption on the CPU by up to 60%. In an overall system, such as a mobile phone, this can equate to a 20- 25% system savings (Fig. 3).
IEM implements advanced algorithms to optimally balance processor workload and energy consumption, which maximizes system responsiveness to meet end-user performance expectations. It works with the operating system and applications running on the mobile device to adjust dynamically.
Indirectly, IEM reduces leakage as temperature on the SoC reduces. It can equally be applied to any hardware block to increase battery life.
To solve the problem of leakage in mobile devices, the onus is on IP vendors, semiconductor manufacturers, and mobilephone designers to deliver intelligent solutions to bring new technologies such as TV-onmobile to a mass audience.