Emerging Applications Spell the End of the Battery’s Life

New mobile applications, such as wearables and mobile gaming, mean there is a shift in power management at the system level, seeing alternative ways to meet performance levels and design challenges.

Applications such as autonomous driving and artificial intelligence are driving a demand for higher performance, high efficiency processors. These rely on processing images in real time and dissecting each image to identify and locate an element within that image to detect objects or to learn behavior.

Image processing and image recognition offer the possibility of creating new operations in other markets. Nick Pandher, Director of Market Development for Radeon Professional Graphics, at AMD, believes we are only seeing the tip of the graphics processing iceberg for deep learning. For example, he suggests, they could be used in financial institutions and organizations to model a training framework in a financial data set. This would perform the analysis usually done by someone with a financial background. It can also be used to look for anomalies in staff log-ins to flag issues and highlight at-risk areas in an organization’s operations.

There are also medical uses for treatment, where anomalies can be quickly identified, and in research, where patterns can be detected across multiple data frameworks to link symptoms.

For wearable devices, the same restrictions on weight, size and power apply as in mobile gaming processors. They have to be light enough to be worn during fitness activities or light and unobtrusive if used in medical monitoring.

Power Adopts a Game Face
AMD has based its latest Graphics Processing Unit (GPU), Radeon Mobile, on a 14nm FinFET technology, to meet the form factor and power demands of emerging markets, such as mobile gaming.

Mobile gaming places different demands on a GPU than a desktop application does. For example, for mobile, the GPU has to be light in weight and small in size to integrate into mobile devices or Virtual Reality (VR) headsets. It has to be thermally efficient, as a fan will add weight and space restrictions, yet have a laptop’s performance.


Figure 1: Mobile gaming is expected to account for half of the revenue generated by the video game industry worldwide by 2020. Picture Credit: AMD

FinFETs are 3D Field Effect Transistors (FETs), named after their fin-like structure rising above the substrate. The transistor’s gate wraps around the fin to reduce the amount of current leaking when the device is in the off state. This approach lowers threshold voltages to improve power consumption without increasing the die size.

Scott Wasson, Senior Manager of Technical Marketing, AMD, confirms the reason for the choice of transistor: “The key thing for anyone building a chip like [Radeon Vega Mobile] is to keep voltage as low as you can. . . . Radeon WattMan [AMD’s power management, based on Radeon software which controls GPU voltage, clocks, fan speed, and temperature], can be used to tweak and tune the voltage,” he says.

By adding “a few bits of special sauce” to the earlier Polaris architecture, the company has improved switching speed and performance in the Vega mobile architecture, explains Wasson. “It is very important to be always the refining power management algorithms we build into hardware and software,” he says. “The essential strategy is to provide performance when needed and to turn down the clocks, and the power, when you don’t need the performance, in order to conserve power,” he adds.

While the Vega Mobile, announced at CES last month, is not VR-ready yet, it is built to be small and relatively low power to meet the benchmark for VR in anticipation of what AMD’s partners will develop for VR and mobile gaming.

Wearable Challenges
For wearable devices, the same restrictions on weight, size, and power apply as in mobile gaming processors. They have to be light enough to be worn during fitness activities or light and unobtrusive if used in medical monitoring. For both, they should be wireless too, so that the wearer can record or gather data without being tethered.

Products such as watches, trackers, and monitors rely on a battery for power, but the processor’s power system must be able to regulate voltage from the battery. The problem is that the battery runs down, so the system has to manage a source with a declining voltage output. Some wearable device functions need a higher voltage than the 3.2 to 4.2V typical of a rechargeable Lithium Ion battery. Many wearable products use main power rails that are below the minimum charge of a single cell Lithium Ion battery, so the rails are sourced by a step-down regulator, possibly more than one.

Maxim Integrated introduced the MAX14690 battery charge device last year (Figure 2), targeting low-power, wearable applications. It has a linear battery charger with a smart power selector, two low- power buck regulators, and three low-power, Low DropOut (LDO) linear regulators.

Figure 2: The MAX14690’s level of integration minimizes footprint for power management in wearable devices.

If the device is connected to a power source, the power selector allows the device to operate when the battery is dead. The input current to the selector is limited, based on an I2C register, to avoid overloading the power adapter. If the charger power source cannot meet the supply needs for the whole system load, the smart power control circuit can supplement the system load with current from the battery.

To conserve power during periods of light load operation, the synchronous step-down buck regulators have a burst mode option and a fixed frequency Pulse Width Modulation (PWM) mode to regulate the load. The output can be programmed using I2C bus.

The LDO linear regulators can also be programmed via I2C and configured to operate as power switches to disconnect the quiescent load of the system peripherals for power management.

This is all packed into a 36-bump, 0.4mm pitch 2.72 x 2.47mm Wafer Level Package (WLP). Maxim Integrated also offers the MAX14690 Evaluation Board, an assembled and tested circuit for evaluating the device.

Battery Management in the IoT
The nature of the Internet of Things (IoT) means it has its own low-power requirements. Devices are wireless, mobile, often located remotely, and rely on batteries. In many designs, the battery is the sticking point. The battery can be expensive to replace. Additionally, the remoteness of the IoT node, either geographically, or in a hard-to-reach spot in a building or factory, can make battery replacement a time-consuming exercise. Hence the concentration by many Power Management Integrated Circuit (PMIC) manufacturers to take a system-level approach to power management.

The vision of the IoT is for hundreds of billions of nodes to be still active into the next century, despite being located in hard-to-reach, inaccessible, or hostile locations. For Dr. Peter Harrop, Chairman of market research firm IDTechEX, this means batteries will have to go (Battery Elimination in Electronics and Electrical Engineering 2018-20128). In a series of reports, he points out that batteries have “serious limitations of cost, weight, space, toxicity, flammability, explosions, energy density, power density, leakage current, reliability, maintenance and/or life.” He is clearly not a fan. He continues “Lithium Ion batteries will dominate the market for at least 10 years, and probably much longer, yet no Lithium Ion cell is inherently safe and no Lithium Ion battery management system can ensure safety in all circumstances.”

Energy harvesting is an alternative, practical approach to eliminating batteries. A Power Management Unit (PMU) converts DC power from one energy source to another. For example, the ADP5091 and ADP5092 (Figure 3), by Analog Devices can be used in PhotoVoltaic (PV) cell energy harvesting, ThermoElectric Generators (TEG) energy harvesting, industrial monitoring, and self-powered wireless sensor devices as well as portable and wearable devices.

Figure 3: Analog Devices takes a system-level approach to IoT battery management

The PMUs harvest from 6µW to 600mW, with an internal cold start circuit which allows input voltage down to 380mV. In both devices, the charging control function protects the rechargeable energy storage by monitoring the battery voltage with the programmable charging termination voltage and the shutdown discharge voltage.

There is the option to connect a primary cell battery, managed by an internal power path management control block. This enables the power source to switch from the energy harvester, rechargeable battery, and primary cell battery.

Figure 4: The 24-lead LFCSP ADP501 by Analog Devices could eliminate batteries in the Industrial IoT

The company also offers the ADP509 evaluation board, based on the ADP509PMIC and the Alta Device PV cell. It includes a PV panel and power management to enable devices to be powered by energy harvesting.

Caroline Hayes has been a journalist covering the electronics sector for more than 20 years. She has worked on several European titles, reporting on a variety of industries, including communications, broadcast and automotive.


Share and Enjoy:
  • Digg
  • Sphinn
  • Facebook
  • Mixx
  • Google
  • TwitThis
Extension Media websites place cookies on your device to give you the best user experience. By using our websites, you agree to placement of these cookies and to our Privacy Policy. Please click here to accept.