Power management for the Internet of Things

BY ANAND BALAKRISHNAN
Systems and Applications Engineer
Freescale Semiconductor
www.freescale.com

The Internet of Things is a broad subject as made clear by the clever use to the word ‘things’ in its name. That makes the task of writing a piece focusing on a specific aspect of it is somewhat tricky. Here is my take on its system needs from a hardware power management perspective.
I see the Internet of Things as an amalgamation of two diverse but interconnected advancements:
1. The Internet, its infrastructure and its evolution: What has thus far been a means to communicate from end to end is being transformed to a carrier of untried amount of data (aptly named Big Data) generated from billions of interconnected devices
2. Transformation of everyday objects from dumb to smart: Things go from dumb to smart with embedded processing. As an example, the previously unsophisticated thermostat now has a powerful processor that learns about your preferences, communicates over the internet and in the long run saves you money by automatically optimizing your home energy usage. Such connected thermostats, speakers, cars and watches are becoming common place and as system engineers stretch the boundary of what constitutes the ‘things’ in the Internet of Things, one crucial enablement needed is flexible power management.
In this article, we will go over important features needed for system power management for the Internet of Things (IoT).

Creating an ecosystem of smart objects
Most smart objects (will) have certain key common blocks that can be categorized broadly as below:
 •  The embedded processor
 •   Memory
 •  Sensors
 •  Connectivity
 •  The user interface
These blocks are tied together using application software typically running on top of an operating system.
Consider a smart refrigerator, a smart watch, or a smart music system; it is easy to see that the hardware core of these can be broken within the above categories. Companies are vying to develop an ecosystem of smart, connected objects that will encompass every aspect of our lives. It helps that the hardware core is similar as that allows faster, scalable development and subsequent proliferation of these products. Hardware engineers can design intelligent platforms from which, with minor tweaks in hardware and software, different smart objects can be churned out.
From a power management perspective, it thus becomes crucial to have scalable, flexible solutions that can adapt to multiple smart systems.

Power management needs of a smart object
Traditionally power engineers are tasked to design power solutions meeting the following criteria:
•    High efficiency: High efficiency is important as that not only helps with energy savings but also helps with thermal management.
•    Low cost and size: It is easy to achieve high efficiency by over-designing a power converter. But engineers are also required to provide cost optimized solutions. Power devices, components such as inductors and capacitors need to be sized to strike a balance between efficiency, cost and size.
•    Voltage accuracy: Loads (microprocessor, memory, peripherals etc.) have strict voltage tolerance specifications. It is the duty of the power converter to control the output voltage within the acceptable tolerance levels under static and dynamic conditions.
•    Fault protection: Power converters are expected to protect against faults. Faults could occur due to damaged components, due to accidents (dropped your cell phone in the swimming pool anyone?) or other circumstances such as overheat. The power converter is required to protect the load and itself during such fault events.
Hardware designers for the IoT need to think about power management from a platform perspective. As discussed earlier, a common platform can be tweaked to form the core of multiple smart objects. 
Features required of a power management solution for a smart platform are:
•    Flexible power up timing and sequencing: Startup voltage of the different rails needs to be tuned depending on the processor operating frequency, type of memory and type of peripherals. The timing and sequence of the rails also needs to be tuned depending on the use case. For example, portable battery operated systems could have inrush current limitations which will require output voltages to come up at a slower slew rate and more spread apart in time. These limitations wouldn’t apply to an automotive system where faster the power up, the better. Typically the voltage, sequence and slew rates are controlled using external resistors and capacitors.
•     Dynamic voltage scaling: The voltage required for a processor can be lowered when it is running at lower frequencies to lower power consumption. Similarly it needs to be raised up when there is heavy activity and the processor frequency needs to be increased. This brings the need for a power converter to have the ability to change its output voltage on-the-fly. A change in voltage can be triggered either through a specific pin or through communication using I² C or SPI.
•    Event reporting: The power management solution is expected to monitor and report faults to the microprocessor. Going one step further, it is beneficial if it can warn the processor even before a fault occurs. Most power ICs have thermal shutdown protection whereby they force shut down if the die temperature exceeds a given threshold. Rather than shutting down, it would be beneficial if the IC can report to the processor prior to getting to the shutdown threshold. For example, if the IC is designed to shut down at 140°C, it could inform the processor as its die temperature crosses 110°, 12°, and 130°C so the processor can appropriately decrease its electrical loading thereby letting the user still operate the device albeit at reduced capability.
•    Built-in low-power modes and high light load efficiency: Smart objects spend a large portion of their time in lower power modes. There are short durations of high activity and long durations of low activity. The power management system must work hand in hand with the processor and optimize its own operating mode to decrease overall consumed energy. High light-load efficiency is also important for improving system reliability. Energy lost as heat raises ambient and die temperature. An increase in temperature speeds up aging. Considering that smart objects are required to operate for anywhere from 2 years to 10 or 15 years, even a small increase in temperature could have significant effects on life time reliability.
•    Internal compensation: Voltage regulators need to be compensated and are typically compensated using external resistors and capacitors. External components don’t help with flexibility. Voltage regulators with internal compensation are preferred when considering smart objects.
•    Scalability: The power requirements of a quad-core processor are different from that of a single-core processor. Most smart objects have software drivers handling the processor interface with the power management system. While designing the platform power solution, scalability is needed not only in terms of hardware but also software. A large portion of the cost of development of a smart object is software development. It is beneficial if a single software driver can be used across a family of power management solutions (read common register map).
Comparing discrete power converters to Power Management Integrated Circuits (PMICs), the above mentioned requirements to address overall system power for the Internet of Things easily tilts the balance towards PMICs.

Power fusion PMIC
Freescale Semiconductor’s power fusion (PF) series of PMICs brings advanced levels of configurability and programmability in a system-level PMIC solution, enabling a single device to be easily configured to provide power to a wide range of processors and peripherals. These PMICs have internal compensation and regulator output voltage is controlled internally through digital-to-analog converters compared to traditional converters which require external resistor dividers.

Figure 1 shows a high-level block diagram of the MMPF0100 PMIC. One time programmable (OTP) memory is used to store startup voltage, timing, sequence, and regulator configuration. This enables the PMIC to be used in multiple designs without any change in the bill of materials. The try-before-buy feature enables an engineer to test different configurations prior to programming the OTP device function.

The PF series of PMICs is ideally suited for the i.MX 6 series of applications processors though flexibility of the PMICs, thanks to OTP, allows them to work for a broad range of processors and systems. For more information, visit www.freescale.com/webapp/sps/site/taxonomy.jsp?code=CPMICSPFS

These PMICs are featured in a number of i.MX 6 reference designs that contain the necessary software drivers to handle different power modes. For more information, visit www.freescale.com/webapp/sps/site/taxonomy.jsp?code=IMX6X_SERIES&tid=vanIMX6

The MMPF0100 PMIC is featured in the Revolutionizing the Internet of Things (Riot) development platform. The RIoT board is a great starting platform for your Internet of Things application. For more information, visit www.riotboard.org/