ARM Cortex-M4-based microcontrollers bring energy efficiency and high performance to intelligent applications

3Many designers are on a quest to embed intelligence into even the most mundane objects as they try to equip smart homes with intelligent connected devices and sensor networks that link to the burgeoning Internet of Things (IoT). Other embedded designers focus on increasing the capabilities of already-intelligent consumer electronics products or on making portable versions of sophisticated sensing devices that require energy efficiency. Though these intelligent devices often need to run on battery power for a long time – sometimes for a period of years – many also require high performance due to their complexity.

Today’s Microcontroller (MCU) market is abuzz with talk about ultra-low energy consumption measured in Nanoamps (nA). All of these power-sensitive embedded applications need MCUs that will function reliably. This trend has culminated in the recent launch of the ARM M0+ processor, a stripped-down, ultra-low-power processor core intended for just these applications.

However, MCUs based on the more powerful ARM Cortex-M4 processor with an integrated Floating Point Unit (FPU) are also being used. This core offers more performance, including floating-point and DSP instructions. Most MCU vendors position their M4-based products as high-performance solutions supported by complex software libraries, assuming that truly energy-sensitive applications will require a stripped-down processor.

So why would anyone choose a low-energy MCU equipped with FPU and DSP functions? The first reason is that, perhaps counter intuitively, such a processor may in fact prove to be more energy efficient than a less powerful device. The availability of floating-point and single-cycle multiply-accumulate instructions often allows the designer to reduce execution time, or reduce the clock frequency to accomplish the same workload. Put simply, expending 10 percent more power for 20 percent less time represents an energy saving overall.

21
Figure 1: The saved energy (blue area) increases when an application is optimized by increasing the processing speed and reducing the active and sleep current.

This effect is accentuated in devices with well-designed sleep modes that deliver very low power. For instance, the Silicon Labs Wonder Gecko MCU has five distinct low-energy modes, including a 20 nA shut-off state and 950 nA deep sleep mode (running real-time clock, full RAM and register contents retained and brown-out detector enabled). The bigger the difference between active and sleep mode power consumption, the greater the benefit will be of a rapid return to a low-energy state.

Ultrasonic water metering is one example of such a use case that, in addition to signal processing, also requires outstanding sleep mode performance, as well as battery life measured in years. In such an application, an ARM Cortex-M4 based MCU can be configured with the CPU in sleep mode and its peripherals set up as a kind of analog state machine that wakes the Cortex-M4 processor only when water is flowing. The availability of DSP instructions allows the designer to build sophisticated filtering functions that come into action when there is water flow to measure, eliminating the need for expensive ultrasonic transducers.

High performance

Some applications simply need the processing horsepower of a DSP. Consider, for example, a security device that senses glass breakages using acoustic analysis. The sensors rely on both audio and shock, or vibration, of shattered glass to determine if someone is breaking in.

For such sensors to reliably report break-ins and not be triggered when keys fall to the floor, the telephone rings, or even when a drinking glass breaks, the sensor logic needs to perform numerous complex operations. To positively detect and confirm a window breaking, the sensor also needs to analyze pre-break actions: was there an impact or flexing of the glass prior to shattering? Secondly, the sensor needs to consider the shattering frequencies, the audio of breaking glass, within a defined time frame. After filtering the frequency information, the duration and amplitude of the signal are checked to provide further verification of a valid alarm condition.

Low power without sacrificing performance

The MCU used in a glass break detector performs fast Fourier transforms on the output from a wideband sound transducer to determine whether a breakage has taken place. The use of a DSP-enabled MCU allows glass break analysis to be performed much faster than software-based solutions. And with a sensor interface that only wakes up the processor when actual glass-break frequencies are detected, the total energy consumption can be reduced as well. This in turn enables sensors to become wireless and battery operated, making them more tamper-proof and easier to install and retrofit into existing security systems.

22
Figure 2: Applications save energy when staying in sleep energy modes while FPU- and DSP-enabled MCUs solve tasks faster. A 512-point FFT is 3x more energy efficient on the Cortex-M4 to the right.

Because the glass break detection system needs to run for a period of years from a single battery, it must be designed so that the MCU wakes up and starts processing only when a possible acoustic event has occurred. The chip’s fast wake-up capabilities facilitate this energy efficiency, and its DSP capabilities minimize wake time while allowing the system designer to choose a detection algorithm that is robust against false alarms.

Portable medical equipment provides another typical example of a power-sensitive application that requires long battery life without sacrificing performance. Battery-powered ECGs are becoming increasingly popular portable medical devices, but the fact is, to obtain a really accurate read-out of cardiac health, the patient often needs to visit a hospital or health clinic for a comprehensive ECG. The processing power of many current portable devices, including ECGs, is restricted by the power and energy budget imposed by battery-based operation. As a result, designers compromise system performance by reducing sampling rates. DSP-equipped low-energy processors present a practical solution to this design challenge by enabling higher system performance without significantly reducing battery life.

Achieving the right balance, for the right price

Power-sensitive applications like glass break detectors and portable medical devices demonstrate that low-energy operation is not always about paring processing power to the bone. Tomorrow’s energy-efficient products require high-performance, FPU-capable MCUs offering the right balance of processing capabilities, low active power consumption, well-designed sleep modes, and optimized, autonomous mixed-signal peripherals.

Such considerations also come into play in the field of smart sensing and the IoT. As MCU prices have dipped below the dollar mark, it has become increasingly possible to put intelligence into everyday objects. A DSP-equipped MCU provides a special type of intelligence that can be used for signal conditioning purposes. By siting such processors at the location where each signal is captured, designers can choose lower cost sensor types, increasing the range of applications that can cost-effectively be addressed.

Rasmus Christian Larsen, Applications Engineering Manager, Microcontroller Products, heads Silicon Labs applications engineering teams for EFM32, EFR and Ember ZigBee solutions. Based in Oslo, Norway, he joined Silicon Labs in July 2013 with that company’s acquisition of Energy Micro, the leading supplier of energy-friendly ARM based solutions. As one of the designers of Energy Micro’s first microcontroller series, the EFM32 Gecko, Rasmus has in-depth knowledge of the complete EFM32 product family. He previously worked as a digital design engineer for Atmel AVR and has a master’s degree in electronics from The Norwegian University of Science and Technology (NTNU) in Trondheim, Norway.

Silicon Labs

www.siliconlabs.com