Wearable electronics market owes its rapid growth, primarily on the processor architecture ARM. Simple devices, such as fitness and smart bracelets watches Pebble use processor ARM Cortex-M3 (here and below the words “processor ARM Cortex-…” we will have a view of chipsets using these nuclei – approx. translation. ) for processing of information from the sensors, as well as auxiliary Cortex-M0, which acts as a Bluetooth-controller and consumes very little power.
In the premium segment of wearable electronics line chipsets ARM Cortex represented more broadly, including models Cortex-A5 and Cortex-A7, dealing with complex calculations, graphics cards, ARM Mali, as well as processors that handle video and display it on the screen – they are important for devices with large screens. In this energy-efficient Cortex-M0 processor is often used as a hub of information processing sensors.
Importantly, ARM as no other company feels wearable electronics market growth because it gives her and partners, including ST Microelectronics, Nordic Semiconductor, TI, Freescale, Qualcomm and other new opportunities for development. Bought or intend to buy wearable gadget? Most likely, it uses an ARM Cortex-A lines or Cortex-M. These chipsets use real-time operating systems (RTOS) in the case of low-power processors and more advanced operating system in the mid-and top-end processors segments.
Knowing processors Cortex-A and Cortex M, which play an important role not only in wearable electronics, but entry-level smartphones, you can better understand the purpose of their use in various devices: chipset Cortex-A7, which will be equipped with wearable gadgets and smartphones, are often quite different electronic circuits. Nevertheless, we want to start with a modest Cortex-M.
Embedded Processors line Cortex-M – the most energy efficient and are used virtually everywhere, often in a general-purpose microcontrollers, sensors and controllers, USB and Ethernet. These can be found in the network and office equipment, alarms, displays, power supplies, hard drives, and medical devices. And this is only a small piece of equipment that uses the processor ARM Cortex-M, volume of production in the billions per year.
Despite the fact that in the line of Cortex-M presented only four chipset, they have ample opportunities for individual configuration, which is based on the frequency of the processor and its use scenarios – ARM announces more than 3000 available configurations that offer its partners.
Second from the left in the above image processor, Cortex-M0 +, is the most energy-efficient chipsets of ARM. Based on the architecture ARMv6-M, Cortex-M0 + uses a two-stage ordered system of power, and compares favourably with the Cortex-M0 presence of block memory protection (he also has a Cortex-M3 and Cortex-M4) and single-ended input bus-input helps further reduce power consumption. Furthermore, the chipset has tiny size, and is tailored to its holdings of 12,000 logic gates.
Cortex-M0, meanwhile, uses a slightly more complex three-stage power supply system, which increases the computational capabilities of the chipset, but at the same time, increases power consumption. More simple design makes Cortex-M0 affordable chipset in the family ARM Cortex, which can be used for general computing and control input-output. There is also not shown in the diagram Cortex-M1, but it was designed specifically for FPGA (FPGA) chips.
Chipset Cortex-M3 and Cortex-M4 more have more powerful central (CPU) and signal (DSP) processors, effectively cope with the calculations of floating point numbers. Both chips are based on the architecture of ARM v7-M and can operate at speeds up to 200 MHz – it all depends on how fast they want to make the manufacturer.
Still, as far as these small and energy-efficient chipsets? ARM reveals rare maps, performance issues in this series, primarily because manufacturers can flexibly configure the chipset used by the kernel. Despite this, make a rough impression of the family ARM Cortex-M can be on the table above.
Made on 40-nanometer process technology Cortex-M0 consumes 4 microwatts per megahertz, and the transition to the Cortex-M0 + this value is reduced to 3 mW. Frankly, it is very small: it is enough to recall that microwatts – is a millionth of a watt. This way, when operating at 100 MHz (which is a lot) Cortex-M0 + consumes one ten-watt, with an area of the CPU 0.009 mm ². The smallest area of the chipset available from Freescale Kinetis KL03 and is 1.6 x 2 mm at a frequency of 46 MHz – at such a scale chip can fit in a recess ball golf.
Optimization of ARM Cortex-A for wearable electronics
It has already been said that the Cortex-A processors for smartphones and wearable electronics – two very different things. ARM provides its partners how to use the various IP (CPU registers) for each device, and it is not simply a change in the clock frequency or voltage drop, as one might think. In the end, wearable devices for stand-alone operation is much more important indicator than performance, so you have to use other technologies.
Shown in the image above combination of Cortex cores and Mali graphics accelerators are used in three types of devices: smartphones middle segment, budget smartphones and smart watches. The key point to note: processors Cortex-A7 can be used in all three types of gadgets that achieved by reducing their energy in different ways. But how exactly is this done?
The modular structure of nuclei chipset Cortex-A series allows manufacturers to design them according to high customization capabilities, especially in regard to energy consumption. When simplified, then, for example, processor cache is beneficial for productivity, but has a negative effect on consumption. Most chipsets for smartphones using core Cortex-A7, have a 32KB cache L1 (first level). Reducing its volume up to 16 KB can not only reduce the required chip energy, but also to make it more compact dimensions. In addition, these actions do not greatly affect the performance, especially for portable devices.
While the reduction in L1 cache able to lower power consumption chipset, it increases the load on the L2 cache (the second level).Nevertheless, continues to work with partners to reduce the size L1 cache is because the performance of portable devices hardly drops even when using a minimum of the second-level cache. Therefore, the development of chipsets for mobile devices, it is necessary to rely on reducing the cache to achieve low power consumption.
What else can make partners ARM? For example, they may refuse the use of an optional instrument NEON SIMD, harder to reduce the overall size of the chip. Despite this, such a decision is likely false economy, because it at least allows to reduce the area of the chipset, but adversely affect its performance due to integration of this tool with nuclei.
In addition, partners can consciously choose optimized, energy efficient manufacturing process. The implications of this are deep enough: as truncated potential frequency processor core does not turn out to disperse up to several hundred megahertz and maintain super low power drain (so called energy wasted during idle CPU) which can be reduced to 95%.
As has become clear in creating wearable devices must take into account the cache architecture and workflow. As a result, can achieve high efficiency and small size of the chip which would not be possible by simply changing the operating frequency and processor voltage.
Modular technology ARM allows its partners to create processors that are suited for specific tasks, so the chipset on the Cortex-A7 in the smartphone may be different from the same chipset in a handheld device. Moreover, there are no universal solutions.Innovative companies that conduct research, can create fundamentally better products, in comparison with ready “box” solutions.
As we will soon see that the architecture behind wearable devices evolves with them. Alas, there is no magic technology that can increase the battery life of smart hours to several weeks or months, but careful work on chipsets and new technologies allow manufacturers to endow their devices the ability to work longer between recharges all, while maintaining their performance.