Richard York was appointed director of embedded processor products in December 2007. Richard is responsible for the team marketing ARM’s embedded and microcontroller CPU products including the Cortex-M and Cortex-R processor series. He is responsible for the embedded roadmap and surrounding … More »
The Growing Role of Analog-Digital On-Chip Integration in Saving Energy
August 20th, 2014 by Richard York
Mixed-signal silicon design, bringing the worlds of analog and digital technology onto a single die, has never been an easy task. Formerly, the analog and digital teams would work independently on their designs, leaving the place and route team with the thankless task of integrating everything onto a single chip. A microcontroller design, with all of its carefully thought out peripherals, would be routed leaving analog-sized holes for the oscillator, ADC and transceivers needed to complete the design.
The first real test of the design would be undertaken with the first silicon on the laboratory test bench, a potentially risky process that inevitably resulted in one or more metal fixes before mass production could ramp up. Thankfully, with the advances in sophistication of EDA design tools, with their sophisticated integrated mixed-signal analysis capability, the chances of finding potential issues when merging digital peripherals with analog blocks has significantly increased. In turn, the risks, and potential costs due to errors, have dropped significantly, making the development of silicon devices with peripherals and features highly attuned to vertical market needs, considerably more attractive.
Another factor driving such deep levels of integration is the quest for ever-higher levels of energy efficiency of applications and appliances. Within the EU, the 92/75/EC Directive on Energy Labeling has raised the awareness of energy consumption for consumers, providing easy to understand comparisons of the energy needs of various home appliances, air conditioners, lighting and motor vehicles. The United States of America operates a similar program under the Energy Star standard for products originating there.
Computer power supplies are also voluntarily certified against the 80 Plus standard, with some very demanding efficiency levels for the highest, Titanium, level of certification, as well as a better than 0.9 power factor performance. This focus on energy usage results in white goods manufacturers taking a very careful look at the microcontroller products they use. Who would have thought that a mains powered refrigerator design would demand a microcontroller with low-power sleep modes and semi-intelligent, autonomous peripheral modules to help contribute to an A+++ energy labeling rating.
All this means that traditional methods of circuit design, using pure analog or mixed-signal devices and potentially an external processor, can no longer be the method of choice to tackle such challenges. In order to shave further energy losses out of a design, the processor has to be integrated into the system with semi-intelligent and/or semi-autonomous peripheral modules in order that the right control choices can be made based upon the best possible measurement information available for the application.
The 32-bit ARM Cortex-M0 processor is well placed to provide the necessary processing intelligence to such systems. At 12,000 logic-gates, the die area required is negligible compared to the more area-intensive adjacent analog blocks (Figure 1). Thus, adding this area of intelligence to a design costs little, whilst adding an enormous amount of value to the end user, enabling them to develop products with a stronger value proposition. With the processor’s power consumption starting at just 16μW/MHz (90LP process), and a single cycle 32×32 multiply option, the Cortex-M0 core is well positioned to deliver an exceptional ratio of system-level power efficiency to power consumed.
Clever integration of processing core and semi-autonomous peripherals is displayed by Analog Device’s family of ADSP-CM40X devices (Figure 2). For example, for the control of PMSM motors, these mixed-signal control processors feature dual 16-bit A/D converters with 14 bits of accuracy. This provides an excellent starting point for accurate measurement of the current flowing into the motor. But accurate measurements alone are not enough; the moment of measurement is equally important to ensure an accurate picture of the status of the motor being controlled.
Figure 2: ADSP-CM40X block diagram
Firstly, the dual A/D convertor ensures that the two measurements taken are simultaneous, leading to greater control loop accuracy and enhancement of performance. Furthermore, the A/D is synchronized with the PWM module, ensuring that sampling occurs at the midpoint of the zero vector, providing the instantaneous average current that effectively suppresses switching ripple. The on-chip Cortex-M4 processor, with its floating point capability, can then use this accurate information to implement complex control algorithms leading to highly power efficient control of the motor.
Another case in point is Infineon’s XMC4000 series of MCUs (Figure 3). Targeted at, amongst other things, solar inverters, SMPS, UPS and motor control application, these devices feature the CAPCOM capture and compare unit, 12-bit ADCs, Delta Sigma Demodulator and PWM modules. In isolation, these modules are not of any special significance, but when combined with the integrated “connection matrix”, the modules can semi-autonomously undertake many control and measurement tasks, leaving the ARM Cortex-M4 core with DSP extensions, supported by the CMSIS DSP library, with the information and opportunity to enable highly power efficient solutions. And Dialog Semiconductor, as a new licensee of the ARM Cortex-M0 processor, plans to further expand its range of PMICs and battery management devices with integrated processing power.
Figure 3: XMC4000 series block diagram
As interest grows and the technology develops, the Internet of Things becomes a growing reality, promising further opportunity for advancements in energy efficiency. On the one side, IoT promises to make consumers more aware of their energy consumption by appliances sharing energy consumption information with smartphone and tablet applications. On the other, appliances, power supplies and motor control systems attain the ability of communicating with one another, potentially leading not just to unit-level power savings, but a collective power saving of an installed system of many units. An integrated processor is an essential element to enable such a future. One thing is for sure: the demand for carefully crafted mixed-signal silicon devices will continue to grow and with it, the need for competent, power efficient processing cores to complement them.