First, the good news. Mixed-signal design is everywhere, enabling exciting new applications in medicine, communications, transportation, solar power, and the "smart" energy-efficient home. The bad news: As we scale down to advanced process nodes, design at the analog/digital interface is more challenging than ever.
This good-news, bad-news situation was explained in depth by Prof. Terri Fiez (right) at the Mixed-Signal Technology Summit held at Cadence San Jose headquarters Oct. 10, 2013. Fiez, professor of electrical engineering and computer science at Oregon State University, gave an "academic keynote" titled Challenges in Emerging Mixed-Signal Systems and Applications. Following her keynote, an industry keynote was given by Geoff Lees, senior vice president and general manager for microcontrollers at Freescale (see Brian Fuller's blog post for more information about the Lees keynote).
Fiez opened on an optimistic node, telling the audience that "many of the things you are working on have changed the world." Existing and upcoming technologies include medical devices such as pacemakers, communications facilitated by wearable computing, the energy-efficient "smart" home, and the pervasive electronics systems in cars.
"Finally, the world of energy is really waiting for the mixed-signal community to dive in, and in many ways it already has," said Fiez, who is a co-founder and past CEO of Azuray Electronics, a startup that was launched to develop smart panel electronics for solar energy systems.
The Challenges of Scaling
With scaling, more and more functionality can be placed on a single chip - and it's not just digital functionality. Full system integration also encompasses analog and RF, and it should extend to high-voltage and high-power circuits. It's becoming more important to integrate passives to get the cost down. "So with Moore's Law," Fiez said, "the biggest challenge is incorporating a lot more technologies in order to get full system integration."
Moore's Law holds that microprocessors will double in relative performance every 1.5 years, but analog-to-digital converters (ADCs) don't follow the same curve, Fiez said. In fact, there is a "big lag." According to a graph she showed, it takes ADCs 4.7 years to double in relative performance.
When it comes to scaling for analog processes, "most of it is not good news," Fiez said. For example, reduced supply voltages make it harder to get the dynamic range that's needed for high-power, high-voltage operations. "What this means is that there needs to be a lot of ingenuity," she said. "We need good models and good tools to overcome limitations."
Making Solar Smarter
Drawing on her experiences at Azuray, Fiez talked in depth about the "smart solar panel." In a traditional solar power system, the solar panels produce DC and this goes through an inverter. The inverter is a separate box that converts the DC to AC, producing energy that can be used in the building or home, or fed into the grid. (I have one of these systems myself - see photo at left).
But what if we could put this intelligence on the solar panels themselves? A smart panel does the conversion to AC right on the panel, and synchronizes with the grid. As a result, you can get information about the panel that was never available before - monitoring how much energy is harvested out of every panel, for instance. And if one panel is shaded, you don't lose the entire row, as is typical today. Finally, an inverter box has electrolytic capacitors that are prone to failure, and the smart panel gets rid of these.
To interface with the grid, however, you need devices that run up to 1,000V. Power electronics have normally been discrete, and have not typically worked with digital intelligence and control. It's also important to minimize the number of passive components in order to reduce costs. So how do we architect all this?
"The Holy Grail is when we can integrate high-voltage devices, and it's all a complete SoC," Fiez said. "But it's not just the analog, digital, or RF circuitry; there's also a software component. So the challenge is, how do we design, simulate and validate the entire system?" Being able to model the grid, and to develop an interface that operates with the grid, is a tough challenge, she added.
Wireless Sensor Networks - Making Their Own Energy
You really don't want to replace batteries on a wireless sensor. For this reason, Fiez and other researchers are looking for ways to "harvest" energy from the environment. Fiez identified several energy-harvesting methods, including RF energy, solar, piezoelectric and vibration, thermoelectric, and acoustic. All of these methods require mixed-signal interfaces.
In her keynote, Fiez focused mainly on RF and piezoelectric energy harvesting. In the direct piezoelectric effect, an applied force produces a charge. In a converse piezoelectric effect, an applied voltage produces a deflection. Fiez showed some results that were obtained from modeling piezoelectric circuits. She also showed how her research group determined that an active rectifier would be more efficient over a broader range of low current than a passive one.
With RF energy extraction, an RF power signal is generated and a harvester converts that signal to DC voltage. However, the available power drops off sharply as the distance from the transmitter increases. Harvestable RF energy is also limited by the diode threshold. But there are technologies that can make RF harvesting more possible. For example, a floating gate rectifier can reduce threshold voltage, and the use of "cascading" to increase the number of rectifier stages allows more energy extraction and long distance operation.
Finally, Fiez talked about trends in ADCs, a technology she has worked with for most of her career. ADCs are essential, she noted, to connect a system to a real-world signal of any kind. But ADC design is challenging due to analog "imperfections" such as non-linearity and mismatch of devices. Fiez looked at challenges and trends in various ADC types including flash converters, pipeline converters, successive approximation ADCs, and delta-sigma ADCs.
In conclusion, Fiez said, work is still needed in technology, tools and education. "I think that tools are the biggest challenge because it's a moving target," she said. "Every time Cadence gets it figured out, there's a whole new technology. So they have a lot of opportunity in the future and they have done a great job already. I think it's an exciting space going forward."
Related Blog Post
"The Last Simple Node" and its Profound Consequences