The art and science of RF and mixed-signal design
November 01, 2014
Over the past few decades, mixed-signal integrated circuit (IC) design has been one of the most exciting and technically challenging segments of the s...
Digital and memory ICs constitute about two-thirds of today's roughly $320 billion global semiconductor market. These ICs are driven by Moore's Law and cutting-edge CMOS process technology, which reduces the cost and increases the integration of semiconductor devices every year. Discrete and analog semiconductors account for approximately one-fifth of the global semiconductor market and are mainly served by older semiconductor process technologies as core analog components are costly to manufacture in newer process nodes.
Mixed-signal ICs account for about one tenth of the global semiconductor market. This estimate depends on how you count a mixed-signal IC, which can be defined as a semiconductor device that integrates significant analog and digital functionality to provide an interface to the analog world. Prime examples of mixed-signal ICs include system-on-chip (SoC) devices; cellular, Wi-Fi, Bluetooth and wireless personal area network (WPAN) transceivers; GPS, TV, and AM/FM receivers; audio and video converters; advanced clock and oscillator devices; networking interfaces; and, more recently, low-rate WPAN (LR-WPAN) wireless MCUs. Highly integrated mixed-signal IC solutions often supersede legacy technologies in established semiconductor markets when the required functionality and analog performance can be achieved at lower cost than is possible with discrete or other analog approaches. Even more important, a high level of mixed-signal integration greatly simplifies the engineering required by system manufacturers, enabling them to focus on their core applications and get to market faster.
Designing mixed-signal ICs
Mixed-signal ICs are not easy to design and manufacture, especially if they include RF functionality. A large standalone analog and discrete IC market exists because analog integration with digital ICs is not a simple, straightforward process. Analog and RF design has often been referred to as a "black art" because so much of it is done generally by trial and error and very often by intuition. However, modern mixed-signal design should always be considered more science than alchemy. "Brute force" analog integration should always be avoided, as trial and error is a very costly process in IC development.
The real "art" in mixed-signal design must result from a deep understanding of how the underlying physical interactio• phenomena manifests in complex systems combined with a robust and elegant design methodology founded on a digital-centric approach. The ideal approach unifies mixed-signal design and digital signal processing and enables the integration of complex, highly sensitive and high-performance analog and digital circuits without the expected tradeoffs. The powerful capabilities of digital processing in fine-line digital CMOS processes can be used to calibrate and compensate for analog imperfections and mitigate unwanted interactions, thus improving the speed, precision, power consumption, and ultimately the cost and usability of the mixed-signal device.
Moore's Law has been remarkably consistent for digital circuit design, doubling the number of transistors in a given area every two years, and it is still partially applicable in the era of deep sub-micron technologies. However, this law does not generally apply as well to analog circuits, resulting in a significant lag in the adoption of scaled technologies for analog ICs. It is not uncommon for analog devices to be still designed and manufactured o• 180 nm technologies and above. The reality is that the scaling of the process technology only partially drives the area and power scaling in analog circuits and sometimes even becomes a design obstacle. Actually, analog scaling is more often driven by the minimization of unwanted effects (such as statistical device mismatch or noise resulting from imperfections at materials interfaces), which is the result of quality improvements of the process itself. For this reason, mixed-signal designers prefer to rely on processes that are few steps behind the cutting edge of process technology, which can still improve device quality by relying on some of the latest technology advancements. In other words, the analog aspects of Moore's Law fall behind the standard digital approach. The situation is more dynamic, and the digital/analog technology gap can be partially compensated if it is still worth the investment of the IC technology suppliers (Figure 1).
The most suitable manufacturing process node for mixed-signal IC design lags behind the bleeding edge of process technology, and the choice of nodes is a trade-off of several factors, which ultimately depends on the amount of analog and mixed-signal circuitry included in the device. More precisely, a more digital-centric mixed-signal design approach enables the designer to leverage more advanced process nodes in order to resolve one of the most challenging commercial issues with analog circuit integration – the ability to integrate analog to reduce cost while increasing functionality. Design engineering teams at many leading semiconductor companies are actively pushing the boundaries of mixed-signal design and trying to address this challenge with novel solutions in which logic gates and switching elements are replacing amplifying and bulky passive devices.
The IoT and mixed-signal design
The Internet of Things aggregates networks of IoT nodes, i.e., very low-cost, intelligent and connected sensors and actuators used for data collection and monitoring in myriad applications that improve energy efficiency, security, healthcare, environmental monitoring, industrial process controls, transportation, and livability in general. IoT nodes are projected to reach 50 billion devices by 2020, and possibly reaching the one trillion threshold just few decades later. These astronomic market numbers pose serious constraints in term of engineering, manufacturability, energy consumption, maintenance, and ultimately the health of our environment. In addition to being available in extraordinarily high quantities, all of these IoT nodes must be very small, energy efficient, and secure, and they are generally not easily accessible to consumers for maintenance. IoT nodes often must be required to operate with very small coin cell batteries for a decade or more, or possibly relying on energy scavenging techniques.
These application requirements make the IoT node the ultimate candidate for very advanced digital-centric mixed-signal design techniques. The ideal IoT node will require state-of-the-art mixed-signal circuits to interface to sensors and actuators. They must include RF connectivity, use very power-efficient wireless protocols and require minimal external components. They also must include power converters to optimize power efficiency and cope with different battery chemistries or energy sources, all characteristics generally obtainable with more mature process nodes. At the same time, these IoT nodes will require moderately complex, ultra-low-power computing resources and memories to store and execute applications and network protocol software, which is better addressed with finer technologies. The current instantiation of such a paradigm is a mixed-signal IC that is widely known as a wireless MCU: an easy-to-use, small-footprint, energy-efficient, and highly integrated connected computing device with sensing and actuating capabilities.
The proliferation of ultra-low-power wireless MCUs is critical to the advancement of the IoT. Wireless MCUs provide the brains, sensing and connectivity for IoT nodes, from wireless security sensors to digital lighting controls. The art and science of mixed-signal design is the key enabler for the development of next-generation wireless MCUs that bridge the analog, RF, and digital worlds and maximally leverage the power of Moore's Law, without compromises in performance, cost, size, or power consumption.