While analog technology has been a vital part of our industry’s growth over the last 40-plus years, one might wonder whether it faces a future of relegation to an “interface driver” role during the next 40 years. We are seeing the opposite: for every analog signal processing function replaced by digital, there are two new analog functions springing up, and we expect this trend to continue or perhaps even accelerate in the coming decades.
One factor accounting for this analog expansion is the wave of new sensors capturing all sorts of signals that were previously missed or ignored. For example, consider the many sophisticated sensors in the new generation of automobiles, ranging from back-up cameras to oil quality detectors. Smart phones are another application in which there has been a virtual explosion of sensor integration. Each of these sensors presents new challenges for signal conditioning, and in many cases, triggers signal processing activities seeking to control systems that previously ran “open loop.”
The nature of signal processing also continues to evolve. When one considers the first 100 years of electronic signal processing, we find that in a great many cases, the fundamental challenge was extraction of a signal from “the noise floor.”
The signal transmission problem in the transatlantic telegraph cable is a classic example. While this sensitivity challenge continues to be important in many situations, increasingly, we see the signal processing problem focusing on extracting the signal from the heavy traffic of strong interferers. This creates the need for both enhanced sensitivity and selectivity. In conventional architectures, selectivity would be realized through filtering.
As systems move more toward broadband architectures to accommodate flexibility, narrowband filtering is no longer a viable means of blocking interferers., Advanced signal processing circuits need much greater linearity/dynamic range and at the same time must provide much greater bandwidth. Analog and mixed signal circuits need to meet these demands in order for the next generation of “all digital” systems to function.
What about the threat of “hitting the wall” in performance, or the end of Moore’s Law as we know it? The semiconductor innovation engine is not derailed that easily. For many generations, advancements in semiconductor technology were exploited in processor speed, but we have already seen the direction of innovation shift from one processor running faster and faster to the integration of multiple processors.
The theme of parallelism in signal processing is pervasive, and extends from MIMO (multiple input/multiple output) radios for wireless LAN devices to microprocessors in laptops. In many applications, architectures are moving to massive parallelism, where signals from hundreds of channels are combined to extract information. Phased-array radarsand medical imaging systems, such as ultrasound, are two notable examples. Again, these new architectures present new challenges for mixed-signal processing. Obviously, the number of signal paths increases (in some cases dramatically), but for these systems to work effectively, there are often requirements on the matching between paths.
The digital revolution is not driving analog signal processing to extinction. In fact, just the opposite is occurring: as the digital world continues its exponential expansion, the analog and mixed-signal functions are growing in number, performance and diversity. As always, the challenge of providing “more” (functionality, bandwidth, dynamic range) for “less” (power, size, cost) will drive the technology forward.
Click on image to enlarge.
is vice president for analog technology for Analog Devices Inc.