|
||||||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||||||
|
||||||||||||||||||||||||||||||||||||
|
|
Reinventing The Role Of Analog/Mixed-Signal by Harry Davoody, Vice President Mixed-Signal Products, Texas Instruments, Inc. Not long ago, analog and mixed-signal functionality were treated as though they were an afterthought in the system design process. In emerging markets, systems are designed by first picking a central processing unit with the capabilities and performance needed, and then working outward by designing, with off-the-shelf components, the peripherals that surround the CPU. As markets grow, and reach a mid-life stage, developers begin to reinvent architectures by optimizing system building blocks and integrating functionality into fewer, more application-specific components. One outgrowth of this is a move towards mixed-signal technology which combines analog and digital functionality. Then, as markets become bigger still, the stakes increase dramatically. End-equipment designs become more sophisticated and user preferences change more rapidly. As a result, suppliers have less margin for error in their understanding of market dynamics as well as the product development process itself. At this stage, system optimization, not just CPU optimization, is of critical importance. Products that miss the mark often lead to missed-market opportunities. Missed-market opportunities all too often lead to disaster. At this point designers realize that the performance of peripheral functions, like analog and mixed-signal, is just as critical as CPU execution speed because the throughput of a system's peripherals can act as a gating function on CPU performance. Reinventing the Modem The evolution of the modem market provides a perfect example of this market-driven trend toward system architectures that are highly optimized and highly integrated at the component level. As with most electronics markets, the very early stages of the modem business didn't resemble a market at all. The existence of a market for modems had not been proven because users had yet to vote with their pocketbooks. Despite this, time-to-market was of the utmost importance during this early stage because suppliers were trying to stimulate the development of the new market. As a result, off-the-shelf, discrete, components tended to be used and system design focused on core technical issues: like processor functionality and performance. Analog and peripheral functionality was designed secondarily to the central processing unit. Cost, as always, was important, but it was not a paramount consideration. The pre-emergent days of the modem market occurred in the late 1970s and early 1980s. Back then a modem consisted of two printed circuit boards that were chock full of discrete integrated circuits. Some of the base modem designs were dubious at best and data transfer rates would creep along in the range of 300 to 1200 bits per second (b/s). Despite the relative shortcomings of the technology, users were not in short supply and the market took off. In order to win market share, OEMs soon began the process of optimizing their systems and integrating components, resulting in a doubling of modem speeds every two years. Mid-life Crises As electronics markets evolve, growth in volume removes all doubts about the market's existence. The increased dollars passing back and forth in the marketplace justifies a fresh look at system design to better meet user demands. Designers begin by looking at the existing building blocks and devising ways of combining these blocks to optimize on certain market-driven product goals. In many cases, such as the modem market, analog functionality is closely examined much earlier in the design cycle because now at the mid-life of a market all aspects of the system must be optimized to remove any and all bottlenecks. Mixed-signal technology is considered as a way of reducing chip count and improving reliability and performance. At this point cost takes on greater importance as competitors vie for market share with the assurance that a true high-volume, fully-customized market is just around the corner. High-level system optimization has not yet emerged as the ultimate market differentiator because most users still aren't tremendously sophisticated. Users are more concerned about functionality at a reasonable cost. The modem market of the early 90s presented an apt example of a market at mid-life. At that time a distinct transition-point or turning-point was reached in modem technology. The modems of the early 90s were hardwired solutions and as they approached the 9.6 kilobits per second (kb/s) performance level, they were running out of steam. Pressure was mounting to move on to the next performance node and the old designs using discrete components could not make the leap. The more complex software needed to send high-speed data over twisted-pair copper transmission lines would simply not run on the old hardwired modem architectures. To move up to 14.4 kb/s performance, TI and US Robotics re-invented the base modem design by using programmable DSPs and a data-pump architecture. This new breed of modems featured higher levels of component integration, a closer working relationship between the system's central processor, which was a DSP, and the system's analog/mixed-signal functions, as well as much more complex software that could compensate for the shortcomings of twisted-pair copper lines. Finally unfettered by limited hardwired designs, the modem market quickly went through several iterations of reinvention. Between the 14.4-kb/s performance node and today's fastest 56.6-kb/s x2 modems were seven generations of modem designs. Reaching a new performance node usually represented a new level of component integration and optimization. Along the way, the modem's standalone microcontroller, bus-interface circuitry, logic and memory were all integrated into one custom DSP (cDSP). We can expect this sort of trend to continue as modems approach another turning point in technology and move into the next phase of market development. Highly Customized Solutions Ultimately, markets grow and achieve a certain size that warrants total system optimization, customization and component integration. At this point, every aspect of the system, including analog and mixed-signal functionality, is examined and reinvented in light of market-driven goals and priorities During this phase of market development, competition is fierce. The size of the market has attracted new and substantial suppliers. Buyers have become more sophisticated; they know what they're looking for among end-equipment products and it's up to the OEM suppliers to provide it. This situation places tremendous pressure on all of the building blocks in a system. Eventually, those building blocks either move up a notch in performance, bringing the system's overall performance along with them, or the system architecture is reinvented once again. At this point another market dynamic comes into play. What has come to be known as the value web or technology complement takes on great importance. Quite simply, the market has reached such a large size that nothing can be left to happenstance. Everything in the system is related and must be treated as critically important. A value web is formed around successful platforms and the competitive landscape changes to a web vs. web model as opposed to a mere device vs. device competition. Another Turning Point Today, the modem market can be characterized as a highly-customized market which is fast approaching another turning point. Getting beyond 56.6 kb/s in performance is requiring another reinvention of the base modem architecture to take advantage of xDSL technology. xDSL technology -- with a performance range beginning at 384 kb/s and continuing up to several megabits per second -- holds great promise, but it makes great demands on all aspects of the system's architecture. System performance will suffer if designers concentrate solely on the raw horsepower of an xDSL modem's DSP central processor. Certainly DSP power is critically important, but that power can be mitigated by less than optimum mixed-signal functionality in the system's codec and the analog portions of the architecture, such as line drivers, if optimum trade-offs are not made up front. In addition, DSP design and embedded-software development for xDSL modems obviously go hand-in-hand. The development of xDSL-software algorithms is the new frontier for creative algorithmic optimization. The performance of the mixed-signal codec is especially critical to the performance of an xDSL modem. The codec's sampling rate must be fast enough, and its dynamic range precise enough, for the codec to keep up with the bandwidth demands of an xDSL modem. Without an advanced codec even the industry's highest performing DSP will have little bearing on the throughput of an xDSL system. The dynamics we saw at work during the modem market's mid-life stage will be seen again as the modem industry passes another turning point and moves from today's data-pump architectures toward xDSL designs. The pattern of successive iterations of reinvention will occur again as xDSL architectures are refined through customization, optimization and component integration. The drum beat of market emergence, reinvention and customization will accelerate and market discontinuities will occur at a faster pace as we head into the next millennium. HARRY DAVOODY is the Vice President of Mixed-Signal Products at Texas Instruments, Inc. Analog Main | Product of the Week | Columns | Editorial | Tech Notes
|
|||||||||||||||||||||||||||||||||||
|
Copyright © 2003 ChipCenter-QuestLink About ChipCenter-Questlink |
||||||||||||||||||||||||||||||||||||