Sometimes we forget how good we have it.

I must confess there are moments of weakness when I envy our digital brethren. The excitement in their eyes as they peel the shrink-wrapping off the latest and greatest design tool reminds me of a kid opening that big, bright present that's been sitting under the Christmas tree for weeks. But then, a saying from my youth tumbles me out of my reverie and snaps me back to reality. That saying is, "Be careful of what you ask for. You may just get it."

Innovation can't be shrink-wrapped

Over the last 10-15 years the tools for both digital and analog designers have made great strides. Although mixed-signal designers use these very same tools to design the digital portion of their mixed-signal designs, in our weaker moments, some of us may look at digital tools like VHDL, Verilog, Synopsys and others, and long for the day when analog design might be so well automated and standardized. We sometimes forget that the very limitations of our tools instill in us a tremendous opportunity for innovation.

One can only wonder how much room is left in the digital design process for innovation when practically all designers use the same automation tools and the same design processes. We have no basis for presuming that several digital design projects with identical objectives would not all end up generating identical designs. Digital design innovation is increasingly moving away from the real world interface and up to higher levels of abstraction, e.g. architectures and algorithms. Digital design tools enable digital design teams to work productively at higher levels of abstraction..

Even though standardized operational procedures may be a cherished goal of engineering management, the fact remains that products are bought not because of how much they are alike, but rather, consumers buy one product over another because they are different. More specifically, the buying decision is made on the basis of how much better a particular product is perceived vis-ý-vis all other products in that category.

So, we are left with the question: Where in the analog and digital design flow does the creation of this perception of being better, of having greater value, come from? Real innovation, it seems to me, begins at that point where the capabilities of our analog design automation tools fall off, and human ingenuity and talent pick up.

We've come a long way

Don't get me wrong. Design tools and the analog design process have come a long way in the last 10-15 years. Gone (and thankfully so) are the days when we'd leave work with our forearms covered with every color of the rainbow from leaning over and manually checking room-size layout plots with colored grease pencils. In addition, the portability and scalability of SPICE models have improved tremendously. It's difficult to imagine today, but there was a time when SPICE was hard-pressed to handle designs over several hundred transistors. And moving a model from one computer to another was a nightmare, if not impossible.

Simultaneous with improving software tools, the raw power of personal computers and workstations was exploding. Some have said that compute power has increased by a factor of 100 over the last 15 years and I tend to agree with them. Thirteen years ago in 1985, we were very happy when we spent $60,000 to equip a design team with a ý286-class workstation running the latest design automation tools. Today, for less than $20,000, we can give an engineer a workstation with 10 times the power and capabilities of that ý286 system.

As a result of the progress we've made with automation tools and the power of the platforms they run on, we are today able to describe designs involving hundreds of thousands of transistors. And through the use of macros written in C we can describe entire functional blocks, like memory or analog-to-digital converters (ADC). Analog innovation is increasingly directed toward developing circuit topologies, and optimized implementations that are effective at delivering the required performance within the constraints of process technologies and cost objectives.

Reality Bites!

We were sailing right along, making huge strides in analog and digital design processes, when we abruptly stubbed our collective toe on the real world. Although we never completely forgot about the real world, our automatic design processes may have conveniently neglected it for a while. Once again, we were reminded that our tools couldn't replace human know-how -- that the characteristics of the environment within which the system operates must be taken into account. For example, at one point our tools couldn't take into account the fact that the system we were designing might control a motor that could kick back an electrical charge and damage the system. None of the tools in use at the time would tell the designer to beware of this circumstance. He or she had to know about the problem and design-in bypass transistors or specialized circuitry to protect the system.

Analog design includes the all-important real-world interface that connects the idealized world of bits and bytes to the limitations and hostilities of the physical world. Analog and mixed-signal design must contend with issues of impedance, bandwidth, noise and distortion. It must overcome challenges of electrostatic discharge (ESD) and the potentially damaging effects of the electrical over-stress (EOS) caused by the requirements of driving motors and other inductive loads. These mixed-signal design problems cannot simply be modeled with today's design tools. They require a special combination of knowledge and experience that comprehends not only circuit design, but also device physics, as well as system architecture requirements.

Recently, some of the more advanced tools have, to a certain extent, come to terms with the real world context of our designs, but in the end the individual designer must use his experience and expertise to decide what capabilities the system will need and which precautions should be taken. More often than not, an analog designer is best suited to make these calls.

Adding analog to the mix

The dynamics at work in the electronic marketplace over the last five years have led to a greater emphasis on mixed-signal design. While tools were progressing to where we could handle functional blocks, like ADCs, we were also realizing that computers had to interface to the real world because that's where we live. As a result, the device integration we have observed in the digital world began to take place between digital and analog devices, resulting in advanced mixed-signal technology.

Mixed-signal technology is decidedly more complex than purely digital or analog design. Our tools had to catch up with our mixed-signal goals, and they have to some extent. Many of our simulators can now model mixed-signal designs, but, the designer must ultimately make a judgment and decide what will be the most effective design, given the conditions of the marketplace. In most cases, the mixed-signal designer who is making these crucial design decisions has come from the analog rather than the digital world. Analog designers simply understand better many of the critical issues such as how to distribute power across the system, how analog signals are captured and how digital information is re-converted into analog signals for interfacing to real-world applications. Many digital designers have probably never confronted issues such as these.

System-on-a-chip

As we followed this evolutionary path, we came to a point where we were no longer just combining analog and digital functionality in mixed-signal devices. More than anything else, our designs have come to resemble systems-on-a-chip. Here again we began to strain and stretch the capabilities of our design tools, only to realize what we knew all along. Design tools are essential to the process, but they can take us only so far. Beyond that point the creativity, innovation and simple savvy of the individual designer takes over.

Our mixed-signal designers frequently use a two-dimensional solution to the system-on-a-chip design challenge. They must model and simulate logic circuits at the transistor level and analog blocks at the behavioral level. At the fulcrum point in the process of developing optimized system-on-a-chip architectures is an engineer with a firm grasp of system-level analog/digital tradeoffs. He or she will know the bugaboos to watch out for. He or she will know how to wring out that last ounce of performance. He or she will know the difference between a merely acceptable end-equipment system performance level and one that will provide the customer a competitive advantage in the marketplace. That is the value our OEM customers demand and that is what our mixed-signal designers frequently deliver to them.

As we make further progress, our digital future will require a cooperative analog, mixed-signal and digital design environment and only this type of cohesive development environment will enable the advent of the system-on-a-chip era. In fact, we've already seen that optimized systems-on-a-chip architectures increasingly involve judicious tradeoffs between front-end analog and mixed-signal circuit precision on one hand, and downstream digital processor sophistication, such as that found in DSP engines, on the other hand.

HARRY DAVOODY is the Vice President of Mixed-Signal Products at Texas Instruments, Inc.