“What can you do with bits?” asked Dr. Venu Menon, VP of Analog Technology Development at Texas Instruments at the beginning of his keynote talk at last week’s ISQED. Then he answered his own question: “Nothing without analog,” because analog is the technology you need for real-world interfaces and for power management. “You’ve been hearing about 20nm [from the previous two keynote speakers],” said Menon. “Analog is a whole different world.” Then Menon paraphrased Mark Twain: “The demise of analog has been greatly exaggerated.”
Think that analog is a “niche” market? Take a look at this slide from Dr. Menon’s keynote speech:
Analog IC unit growth is outpacing overall semiconductor unit growth. Why? By providing new solutions to old problems (power tools, appliances), opening up and enabling new analog applications (solar, battery-driven bicycles, and other energy-efficient devices), and opening whole new markets that put electronics in places electronic systems have never been (building, bridge, and structure sensors; food-quality monitoring; portable medical instrumentation).
However, said Menon, the real reason analog is growing faster than ever before boils down to one word: sensors. There are more sensors being deployed in more systems than ever before. For example, smartphones and other portable, handheld devices such as tablets and gaming platforms incorporate more sensors each year: gyroscopes, accelerometers, image and light sensors, etc. The same is true for automobiles. “There are lots of sensors in automobiles,” said Menon, “’Sensorization’ has driven analog content.”
Then Menon gave some additional, interesting statistics: 60% of the analog devices shipped are application-specific. Only 40% are standard linear parts.”
Texas Instruments now has a catalog of more than 40,000 analog parts. The company introduced 900 new analog ICs in 2011. That’s an average of almost two new analog parts per day, every day including weekends and holidays, all year long. Even with this rapid pace of new product instructions, said Menon, analog IC designs tend to live a very long time. Product additions are generally cumulative. That’s how you get to 40,000 parts in the analog IC catalog.
Turning to the IC processing used to fabricate these analog parts, Menon pointed out that analog process engineers care about very different sorts of things than do digital IC process engineers. For example, A/D converters need high-precision capacitors to reduce error noise. If you can develop a good analog IC process technology with enough dimensional precision, you can build better A/D converters than your competition. On the other hand, D/A converters need matched resistors to reduce converter error. Better resistor matching eliminates laser trimming and thus reduces part cost. When was the last time you heard a digital IC process engineer worry a lot about precision resistor matching in a process technology?
To enable all of the different types of analog parts that constitute that 40,000-item catalog, TI has developed and maintains four major process technology classes, as shown here:
Each IC process offers the analog IC designer a different set of key characteristics and each process can be tweaked as needed for special requirements. Because the chips themselves tend to have a long lifetime, the processes must have a long lifetime too.
That’s why trailing-edge digital IC process nodes tend to become leading-edge analog nodes. The fabs and equipment for trailing-edge digital IC process nodes are fully amortized and continued use of that fabrication equipment changes capitalization and financial models that determine chipmaking profitability. The following slide throws more light on the situation:
The 250nm node appears as a dark red line in this image. Work starts on the node in Year 1. By Year 3, the node has reached its peak use as a digital node and then usage slowly falls as the 180nm digital process node (shown as a pink line) comes to production. By year 4, the 250nm node has fallen off in usage as a digital process node just as the 180nm process node is reaching its peak usage as a digital node.
Then, eight years after development work started on the 250nm digital IC process node, just as 250nm production has fallen almost off the chart, the 250nm node revives as an analog node. This time, however, the analog version of the 250nm node is running on fully amortized equipment in fabs that have already been built and capitalized. Note that the 250nm process node’s analog peak is also wider than the 250nm digital peak. That’s because it takes about two years for the new generation of analog parts to catch on with analog system designers, said Menon, and the resulting parts are used for a much longer time.
The whole cycle then repeats with the 180nm node.
So if analog IC design doesn’t need dramatic scaling in new process technology, what does it need? Talent. Dr. Menon made it clear he was looking for design and process engineering talent in the ISQED audience. He offered some attractive incentives in the form of four sets of challenges and opportunities.
- Analog IC designers are not chained to the scaling locomotive called Moore’s Law—the digital IC industry’s roadmap. There is considerable opportunity for other types of creative differentiation. Creative ideas are welcome!
- The large diversity in analog IC process technology is a challenge, he said. TI has to hire retired digital IC process engineers who solved a node’s problems years ago and still remember where the pitfalls are.
- The analog IC industry has accumulated years of process and component IP and it’s a challenge to keep that wealth of IP working for the company across its several analog fabs. Maintaining and updating processes, SPICE models, PDKs, and documentation is a continual challenge.
- Managing several analog IC process technologies while leveraging older equipment and factories presents many challenges. You can’t “copy exact” with the diversity of installed, fully amortized equipment, you must “copy smart” instead.
In the end, Dr. Menon characterized the analog IC business as being very different than today’s digital IC business. “We build [many, small] speedboats, not [a few, large] aircraft carriers,” he said. As a result, there might be only three designers for an analog IC, as opposed to 150 designers for a digital IC. That’s how you can manage to introduce 900 products per year.