I'm stealing and expanding on Mark Twain's often-cited saying, "reports of my death are premature" to say that "reports of 'analog's' death are not only premature, they are completely wrong."
How so? As everything goes digital, and digital goes everywhere, analysts who don’t know what they are talking about (or who should know better) periodically pick up on the theme that there's little or no future for analog circuits and ICs. Many of these "know-nothings" are Wall Street types for wouldn’t know an op amp from a logic gate, but that doesn't stop them from opining with confidence.
Here's the reality: analog-based transducer I/O is expanding its markets, in parallel with the growth of digital-based everything into all sorts of application nooks, as well as higher-visibility applications such as multimedia, smartphones, and more. Even "mundane" items, such as the ubiquitous $5 combination lock used on gym lockers, have gone electronic. [You can make your own list showing the increasing pervasiveness of electronics into everything, no need for me to do it for you.]
While it's risky to generalize, analog for I/O, transducers, and signal conditioning has gone on two paths. First, there are the improvements in speed, accuracy, power dissipation, distortion, and other key parameters, which has both expanded applications for the end products, as well as been a driver of the improvements themselves.
But we also see some new, innovative architectures for analog-centric ICs. Check out the configurable analog front ends (AFEs) just introduced by National Semiconductor Corp., reported on by my colleague Paul Buckley here. Their LMP91000 and LMP90100 AFEs—combined with their Webench design tool which selects specific make and model numbers of real-world sensors from leading transducer vendors—changes the context of transducer and signal-chain design.
Here's why: the sensor world is highly fragmented in terms of application, transducer sources, signal types, constraints, and priorities. You'll find a bewildering universe of sensor types, even for a simple-sounding parameter such as "temperature". (If you don’t understand the complexities and subtleties of measuring temperature in the diverse applications out there in the real world—well, all I'll say is that you have an awful lot to learn.)
Once you get past your sensor selection, you need to condition its signal and make it system friendly, often despite an electrically unpleasant and even hostile environment. So your challenge is with both the sensor and the electronics that makes the sensor signal valid and viable.
Analog vendors know this, and have developed a large range of solution options, starting with basic building blocks (op amps, instrumentation amplifiers, A/D converters), through system-on-a –chip (SoC) ICs and even higher-function modules. But these approaches often lack flexibility once in place, and require a unique bill of materials (BOM) for sensors which may differ only slightly.
That's why developments such as the National's LMP91000 and LMP90100 AFEs are so interesting. They simplify the selection and integration process, while offering flexibility in application. At the same time, those who prefer to "roll their own" or have super-special needs can take advantage of the incredibly good amplifiers and A/D converters now on the market, if that's their preference.
In short: these are good times indeed for transducer-interface analog-circuit designers and users, as well as vendors. Don't let the negative prognostications of the analysts mislead you. ♦