|
...you could only get a small number of switches on a chip.
|
With each technological step, we can do so much more. It took many square inches of relays to implement a simple AND gate. It took perhaps half a square inch to implement a TTL DIP-chip gate. Today, tens of millions of logic gates can fit, literally, in less space than your thumbnail. In effect, with each technology generation, we build by using ever more complex, individual Tinker Toys. This lets us incrementally step back, and look at a bigger picture of what we're trying to accomplish, rather than getting bogged down in the details. Today, even simple projects would be a major effort (if realistically possible) using discrete TTL logic chips, and it won't be many years before todays best-of-breed controller capabilities become a similar historical footnote.
Last year, we explored some of the early work that will lead to the next generations of control logic ("Think Small, Really Small"). However, given the double-exponential growth of technology, a year is a very long time, so lets get an update.
As traditional semiconductor manufacturers move to a 65-nanometer process, ...the smallest modern transistors are [now] no more than a handful of molecules across, wrote John Markoff in a Dec. 29, 2005, New York Times article. As research marches on, some of last years speculative ideas are solidifying into devices that may be in our toolkits in as little as nine years; ...The industry's roadmap [has] growing confidence in new technologies that make electronic switches from single molecules or even single electrons.
Indeed, Intel already is planning for 10-nanometer processes that may create chips with a trillion switches. However, this is seen by the International Technology Roadmap for Semiconductors, 2005 Edition, as the most optimistic projection for continued scaling of traditional transistors. So here lies a quandary: how can the growth curve of Moores Law continue as our silicon processes reach fundamental physics limits?
The answer is that, as has always been the case, scientists simply refuse to be held back by any limit. They either find a way of extending the limit, or find a way completely around the limit by thinking outside the box and changing all the rules. This is just what theyre beginning to do now, as they approach what is commonly seen as a limit to how small conventional transistors can get before they will no longer work.
Unsurprisingly, this is not a smooth road. As experimental transistors shrink towards individual molecules and atoms, the rules of physics that have sustained high school and college physics and all of our manufacturing efforts to date simply refuse to play by the established rules. For example, they defy traditional beliefs that a switch must be at a specific, discrete value (such as on or off), and insist on simultaneously holding values of one, zero and every value in-between! (Don't ask; although you can begin to explore this contradiction to common sense by reading "Quantum Superposition." Imagine trying to design with something that can represent all values all the time!
Again, as in the past, scientists are working towards taming such quantum switches, so that, from the end-user (our) perspective, we simply move to yet another level of Tinker Toy complexity that hides the interior details. This is the same way we evolved from designing with relays to single-gate logic chips to PLCs.
Still, how can we visualize what this means to us? At this stage of the game it's virtually impossible to imagine the details of what the next generation of Tinker Toys will be able to do, or how we'll combine the new Tinker Toys into functional solutions. Back when we were happily designing products with TI 7400-series chips, imagine how impossible (for most of us, that is) it would have been to foresee the PLC. Now, again, we sit on the brink of another generation of control logic whose capabilities will be so far beyond that of today's premier PLCs that those devices will eventually feel like the TTL logic chips of older days.
We do know that these new devices will be far faster (due to the smaller distances between logic elements on the chip), and will be far more complex (due to the vastly larger number of logic gates that will fit in a given space.) They also will likely be non-volatile, will draw far less power, and generate far less heat, opening up completely new classes of applications. Well see development of switches that rely on the spin of an electron, on its magnetic state, and on its phase state. And some of our future logic seems likely to blend the realms of the inanimate and the animate, as we hijack natures biological creations. Already, single-protein, wet biotransistors have been demonstrated, and future organic molecular switches hold the potential to switch thousands of time faster than todays. (In-depth insights into additional potential future logic technologies can be found in "Emerging Research Devices.")
So, as we begin 2006, let's remember that if a project seems too complex or otherwise impossible, most of today's projects would have seemed equally impossible when viewed at the dawn of the PLC. The next generation of control logic might similarly make today's impossible projects quite possible, even easy. And just the beginning, so don't blink!
