Boffins draw finer lines on chips

Please post more stories like this.

I know it can be difficult but you can amuse AND educate us peons who read your stuff.

Go on ... give it a try ... please?

Still some life in the ole hoss yet.

But what we need is a new paradigm. Time to figure out how to use more than saturated binary logic. Just think how many logic states there could be between 0v and Vcc?

I mostly work in an analog world as a chemical engineer and wonder why someone hasn't taken on the challenge of multistate logic. Must be because most electrical engineers are simple minded, on and off is all they can handle.

And yes, I did watch Star Trek when I was a lad at university.

Just in time for AMDs Bulldozer CPU ;-)

Its not as simple as you present it. Each change in voltage in one line causes induction in nearby lines, effect rising exponentially by the proximity of lines (I suspect you know this already), voltage and frequency of change (which is why you still don't have THz electronics). In binary technology this induction noise is catered for by sufficient Vcc elevation and frequency regulation to get in a state where even with a maximal noise you still have 100% accurate 0 or 1 readout. If you split the analog Vcc span to N states than each two states would have to be separated by a full voltage span to compensate the induction . By doing this you will get higher than binary voltages in lines which will again increase induction which will require even higher Vcc spans. Eventually it will equalize and a stable voltage equation would be produced but then the physics kicks in; the new top Vcc and amperages it causes through tiny conductive lines would simply burn them to crisps, forcing the use of wider "wires" and spacing them more apart, insulating them more, etc, etc, which will effectively produce some kind of newage Eniac.

Binary might seem limited, but its a perfect balance and compromise between physics and speed, until such day comes that new materials and new physics discoveries would allow us to change this (for example quantum computing). However high voltage analog electronics is not a way to go for computing.

Great posting.

Doesn't someone say something very similar to this every year?
"Boffins____new technique________
smaller circuits_________years away"

Coments on problems with muli-state logic are correct; another way of looking at it is that from 0 to Vcc you have Vcc/2 margin of error; if you divide it into more states you are decreasing the margin of error; also know as noise margin. In binary logic anything less than Vcc/2 is a logic 0 and anything over Vcc/2 is a logic 1; simple as pie: Vcc/2 of noise margin.

Also another important considerating is that you create transistors with two kind of devices: pmos transistors and nmos transistors. There is always one of this devices in off state preveting a direct current path from Vcc to ground; if you were to have a say Vcc/2 state (or any analog kind of logic); then both the pmos and nmos will be at least partially on which will mean a very hot chip; that's good for making toasts but not for cpu design.

By the way analog math circuits do exists (look for op amp design); not the same as multi-state digital circuits; but is not as reliable as digital systems ...