Posted on 04/30/2008 7:09:12 PM PDT by Aristotelian
CHICAGO, April 30 (Reuters) - It took about 40 years to find it, but scientists at Hewlett-Packard (HPQ.N: Quote, Profile, Research) said on Wednesday they discovered a fourth basic type of electrical circuit that could lead to a computer you never have to boot up.
The finding proves what until now had only been theory -- but could save millions from the tedium of waiting for a computer to find its "place," the researchers said.
Basic electronics theory teaches that there are three fundamental elements of a passive circuit -- resistors, capacitors and inductors.
But in the 1970s, Leon Chua of the University of California at Berkeley, theorized there should be a fourth called a memory resistor, or memristor, for short, and he worked out the mathematical equations to prove it.
Now, a team at Hewlett-Packard led by Stanley Williams has proven that 'memristance' exists. They developed a mathematical model and a physical example of a memristor, which they describe in the journal Nature.
"It's very different from any other electrical device," Williams said of his memristor in a telephone interview. "No combination of resistor, capacitor or inductor will give you that property."
Williams likens the property to water flowing through a garden hose. In a regular circuit, the water flows from more than one direction.
But in a memory resistor, the hose remembers what direction the water (or current) is flowing from, and it expands in that direction to improve the flow. If water or current flows from the other direction, the hose shrinks.
(Excerpt) Read more at reuters.com ...
Imagine how this will enable quantum computing ... we may be on the verge of touching the ‘temporal’ barrier hiding future and past actual events.
I don’t understand your relation to quantum (theoretical) computing here.
A theoretical advance that has been around for 40 years before realization is a great indicator of future worth.
I hear MS is hiring millions of people to code the new operating system to work with the memristor. They have been challenged to write the OS large enough that even with the memristors, it takes even longer to boot than any previous OS. Their goal is an amazing 20 minutes! And I bet they can do it.
Most computers take a long time to boot up. Not just devices running a Microsoft OS. A fully featured Linux takes a long time too. The Sperry/Univac 1100/64 took 20 minutes to load the microcode before you could even start the boot process. The Western Electric 1ESS required 3 1/2 weeks to load the code store from magnets arranged on aluminum plates (the "bits") into the core memory. It took a couple 53' trailers to hold the racks with all the software. It's just a matter of perspective.
“The “memristor” sounds like a “leaky” diode or a diode with a resistor in parallel to model the backward direction. I’m not convinced it is much of a “breakthrough”. “
That is my take as well.
I said leaky because it will conduct the “wrong” way, it’s just not so good at doing so. (And, if I understand correctly, if you keep doing that it trains to go that way). A flip-flop requires but two transistors, and if you make it with MOSFETS (ideally, with infinite gate resistance) the thing can remember its state on power-down - essentially nonvolatile memory.
A slight flux of current on the other side of it flips the diode. Diodes are baked in at the factory. They aren't persistently reprogrammable in times and with power less than it takes a single transistor to switch.
It will be interesting to see what the new class of device brings to the table. I suspect it will hold value for creating devices with lower power consumption.
The logic element is a gate. It's composed of a titanium dioxide junction in a cross configuration. The current is composed of 2 carriers, electrons and ions. The Ti4+ is the most mobile. The ionic current sets up a space charge, which results in a change in the electronic resistivity in both directions. The current through the junction creates the effective diode.
I'd assume a change in spin states was involved here, because memristance(M) is dφ/dq, where φ is the magnetic flux. ie. L=dφ/dI. The equivalent of ohms law for the device is v(t)=M(q(t))i(t), so the resistance(memristance) depends on the integral of the current through the device.
“No — not just a leaky diode, in parallel or not with a resistor. It’s a dynamically reversible version of that. A very small, very fast, persistently reversible version.”
Hey, don’t hurt me. I’m a mechanical, not EE. But it has to relate to the thermo in the sense of it’s being some variation of a mass, spring, damper system. That is, there must be an energy, mass, and entropy component - there isn’t a whole lot else unless you reinvent physics.
The damper is friction, i.e. the resistor.
The mass with momentum, is I believe a capacitor, storing kinetic energy.
The spring also stores energy, essential an inductor in the model.
E=IR would be the mass at velocity experiencing a resistive force. Dissipating energy E=I^2R would be a mass accelerating or decelerating.
So I guess you might be missing an entropy term, that would be the information energy of the system or the memristor.
But I’m just blowing some smoke, my Laplace transforms are a bit fuzzy.
After having waded through as much of the math as I could find summarized from their papers, I have to say you’re right.
There’s some neat kinks in the fab, but fundamentally we’re talking hysteresis and flux here.
But what I'm saying is pretty easy, I think. A diode is an electrical device that lets electricity flow one way, but blocks it the other way. Once a diode is manufactured, it's fixed in which way it lets electricitly flow, until the day you fry it, and then it's just a burnt smelling ugly resistor.
This memristor device is (very) small, (very) fast, reprogrammable diode. It's "as if" it had this little switch on the side, and you could flip the switch to reverse the diode. But that switch can be smaller, faster, and lower power than a single transistor.
I guess the mechanical analogue would be a ratchet. Plain old diodes are like ordinary ratchets -- once manufactured the direction they can turn freely is fixed, until the day you strip off the teeth on it. This is a ratchet with instantly reversible teeth, that will be smaller, cheaper, faster comparable technology (transistors) made in the same time frame.
Basic semiconductor devices, such as resistors, capacitors, inductors, memristors, and transitors, are each conceptually very simple. Similarly nails, screws, rivets, tape and glue are conceptually simple.
But each such basic device, in myriad forms, can end up being an essential element of more complex constructs worth enormous sums.
Now we’re talking of using a passive device that (from the sounds of what I’m reading) stores the state with a magnetic field; the memory is in the ability of the device to remember, via magnetic flux, the level of voltage last applied to it.
This could be (and I’m talking completely off the top of my head here, because I haven’t read their paper in _Nature_) that the device can remember more than two states (ie, “on” and “off”) - if the device has a linear response in magnetic field and flux to a change (and polarity) in voltage, then you could store umpteen states as values in the memory. All you might need is the equivalent of a reference to compare against.
Disclaimer: Opinions posted on Free Republic are those of the individual posters and do not necessarily represent the opinion of Free Republic or its management. All materials posted herein are protected by copyright law and the exemption for fair use of copyrighted works.