The Memristor: L, R, and C have a New Buddy
Tarantio
Registered User regular
Registered User regular
Hewlett Packard just announced that they've created the first Memristor, a new elementary electronic component. Its resistance goes down when charge is driven one way, and its resistance goes up when charge flows the other way, and it remembers this state, without power.
Obviously huge for digital computing, the first story seems to be that it will allow for computers that boot up instantly, and eventual superior memory storage- replacing both RAM and hard disk. I would not be at all surprised if a great deal more than even that came from this, in analog electronics as well as digital.
Particularly cool note: the memristor was theorized to exist in 1971 by electrical engineer Leon Chua, derived from mathematical equations.
Now is a good time to be an inventor in the field of electronics, methinks. Anybody have any world-changing ideas yet?
[kurzweil]Does this mean the advent of the Singularity?[/kurzweil]
Links:
HP Discovers Potential "God Particle" of Electronics
Scientists Create First Memristor: Missing Fourth Electronic Circuit Element
Obviously huge for digital computing, the first story seems to be that it will allow for computers that boot up instantly, and eventual superior memory storage- replacing both RAM and hard disk. I would not be at all surprised if a great deal more than even that came from this, in analog electronics as well as digital.
Particularly cool note: the memristor was theorized to exist in 1971 by electrical engineer Leon Chua, derived from mathematical equations.
Now is a good time to be an inventor in the field of electronics, methinks. Anybody have any world-changing ideas yet?
[kurzweil]Does this mean the advent of the Singularity?[/kurzweil]
Links:
HP Discovers Potential "God Particle" of Electronics
Scientists Create First Memristor: Missing Fourth Electronic Circuit Element
Tarantio on
0
Posts
Edit: Besides the insta-booting computers I mean. I'm just not really clear on what the technology is.
Heard about this a week or two ago. It'll be nice in 2020. :P
We're going to need more atoms.
Well, I just finished my Electronics for Scientists II final today, but I'm certainly no expert. It's kinda tough if you don't know much about electronic circuits, or computing.
Here's an example- do you know how computer memory works right now? In most cases, there's a hard disk (magnetic ones and zeros, lots of space, slow to write/read) and RAM (random access memory; it's what the computer actually uses for what it's currently doing, but as soon as the power turns off, it loses everything).
Memristors can be used to store data within a circuit (so it can be used for what the computer is actually doing) without power.
I feel like a bit of a fraud here, as there's not whole lot I know that isn't in the articles in the OP, and there are definitely people here who know more than me. But hopefully helps a little.
--
Wait. That article suggests that it does something similar to my old "wavelength" idea because binary recording is lame. That would be nice. Also I expect royalties.
Ooh, new links:
http://en.wikipedia.org/wiki/Memristor
http://www.eetimes.com/news/latest/showArticle.jhtml?articleID=207403521
Second one talks about how part of the reason making these took so long is that people thought the fundamental relationship of electronics was between voltage and charge, and now they're saying that it's actually between flux (change in voltage) and charge.
That's an interesting aspect- if it's important early on, though, it would represent a very fundamental change in the way computing currently works.
Said simply, we use binary because it easier. Analog electronics are fidgety- they don't always work quite right, and they're a lot harder to get working "perfectly"- digital electronics you pretty much just have to turn on.
Here's an analogy my professor used: Imagine you and a friend are trying to send information to each other down a long hallway, with lamps. You've each got a lamp with a variable aperture, with ten settings of brightness. You can use that to send numbers- 9 is brightest, 1 is dimmest, 0 is dark.
Now, in perfect conditions, you might be able to use this system to send, say, your phone number. But what if the hallway's filled with fog, he's really far away? It's going to be very difficult to tell the difference between, say, a 7 and a 6. Mistakes are going to be made.
Now say you decided to just use the brightest lamp setting and off to send information. You have to flash a lot more times for the same amount of information, but it's a lot harder to confuse. This way, you can actually get stuff done.
(by the way, my professor is the guy who invented those security tags that beep when you pass through a gate)
If they can eventually shrink these memristors to a scale similar to that of neurons, then maybe the variability they have can be put to real use. Hell, it's possible that this is the secret to real artificial intelligence, so I'm not going to make any predictions.
(this in no way implies I do not find this idea extremely cool, in fact I am highly excited by the possibilities here).
EDIT: I don't think these are a revolution in making computers work "analog" though, however they certainly give us some interesting new ways to go A2D.
EDIT 2: Actually thinking about this more I guess I can see their point since these things are kind of classically ohmic circuit elements, but still.
Nonetheless, my questions are (1) when will they be available at Jaycar and (2) what sort of range does the resistance go to? If you can go from like 0.1 ohms up to 1Mohm+ that'd be pretty damn rad.
As I understand it, it's the fourth passive circuit element... ever. It does something, just with current applied, that has never been done before.
EDIT: Reading the paper the argument is actually simply that you have 4 basic circuit values - charge, current, voltage and magnetic flux, which are connected by 6 mathematical relations (I don't fully understand where these come from) and such should give you 4, not 3, passive elements (and 2 innate relations that are a property of electricity).
I forget who said it, but it's true:
Here's the big problem with all this singularity horseshit: Hardware may keep getting better and better, but software has been stuck at "fucking awful" for decades, and it ain't getting better anytime soon.
Truest truth ever.
EDIT:
Ok, I finished the reading the letter. What HP have actually accomplished is a mathematical model of a memristor which has physical relevance i.e. they can actually relate it to physical effects within a semiconductor, which in turn has allowed them to explain a metric fuckload of oddities in solid-state nanoelectronics as being memristive effect.
The main issue is, memristivity seems to be limited strictly to nanoscale devices - their description is dependant on the fact that when you shove voltage into something that's a couple of nanometers big, the actual ionic impurities used to dope it can start moving around and as such change the observed resistance. While I'm finding the concept still interesting I'm not actually super-convinced that calling it "the missing passive" really makes a hell of a lot of sense. That isn't stopping me from jotting it down as a potential Ph D project application though
I can definately see its use in memory applications, but like DRAM, it would need to be refreshed after every read. Unlike DRAM, however, it wouldn't need constant refreshes/power to maintain state.
It's not going to bring about an analog revolution for "general purpose" computers, though. Mostly due to the fact that you can't get rid of the instabilities/noise in any system. So it is tough to get guaranteed, repeatable results if you have a long, multi-element analog path. Digital avoids this by limiting the possible choices to just two. Analog is great, but digital is easier.
Can't be. The math actually works.
Perhaps the return of the video game cart. :rotate:
In the end, is this a quantitative of qualitative thing? Or is it too early to say?
WHY HASN"T ANYONE ADDRESSED THIS IMPORTANT ISSUE!?
This... this has pretty much been the entire history of computing.
Well, yes. But nobody gets terribly excited when Intel busts out their new Pentium 5 that tops out at 4GHz instead of 3.5GHz and contains 10M transistors instead of 9M. Whereas this seems to be generated some hullaballoo.
Basically, I'm trying to find a reason to get all giddy. I mean, it's conceptually cool, but I'm trying to connect the discovery to something really awesome, rather than being the 2,354th claim that we're just around the corner from advanced AI and neural networks.
Massive storage available at RAM speeds sounds pretty cool.
Yeah, I'm pretty sure there are 2 more. Give me a while to figure out what they should be....
That's pretty much the best instance of wikipedia vandalism I've seen.
This is a huge discovery in electronics, I'm absolutely blown away by it.
Even the guy who theorized this components existence in the seventies is amazed that HP have actually gone and built one. It's incredible, and could have repercussions bigger than the atom bomb or genetics. Seriously. :shock:
It's revolutionary if you're an electrical engineer; flip-flops and a lot of gates will probably be redesigned much more efficiently. No, it won't make your computer fly or bend space. I'm not sure what sort of revolution you imagine you might see as a consumer other than better data processing. Digital hardware is designed to realize a Turing machine. It can do it better, but it's still a Turing machine. Even if quantum computers were introduced you would still only see a faster, more powerful computer performing the same tasks.
Care to explain?
Don't praise the machine!