Re: [beam] Re: Memristor
- I was looking into memristors again, today. I saw a silver based
circuit pen and it led me back to these ideas. Has anyone determined
if a metal like silver can function instead of aluminum for the
purposes of the point contact end of a memristor, or any other
materials. I learned how to precipitate copper nano particles out of
copper sulfate using ascorbic acid. You can make it into a paint using
a little gum arabic. Only issue, is we can already make the copper
traces. Aluminum would instantly oxidize, and I'm not sure if oxidized
aluminum nano particles will conduct well enough or not.
I repeat my old theoretical process in the cheap pic above. Could a
silver based circuit pen take the place of the aluminum for the
contact. I think this idea could have potential... It's similar to how
planar transistors completely replaced point contact germanium
transistors, and ultimately led the way to integrated circuits. The
scale is obviously bigger, but the idea is the same. I might have to
just suck it up and buy a circuit pen or two, maybe different
formulas. It'll have to wait though, as I think I need to pay rent on
Begin forwarded message:
> From: Richard Piotter <richfiles1@...>
> Date: July 10, 2012 3:11:33 PM CDT
> To: email@example.com
> Subject: Re: [beam] Re: Memristor
> Reply-To: firstname.lastname@example.org
> Ever used press-n-peel rub on pcb dry transfer decals to create etched
> circuit boards?
> I've used these before and they can produce some great results.
> I've been wondering. If one were to place a pad over the Copper/
> Sulphur layer, I wonder if having that uniformly sized hole over it
> would allow a more consistent surface area and more consistent
> characteristics from one point to another. I know the dry transfer
> decals are intended to be scrubbed away after the copper clad board is
> etched, but if you use the circuit pen technique to paint the decal's
> center hole to an adjacent copper pad, and then drop a dab of epoxy
> over the whole lot, it may protect the whole thing.
> Here's my most recent concept:
> Since the dry transfer decal is exactly that, a dry transfer, I doubt
> it'd have any effect on the surface. As far as I know, the material is
> an insulator, and I suppose one could even get away with stacking two
> on top of one another for added thickness.
> The only concern is hole size. Most are made for through hole
> components. Some pads in the kit may have very small holes meant to
> start a drill bit, but I don't know how many of those are included in
> a typical kit. I don't know if Radio Shack still carries these, but I
> think they can still be found online.
> One could also consider toner based art, using a transfer sheet, of
> course that means more work aligning it all.
> Richard Piotter
> Begin forwarded message:
>> From: "Andrew Hill" <curly.drew@...>
>> Date: July 10, 2012 1:40:13 PM CDT
>> To: email@example.com
>> Subject: [beam] Re: Memristor
>> Reply-To: firstname.lastname@example.org
>> I got mine on eBay - it's sold for a few purposes and in pretty much
>> any quantity.
>> Your right - remember to put sulpher on copper until a reasonable
>> amount of corrosion builds up (for me this took about a week) - you
>> can wipe it off and if there isn't enough corrosion just add some
>> more sulpher.
>> My goal is to more to control the contact point (size, pressure,
>> thickness, etc) a bit, so that we have a reasonable range of
>> accuracy. Nontheless the way you describe should produce a
>> functioning Memristor :)
>> One thing is that you only need a tiny contact area ("sharp end of a
>> pin" tiny) of aluminium.
>> If you are still in full time education I suggest getting the
>> science department involved - I'm sure that one or more staff would
>> be very interested in helping you with this. Plus they will probably
>> have the necessary supplies and test equipment to hand.
Well, fsck me sideways with a broom! All these years I've been hallucinating that "R" in memistor. You are, of course, right.
Still, it seems to me a 3-terminal device would be handier than a 2-terminal device because then one could use the "neurons that fire together wire together" algorithm to automatically adjust the resistances of memistor synaptic weights.
During normal operation the memister works like an ordinary resistor, but during the adaptation mode (between firings?) inactive-exicitory and inactive-inhibitory memistors are left alone, while each memistor element that was active-exicitory when the neuron fired gets a dose of plating current to decrease resistance, and active-inhibitory memisters get reverse plating current to increase resistance.
There is the added complication that the old memistors used AC current during normal resistance mode (presumably to prevent plating action), but with modern semiconductor circuits it might not be necessary since the plating control electrode could be shorted to one end of the resistance with a transistor switch. Just thinking out loud, here.