Hacker News new | past | comments | ask | show | jobs | submit login
An experimental open source memristor / programmable “diode” (bigattichouse.medium.com)
257 points by psawaya on Jan 5, 2021 | hide | past | favorite | 65 comments



This and all the examples linked elsewhere in the thread seem to be electrochemical batteries where the anode or cathode are used as a resistor in another circuit. A current is applied and a metal oxidizes. It’s pretty well known that as a metal oxidizes, it’s resistive properties will change.

As most people know batteries have a limited number of cycles. Every reaction cycle adds some entropy/side reactions, and eventually it will become irreversible. Magnetic storage reorients a crystal or metal, which is more repeatable than a chemical reaction. Our neurons have similar systems to “clean” themselves so they can reset.

I imagine finding a viable memresistor is more about its ability to cycle.


(OP here) - the backplane is already a resistor, we're plating aluminum differentially along the resistive backplane to alter resistance depending on current flow. The resistor is X Ohms/cm.. when we plate out Aluminum near the negative terminal, that shorts a section of the resistor, lowering resistance there (making a "shorter" resistor). The aluminum is forced back and forth depending on current flow.


Isn't this just what dexwiz was saying, restated?


No.


Related to this, but why do people claim that the memristor is a "fundamental" circuit element? It seems to me that it's just a generalization of a resistor where the resistance depends on the integral of the current. There are already all sorts of circuit components that have the property where the resistance is a function of X, so why is the memristor particularly deserving of its own class?

Unless I'm missing something the flux-charge equation that "dPhi = M dq" seems to be the same as that of the resistor as well (since dPhi = vdt we have vdt = Mdq => v = M*i which is just ohm's law assuming a constant M).


https://en.wikipedia.org/wiki/Memristor#The_memristor_as_a_f...

> Chua in his 1971 paper identified a theoretical symmetry between the non-linear resistor (voltage vs. current), non-linear capacitor (voltage vs. charge), and non-linear inductor (magnetic flux linkage vs. current). From this symmetry he inferred the characteristics of a fourth fundamental non-linear circuit element, linking magnetic flux and charge, which he called the memristor. In contrast to a linear (or non-linear) resistor the memristor has a dynamic relationship between current and voltage including a memory of past voltages or currents. Other scientists had proposed dynamic memory resistors such as the memistor of Bernard Widrow, but Chua introduced a mathematical generality.


The basic idea is that you cannot "implement" (emulate) memristance using resistors, capacitors and inductors and that is what causes it to be fundamental property of the circuit element.


Since memristance is just the charge-dependent flux I'm not sure you couldn't get it by just throwing a resistor capacitor and inductor into a single circuit. In fact memristance is just the voltage divided by the current, which is impossible to be 0 for any circuit with some voltage. So by this definition almost any circuit has memristance.

Note that it's not unfair to consider any circuit with nonzero memristance as 'containing' a memristor, after all the whole point of fundamental circuit elements is that you can add them to model stuff like wire resistance and stray capacitance. However while you can't create resistance, capacitance and inductance by combining any of the others you can have memristance by combining all 3 (technically you can create any of the 4 by combining the 3 others, but 3 of them can be built easily so those make more sense as fundemental circuit elements).

So while a linear memristor would be rather unusual, I don't quite see what's so fundamental about it as a non-linear circuit element.


> In fact memristance is just the voltage divided by the current

In fact, voltage divided by current is resistance, BY DEFINITION.

Hey, you can also take a bunch of resistors, wire them in a loop and proclaim you have successfully emulated inductor with just resistors. That is what you get if you decide to set aside rules.

Important part of the description of what is memristor is its dependence on past charge flown through it, you can't decide to ignore parts of definition whenever it suits your argument.


>In fact, voltage divided by current is resistance, BY DEFINITION.

For a linear resistor sure, but otherwise resistance is better defined as the slope.

And well if you don't like that definition of memristance I suggest you update the corresponding wikipedia article.

> Hey, you can also take a bunch of resistors, wire them in a loop and proclaim you have successfully emulated inductor with just resistors. That is what you get if you decide to set aside rules.

A circle of resistors has 0 inductance by definition. In practice it's pretty tricky to build such a circuit using 0 inductance wire in an impermeable environment so it remains an approximation.


Key to a memristor is hysteresis -- the cycling of resistance, based on past current. This makes them fundamentally oscillatory.


Wikipedia has a pretty good explanation to your question how the memristor completes the quartet of inductor, capacitor and resistor among the differentials I, V, Φ, and q.


The equation for the memristor seems to differ from that of the others though in that R, C, and L are constants whereas M is a function of charge. The explicit charge-dependence of M isn't noted in the diagram, but without it you would just get a resistor.


No, it's no different than the capacitor in that respect. Capacitance is dq/dV.

A (fixed) capacitor has a constant capacitance C just like a fixed memristor has a constant memristance M. Both M and C can be used in differential equations involving charge.


What I wonder is: if a PID can be modelled with resistor, capacitor, and inductor, what’s giving us the memristor in practical control terms?


Hysteresis?


But the integrative part already takes care of the signal “history”, doesn’t it?



OK, we've changed to that from https://www.reddit.com/r/electronics/comments/kqbgnn/i_belie.... Thanks!


No worries. Thanks dang.


Quite cool! This reminds me of other homemade memristor endeavours that have been presented:

http://sparkbangbuzz.com/memristor/memristor.htm

https://www.eevblog.com/forum/projects/making-a-memristor/


Whenever I hear about memristors, I think back on this guy I came across online, whose company blog publishes stuff like this: https://knowm.org/thermodynamic-computing/

He patented a memristor design, and seemingly has some wild ideas stemming from his work with them.

> My statement stems from our work with AHaH Computing. When two energy-dissipating pathways compete for conduction resources, a Knowm synapse (aka kT-Bit) will emerge and it can be shown that the pair maximizes power dissipation while driving Hebbian or Anti-Hebbian learning. We see this building block for self-organized structures throughout Nature, for example in arteries, veins, lungs, neurons, leaves, branches, roots, lightning, rivers and mycelium networks of fungus. > > We observe that in all cases there is a particle that flows through competitive energy dissipating assemblies. The particle is either directly a carrier of free energy dissipation or else it appears to gate access, like a key to a lock, to free energy dissipation of the units in the collective. Some examples of these particles include water and sugars in plants, ATP in cells, blood in bodies, neurotrophins in neurons, and money in economies. In the cases of whirlpools, hurricanes, tornadoes and convection currents we note that although the final structure does not appear to be built of competitive structures, it is the result of a competitive process with one winner; namely, the spin or rotation. In other words, a hurricane is a ‘collapsed kT-Bit’.


I wonder how small this could get. If it could be shrunk to fit in, say, a 0603, you could fit a million of these on a decently-small pcb.

And case of those could have a billion memristors.


Time to be pedantic: A 0603 footprint is about 3.2mm by 1.4mm with extremely dense spacing. A two sided board with nothing but 1,000,000 0603s would be 2.24 square meters. Stacking those boards at 10mm height intervals, that would be a 10 meter tall case.


Oh, you're probably right there. Is that metric or imperial 0603? Anyhow, if it could be shrunk, presumably it could be packaged efficiently. A room sized computer, but with a billion memristors might be worth it.


Imperial. In any case, this is helps visualize how incredibly modern IC fabrication and packaging is. The biggest CPUs have up to 40 billion transistors and the biggest flash chips have 2 trillion! Following our naive 3.2mm by 1.4mm by 5mm spacing, that would be 896 and 44,800 cubic meters respectively!


This is a 50 000 liter tank to give an idea.

https://www.directwatertanks.co.uk/50000-litre-vertical-grp-...

I wonder if this means it is undoable.


No that would be a 50 million litre tank.


To be fair those flash cells stack directly on top of each other so in some sense you can shave off a factor of ten.


Chemistry definitely changes with scale, but that could be a good thing. This effect may be more powerful in the Nanoparticle size range, and then we can make memresistors only a few atoms thick.


How is this not simply an electrolytic capacitor operating near its breakdown voltage?


(OP/Author here) In a way it is, but there's a resistor in a U shape between the two plates (so the open side of the U is our connection points, and there's a voltage gradient due to the resistor, which causes a charge/pH gradient in the electrolyte, which causes differential plating/stripping of Al on the resistor - altering the "connection point" as you run a charge through it.


I have experimented a lot with memories, for now the most interesting path is to use just LEDs as 1 bit of dynamic memory.

1. Cover the LED so it cannot be hit by light 2. Charge it in reverse bias to write a 1 or do nothing to write a 0

inverting the polarity and reading the + you detect a voltage for up to 3-5 seconds depending on the diode if a 1 was written in earlier :)


That sounds remarkably like the Williams-Kilburn tube:

https://en.wikipedia.org/wiki/Williams_tube

> The Williams tube works by displaying a grid of dots on a cathode ray tube (CRT). Due to the way CRTs work, this creates a small charge of static electricity over each dot. The charge at the location of each of the dots is read by a thin metal sheet just in front of the display. Since the display faded over time, it was periodically refreshed. It cycles faster than earlier acoustic delay line memory, at the speed of the electrons inside the vacuum tube, rather than at the speed of sound. However, the system was adversely affected by any nearby electrical fields, and required constant alignment to keep operational. Williams–Kilburn tubes were used primarily on high-speed computer designs.


What is the advantage of using the parasitic capacitance of a diode instead of just using a regular capacitor in this application?


It's the memristor part that should have been in quotes in the title.


I fail to grasp the magnitude of this.

Is this the first feasible memristor?


No, it's just a DIY accessible one.


But is it really a memristor? The wikipedia article [1] quoted elsewhere says the memristor deals with charge and flux, whereas the article speaks of voltage and current (and some memory effect in the relation between V and I). I don't think any resistor with a memory effect should be called a memristor just like that.

[1] https://en.wikipedia.org/wiki/Memristor#The_memristor_as_a_f...


Can someone update the URL to the "Old" Reddit which loads much faster and has more information density: https://old.reddit.com/r/electronics/comments/kqbgnn/i_belie... ?



If you want to avoid the 'this chrome extension has been sold to a malicious actor' fiasco that hit me with the Great Suspender yesterday, you can load this as an unpacked extension.

Grab the zip from github, unzip it, and point chrome to it as an 'unpacked extension'. Works perfectly and takes 2 minutes.

The github is here https://github.com/tom-james-watson/old-reddit-redirect

You lose automatic update but that's a feature in this case.


What happened with Great Suspender?


Looks like there's discussion happening here: https://github.com/greatsuspender/thegreatsuspender/issues/1...


Whoa that is heavy.


Double-"huh?"

- This is a Firefox Extension

- Why not just ... disable automatic updates?


Teddit is also worth mentioning:

https://teddit.net/comments/kqbgnn/


Bite me, I like the new Reddit because of dark mode and its information density is perfect for my 4K display.


You can just change your New Reddit profile settings to get it back to the Old Reddit information density and layout. You can also make this apply across the site, overriding subreddit-specific stylesheets that subreddit mods create.

I don't get why people keep complaining about New Reddit when you can just do this once and problem solved.


To do this, you're required to log into an account which I guess a decent size of HN browse Reddit without one.


Because it's not once, it doesn't stick, and certainly not across devices.


I suppose all that is true if you browse Reddit without logging in, as alskdj21 points out above.


I see HN asking all the right questions.


Why?


Not trying to be snarky, it really is faster and has better information density


Because it loads much faster and has more information density.


No one in the word prefers the new interface.


Sorry you are getting downvoted for saying the truth.


[flagged]


We detached this subthread from https://news.ycombinator.com/item?id=25641686.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: