The first fully functional non-silicon ARM Processor, PlasticArm | Hacker News Reader

The first fully functional non-silicon ARM Processor, PlasticArm

twitter.com

63 pointsksec16 comments

Useful details on fab side: the material is indium-gallium-zinc oxide deposited onto a flexible polyimide substrate using the fab-in-a-box manufacturing system (https://www.youtube.com/watch?v=D9A2yYZ-NAI)

>Thin-film material deposition is achieved through a combination of physical-vapour deposition, atomic-layer deposition and solution-processing (for example, spin-coating). Substrate processing conditions have been carefully optimized to minimise film stress and substrate bow. Feature patterning is achieved using a photolithographic 5× stepper tool.

This is big, very big. Completely blows up lots of assumption I have in tech along with cost model.

>Although PlasticArm is an ultra-minimalist Cortex-M0-based SoC, with just 128 bytes of RAM and 456 bytes of ROM - it is twelve times more complex than the previous state-of-the-art flexible electronics!

There are lots of folks on HN working on embedded systems. I am wonder what are their thoughts on it?

3mm radius is very impressive. My former workgroup was able to get almost the same curvature accidentally on a Nanodimension PCB printer by forgetting to UV/heat cure the substrate properly, but that was PCB scale (mm instead of um) and the resistance was then far too high to be usable.

As I'm a MechE before all else, I'd be most interested in structural performance: usability after n strain cycles, resistance to environmental exposure, allowable strain/shear/bending in all dimensions (polyamide, so not much normal strain), heat transport, whether it can be operated inside a fluid/gel, etc.

There are some questions making me wonder what is more likely between an FPGA, Arm Cortex (or other RISC), or a fully nonprogrammable IC being most successfully deployed in a flexible form factor:

How would this be packaged for an end user? Should it be programmable with on-board NVM and a debug port? Or is it programmed with wafer probes during manufacture? What would then be the standard for flexible debug ports? Ideally something easy to connect using automation but (a) small enough to leave on and ignore, or (b) able to be cut off completely for production. Would this still be CoreSight-compliant e.g. having a discoverable ROM table, so anyone with a USB wiggler can attach?

In short, how technical must a user be to program this? Is this then attached to a flexible PCB using BGA? Electronic fabric using thin conductors?

Present-day developers have the alternative of buying a KL02 (2mm x 1.9mm; decidedly better specs/capability) and putting it on a flex PCB. More progress on bendable/non-silicon CPUs would be awesome, and maybe one day they will have a clearer purpose or better economics to displace existing tech.

I'd reserve any final judgment until a test unit or third-party publication appears.

In my eyes, there are gains enough from having an alternative library for the cell architecture. Arm Ltd. involved in the testing suggests a shorter road to market than if undergrad Joe in Dr. Smith's lab published this.

> There are some questions making me wonder what is more likely between an FPGA, Arm Cortex (or other RISC), or a fully nonprogrammable IC being most successfully deployed in a flexible form factor:

I believe it will be some display driver IC for the fad flexible displays.

RFID cards are flexible, but they just use the most tiny chips possible.

Even tiniest chips made with relatively recent IC tech will blow any printed electronics out of water.

> I believe it will be some display driver IC for the fad flexible displays.

There's also a (currently much smaller sized) research market for artificial skin, which is the input version of flexible display.

Deleted comment

OK, now when we say "shipped", we're talking:

https://duckduckgo.com/?t=ffab&q=paper+boat&iax=images&ia=im...

> There are lots of folks on HN working on embedded systems. I am wonder what are their thoughts on it?

Silicons is already super cheap enough to be used in RFID tags. RFID tags are by far not the simplest ICs around judging by the gate count needed to implement the amount of memory they have, high performance encryption, channel coding, complex clock trees, some are made with quite modern litho, and IP cores of the same grade as Cortex M0.

@dang. The real link is not twitter but this: https://community.arm.com/developer/research/b/articles/post...

There is also the official paper in Nature, here: https://www.nature.com/articles/s41586-021-03625-w

Oh Thank You. I think the paper in Nature is much better. Would be nice if Dang could change it.

128 bytes of RAM and 456 bytes of ROM. Looking at the paper, we see that 64 bytes of ram are used for the register file, and the clock speed is 29KHz.

I've written and shipped useful code in much less. This might already have applications!

That being said, consuming 21mW at 29KHz is a bit much. Generally power will scale superlinearly with speed, so it will not take much more speed for this thing to melt itself.

Still, a cool demo

It’s all static power, not dynamic power, so clock speed is irrelevant. Still it’s an NMOS technology rather than CMOS and that is going to mean big issues for power consumption (there is a reason why CMOS won).

By definition it cannot be ALL static power :) Physics applies. Moving a wire from a high to a low and back will incur losses. They claim 99% static. Then, assuming just linear scaling, taking this to 64MHz would be (21 * .01 / 29 * 64000 + 21 * .99) = 484mW . Half a watt is a bit much for a 64MHz micro. But still a cool start

Well indeed, I was dramatizing 99% as all. But you know as well as I do that TFT (even CMOS) will never run at MHz due to the awful RC constants, so it's a bit moot.

I am dubious that flexible circuitry is needed to make flexible things...

When you can pack millions of transistors into the size of a grain of sand, why can't you just embed that grain of sand into a sheet of rubber to make your smart flexible things?

All that's really needed is a good way to electrically connect a flexible and a non-flexible thing without repeated bending breaking connections.

Most chips today use gold bond wires in air to connect to the rest of the circuit. They are pretty okay with a little flexing. Put your grain of sand in a little air pocket in the rubber, with connections made with gold bond wires, and you should be good to squish, stretch, and shake the rubber as much as you like without those gold bond wires failing.

Reply on news.ycombinator.com