As microcontroller prices drop, they appear in more things. Today you will find microcontrollers in your car, your household appliances, and even kid’s toys. But you don’t see them often embedded in things that are either super cheap or have to flex, such as for example a bandage. Part of the reason is the cost of silicon chips and part of the reason is that silicon chips don’t appreciate bending. What if you could make CPUs for less than a penny out of flexible plastic? What applications would that open up? PragmatIC — a company working to make this possible — thinks it would open up a whole new world of smart items that would be unthinkable today. They worked with a team at the University of Illinois Urbana-Champaign to create prototype plastic CPUs with interesting results.
This is still the stuff of research and dreams, but a team of researchers did work to produce 4-bit and 8-bit processors using IGZO –indium gallium zinc oxide — semiconductor technology. This tech can be put on plastic and will work even if you bend it around a radius as small as a few millimeters.
The key problem, it seems, is yield. When ARM put the 32-bit M0 CPU on plastic, the yield was poor because of the high gate count. These new processors are simplified 4-bit devices that account for the 81% yield. At that yield, the devices could cost less than a penny to produce.
The final design fits on a 5.6 square millimeter die and had about 2,100 devices — comparable to an Intel 4004. The M0, by contrast, contains over 56,000 devices. The 8-bit plastic CPUs also worked but had a correspondingly lower yield. What would you do with a flexible CPU that costs a penny or two? Is it going to require more than a simple 4-bit processor?
Of course, the University of Illinois Urbana-Champaign is the home of a very famous fictional computer and some not-so-fictional ones. By the way, although the M0 was put on plastic, it wasn’t without significant compromises.
25 thoughts on “Plastic CPUs Will Bend To Your Will”
Woohoo, microwave ovens are gonna be a buck cheaper because they can stick the uC in the touchpad.
But seriously though, all sorts of exciting applications open up, like DRM for your water filters (Boooo) but stuff like crypto wallet gift cards in your Birthday Cards would be nice, and we all got a calc in our pocket, but one that pulls out like a roller blind from a pen would be cool.
More like a buck more expensive, because… new tech.
It’d be funny to have a flexible CPU/uC, but the passive components are still the standard ones. Unless they make matching passive components, then sure the tech might kick off.
To be honest, nothing needs to be flexible. If you’re using a small enough BGA and passives, then when you mount them on a flexPCB they basically disappear.
Yah, to get them off you’d have to be bending at an angle sharper than the fatigue point of the substrate. It’s like worrying about building an igloo shelter on an iceberg “Because it would melt in the tropics”
Durability? Bend. Bend. Bend. Bend. Bend. Bend. Bend. Bend. Bend. Bend. Bend. Bend. Break?
From the linked article it says that you have to start somewhere but 456 bytes of program store and 128 bytes of RAM,? Other compromises were made, such as no internal registers — they are mapped to the external RAM — and the CPU runs a lot slower than we’re used to, topping out at 29 kHz (note: k not M).
Most people don’t keep their band-aid on for days. I would think those foldable CPU are meant to be disposable such as DRM chip, band-aid, limited use testing kit, etc. Silicon CPU would still be used for things that needs to last longer like kid’s toys and other electronics
DRM’d plasters is the technology we don’t need… I’m sorry, I can’t stop bleeding, the chips in my body aren’t recognising the chips in the plaster. Great fun when you send aid abroad!
It’s not about DRM on bandages/plasters, but you can start to do things like lab-on-a-chip and have it monitor some stats about how you’re healing, or even just relay raw data to your phone and your phone then does the intensive bit.
Or you could have under-skin glucose monitors, or a heart rate monitor, or you could go to food; you could have a a bunch as stickers on a reel and when the good is about to expire it pings the stock room and you don’t need to record it manually or have staff check dates.
Or even go a step further and put one on the inside of your milk and it will send you a ping when it’s got only a couple of days freshness left.
When I was at uni, someone was doing something similar to record bee habits, and someone used something similar with some carbon nanotube sensors to create a stress sensor that could have been used in glass panels for cars, skyscrapers, whatever.
These things really aren’t designed for your average bit of pocket/consumer tech.
The technology is amazing but how most corporations are likely to use it is going to be horrible.
Microcontroller prices drop? Where? :)
Sure, I’m fine with more fabs existing, that sounds good for overall supply chain bus factor. And using a non-silicon substrate could have interesting environmental impact implications, depending on how it changes the production byproducts and energy costs involved.
The flexibility is a nice feature, it’ll be interesting to see how well they pair with existing flexible PCP technology. And whether flexible or not, putting logic directly in plastic seems like it could allow for integrating chipset logic directly into the PCB, which could further reduce costs.
As for what else I can think to do with a bunch of super cheap 4 or 8-bit CPUs off the top of the top of my head, maybe take a TTL-based layout like the Gigatron’s as the starting point, then work in the CPUs as support chips? Something like a little mainframe-on-a-board.
MCU size in 3x3mm also black-blob, no point sacrificing reliability!
“As microcontroller prices drop”
2010s have ended 2 years ago.
They are dropping from their precipitous highs. Definitely nowhere near pre-Covid prices.
I bet you could form these into bristles and make some really interesting sensors. Micron level surface texture mapping, or airborne particulate sensor maybe.
Probably a silly question but are the PN junctions of this indium gallium zinc oxide device, mounted on plastic, activated by light in the 380 nanometers (3.26 eV) to 750 nanometers (1.65 eV) like normal silicon based PN junctions ? (Is visible light going to effect all the PN junctions of the device, does this “chip” need to be put inside a light proof package) ? How about UV and IR ?
Found the answer on wikipedia on the page about Indium gallium zinc oxide: “The transistors are slightly photo-sensitive, but the effect becomes significant only in the deep violet to ultra-violet (photon energy above 3 eV) range, offering the possibility of a fully transparent transistor.”
So no good in money (people often check cash with UV lights) or nightclubs (black lights: UVA).
Literally just coat it with something that absorbs UV. Bam.
It’s an interesting development and I certainly hope this tech will become smaller and the yield will go up. The size and flexibility will have some interesting possibilities where it results in high reliable circuits. I don’t believe in the dream of using he technology to make cheaper chips as we can already make silicon chips dirt cheap.
This tech also has some interesting connotations for device security as decapping will likely be a lot easier for flexible chips. Also the low gate count combined with the large feature size will be easier to reverse engineer and attack.
There is even a “flexible6502” !
As micro-controller prices drop? What!? A Raspberry Pi is over $150. Teensy 4.1’s are out.
Personally I wonder what the appeal of more “smart” stuff even is…
A lot of things have RFID tags embedded in them already, and there is plenty of other odd chips out in the field already. And the vast majority of these goes into the trash bin within the first few hours after the customer has gotten their hands on the product.
To a degree, it seems like a rather large waste of resources for little to no gain.
However, I do see flexible chips as an interesting field, and there is plenty of useful places where a bendable chip could be superior.
I can’t wait for my soda can to come with a tracking device
Oh, another thing: I wonder how well this substrate would work for solar panels compared to existing thin film cells. If it offers comparable efficiency to silicon cells at significantly lower cost per unit area, that could be very useful.
Another comment points to photosensitivity only in the UV, so mountaintop solar maybe, rooftop? Only for Coloradans?
Please be kind and respectful to help make the comments section excellent. (Comment Policy)