Visit the forum instructions to learn how to post to the forum, enable email notifications, subscribe to a category to receive emails when there are new discussions (like a mailing list), bookmark discussions and to see other tips to get the most out of our forum!
Microprocessors manufactured locally - is it possible?
-
22 Comments sorted by
-
Obviously such a factory would not be in every community. Perhaps there would be 1 or 2 for each state. Still, that's a huge improvement over having everything in Asia.
One thing about microcontrollers - even if you could make one, you'd have to do a total professional job up to current industry standards. Because after you put the circuit together, if one of those tiny transistors is defective or gets damaged, it could either totally stop your machine from working, or much worse, have problems that only show up intermittently. What if a few bits of your memory are corrupted? It will work 99% of the time, but when you compile your program in a certain way one of the bits in the program gets flipped with possible catastrophic consequences - perhaps a 0 meant the motor should go forward and flipping the bit made it go in reverse. Maybe a high pressure hydraulic value should have closed but instead it opened. Maybe your emergency shutoff fails to work.
Also, another issue is robustness. Can you guarantee that if you hit it with 15 kV its not going to blow? What if there is a current or voltage spike? What if it gets hot? Cold? Humidity? Shock and vibration? Can you reflow it with no issues? Are you prepared to make 10 - 20 microprocessors every time you need one so you can destructively test the majority of them to make sure this batch was good?
I'm not saying it can't be done, but what I am saying is that it's got to be done right.
-
My comment may not be too informative, but I am pretty sure that machining the parts required to make the processing factory would be a painfully complicated process, perhaps prohibitive. Then there is the additional issue of the need to remove all pollutants in the area (so something like hairs or skin cells don't short the circuits). With what little knowledge I have I would say it is possible but not productive. I think there are some elements of modernity that cannot be decentralized without some significant technological advances.
-
*rolleyes* Of course you can make microcontrollers. There are already open source microprocessor chip designs. As for the manufacturing, you people are assuming that open source has to be poor quality or that it has to be done by amateurs, which is completely wrong. Almost all open source software is done by pros and the quality is often superior even in demanding applications e.g. scientific computation, web servers, etc. (although I have noticed that unfortunately on this project most of the work seems to be done by people like myself with more enthusiasm than expertise, this can presumably be solved if we stick to it and open source culture spreads as it has within software engineering).
Read up a bit about how to make semiconductors and think about it for a while and you should realize that this most certainly be scaled down.
There is no reason to assume, either, that we cannot do something like micron precision machining. There are manufacturing techniques that add accuracy, producing parts more accurate than the parts used to make them - making a lens using a spinning mandrel or a spinning pool of liquid in paraboloid shape, or a pressurized membrane is one example. How do you think humanity got from stone axes to semiconductor fabrication facilities - the precision fairy came along and did it for us?
Once we have a fab lab it will "just" be a matter of bringing appropriately skilled people onto the team, and coming up with a production process which can be followed, which will become simpler and easier as the fabbing equipment becomes more advanced. Whether a good process will necessarily be too time consuming to follow is up in the air until more work is done on it.
What you fail to realize is that the manufacturing fairy does not do these things in large scale processes, we, the middle and lower class do them. Collectively we know everything known about them. And we have all the skills we need to scale them down to a more practical level and open source them.
-
The facts are that there are very few chip manufacturers in the world. There are very good reasons for that. It's not impossible to imagine doing this but why would anyone even try? What could you possibly gain? I can buy a decent programmable "PIC" microprocessor for less than 10 bucks. At volume they are a couple dollars.
So there is no reason to go there price wise. Again, is there any other reason someone can think of that might be of interest?
The Dawg
-
@Dawg
Well, in a perfect world where anything is possible, I can see some real value in making your own electronic components. First of all, if you want to make an electronic project, you've got to get dozens of parts ordered from a number of online suppliers. It would be nice if you could make the exact parts you need and test them out immediately, without having to order them. Just download the design.
Another really cool thing would be custom systems on a chip. Instead of using a microcontroller with the surrounding circuitry you need, why not just make a custom IC? No need for discrete AND gates, Op Amps, OR gates, diodes, etc. Just put it all in a custom chip. Lower cost, more reliable, lower power, smaller, etc.
However the technology I'm talking about isn't here yet (or so I think).
-
If expedience is a justification for making it, I would assert that ordering online and paying for one-day shipping would be far easier. I remember something frustrating I was once told, but it was accurate, "Adam, just because you can make it, doesn''t mean you should." I do not recall what kick I was on when this was said to me, but it was a surprising simple truth.
-
for what it's worth see the response I got to the same question over at the reprap forum.
My reason for asking this is question is not to assess manufacturing processors w/n the context of our existing society, but to explore the limits, if there are any, of the decentralized automation of manufacturing.
-
-
Also I think this is already referenced on the Wiki:
http://vimeo.com/channels/26257
It can be done, but is it economical? -
Are you meaning economical in money, or in man-hours?What you've got is always cheaper than what you don't got.And once you form a community of some size, what you've got - if nothing else - is man-hours for the cost of food and shelter. Money maybe not so much.Be sure to calibrate your economical calculation to the cost of resources available!
-
Jason
To your second point, technology exists that allows you to build a custom System on a Chip. It is called FPGA or Filed Programmable Gate Array and it reconfigures the logic gates on the chip based on your programming it using something called a hardware definition language (see http://en.wikipedia.org/wiki/Fpga).
As to the broader point, I agree with those that think this unrealistic in the near term, at least the idea of creating a useful microprocessor. The transistor video is cool and would be a good learning project.
-
In 2011, four universities prototyped electronic circuits made from chemicals extracted from vegetables.
http://eng.auburn.edu/news/2010/09/fergusedibleelectronics.html
I think this increases the feasibility of local production, or at least it might when that technology reaches the point of microprocessors.
Also, how about some libre memristers... https://secure.wikimedia.org/wikipedia/en/wiki/Memristor
-
We can't manufacture electronic components and probably not for decades. There may be local ways to do this with more advancements in lithography and there may be tradeoffs available that the conventional chip makers will not explore.We can localize electronic manufacturing by making designs that can be cnc etched. We can minimize the logistics of needing to order parts for everything by deciding a common set of components and connectors to use in all designs.
-
Lithography is like injection molding for plastics. Infrastructure intensive, but good for making thousands of parts cheaply. They have smaller machines that can do it, but they cost a great deal.
What we want here is the equivalent of a reprap 3D printer in comparison to the molding machine. While you may end up with laser lithography or something, it doesn't have to be lithography. Controlled ion deposition maybe? It does a complicated job, so don't worry too much about keeping it simple.
Look up Nanosolar. They print their solar panels out on a commercial printing press. Could the process be adapted to an inkjet type system? It might just work for simple discrete components, but that's a start. If you can print out a solar cell, I"m sure an LED wouldn't be too hard, and if you can print an LED, a regular diode can happen. Only complicated thing to make would be an electrolytic capacitor, but I'm sure there are ways to work around that.
Lithography is only of interest because we know the process works. Find something else. This is a project I will be working on eventually, it's one of my goals to be able to put three machines on my workbench, and get a circuit board, discrete components, and chips out of each of them, to put together.
-
I think you have to look at this issue like you would the development of a civilization. This would come long down the road and require some evolutionary processes. It would probably require that you build very simple processors first, then use those to build more complicated, etc all while using the already gained knowledge to speed it up. This would be Moore's Law on crack!
-
I originally saw this posting a few months back which is what inspired me to get back on the board to share something. (but didnt get a chance to until now, note I havent even read the rest of this thread just responding to the original poster so far :)
Yes it IS possible, all you have to do is go back in time a little bit and change your expectations some. "Fabs" were common all over the US originally like back in the 60's. Originally microprocessor fabrication used off the shelf scientific equipment... it was only when feature sizes shrunk and complexity radically increased, that more specialized equipment was needed, and so those without the new tech went out of business, and the cost of tech needed for the precision equipment went way up because it all had to be custom designed and built and made ever more precise. The cost to start a semiconductor factory now is like several billion... but that is assuming current levels of technology and for profit capitalist supplied equipment for 45nm CPU's and similar. If you are just trying to make an 8088 and not a Phenom Quad it's alot easier, and if you are just trying to make things even simpler than a cpu on a single chip like logic gates it's easier yet.
Well, easier doesnt mean EASY... but my point was to share the parable of looking to the past... if the problem looks too complex now ask how they did it at the birth of that industry. Manufacture of small scale integrated circuits similar to early 1960's chip technology should be doable and even within reach of a moderate sized community budget I should think, given the right design. Getting up to the first 8 bit computer is alot tougher. So my questions to think about the issue more completely are... what level of chips are most critical to the functioning of society and what level of precision scientific equipment is it possible to home fabricate? (either directly or first making precise tools which let you design and build even more precise tools)
Just for note there is off the shelf electrical engineering software available down to at least the Core 2 Duo level of complexity/65nm. I dont know of any open source, but all the design rules and so forth are _off the shelf_ for that level. Yet the first cpu's like the 8088 were still "taped out" manually on a huge drafting board before being photo-reduced, so even engineering a simpler cpu by hand is possible in a pinch. I'm pretty sure the 6502 and Z80 were hand taped too, computerized design was used by the era of the 68000 16bit and 80286 though for sure.
I have a fantastic PDF (its publically available but I cannot remember the title or where from) 250 miles from here, the next time I go out of town i'll try to find it and share it, which talks about the history of semiconductor fabrication. Google searches are not turning up the tiny obscure hole I found it on originally. :(
Perhaps a better question is not whether it's possible but whether it is a high priority and worthwhile... eventually probably yes, but there may be other and better ways either for the meanwhile or higher priority projects even if I can also think it would be nice to insure the technology never gets lost to the future. A few examples:
- better reuse of existing microprocessor stocks (eg - a code database that lets you salvage something from an older computer of nearly any make, and stick a standardized OS on it to make it do useful things... whether cpu's and microcontrollers stripped out of mobile phones, microwaves, or whatever... and repurposed to run a CNC machine... we have billions of chips in the world, as long as EMP doesnt wipe them out I could see intelligent reuse being worth supporting.
- creation of pneumatic/fluidic microcontrollers... especially solid state (which is supposed to be possible). You can create logic gates and other simpler devices using fluids instead of electronics... although it is not as small/dense it may be easier to make. Electronics requires the specialized silicon purity for chips and fancy equipment to lay down photolithographic traces... but fluidics could be done out of plastics or printed out of a RepRap machine. It would never be as dense and efficient and long lasting as a cpu, but if you can print it from cents of plastic it's almost like "who cares?" Simply having certain minimal functionality, over the long term (multigenerationally no matter what happens in outside society) could matter more... it could enable more simplified manufacturing and automation to at least exist, until you could get back to a level of technology allowing the automated creation of electronics, electronic components, and chips again. For me I DO consider it vital and important to have "automation" and mass manufacturing capacity as essential for society... but you do not absolutely require digital or electronic computers to do this. Earlier numerical control machines could do the same operations with the same precision, you might recreate this ability even if it means going back to punch cards, vacuum tube electronics, or... pneumatics.
- stockpiling for the future and simply keeping the cheaply manufactured chips stored safe... if you have a thousand of them manufactured at $1.50 a piece you probably have several lifetimes worth of all the chips you'll ever need for a small community even if they periodically die, burn out, get damaged or whatever.
- the application of open source cores and engineering to come up with our own multipurpose IC design, paying some existing fab overseas to manufacture thousands of in some kind of mass community group buy. Although due to the amount of engineering required I am not sure whether the money saved would ever present itself... use of cheap off the shelf microcontrollers and cpu cores may make more sense, along with open source hardware and software design to best use them. Afterall conventional chips are often made in the MILLIONS...
-
Okay, finally figured I should read other peoples comments and comment on them myself...
There are potentially separate issues being explored here - manufacturing your own electronic components is one thing. That should be doable without excessive difficulty, a look at 1920's radio for instance sometimes had people making their own components from scratch... the bigger issue is MASS manufacture. You can handmake radio tubes, you can handmake resistors and capacitors but it is not time-efficient. The creation of mass reproduction (even if a smaller scale, like thousands instead of millions) is important.
The creation of your own microprocessor - does NOT have to be on one single chip. They werent originally anyways. Whereas I dont personally want to rebuild ENIAC out of handblown tubes it would theoretically be possible to still create a computer like that if you were eager enough and put the time in. >_> So the more important issue is the reduction of cost of the making of generalized electronics.
Robotic assembly of (homemade) electronic components onto IC boards of some sort is perhaps a more feasible approach... I could see it being doable within two years if people were dedicated. The components can be made, the process just has to be automated. RepRap and similar 3d "whatever" machines could just as well place individual components onto a board... and you should be able to print the traces with near future technology, or use conventional board making methods using photoetch to make the boards. If a robot puts the pieces on and solders them in place, the cost to make one board is reduced substantially... even if you still need a thousand boards for a computer. It is also possible that robotic assembly could be done using pneumatically driven numerical control and not require electronics itself.
Unconventional IC manufacturing is possible - the first IC's were handmade, not photolithography. Photolithography just enabled mass production with minimum time. Creating tiny components and soldering them to one another precisely, you could build up a 3d block of components which are then wired to other blocks of components with better use of space than hand wiring everything together. Manufactured transistors should be makeable with the right source materials... if you want to independantly produce the source materials (very pure silicon for instance for the right electrical properties) from scratch it gets alot harder and I know little in that direction... but they are all known industry processes that should be replicateable by someone that wanted to study and figure out how to do it on the simple and cheap. Put another way the materials should be more of a problem than the actual manufacturing of something... we should be able to have a more precise RepRap able to place and churn out some pretty cool stuff, even if not full IC chips in one swoop.
The creation of machines which themselves allow ever more precise work allows you to shrink all of the above down as well... make miniaturized transistors, use surface mount components, or even directly machine or 'inkjet' down connecting traces of conductant... furthermore this kind of making of precision machines is in itself necessary for the steps of photolithography. Early IC manufacture used general purpose scientific gear. If we can machine something suitable to be 1960's scientific equipment we could get back to that minimum stepping stone.
Note however that things like a Core 2 Duo level microprocessor, or even an FPGA is reasonably, beyond both manufacturing and materials reach of anything youre likely to find on these boards. There are orders of magnitude level improvements from an 8088 to an AMD64. But I would think eventually recreating even a 4004 or 1k RAM chips of the mid 70's to be within the eventual reach if a group were dedicated enough, and did the baseline manufacturing of extremely precise scientific equipment (you can have as many as 300 steps in modern cpu production and many layers... you have to get the features to line up EXACTLY on each pass, by the 1970's we were talking feature levels in the 10 micron range or about 1/10th the diameter of a human hair and only then were computers even used in any part of the fab process, and that may have been more for convenience than absolute necessity), and had access to the pure materials like silicon and the chemicals used in each step of the process. The use of multiple cores or multiple chip computers should be doable to get back to a level on par with say a Commodore 64 again if you were determined.
Designing the front end design of the chip is easy... manufacturing the chip itself is the hard part.
Reasons not to do it include cost of materials and man hours vs the outcome - without the large wafers and many chips from each wafer the entire process becomes very expensive to replicate. It would be cheaper to work a few hours at 7/11 and just buy the chip you need from china most of the time. My primary interest is if for some reason mass social upheaval occurred, would it be possible to get back to some of our current technology in less time than it took to enter the early information age, like a few decades... I think yes with the proper foreplanning.
-
Have you guys seen this latest on printing your own circuit board with a Mendal RepRap equipped with a MakerBot Unicorn style pen plotter and a micron pen?
http://openmaterials.org/2010/12/14/diy-printed-transistors-botacom/
Alastair
Howdy, Stranger!
It looks like you're new here. If you want to get involved, click one of these buttons!
Categories
- All Discussions1,013
- General Discussion895
- ↳ Introductions145
- ↳ GVCS Development43
- ↳ GVCS Replication14
- ↳ Similar Projects, Partnerships and Open Culture53
- ↳ In the News14
- ↳ Education15
- ↳ Food18
- ↳ Energy42
- ↳ Health2
- ↳ Sustainable Architecture21
- ↳ Transportation10
- ↳ Household6
- ↳ IT, Web Infrastructure110
- ↳ Shared Personal Notes1
- ↳ New Communities10
- ↳ Other Languages34
- Project Management4
- Proposal and Development Status10
- Organizational Development26
- ↳ Team Logs4
- ↳ OSE Core Team1
- ↳ OSE IT - Core Team2
- ↳ OSE Collaboration Platform - Core Team6
- ↳ Resource Development3
- GVCS Technical Development52
- ↳ Open Source Car10
- ↳ Open Source Tractor1
- ↳ Gasifier Burner1
- ↳ CNC Torch Table5
- ↳ Agricultural Microcombine4
- ↳ Dimensional Sawmill1
- ↳ Documentation and Instructionals18
- OSE Dev0
- ↳ Trucktor0
- ↳ Microtrac0
- ↳ CNC Router0
- ↳ CNC Torch Table0
- ↳ Bulldozer0
- ↳ Backhoe0
- ↳ Car0
- ↳ 3D Printer0
- ↳ 3D Printer Laser Diode0
- ↳ Tractor0
- ↳ Powercube0
- ↳ Brick Press Controller0
In this Discussion
- Alastair December 2011
- Anderxander June 2011
- aneiren June 2011
- ARGHaynes June 2011
- DanielKey July 2011
- DavidIAm June 2011
- Dawg June 2011
- dorkmo December 2011
- gregor June 2011
- Jason June 2011
- jerryshaw November 2011
- JMBT June 2011
- kunkmiester July 2011
- Marcus June 2011
- PatrickGibbs June 2011
- robotjosh July 2011
Tagged
Loading