Want to read Slashdot from your mobile device? Point it at m.slashdot.org and keep reading!

 


Forgot your password?
Close
typodupeerror
×
Hardware Build

This 22-Year-Old Builds Semiconductors in His Parents' Garage (arstechnica.com) 84

Posted by EditorDavid from the raking-in-the-chips dept.
Wired reports on 22-year-old Sam Zeloof, who builds semiconductors in his family's New Jersey garage, "about 30 miles from where the first transistor was made at Bell Labs in 1947." With a collection of salvaged and homemade equipment, Zeloof produced a chip with 1,200 transistors. He had sliced up wafers of silicon, patterned them with microscopic designs using ultraviolet light, and dunked them in acid by hand, documenting the process on YouTube and his blog. "Maybe it's overconfidence, but I have a mentality that another human figured it out, so I can too, even if maybe it takes me longer," he says... His chips lag Intel's by technological eons, but Zeloof argues only half-jokingly that he's making faster progress than the semiconductor industry did in its early days. His second chip has 200 times as many transistors as his first, a growth rate outpacing Moore's law, the rule of thumb coined by an Intel cofounder that says the number of transistors on a chip doubles roughly every two years.

Zeloof now hopes to match the scale of Intel's breakthrough 4004 chip from 1971, the first commercial microprocessor, which had 2,300 transistors and was used in calculators and other business machines. In December, he started work on an interim circuit design that can perform simple addition....

Garage-built chips aren't about to power your PlayStation, but Zeloof says his unusual hobby has convinced him that society would benefit from chipmaking being more accessible to inventors without multimillion-dollar budgets. "That really high barrier to entry will make you super risk-averse, and that's bad for innovation," Zeloof says.
This discussion has been archived. No new comments can be posted.

This 22-Year-Old Builds Semiconductors in His Parents' Garage More

This 22-Year-Old Builds Semiconductors in His Parents' Garage

Comments Filter:

  • Hmm (Score:5, Interesting)

    by backslashdot ( 95548 ) on Sunday January 23, 2022 @06:41PM (#62200471)

    It is great we have people doing stuff like this at the hobbyist level, even though many academic labs and courses will give you the knowledge to do this. In case there was a mistake in the early evolution of microprocessor tech, he might find it. I guess that is why, besides for just fun, he may have chosen to learn this way instead of at a university where, in my opinion, he would develop ways to make existing CPUs better by thinking along the edge of the current evolutionary tree. By going back to practicing things the old way .. there is a chance of getting ideas to do things totally differently. Maybe find a way to make better 3D CPUs .. or use non-binary logic. Who knows?

    • Fixing the global semiconductor shortage, one hand-made chip at a time...

    • What a great achievement by this lad. First of all, I had never imagined one could create chips outside a fabrication factory. It just blew away my head reading the article. I would love to create semiconductors devices at home - I always loved understanding how to build things using first principles. As a engineer hobbyist, when come circuit fails, the amount of time and effort in debugging the root cause and fixing takes so much time. Now imagine doing chip fabrication and one bit wrong or improper at one

    • Re: (Score:2)

      by sjames ( 1099 )

      It's not really a matter of him discovering an early mistake. It's more moving it from expensive ivory towers to the family garage. It's early days for home semiconductors, but it's in the same theme as moving computing from big expensive rooms in a corporation with raised floors and secured glass rooms to the living room that kicked off in the '70s.

      I'm not holding my breath for him to start cranking out ARM chips, but it's a good effort at more basic parts.

  • This is really cool, however... I'm holding out hope that organic semiconductors get to the point where it will be practical to print several thousand transistors with a printer that costs a few thousand.

    Long way to go, but it might happen.

  • Very cool and I doubt I would have the patience or skill to do this, but of course his growth rate can easily outpace the early semi conductor industry, he doesn't need to invent/innovate/productize/mass produce anything.
  • No bunny suit? Does he sweep out the garage regularly?
  • I understand the result isn't as important then the process for this journey, but is there a practical advantage to doing this at home vs just using FPGA? Without any real knowledge, it seems that approach would be much more practical for the enthusiast.

    • Re:Comparison to FPGA (Score:4, Interesting)

      by narcc ( 412956 ) on Sunday January 23, 2022 @10:15PM (#62200923) Journal

      It's neat. Why isn't that a good enough for you?

      I'm guessing jealousy.

    • Those are almost entirely different skillsets. Both worth having and exploring.

      FPGA - The design of logic with synthetic HDL to RTL to gates. Lot of gates to play with.

      Making your own chips in your garage - messing with atoms in super pure silicon using chemistry and light to make transistors.

      He'll do fine. I'd employ him if I had RCG reqs to spare.

    • I understand the result isn't as important then the process for this journey, but is there a practical advantage to doing this at home vs just using FPGA? Without any real knowledge, it seems that approach would be much more practical for the enthusiast.

      It really depends on what you're an enthusiast in. If you really want to build high gate count digital circuits, then go nuts with an FPGA. If you are an enthusiast in chip fabbing, then an FPGA is completely useless. And an FPGA can really only do digital l

    • Re: (Score:2)

      by hey! ( 33014 )

      I'd guess one advantage might be if you wanted to include some analog features on your chip like A/D. Theoretically an FPGA takes more area and consumes more power than a custom IC, but that's negated by orders of magnitude superior fabrication.

    • Re: Comparison to FPGA (Score:5, Informative)

      by Miamicanes ( 730264 ) on Monday January 24, 2022 @11:46AM (#62202819)

      There are a few issues with FPGAs and CPLDs:

      1. Open-source software to synthestze & program them is nonexistent, and attempting to reverse-engineer it in any useful manner that wouldn't get you sued into oblivion is nearly impossible.

      2. Number 1 notwithstanding, vendors have begun moving the 'synthesis' part ENTIRELY to their own cloud-hosted servers, so even if you were willing to risk a lawsuit, you can't get close enough to the actual code doing the synthesis to analyze it anyway.

      3. Modern chips use signed and encrypted firmware. So analyzing the generated firmware is almost impossible, and taking advantage of anything you learned by independently synthesizing your own is equally impossible.

      4. Free editions of the tools to synthesize code for FPGAs & CPLDs exists... but vendors aggressively remove support for "End of Life" products from those versions to prevent cloned silicon competitors once the patents fall. A company like Sony or Boeing can still get the commercial tools to use EOL'ed chips, but they're so expensive, literally nobody who doesn't HAVE to support legacy products does.

      Case in point: I have a bunch of Mach64 CPLDs and a programmer I bought ~13 years ago. Literally 6 months later, they removed support for it from the free version of their ABEL compiler, the old copy I have only works under Vista, and it still requires online authorization to use that could be revoked at any time.

      People have been trying to bring FPGAs into the Arduino ecosystem for YEARS... and mostly ended up stymied & stonewalled by vendors who impose arbitrary EOLs and make their "free" tools as annoying as they can, while the entry fee to step up to the "real" tools starts around a thousand bucks & increases rapidly from there.

      The relentless deprecation & EOL'ing of "old" chips also means that pretty much any one you can buy today is either BGA or has something like 0.35mm pitch, and has 2.7v or lower I/O. Older chips in larger sizes and at least 3.3v i/o used to exist... but they're no longer programmable with "free" software tools anymore.

      Interfacing 3.3v with 5v is kind of a pain, but 0.9v is another level of 'hurt' entirely, because the latest FPGAs don't even "play nicely" with breakout boards. They're designed for use by companies that literally have staff & manufacturing to do 10-unit prototype builds every few days... and the older chips that COULD be hobbyist-friendly have been "written out of existence" by removing support for them from the "free" versions of software, even if the chips themselves still exist at DigiKey & Mouser.

      Vendors also discourage hobbyist use of older chips for another reason... they have enough "new old stock" warehoused to meet the "likely" needs of "legitimate" past users. If hobbyists discover an old chip that's warehoused, but hasn't been manufactured in 10 years, and start using it for Arduino projects, old stocks will be rapidly depleted, and then the vendor's "real" customers will scream.

      What really needs to happen is for someone like Arduino (the company), Parallax, or RasPi foundation to buy/license the IP to old (early-2000s) FPGAs from someone & resurrect them for hobbyist use. But that's unlikely to happen, because companies like Lattice & Xilinx would prefer to have old chips just go away completely, lest they end up cannibalizing their new chips. Also, big companies like GM and Boeing LIKE the fact that commerciall use of modern FPGAs has high barriers to entry. It's kind of like the relationship between Qualcomm & US carriers (and their "partner" manufacturers). They scratch each other's backs in a cozy relationship, and don't WANT "disruptive" companies to mess up their neat 5-year product roadmaps (say, the way Google did with the Nexus One, which pretty much smoked 4 years of planned incremental roadmap improvements in one single leap and left every other company scrambling with phone designs that were obsolete before they even made it into pre-manufacturing).

  • Intel's response (Score:4, Funny)

    by Gabest ( 852807 ) on Sunday January 23, 2022 @07:41PM (#62200599)

    The guys are on the way with baseball bats. Before this gets out of control.

  • This is news for nerds, (Score:4, Insightful)

    by renzhi ( 2216300 ) on Sunday January 23, 2022 @08:10PM (#62200657)
    Yet so few comments. I guess Slashdot can change its slogan now, I guess nerds are not lurking around anymore. Anyway, this is such a cool hobby.
    • Or maybe there's no need to comment from us nerds. He reinvented the wheel. That's cool and all, but like we all knew it was possible. Just not practical. If he had found a way for us to easily and cheaply make chips that can match, say, mid to late 90 cpu's without having to deal with acids, then it would warrant more discussion. As of now, this is more a social/economical statement about how people approach innovation.
      • Why would you want to make a relatively modern cpu of all things. No this would be for a custom mems device, wavelength scale optics, all sorts of stuff you can't just buy because they are only made on custom order for setup cost of a fortune

      • You are a total wanker. I guess the people who build other things from scratch too are wasting their time because it's already been accomplished.

        • That's literally not what I said. If you want to throw your own context in there, that's your thought, not mine. I'm just saying it's not anything that us nerds would comment about because it's someone who has access to the tools, materials and instructions doing a thing with said tools, materials and instructions. Is it cool? Yes. Is it something to go into deep discussion over and gush over? Not really. Like if, tomorrow, I wrote and OS common to the 1970's from scratch (aka via pure asm), no one would co
      • This is spot on. It is not hard to play around but there is a lot of background needed to actually make developments. Mixing up chemicals and having a go at materials science and EE becomes a lot more useful once some background is in place. It isn't really news for nerds at all. Photolithography can surely be fun but things are so well developed and complicated that the first step is to see what current state-of-the-art is and work becomes imoressive when somebody can match or beat it. But this story is m

    • Re: (Score:2)

      by Tablizer ( 95088 )

      "Dupes for Nerds"

    • His hobby is basically collecting vintage gear, except in this case it's gear for making microchips. The usual caveats (old gear isn't always easy to get, is frequently found in a non-working state, and isn't always cheap) to such a hobby still apply, so this article is a bit disingenuous in claiming this young man is doing anything to make building semiconductors more accessible to the "average garage tinkerer".

      • Re:This is news for nerds, (Score:4, Insightful)

        by serviscope_minor ( 664417 ) on Monday January 24, 2022 @05:39AM (#62201779) Journal

        this article is a bit disingenuous in claiming this young man is doing anything to make building semiconductors more accessible to the "average garage tinkerer".

        Bullshit it isn't.

        It's a decent progression. Jeri Ellsworth managed some homebrew transistors, this guy took the knowledge and extended it further and can now make whole chips. Those are the first two steps along the way. At the moment he uses a mix of old and cobbled together kit. His photolithography machine is a microscope coupled to a projector. Sure he has a SEM, and curve tracers, but turns out you can build them too if you want. Every time a hobbyist makes an advancement and documents it for other hobbyists, the process becomes easier and simpler for other hobbyists.

        It's not going to happen in one step nor overnight, but if I wanted to have a crack at this, it would be substantially easier now given his knowledge than it would be before he did this.

        I swear some people won't be happy unless he's all like "build a 3nm, 69GHz core i11 in your basement with this one weird chip that semiconductor engineers don't want you to know".

        • Re: (Score:2)

          by hey! ( 33014 )

          One thing that has fascinated me is the primitive tech enthusiasts who have attempted to create glass -- just plain old transparent glass -- just from sand they can dig up and using minerals from evaporating ponds for flux. It turns out to be crazy hard to get anything close to decent glass, even if you've read textbooks on glassmaking and have a pretty good knowledge of chemistry.

          The same goes for people who've tried to create bloomery steel from ore they can dig up themselves. Even if you know the outl

          • One thing that has fascinated me is the primitive tech enthusiasts who have attempted to create glass -- just plain old transparent glass -- just from sand they can dig up and using minerals from evaporating ponds for flux. It turns out to be crazy hard to get anything close to decent glass, even if you've read textbooks on glassmaking and have a pretty good knowledge of chemistry.

            I didn't know that, that's quite remarkable.

            This kind of activity has no practical use

            Not yet, maybe not ever, but we may end up

    • What kind of a nerd? Nerds come in many flavours, not all people are electrical nerds, or IT nerds. Don't gatekeep.

      Also what kind of a reply do you expect? It's a cool story. I'm happy to see it. I tinker with electronics at home as well, just not at this level. Should we all feel obligated to jump in and say "attaboy" even if we have nothing of value to add?

      Comment count doesn't reflect people's interest. It reflects people's emotion. It reflects controversy. The comment count is low because there's nothin

  • Microchips on Mars. (Score:4, Interesting)

    by MacMann ( 7518492 ) on Sunday January 23, 2022 @08:56PM (#62200745)

    This looks almost like the sequel to The Martian where some people on Mars need to figure out how to repair some vital electronics until a supply ship from Earth can bring spare parts. Our tiny colony of people on Mars needs to hack together a microelectronics lab to produce a handful of electronic circuits using whatever they can find inside the habitat. in one scene they take the microscope from the biology lab, and the projector from the movie theater, to create something that can produce the fine photo lithography required.

    I expect something like this to show up in some science fiction, if not happen in real life, on Mars. Perhaps it won't be so dramatic as a countdown to a life or death moment. It might just be some people thinking that they could make some money in producing 1970s era electronic chips for some toys. The Merlin electronic toy might come back. https://en.wikipedia.org/wiki/... [wikipedia.org]

    Of course making electronics on Mars would be great for producing calculators, rocket control systems, and so many other useful items. We'll probably see this used to make toys first. That's just a guess based on how things played out before. Computers were toys in the home long before they were considered tools. People won't trust Martian made electronics for life critical functions right away but they'll use them in toys. As the process improves and gets some real world testing then people will trust it more.

    Another thing this shows is how a society could rapidly develop technologically if given the knowledge up front. It doesn't have to take 100 years of development to get from electronics as a theory to the iPhone 13. A sufficiently large population with access to just dirt and rocks could go from cave dwelling and cooking over campfires, to air conditioned homes and meals cooked in microwave ovens, in a very short period of time if they had a pile of books to tell them how to get that done.

    • Yeah and just grab all the purified silicon wafers from out back. Then grab some of that natural Martian acid to etch away excess. Then wire it all together with some wires that grow freely on the Martian surface.

      This guy's work is really interesting but it's still tied to a long complicated logistical base you're only going to find on Earth in relatively developed areas.

      • Building chips on silicon substrate is convenient, but it's far from only viable option. There are lots of ways you can simplify the process to make do with lesser resources. For one offs you can skip photolito entirely and etch with electron beam for example.

      • Any colony on Mars is going to come with the basics of a scientific laboratory and repair shop. They are going to have things like a microscope and oodles of different solvents. If there is a colony that is down to just the dirt under their feet then they are royally fucked, not because they lack silicon wafers and tungsten wire but because they can't eat.

        Nothing in this young man's workshop were items that would be all that unusual in a facility fitted out to provide basic medical care, food preparation,

    • The riddle is simple if you know the punchline.

      A big part of R&D is not knowing what is possible. It took thousands of tries to find a filament that would work for an electric light. It was a long while until we gave up Roman numerals for Arabic numerals and a base-10 system. Newton "wasted" half his life trying to transmute lead into gold; we accept now with hundreds of years of chemistry that this is AFAIK-impossible without nuclear bombardment.... but the point is they didn't KNOW then that it

    • Re: (Score:2)

      by Kaenneth ( 82978 )

      Just buy the game 'Factorio' and install the full set of 'Bob's and 'Angel's mod packs if you want to simulate this.

    • It's an interesting mental exercise to contemplate how quickly someone with modern knowledge could bootstrap ancient tech to "invent" modern goods decades or centuries before they were "supposed" to be discovered, or do it "out of order".

      For example, propane could have probably been obtained in the early 1800s if you knew it existed, how to collect it, and got yourself a supply. Using it as a refrigerant would obviously be dangerous (especially if you made a compressor that used coal-heated steam for mechan

      • Or... building upon "out of order" inventions, develop shitty, primive red, then amber, LEDs circa 1750 and combine them with chemical batteries to go into business with Ben Franklin selling disposable "electronic candles".

  • From 1970 (Score:3)

    by dsgrntlxmply ( 610492 ) on Sunday January 23, 2022 @09:10PM (#62200779)
    Scientific American June 1970 Amateur Scientist feature was on homemade CdS thin film FETs. I worked with the pyrolytic chemistry very crudely, steering clear of cadmium, only trying and failing to get a film of something metallic. Zeloof's I-V curves mark an admirable accomplishment.

  • "Maybe it's overconfidence, but I have a mentality that another human figured it out, so I can too, even if maybe it takes me longer," he says

    I love that attitude!

  • Yes it'd be nice to have that kind of a lab. But this article completely ignores the technology and physical tools necessary for non-dinosaur era chips (relatively speaking). Most inventors aren't going to be getting quality fab surplus nor at obtainable pricing.

    The biggest fear: the crap you'd be pouring down the drain because you don't appreciate the ancillary lab equipment, safety, or a silicon guy wants to play with the fun crystals and starts outgassing As because his equipment is improper. There ar

  • What does the Z2 chip (the one w/ 1200 transistors) do?
    I don't see it written in the Wired article.

    • According to his tweets, it's a transistor array. That's a basic test circuit you might build in process test chips to characterize the transistors, right next to the IDVs and SRAMs.

      Building test logic on flakey test chips to test the logic on the flakey test chip is fun. The test logic needs to be reliable, while the tested logic needs to expose just how flakey or non flakey the silicon is. So the test logic is built with modular redundancy, self correcting logic and other goodies to make it reliable. Stuf

  • Having worked in an 80s wafer fab (Score:4, Insightful)

    by oldgraybeard ( 2939809 ) on Monday January 24, 2022 @12:18AM (#62201201)
    this is neat! The fact that he has been able to get as far as he has is amazing. And kudos to him! you really have to wonder where he will end up at. Having some resources to work with is one thing. But actually the having the drive and determination to produce an end result. Like I said, amazing.
  • This is on a totally different level compared to making vacuum tubes. https://www.youtube.com/watch?... [youtube.com]

  • Nice. I like it. A lot. (Score:5, Interesting)

    by Qbertino ( 265505 ) <moiraNO@SPAMmodparlor.com> on Monday January 24, 2022 @03:35AM (#62201531)

    Now that's an all-out classic Nerd hardware hobby if I ever saw one this millenia. Very nice. We need more of this. Imagine building your own Open-RISC Vs in some Microfab in the basement, in 20 years or so. I'd join a local crew doing that.

    • That ship almost sailed few years ago. Almost.

      The 3D printing of chips was supposed to be one of the applications of a multi parallel e-beam lithography developed in the Netherlands.

      Unfortunately the technology was purchased wholesale by ASML and retooled as an inspection method rather than a production one.
      Thus, the possibility to buy relatively cheap, but still fast enough customizable on the spot lithography machine for experimental and low volume unique chips is lost. Probably forever. Sad.

  • Comment removed (Score:5, Funny)

    by account_deleted ( 4530225 ) on Monday January 24, 2022 @05:43AM (#62201803)
    Comment removed based on user account deletion
  • He has a YouTube channel too.

  • Did he use the Jeri Ellsworth process?

    https://www.youtube.com/watch?... [youtube.com]

  • A MOS 6502 CPU has 3510 transistors and the largest logic chip, the VIC-II, has around 5000 transistors.

    • Re: (Score:1)

      by ebvwfbw ( 864834 )

      A MOS 6502 CPU has 3510 transistors and the largest logic chip, the VIC-II, has around 5000 transistors.

      Nah, make an Atari 800. Then he can play missile command and other fun stuff.

      Either way, I think that's amazing.

  • > His second chip has 200 times as many transistors as his first, a growth rate outpacing Moore's law, the rule of thumb coined by an Intel cofounder that says the number of transistors on a chip doubles roughly every two years.

    That's a silly observation. Moores law doesn't apply when you are using salvaged/modern equipment to build chips that are roughly equivalent to the late 1960's in complexity. It sounds like he just built a really tiny chip the first time. He didn't have to spend time developing

    • Re: (Score:2)

      by 3vi1 ( 544505 )

      And I don't mean that as a slight to what he's achieved. I mean it to the author who's trying to imply this guy has somehow achieved something tech giants could not - when the situations are not at all comparable.

  • In the 1980s there was a project hosted at ISI called "MOSIS", which supported researchers around the country who wanted to create their own experimental chips according to their own specifications. They would send chip descriptor files to MOSIS, and MOSIS would grind out a run of chips according to those specs.

    Unfortunately, some of those researchers were connected via CSNET, the world's first ISP, where I worked at the time. CSNET did provide reasonably high-speed IP access to customers who could afford

  • Keep Working Hard... recruitment agency Dubai [markwilliams.ae]

Slashdot Top Deals

Two can Live as Cheaply as One for Half as Long. -- Howard Kandel

Close