Make Your Own Integrated Circuits At Home

The Nyan Cat you see above is only 600 micrometers from head to tail. To put that into perspective, that’s about 10 times the diameter of a human hair. Also, that Nyan is etched into 200 nanometer thick copper foil and is the work of the HomeCMOS team, who is developing a hobbyist-friendly process to make integrated circuits and MEMS devices at home.

The project is far from complete; HomeCMOS has yet to produce a working IC but a few experiments – getting wet etching down pat and even building an almost working quantum qbit – are remarkable given the small amount of equipment and tools involved.

The HomeCMOS team has yet to actually make an integrated circuit or MEMS device, [Jeri Ellsworth] has shown this is possible by making transistors and integrated circuits at home. While there won’t be chips with millions of transistors coming out of the HomeCMOS lab anytime soon, it’s more than possible to see a few small-scale integration-level tech such as a few logic gates or a regulator.

25 thoughts on “Make Your Own Integrated Circuits At Home

  1. Its always cool to see someone taking something is normally really expensive and find a way to do it at home on the cheap.

    Sure, no-one is going to be producing chips at a 22nm node size (or even 22um) in their back shed anytime soon but its still cool that someone has a way of (hopefully) building a functioning computer chip without a massive expensive fab setup.

  2. Very cool! Although, to be impressed, I had to look up a conversion to mils (turns out to be about 23 mils), but once I got that number this was very impressive!

  3. A few years ago you could purchase actual factory gear at scrap-metal prices, and it would have been easy (well, easier) to do 70’s vintage IC fabbing.

    Most of the gear got BRICKed. He’s within an order of magnitude of where he needs to be to make the process work, so that’s something.

    It would probably be a good idea for this guy to hook up with some of the guys reverse-engineering chips and ask them to try and come up with the masks and artwork for the 4011 NAND Gate.

    Being the first to build a 4011 will provide all the basic knowledge and process to allow him to build an 1802, which could also be reversed.

    If he doesn’t have access to an SEM or similar device (they can be used for etching – sort of a rep-rap for silicon), he’s stuck with optics and masks and lots and lots of micro chemistry… along with all the requirements for cleanliness, procedure and attention to detail.

    Etching nyancat? pssht. He could be etching analog circuit components on 1D copper, meaning inductors, resistors, and capacitors with the two stage tech he already has. Presto – analog flip-flops and multi-vibrators… but instead, we get nyancat. That’s fine, but dude – make something.

    You can shorten the development process by reading 1970’s chip fab/silicon trade magazines and obtaining excellent process overviews and even discover techniques that won’t be available in google or some text book.

    As for the almost working quantum qbit… it was actually a sculpture of a working quantum qbit, which has the same relationship to the word “working” as the sculpture of themis has to justice.

    But again – 10,000+ points for alacrity.

    1. The reason I haven’t made anything bigger than nyan cat is that I ran into a brick wall with field of view – I was doing projection lithography through a microscope and my FOV was just too small.

      I’m currently in the process of designing and building a contact mask aligner for 2-inch wafers that will use commercially made 8000DPI masks from laserlab.com (about $50 per 12×18 inch sheet; this is enough for a full set of masks on a 2-inch wafer). If all goes according to plan at that point I’ll be able to do full-wafer lithography at the 12.5um node.

      I’m involved with John McMaster and the Silicon Pr0n team and we have no shortage of 74xx / CD4xx samples around to work with.

    1. I haven’t done any yet (been focusing on improving lithography) but I will be using doped spin-on glass from Emulsitone (http://emulsitone.com/alphacatalog.htm). The process looks something like this:

      * Spin coat doped oxide, bake to densify
      * Spin coat photoresist, soft bake
      * Lithography for dopant pattern
      * Etch in 3% HF
      * Backfill with undoped SOG
      * Put in furnace and do diffusions

      This is based on Jeri’s FEOL process. I’m going to be shrinking her process using my lithography techniques and then putting a BEOL process (possibly copper damascene) on top of that.

  4. I wonder if you could use a dvd burner to generate a mask (on a dvd disk). the patterns you write on a dvd are pretty
    small. you would need to develop a custom driver for it, though.

    1. Laser direct-write lithography is on the long term TODO. I’d probably use a blu-ray diode since 405nm wavelength is right in the middle of the maximum sensitivity range for my photoresist.

      I’d probably also mount the wafer on an X-Y table and move it under the diode rather than reusing the spinning platter. Most chip layouts are more amenable to Cartesian than polar coordinates.

  5. A friend and me are planning on something similar yet completely different – our goal is to copy the 6502 or similar Chips while using self-built machinery where possible. The efforts of HomeCMOS-Team come in very handy by solving a few problems we thought about.
    So far, we are still planning on the devices we want to use, but if we can figure some things out, we will build a small website for all of you who are interested.
    Our plan at the moment is to build a modular free hardware vacuum system which we can use to try different types of electron-etching.
    we know that it sounds quite unbelievable, but we have no pressure to succeed and we are willing to experiment a little bit despite the fact we will probably fail…

    And by the way, sorry if my english sounds bad, it’s not my native language.

  6. This is exciting. However, I think you should be looking into fab’ing analog chips. There’s a big market in boutique op amps for audio applications. If you manage to finally fab the ‘chip that really sounds like a tube’ you’d be rich beyond your wildest dreams.

    Given some segments of the audiophile world’s proclivity to believe BS, you might not even have to achieve that milestone, as long as you can convince a few well-heeled or influential members that you did.

    1. About that…

      The market for audiophile electronics is a lot smaller than you’d think. I know a guy who makes a tiny bit of money building retro-tube stereo systems, but the truth is his production barely pays his costs in time.

      I used to have the setup, and yes, it sounded great. When I bought my first integrated modern receiver, I felt ripped off, and it was a $400+ unit that gets rave reviews. It was absolute junk compared to my discrete setup.

      But after 3 or 4 days, it sounded fine. Our ears become accustomed to sound quality, even if it’s bad – many .mp3 audio files are of rather low quality, but nobody even notices.

  7. U could combine the two. Jeri used acid
    Etchant to etch the die, which is harder to make smaller. But what if u make it a multi step process using a dvd burner to etch the holes in the die, then spin on dopant (potentially while still in the drive!) remove the disc and bake, etch off extra dopant, reinsert disc into drive, burn gate dielectric holes, remove, bake again, and ur done.

Leave a Reply to John FendonCancel reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.