Fun times. It is fortunate he has the means to pursue this interest, I would have been so jealous of him when I was in HS!
I wonder what he is using to do metalization. There is the sputtering setup and the etching setup. Some nasty stuff involved in that. You can of course do quite a bit with polysilicon conductors but if he wants to do more useful things (as opposed to 'look it wiggles!' sorts of things) I'm guessing he will need some permits from the city :-). There is a reason nearly all the old fab sites in the bay area coincide with superfund cleanup sites :-(.
I worked in small scale fabs back in the day. One and a half inch wafers. It’s not nearly as hazardous as you make out, but it does require training in chemistry. I’m a physicist, but I also did a pre-med, so I had the necessary chemistry background to be safe.
Sputtering requires a small amount of high vac equipment - diffusion pump plus LN cold trap is fine. Sputtering titanium plus gold is straightforward and non-toxic. Etching the pattern requires small amount of aqua regia, which is, of course, a strong acid, but can be easily neutralized to non-toxic by-products. Aluminum evaporation and etching is even easier. The complexity and cost comes in four areas: nanoscale features, large wafers, high throughput, and a process that doesn’t need professional chemists to perform.
Yeah, but doing it without a lab makes it a lot more difficult to do safely and with multiple likely dangerous (toxics, oxidizers, mutagens) byproducts (including the Si dopants) doing it at home is sketchy. Could one person do it in something the size of a garage? Yes, I think 20-40 probably do every year in a university clean room. But building that for a single person is an exercise in both chutzpah and perhaps (toxic) waste. It's a lot more tempting to cut safety corners when you're doing it for the first time, and haven't made THAT mistake yet.
There is a manual for it though (thanks abecedarious).
Absolutely right. This isn't something anyone should even contemplate in the kitchen! If they are thinking about it, then they don't know what they don't know. Experience in chemistry is needed, and that's unfortunately a lot rarer now in the US than it was 50 years ago. People (and governments) have developed a (mostly unwarranted) fear of it.
The minimum processing requirement (aside from the furnaces and metallizing equipment etc.) is a polyethylene sink hood with at least 100 ft per min flow and a neutralizing tank connected to the drain line, along with fume exhausted chemical storage below. If you can't acquire or build that, you shouldn't proceed further. I did a lot of commercial work in a three foot hood with a large Corning hotplate.
You also need an arrangement to recycle your organic solvents.
People are not this dumb- they have come to associate large scale chemical (aka chemistry) processing with environment and health issues- and although nobody in the various industries likes to hear it- they are usually right. Every city has some paved over "dead" zone from the 50s and 60s, before we learned to offshore the burden to third world countrys.
Not to be a grinch, but I followed a lot of his videos, and I think he has not yet actually put 2 transistors on the same die, it's the technological threshold for an IC. I think he is working on making patterns (and that's interesting, because nobody is working on cheap masking), he made some individual transistors with masking tape, but not combined yet.
But his videos are very cool, and I'm looking forwards to more. I didn't realize he was that young.
Caltech had a freshman course like this in the mid-80s, using a homebrew low-tech process. The dire warnings about hydrofluoric acid have stuck with me ever since. It was fun to try to lay out your gate to take as little area as possible -- fun you can revisit at http://www.zachtronics.com/kohctpyktop-engineer-of-the-peopl...
I was 17 at the time too, but doing it on your own is something else.
HF is dangerous and you definitely need proper PPE and calcium gluconate + a shower accessible but lots of us, myself included, work with it routinely.
warning: I'm not advising you go home and start etching silicon oxide and cleaning wafers with vats of HF without doing adequate research and preparation to do it safely.
Working on a small scale, say 40 mm wafers, you’ll be dealing with small amounts of HF - approximately 250 ml. But.
There are two problems with HF: first, it doesn’t give a burning sensation like other acids, so there is a tendency to discount the seriousness of a major spill once you get past the shower off stage. Second, it is rapidly absorbed transdermally and causes an insidious necrosis, but, more importantly, it can affect blood potassasium balance and cause heart failure after some delay. While the emergency rooms on the peninsula are aware of the risk and the treatment, that may not be the case in rural Iowa. So do the homework, have a supply of gluconate on hand, and get to the emergency room right away (with the msds) if you wet, say, more than 50 sq cm of skin with HF. Not all people are susceptible, but EKG monitoring is essential following major bodily contact.
Always wear gloves, and test them. Since HF doesn’t give a burning sensation, it’s easy to get a small puncture and an HF burn. I remember one older tech who was lax about this and he lost fingernails over time. He would joke about soaking his hand in a bucket of water by his bed so he could get to sleep. Pioneer days!
Yep, Whink brand rust remover is 4% HF. It's a somewhat useful source of hydrofluoric acid in a pinch. I recall a procedure for making uranium tetrafluoride from uranium trioxide (which in turn was extracted from uranium ore) using that brand of rust remover as the acidic fluoride source.
I guessed correctly that he was inspired by Jeri Ellsworth, and that the article would mention her too. Great to see he went beyond what she did, and although trying to clone a whole 4004 might seem far-off, it's really a matter of replication once smaller structures can be made.
Another relatively "simple" early microprocessor that might be doable is a 6502, it also has the (reverse-engineered) mask layers available: http://visual6502.org/images/6502/index.html
I wonder if he could make something like the totally unobtainable Curtis filter chips, or other devices which are long out of production but still desirable? I'd think there would be a business in there if it is doable at a sane price. Obviously I have no idea how complex the internals of such a chip are - is an 'analogue' chip significantly different internally to the CPU talked about in the article?
In general, analog chips are easier to fab because less transistors are involved and thus minimum feature size isn't as important. Typical parts have merely a handful of transistors.
Cutting-edge analog ICs are a different beast, though. Tightly controlled processes are the norm and post-fab trimming is common.
I've always heard that analog is harder because you need to characterize the full behavior of the transistor as opposed to just setup and hold times for digital.
Then again I only tinker with discrete components so it's an area that's somewhat foreign to me.
It all depends on what you are trying to do. Some good analog parts use a BiCMOS process with both BJTs and FETs, which is exotic and complicated. The margins of any process are usually complicated, though. Just look at how many billions it takes to develop a cutting edge digital process.
In my opninion it would be easier to do interesting analog parts in a home lab than an interesting digital part. Making an opamp would be easier than a processor and more interesting than an XOR gate (around same # of transistors). Also, you can do some analog stuff with only NMOS or PMOS, which simplifies things by at least 50% compared to CMOS.
For a home lab, you care about process and characterization, but variation isn't as much of a concern because you won't be making that many. Also, just look at the development of the semiconductor industry. Early factories (making analog parts) didn't look that much different from what you could cobble together in a garage. By the time they were doing integrated processors, things were a lot more high-tech.
I have worried for years that budding electronics folks don't get an opportunity to tinker any more, now that everything is integrated. Apparently I need not have worried.... Godspeed to this new generation.
Wasn't surprised to see it was one of Jeri Ellsworth's videos that inspired him. The production quality of her videos seem stuck in the early 90s but the actual content is top notch inspiration. She has a great one on the importance of failure that anyone struggling with not being perfect could benefit from. Nice to see younger folks getting value from her sharing.
He should work with a lab like SNF/ANFF to make more advanced devices, you should be able to get access to one of the Raith E-beam machines for an acceptable price.
E-beam litho is...uh...not cheap. The system in the shared fab I use (WNF) costs around $450/hr with various discounts applied. Also, you're not going to get much done in just an hour. Now, maskless laser lithography is reasonably cheap...
However, you can get reasonable quality masks for $150ish, and if you don't care about wafer yield, you can fit all of the layers on one mask. It's 1970s tech - 1:1 contact exposure, but you can hit a few microns.
Fair enough, I guess we have different ideas of what cheap is, $4500 for a small prototype is fair enough for me (especially considering there's no other way to get ~50nm lithographic resolution), but on the other I run a funded company, so I suppose the story is different.
This. Is. Awesome. Also, horribly frightening, if you've ever worked in old fabs, they had some extraordinarily toxic stuff. My brain is torn in two trying to figure out if I should cheer or shriek in fright. It reminds me of the story of the kid who made a nuclear reactor.
Very cool. There's one way to make sure that you're not getting any free 'riders' on your Silicon though it will be quite a while by the looks of it before he'll be able to make a CPU that can do useful work outside of small embedded projects by todays standards.
Wonder how he deals with the various solvents required as well, Silicon wafer production is notoriously dirty.
To clarify, the P in P-type stands for positive free charge. Phosphorous is actually an N-type dopant in Silicon, where the N is for negative free charge.
I have really wanted to do this myself. Buy some older fab equipment, and produce NAND chips. Start selling 4TB SSDs for $100 or so. It's ridiculous how long the price-fixing in the SSD industry has been permitted to persist, and someone needs to take them out. We don't need even 10nm production for SSDs or anything. Go ahead and fill a standard 3.5" drive cage with the device. Since the chips function mostly in-parallel, you get speed increases just from concurrent use. And I'd imagine you might find a subversive company willing to sell you controller logic on the cheap. Just so absurd that such a basic part as NAND chips can even get treated like a premium item despite being in basically every product of any kind created in the past 10 years. Any other component like that is commoditized and dirt cheap, especially when they contain no rare materials.
Does anyone have a single, basic book recommendation that would describe chip fab all in one place?
Also, it seems to me he or people like him should try to make microcontrollers rather than "full" computer chips. I would buy something to drive blinky lights for $5.00 that ran 1/5 as fast as an Atmel chip to support a free chipmaking movement, but I can't see doing something for a general purpose computer chip (with the same ratios at least).
If you take a look at the circuitry on some recent cheap multi-mode blinky lights you can find a likely analog IC fabricated right there on the surface of the small PCB and covered with a drop of epoxy, along with a few discrete components and battery holder.
I found a more advanced digital application where the memory modules on our automatic titrator's exchangeable burets are fabricated on about a 5x20mm ceramic substrate and covered with the epoxy droplet with only the gold-plated connector pads showing. Only a limited byte capacity is needed, enough to contain the serial number and a line of variable text to store a couple user variables.
Small feature size and physical chips off of a multi-IC silicon wafer are not exactly essential for such a simple application.
My high school didnt even offer drivers ed, couldnt have a football team because of drive-by shootings, shop class wasnt an option after my freshman year (or metal shop) because of a fight that occurred, there were no opportunities for this when I was coming up.
This is something high schools need to be investing in, literally this is what is fueling our future, I think it's amazing, good job.
I'm going to disagree. It is really interesting but I think the amount of equipment, space and potentially dangerous chemicals (requiring extensive supervision for high schoolers) and the associated PPE and proper infrastructure to work safely aren't the best use of money/space for a high school. Old equipment also requires a lot of maintenance to keep running. You can learn about these things from a book, watching a plasma etcher or CVD furnace run isn't exactly very exciting after all. Might be worth doing a few of the simple processes but you need so much equipment to do the whole process. The lithography approach mentioned is interesting and could be cool in an undergrad/high school setting.
Right. A shared fab run is better: many designs are merged on one wafer, and the resulting chips are packaged and returned to the designers. The cost is split, each designer gets only a few chips per run. Right now, the turn-around time is lengthy, but, like PCB fab these days, that could change.
props to him and his parents for supporting this. Hope this story will inspire other kids. Personally i'm inspired, adding this to my "bucket list" of satisfying things to do when i will have time (someday...i hope...)
You mean the kid who disassembled a clock radio, intentionally made it look like a bomb in a suitcase, used it as a distractant in multiple classes, eventually got appropraitely disciplined for it, but then got tons of free stuff, scholarships from universities, free money and swag, and personally invited to White House by the President?
Yeah, heaven forbid anyone be treated that poorly.
Woah, I've never heard the story told like that. Could you link me to an article or something? If true that's pretty depressing, I've seen a lot of kids randomly blessed by corporate PR teams that want to look like they're supporting education but I thought at least that story was true. (It's not hard to make a clock and electronics in public places have caused a silly bomb scare before...)
I wonder what he is using to do metalization. There is the sputtering setup and the etching setup. Some nasty stuff involved in that. You can of course do quite a bit with polysilicon conductors but if he wants to do more useful things (as opposed to 'look it wiggles!' sorts of things) I'm guessing he will need some permits from the city :-). There is a reason nearly all the old fab sites in the bay area coincide with superfund cleanup sites :-(.