Awesome video and process, we've been talking about this at work since it showed up here and have a few questions if you don't mind:
How far do you think you'd be from making a simple 8-bit CPU like a 6502?
Have you thought about illuminating the initial film exposure as well as the substrate exposure (on the slide glass) using monochromatic light sources like a mercury UV lamp?
Instead of a vacuum pump to flatten the film during exposure, how about electrostatic charges?
Thanks! I'm very far from making a CPU. Check out Sam Zeloof's work to see some of the other parts of home IC fabrication. Creating the mask is just one part of the process, and dialing in diffusion, oxide growth, metal deposition would take a lot of additional effort.
Yes. My original attempt (years ago) was to use a standard photographic enlarger: replace the lens with quartz, place the paper or plastic master design on a UV light source on the table, then place the microscope slide with photoresist in the enlarger where the film would normally go. One problem is the lens is not setup to cover a whole microscope slide. Another problem is not enough UV light getting through the whole optical system -- exposure would have been hours if it worked at all.
Haven't tried electrostatic. Considered low-tack adhesive (3M post-it).
I looked up the 4004 earlier today, thinking about DIY microprocessors. It was 10 micron process. 20 micron is more like 1960-level technology, unfortunately.
As a member of Foulab, home of Thought Emporium, and as an avid watcher of many channels (yours, Thought Emporium, RedNile, BlueNile, etc) which you contributed ideas, knowledge or/and hardware over the years: thank you for spreading such tech around the hackspaces, youtube and homelab community! The content you inspired, made or indirectly contributed to is what makes YouTube the such an amazing platform.
I love your passion and drive towards making things work. Often without a blueprint to follow. Thanks for shining light (sometimes literally) on so many obscure topics.
If you ever find yourself bored, we could really use a reliable method of removing a camera senor bayer matrix. Bonus points if you can figure out how to deposit microlenses back onto the photosites after removal.
Did you ever get the low temp vantablack working as expected? I’m guessing oxygen is passivating the aluminum surface somewhere in the process, and the paper doesn’t document how careful they were. I’d personally want to do this under a nitrogen atmosphere, but I’ve never sonicated under a controlled atmosphere.
I’d bet the reason you can sort of get aluminum to blacken over the heated beaker has something to do either with the fact that the beaker is effectively positive pressure and pushing out everything but water vapor, or that the condensed watervapor on the surface of the aluminum is consistently oxygen free.
I find the former somewhat less believable than the latter, but it’s been a while since I’ve worked on things like this.
If you used your microscope slide as a mask to do photolithography on silicon, you'd use an optical reducer, right? So the feature size on the silicon could be ~5x smaller?
I believe current tech uses 4x reduction from masks down to actual chips. This works well with sizes needed for quartz lens systems, etc. Older processes used 1x reduction (either direct contact, or very tiny offset). This is cheaper and easier because you don't need any quartz lenses, which are needed instead of standard glass for their UV transmission ability.
Interesting. I've got a 180nm mask from a previous job. It's obviously a larger scale than the finished chip. Could be 4x. I think it is from TSCM, circa 2003, for mixed signal CMOS.
I used to travel with a crown graphic - pretty much the same camera as in the video, shooting colour positive. You can do some great stuff with them, and tilt-shift is trivial, if that’s your thing. Quality on slow film is just crazy - I’ve a shot of an ex, she’s half-frame, and you can see the second hand on her watch - you can get very fine focus with a loupe, and having a Polaroid back to do some instant tests was a godsend - but you can’t get the film for it any more.
I have no idea where he’s finding film so cheap - I stopped when it was costing me £5 per exposure.
It also folded up into a very neat little box, which is irresistible to children - sadly one was handed it by a friend, who promptly destroyed it.
Going to admit that I only scanned through the video.
More so than in smaller formats, B&W is cheaper. Older style emulsions (cubic grain) are also cheaper than newer style (tabular / shell / whatever). Litho film is cheaper still.
B&W is USD$1-$3 per shot in 4x5, and if you process yourself you’re only paying for your own time and consumables. Newer styles like Delta, T-Max go for $2-$3. Older styles like HP5, FP4, Tri-X go for $1.50-$2.50. Budget lines like Arista can be found around $1/sheet.
Ortho litho (what is being used in the video) has always been dirt cheap by comparison. It’s like $0.30 per sheet. Basically, all of the technical innovation that has gone into modern film is simply not used in ortho litho. You just need high contrast and resolution, that’s it. So you don’t need complicated sensitizing dyes or anything of that sort.
I've shot a ton of Kodak Ortho film in 4x5 at EI 3, 6, and 12. It's actually super easy to develop. I recommend stand development in 1:100 HC-110 with occasional agitation. The developer basically exhausts after 10 or 12 minutes depending on the temperature, which happens to be pretty much exactly the right development time, so it's pretty much impossible to over-develop.
Greatly reduced Rodinal, on the order of 1:100, will bring some "normal" tones out of hi-con film. At $0.30 a sheet it's not a bad experiment.
I was gifted a roll of expired Kodak Imagelink microfilm stock and once it's cut down to sheets that's my development plan for it. Similar tonal range and ISO.
There are still photographers shooting 4x5 in DC professionally with the press core. One is David Burnett and the other I can't remember off hand, but he's been shooting 4x5 of the congressional hearings. You just can't beat the quality of large format. With Kodak releasing Ektachrome in 4x5 this last January I've been shooting more and more of it. It's expensive for chromes, but shooting HP5 sheets is pretty cheap for the quality of the output in the end.
Impossible Project has brought back peel-apart 8x10 polaroid film. Currently $18 a shot, so not for the faint of wallet, but perhaps they will bring back 4x5 someday as well.
For color, I think it really depends on your willingness to use expired film. I've shot mostly expired film for about 15 years, and once you've tried a range of emulsions of various ages and provenance, it's not hard to figure out how what's going to perform well under what conditions. Even 20 year old color neg performs pretty well with only modest overexposure. B&W you can still get down to about $1/sheet if you're willing to shoot Arista.
Are you sure you mean a Crown Graphic? A Crown Graphic is about as different as it's possible to be from the camera in the video while still being a 4x5 camera (https://vintagecameralab.com/wp-content/uploads/graflex-crow...). It also doesn't make tilt-shift movements as easy as you're suggesting.
The camera in the video is a Graflex Graphic View, so made by the same manufacturer, but yeah, totally different in use. A Crown does not have the same level of movements. But they [Graphic Views] are relative bargains for view cameras if you can find one in decent shape.
One of my coworkers did his dissertation on photolithography, I grew up in a family printing shop and we followed a remarkably similar process to generate the metal offset plates that were used in the presses. It turns out that the processes are so similar that my colleague and I often frequently remark about it during idle chit-chat.
There's pitifully few good explanation videos about the printing version of this process, but here's a quick video (in Chinese, but if you watch the OP's video this one should be easy to follow) https://www.youtube.com/watch?v=KENIyBWNhyk
The process in the printing industry has gotten very refined as well: you can go directly from a digital image file to a plate in one entirely automated step:
How much work would it take to go from this process to producing a working chip in his garage, like a 6502 or something? Would it be at all feasible? He did point out a ton of defects in the final product, probably due to dust or scratches in his equipment/materials. Would it be possible to clean this up enough to get 1-2 usable chips from a single sheet of 4x5 film? Without turning his garage into a clean room?
Having done some work in a university lab, I have no desire to be involved in silicon work in a home environment. It requires dangerous chemicals, and lots of equipment and dedication.
A short list of what I remember:
- Spin coater
- Some solution of exposure of masks
- Various chemical baths (HH, piranha, etc) in a dedicated area.
- Inspection microscope
- Tube oven
- Vapor deposition
- A way to measure film thickness
I wonder if at some point we'll develop new materials and techniques that make it easier to do this, with safer chemicals and less equipment. It may seem far-fetched but you never know. When I was a kid I would've thought it pretty far-fetched that people could 3D print ventilator valves at home that are actually saving lives during a pandemic.
I can't answer most of this since I don't know, but the first 6502 was fabricated at 8 microns. I wonder if a chip made at 15-20 microns would work just as well?
I don't see why not it wouldn't work, assuming you got the process right and produced them in a clean enough environment. Yields would probably be lower than that of a professional foundary during the 60s or 70s that was processing at the same feature size. Performance of your DIY 6502 would probably be lower as well -- requiring more power and producing more waste heat to run at a lower clock rate than that of an original.
So would it work? Possibly. Work as well? Less likely.
> An antique is usually an item that is collected or desirable because of its age, beauty, rarity, condition, utility, personal emotional connection, and/or other unique features. It is an object that represents a previous era or time period in human history.
The camera in the video doesn't meet any of these descriptions. Metal construction 4x5 monorails are ten a penny and are still manufactured today. This model probably dates from the 50s-60s.
(Ignore all the marketing copy about digital. Anyone dropping this amount of money on a monorail probably would be using it with a high-end digital back, but that's not included in the price, and the camera above could perfectly well be used with a film back too.)
I sorta agree. For me anything made after WWII isn't really antique. "Vintage" maybe. That Graphic View II he used was manufactured until 1967 and likely used professionally for many decades after.
The effortless way he is able to describe complex processes and techniques used in his experiments borders on being Feynman like.
[1] https://benkrasnow.blogspot.com/