Hack Chat Transcript, Part 1

A event log for Homemade Integrated Circuits Hack Chat

What's in your garage?

lutetiumLutetium 08/14/2019 at 20:040 Comments

OK, let's get started. I'm really excited to have Sam Zeloof here with us today. We've been covering Sam's homebrew ICs for a while now, and it's great to have him for a realtime chat.

Welcome, Sam! Can you maybe kick things off with a little about yourself?

Boian Mitov12:00 PM
Hello everyone :-)

David大卫 Lanzendörfer12:00 PM
We're using a laser writer for making the masks

David大卫 Lanzendörfer12:01 PM
The funny thing: The guy making the masks has also the English name Sam, but he's Chinese :D

Sam Zeloof12:01 PM
Sure! My name is Sam, i'm a second year student in college now and a little over 2 years ago I started acquiring and building the equipment needed to make ICs at home

philip.lafountain joined  the room.12:01 PM

Cyrus Tabrizi joined  the room.12:01 PM

de∫hipu12:02 PM
"Lanzendörfer" sounds like a very traditional Chinese surname ;-)

Sam Zeloof12:02 PM
Last april I succeeded in making a diff amp chip, PMOS, with ~150um gate size and I've been improving things since then to now fab <5um ICs


The Fab Lab Next Door: DIY Semiconductors

You think you've got it going on because you can wire up some eBay modules and make some LEDs blink, or because you designed your own PCB, or maybe even because you're an RF wizard. Then you see that someone is fabricating semiconductors at home, and you realize there's always another mountain to climb.

Read this on Hackaday


David大卫 Lanzendörfer12:03 PM
I'm not Chinese, I just was looking for a job opportunity to build cool hardware ;-)

David大卫 Lanzendörfer12:03 PM
I'm originally from Switzerland

adam joined  the room.12:03 PM

Adam Zeloof12:04 PM
Hi Sam

Sam Zeloof12:04 PM
hi adam

So how did your first year of college go?

haday12:05 PM
I'm not familiar with your project, but I'm guessing you are working with Si substrate. Where are you sourcing it?

thethoughtemporium12:06 PM
hi sam, biggest question where do you get your photoresist?

Sam Zeloof12:06 PM
Pretty well, lots of work overall but I had some great experiences and was lucky enough to start working on other stuff right away. Outside of classes, I am working in the nanofab there and have started my own lab in that department to develop new fab tools and processes. Not much info on that yet but hopefully soon I can talk more about it, make a website, etc

Sam Zeloof12:07 PM
Yup, Si substrate. you can get small quantities on ebay or here

David大卫 Lanzendörfer12:08 PM
Sam: Our lambda is 500nm at the moment

David大卫 Lanzendörfer12:08 PM
January we'll get access to an actual fab with a lambda of 250nm!!

ammater77 joined  the room.12:09 PM

Sam Zeloof12:09 PM
thethoughtemporium ! I saw one of your recent youtube vids and was going to reach out and offer you some photoresist, but couldn't find a contact button on your YT page. In the beginning I used PCB resist from Aliexpress (i think) but if you live in the US most companies will give out samples of their resists. Other than that, make friends in the industry...

David大卫 Lanzendörfer12:09 PM
The entire PDK and process flow will be published online

itsallgoodjames12:09 PM
am i able to purchase one of your homemade ic's

haday12:09 PM
I'm guessing you are working with already doped wafers?

David大卫 Lanzendörfer12:10 PM
Yeah, at the moment 4 inch, pdoped, single side polished, prime wafers, but the process is designed

Sam Zeloof12:10 PM
haday, yup. Commonly for NMOS circuits you start with a p-type bulk doped wafer, and the opposite is true for PMOS. For CMOS, you can start with either but commonly p doped are used, then to make P channel devices you implant a N-type well

David大卫 Lanzendörfer12:10 PM
in a way, that you just have to recalculate the dose for the nwell and pwell

David大卫 Lanzendörfer12:11 PM
Exactly, Sam!

thethoughtemporium12:11 PM
Ya pcb resist is just about all I can find. getting su-8 or similair seems to be a monumental task which seems silly considering. Was worried the answer was gonna be "make friends in the industry". Which evidently it is.

G-rout joined  the room.12:11 PM

David大卫 Lanzendörfer12:11 PM
I''l update the document in the next few days:

Stefan Biereigel joined  the room.12:11 PM

Sam Zeloof12:11 PM
you can order small quantities of SU-8 through any school or business

David大卫 Lanzendörfer12:12 PM
Please join next Sunday, 9pm HKT on our Mumble sessions!

David大卫 Lanzendörfer12:12 PM

Yann Guidon / YGDES12:12 PM will be interested :-)

haday12:13 PM
You mentioned you are completing your toolchain for 2 years now. How many tools were required to complete a fist prototype? Can you give us a gloss-over of the procedure?

Sam Zeloof12:13 PM
FETs are easier, in my opinion, to make. You can make them with only 1 doping step, BJT you need 2 and the base junction depth needs to be controlled very precisely

Scott Swaaley12:13 PM
Hey hackadoodles - new project log and video. We did an inertial test and weren't decapitated. I call that a win.

David大卫 Lanzendörfer12:14 PM
We've got tripple well

David大卫 Lanzendörfer12:14 PM

thethoughtemporium12:14 PM
i watched your lab tour video. quite an impressive setup. Was actually thinking of picking up a DLP projector because of your demos of using it

David大卫 Lanzendörfer12:14 PM
and SONOS flash :D

David大卫 Lanzendörfer12:14 PM

Sam Zeloof12:15 PM
thanks, yeah DLP projects work quite nicely for maskless litho. I had great results with a newer 1080p projector and stock Hg lamp

David大卫 Lanzendörfer12:15 PM
wanna have a talk on mumble?

thethoughtemporium12:16 PM
what's the smallest feature size you've managed with the dlp?

itsallgoodjames12:16 PM
will any cheap 1080p projector from amazon work?

David大卫 Lanzendörfer12:16 PM
I'd really love to have another project engineer helping me ^^

Sam Zeloof12:16 PM
Ive made nice 1um lines and spaces, which was really pushing it. But I can't align that far down, XY stage isnt nice enough. Typically you want <20% alignment error, which is 200nm for 1um features.

Sam Zeloof12:17 PM
2 - 4um is "comfortable" on my dlp setup, but of course you can reduce the field of view and just 'get' more resolution, to a point...

David大卫 Lanzendörfer12:17 PM
our node is made to scale

Sarah Vitak joined  the room.12:17 PM

Sam Zeloof12:18 PM
itsallgoodjames any dlp will probably work, beyond that YMMV. I had some results with LCOS

thethoughtemporium12:18 PM
that's not bad at all considering. what resist were you using for that? also what's the smallest feature you've managed with just the cheap pcb resist, just for sake of comparison

logxen12:18 PM
I assume you are using some more optics after the projector to do that?

David大卫 Lanzendörfer12:18 PM
We exchanged LOCOS for STI ;-)

David大卫 Lanzendörfer12:18 PM
Please join our mumble session this Sunday

@Sam Zeloof - Last I saw, the Z1 hadn't been wire-bonded. Were you able to get the gear needed to finish it?

Sam Zeloof12:20 PM
@thethoughtemporium beyond 50um with PCB resist was tricky and not repeatable, but I eventually made a 10um grating. Any of the normal industry resists like SU-8 have great performance on the dlp setup. Just spin a thinner layer on to make smaller features, control the UV dose, etc.

Sam Zeloof12:22 PM
@logxen the optics can be as simple as turning the projection lens inside the projector backwards, so it makes the image smaller instead of larger... or you can make things more complicated and start replacing the optics with quartz to pass deeper UV. But the plastic / boro glass lenses in most projectors pass <400nm OK for litho. I got 10mW / cm^2 at 365nm UV which leads to ~20 sec exposure, usual resist dose is on the order of 100mJ

David大卫 Lanzendörfer12:22 PM
I should maybe mention, that I've got access to Shenzhen equipment and manufacturers, and if anyone actually is willing to improve the resolution of your exposure equipment, instead of wining about it, we could work together and make a better, open source one

David大卫 Lanzendörfer12:22 PM
Just saying

Sam Zeloof12:23 PM
yes! I finally got wire bonded chips this summer. Wire bonding is an art in itself, but the resulting chips look so nice

Stefan Biereigel12:24 PM
@Sam Zeloof Do you already have plans (in terms of target feature size and implemented circuitry) for the 'Z2'?

Sam Zeloof12:24 PM
The wire bonder was graciously donated, but you can find them on ebay as well. getting the top metal layer (aluminum) to "stick" on the chip was a huge challenge, too

itsallgoodjames12:24 PM
@Sam Zeloof approximately how much have you spent on your whole setup?

morgan12:26 PM
@Sam Zeloof how are bond wires 'stuck' on?

haday12:26 PM
ultrasonic welding?

David大卫 Lanzendörfer12:26 PM
@morgan there's at least 10 different ways

Sam Zeloof12:26 PM
@Stefan Biereigel as far as circuitry, I was never interested much in making specific chips, but rather the processes. Although I do have some ideas for FPAA - like structures. As far as the feature size, the practical limit for chips at home (until someone makes a safe ion implanter) is 1 to 2um, which I am close to anyway. I would focus on changing the process, probably using polysilicon gates on the Z2 instead of aluminum

thethoughtemporium12:26 PM
hmm, i mean 50um is still not bad. not super great for chips, but would be fine for doing traces for some simpler projects. certianly enough for a basic microelectrode array

Sam Zeloof12:27 PM
@morgan many ways, my bonder used heat, pressure, and ultrasonic

David大卫 Lanzendörfer12:28 PM
@Sam Zeloof Dude. Lets build a better exposure unit. Are you up for it, or do you wanna hang around on pathetic 50um??

morgan12:28 PM
cool, a video I missed (and I reminder I hate the youtube suggestion system)

Sam Zeloof12:29 PM
@thethoughtemporium for sure, but the good news with chips is that you can make them as big as you want. sure, it may be slower because of increased capacitance, but FET characteristics depend on the gate width to length ratio. so a linear scale-up still gives you good devices, within reason

David大卫 Lanzendörfer12:29 PM
@Sam Zeloof Building shit like that is the reason I moved into a maybe-tobe war zone

thethoughtemporium12:30 PM
i looked at ion implantation a while back. not hard to build, it's just a linac and you can power it with a van de graff, but a nasty bugger to be around.

thethoughtemporium12:31 PM
needs just as much shielding as a fusor

thethoughtemporium12:31 PM
though no neutron shielding so it's technically easier. just some big sheets of lead

Frank Buss12:32 PM
dude, van de graff is so old-school, I'm sure David can give you a shiny Shenzhen high voltage transformer :-)

Sam Zeloof12:33 PM
Yes for sure its doable, you only need ~30kv velocity to reasonably dope Si.

Sam Zeloof12:33 PM
But the problem is in the dopant source, most of them are very very hazardous, phosphine, diborane, arsine

thethoughtemporium12:34 PM
ya those... aren't pleasant to work with.

Sam Zeloof12:34 PM
I thought about it for a while but never tried making one, but I would probably take the approach of "plasma immersion ion implantation" which has no ion mass selection (no linac, just a big chamber)

Frank Buss12:34 PM
but isn't it just a few atoms? like 1:100 million

Sam Zeloof12:35 PM
and would experiment putting small solid sources of dopant material (phosphorus, boron oxide) in the chamber and heating them to produce small and controllable amounts of dopant gas. Of course, the exhaust from the vacuum pump would be out the nearest window...

thethoughtemporium12:35 PM
i've been playing around with some designs for direct ion writing. wanted to mod the magnetron to focus the plume to a beam and then move it around with some deflection lenses. no idea what sort of resolution you can get out of something like that though

thethoughtemporium12:36 PM
who needs a mask when you can just print your traces directly! least that's the dream. probably won't get to try it for a while though

moh.mohx100012:37 PM
See Applied science video on electron microscope for that.

haday12:37 PM
focusing the beam might me the hardest

Sam Zeloof12:37 PM
I have a green book somewhere on the theory of ion implantation, it mentions direct write doping with a technique like that. Ill dig up the title and send it later, its certainly possible but was abandoned in the industry for speed and resolution problems. Focusing a beam of positively charged things is so much harder than negatively, like electrons i think.

Sam Zeloof12:37 PM
its a good idea, and would be so cool if you could mess with it.

moh.mohx100012:38 PM
you also needs a lot of em shielding.

thethoughtemporium12:38 PM
i figured as much. mostly just want to do it for the same reasons i do anything. seemed fun to mess around with XD

Frank Buss12:38 PM
@Chris Gammell talks often about printing ICs on the amp hour, maybe he has some ideas how to do it?

Sam Zeloof12:39 PM
for sure. I have some random plasma related experiments here, that may be of interest. Looking forward to your direct chip writer...

thethoughtemporium12:40 PM
what are you using as your rf source for that rf plasma?

Sam Zeloof12:41 PM
10M ham radio or CB radio, 25W and sometimes with 200W LDMOS amplifier

Sam Zeloof12:41 PM
anything will work, 1khz to 100mhz easily, as long as you have a matching network and more than a few watts

Hah, cool. Is that why you got your ham license? Or did that happen first?

moh.mohx100012:42 PM
Can I use Tesla coil for that?

thethoughtemporium12:43 PM
hmm interesting. i just used a desktop tesla coil as my rf source XD kinda worked and ingnited plasma easily but didn't really strike me as enough power to really do anything. i'll look into a ham transmitter

Sam Zeloof12:43 PM
Actually had the ham license first, but this was a nice use for it!

moh.mohx100012:43 PM
because the coupling between the two coils is weak.