I just cannot believe humanity is arrived to the checkpoint where fabricating an IC at home (in a not 99.9999...% clean room) is documented on UA-cam. I hope you share every details about this in the near future, and we can learn everything. I am so proud of you, most precious channel, seriously.
With no laminar, hepa filtered flow of air, any photolithografic process would be severely compromised. It's alot like Iron Man building sophisticated technologies while held prison in a cave.
Charles Lavender There are other techniques for printing or “drawing” the circuitry onto the die. Of course photolithography is in common practice because it provides the best economy for mass production, but things get a little switched around when prototyping is the intended application.
@andy low - I do not think, that "progress" is the goal here. These kind of ICs are from like 50 years ago. Until now it was: How can I put more transistors on a chip than Intel. Here it is now: How can I put less transistors on a chip than Intel.
I think it's crazier that I'm probably going to end up learning bonding at home rather than at work (where we coincidentally have that same bonder lol)
Damn, I've been waiting for this video ever since Hackaday dropped that article about your homemade IC. Great to see you got the wire bonder and managed to get it running!
Considering UA-cam's AI algorithms have been demonetizing robot fighting videos for animal cruelty, I wouldn't put too much stock into them. Searching for STEM related stuff on Google, trying to get the results I want is like trying to reason with a brain damaged child half the time. Lately I'm finding a great many suggested videos on my UA-cam home page are ones I've already watched. I usually have to flag half a dozen or so. And it's not like it doesn't know I've already watched them. It's got the red progress bar at the bottom of the thumbnail showing I've already watched it. Does UA-cam really think an average user is going to watch a video more than once when people gripe about a video being more than a few minutes long?
You can also have a slightly larger bondpad (not too large to limit the stray capacitance), and use multiple gold wires on the bondpad to reduce the effective series inductance.
RF is wizardry compared to the quantum mechanics of silicon. To what extent do the wire lengths and geometries here matter like they do in traces for chip antennas?
I just found your channel today due to the thought imperium and I wish I'd know about you earlier, but such is how youtube works today... You've made a very beautiful IC there!!! Modern IC's are boring to look at whereas the older ones from the 70's are works of art, but due to cost, power consumption, speed and reliability I'm willing to accept that they're just pieces of plastic with legs sticking out of them, but I have a deeply rooted fondness of old school components! Now I'm going to binge watch all of your videos - which is the only positive thing about discovering a channel that's been operating for years without getting the recognition it deserves... I'm very impressed with the setup you've got and I hope I'll learn as I'm watching your videos how & why you've acquired all of it, but like I said in the beginning I only wished I'd known about you a long time ago - and I'm even actively searching for content like this?! Good information is sadly hidden amongst a vast majority of pure waste of storage and bandwidth today, but it is there and I appreciate those who link to interesting videos and thus providing access to information youtube doesn't care to show us via their "algorithms"... It's only promoting videos that has a potential to get the most views, thus bringing in that sweet ad revenue money. I'm not going to watch home improvements, makeup tips and whatever stupidities people spend their time to watch instead of actually learning a specific scientific subject! The world is slowly turning into what the movie Idiocracy depicts... Funny and sad at the same time... Thanks for making your videos! Just the stuff my brain needs - especially being a retired engineer who's bored and isolated.
You're a God, Sam! I have to rebond the gold wires from my thermal camera microbolometer array to the chip under it. I have fabricated micro soldering heads and tips before but never tackled anything a 10th this small. I have so many questions and found this video. And of course, it's Sam Zeloof. So now I will have to brush up on some of your other vids to get my head around this task. Thank god only three of the wires are broken. I am not even sure if I can fit the assembly under my microscope. I am so much happier since you reminded me gold bonds to gold with compression and oxidation is not an issue.
Brings back such memories. I used a similar ball bonder for one off thick film circuits in the 80's. It definitely was a challenge to get things right when you only had a few ( if any ) devices to fail on. We had to import bare die from the US and it was expensive so it wasn't a case of give it a go and try again until you get it right. Impressive to see this stuff being used in home shops
This is so cool to see. I actually have been working on Wedge bonders for the last 25 years. The company I work for is now a subdivision of K&S. We make Large wedge bonders for aluminum and copper ribbon for high power applications. It is a very similar process, except that the wires are between 125 an 600 microns in diameter. This is used in high power mosfets and IGBT's, both leadframe and also hybird inverter modules. Thank you for sharing your journey into the semiconductor world with the rest of us :)
Since you mentioned copper pads: if you really plan on using it on the chip itself - be very careful - copper tends to migrate into the silicon and destroy the transistors. So it always needs an isolation layer that it cannot migrate through. Also: use some deposition machine for copper exclusively and make sure it is physically separated from all other machines, if possible by a wall and/or airflow. Even the slightest copper contamination can destroy your logic layers! That includes the almost undetectable copper vapor floating in the air around a Cu PVD site or the Cu traces deposited on a carrier or workbench.
@@teresashinkansen9402 Yes, copper has been a problem since the beginning. Even very shallow surface defects can ruin your semiconductors. AFAIK, AMD was only able to develop the first copper based process in the mid 90'ies when they reached about 130nm. Before that it was just "shut up and use more aluminium".
Outstanding work and adopted hobby. I come from the days when most all the chips were ceramic and gold. When they started to use plastic as the encapsulation material, there was a lot of problems with broken bonding-wires which was attributed to different expansion coefficients of the materials. So, if you ran a few extreme heat-cycles, it was not uncommon to have a disconnection between one of the outside pins and the die inside. So, for many years the military tenaciously held on to total ceramic-packages. Just ran across your channel from Huygens and SUBSCRIBED! I'm convinced that you could make some really high performance (and useful) IC's even if they aren't technically complicated or dense--matched transistor arrays come to mind. Seriously--Hat's off to you!
I love your channel. I have watched most of your videos over the years and wish you would upload more often. Without regular uploads it is easy to forget great channels such as yours exist.
I came from your recent lithography video, curious about the wire bonding process because I work for a company that makes those machines.... lo and behold you are using a wire bonder my company makes (or made) what a world
Well thank you, Sir, very much for making this video. It has been made with professionality. It helped me a lot in understanding the basics of this technology!
In the broadband RF/microwave world, manual bonding is still very much the way things are done. Bare die are used rather than packaged parts to reduce the effects of parasitic inductance and capacitance. It also helps with miniaturisation, although this is generally a bit less important. The geometry of some RF & microwave components, such as mmWave GUNN diodes are tiny by even a regular bonder’s standards and so I’ve seen some much larger machines floating on an air bearing for this purpose!
If i carry on doing the crazy stuff that i do on in my work shop for another 100 years , I might become as crazy as you. I didnt even know what wire bonding is , nor did I know what a hp vacuum is either . Thanks alot bro ,👍
Given the size of this relatively simple die, I don't think that would actually be possible, unfortunately. It would have to be made much more compact, and would also be very complicated!
Great to see you posting again man, your work is increadible. Thank you so much for sharing! I might just have missed it, but what IC is that? The last time I saw you messing around with fabricating MOSFETs and now suddenly there's a whole IC soldered to a carrier! I also have a few questions about this process: Since most ICs we use specify maximum storage temperatures of around 85°C (and exhibit quite severe, almost 1ppm, parameter variation after soldering), how does keeping the substrate at 150°C for so long affect the chip's long term reliability and it's engineered tolerances? And finally, I understand why Copper needs a diffusion barrier to adhere to, but why does oxidation of the copper pads interfere with the wirebonding process, but Aluminium (which obviously instantly forms a tenacious oxide layer) does not?
thanks for the kind words. I guess ive been neglecting youtube recently and mostly just posting on twitter and my blog. Yes, I have fully working ICs now! There will be a video soon. sam.zeloof.xyz/first-ic/. As far as the bonding, good questions. When the bonder is dialed in, it should take no more than a couple of minutes to bond the chip. So the die will only experience elevated temp. for a short amount of time. Idk how this effects the long term reliability of the chip, but I know the required substrate temp can be lowered substantially into the 100C range with more optimized bond parameters. This typically means using a higher ultrasonic power and thicker bond pads, which they have no problem doing because they have electroless plating solutions which I do not. The copper oxidation thing confused me as well to be honest, but I had serious trouble bonding to Cu pads until I saw this paper which said it needs a barrier to stop oxidation and it quickly stopped the problems link.springer.com/article/10.1007/s11664-004-0135-5 but I never had such problems with Al pads. Not sure why, they both oxidize at roughly the same rate to the same thicknesses in air (~5nm)
Total amateur hobbyist that knows nothing here... I think I recall reading a lot of the temperature issues post production are related to the chip packaging. Please correct me if I'm wrong. Again I'm no expert but I imagine the die is more robust on it's own and within a controlled atmosphere. How will it degrade physically/chemically under these conditions?
@@SamZeloof Thanks for the detailed response man, I always love reading the original sources for myself as well. Also, thank you for sharing where you post regular updates now, I never knew you had a website, but I'll be sure to check it out now!
@@UpcycleElectronics So, basically, at elevated temperatures (usually > 125C when no power is applied). While the die itself is indeed pretty robust, the post-implantation annealing (if using ion implantation) is one low-temperature process that comes to mind as having a significant effect on the chip's performance. Take a standard planar BJT for example; Even when only using diffusion doping to form the junctions, you will usually end up with a very thin thin base region that contains *BOTH* n-type and p-type dopants. For an NPN transistor, the base region just contains more p-type dopant atoms than n-type dopant atoms, however, the base region is also flanked by 2 n-type regions. Therefore, keeping the die at elevated temperatures could seriously impact the precise base width and the net dopant majority within it. As for the wire bonding itself, some metals (especially Copper and to an extent also Gold) diffuse into Silicon extremely rapidly, thereby ruining your intended Ohmic contact and turning it into a crude diode with random characteristics. One final thing of note: The longevity of silicon devices (which contain small/thin metallization layers) gets *EXPONENTIALLY* shorter with higher operating temperatures due to a combination of factors, with electromigration (which causes the wires themselves to fail, although not usually the relatively thick package-level bond wires) and wirebond delamination (where the intricate connection between the bonded wire, its matching adhesion layers which form the pad it's welded to, and the doped substrate) usually singled out as the chief causes.
I miss micro so much lol. I did manual wire binding (ribbon/wire, wedge/ball) manual component attach, manual MMIC tuning, rework and manual substrate attach. Was in the defense manufacturing field for almost ten years:/
I used to have a Baush and Lomb binocular microscope and it was one of my most treasured possessions... I think it's funny that you get one just as part of some other single purpose machine. ;)
How in the world have I not seen this yet? You did a great job of documenting the process of the internals… But how did you go about making the chip it’s self? Or are they old reused chips that you just cleaned up exceptionally well and re-goldplated while replacing the insides? Amazing work!
Apologies, I think with my third time watching the video I got my questions answered. You referred to “chip carriers“ at one point in the video and I was mistaken in thinking those were the little glass dishes you had all the chips in. But I’m assuming you’re referring to the actual package/chip themselves…? I know the chip is the little wafer that you’re bonding wires to… I guess it’s just a habit for me to call the thing with legs the “chip.”
One part of me is going "Why am I watching this when I know I'm never going to do this?" while the other part of me is like "Amazing" and I just can't stop watching :D
THIS! she made a few-transistor ics in her little kitchen-inspired lab & is maybe looking for a way to bond, package & test those things & do a little further experiments.
Hi Sam, you mentioned at the end of this video that in the industry they wouldnt use such thick bonds etc you mentioned electroless plating with nickelless and soft gold, could you please mention in an upcoming video or posting here some good reeds related to semiconductors and how industry actually does stuff like I can find papers and books about how most of this stuff is supposed to be done in theory but very little (close to nothing) about industry standards and actual practices (well besides the occasional very general "how CPUs are made" youtube documentary) Thanks.
Eu aqui, sonhando com um computador simples pra jogar paciência e o cara tem uma máquina de passar fios em microprocessador, na garagem de casa!!! Meeee.
At the company I worked at, a plasma cleaner was running for a good chunk of each day removing contaminants from surfaces before die attach and bonding. Is this something you have and use?
i wonder if you can stack these chips with wire bonding then bending the wires up so that its above and use a resin to create a multi layer chip for more density.
...got some expensive rare earth elements and low expectations lying around? ;-) LEDs need weird materials, but the process is comparatively simple. Once you made multiple transistors on a single chip I'd imagine LED does not sound like a worthwhile challenge unless you are a materials guy with a fetish for tuning band gaps.
I love seeing the same chip carriers as we use at work for prototyping ASICs and new chips :) I just started as an intern, so I don't know much about the manufacturing steps of ICs yet, so thanks for showing how boning (used to) work ;)
I am here for my own 101 in wire bonding. First, excellent video! Second, I am truly impressed by Sam's bottom line. I just wonder what his hobby projects are :) Else, I wanted to ask him (or anyone here & in the know) if wire bonding works on thick Cu substrates, 500um strength wise or more. Is it enough to 24kAu-plate the substrate? Known issues or advantages? Comments?
Hey, I have been trying to make a gold bond from Cu pad to Au pad using the machine 53xx Deep Ball Access by Delvotec. I scratch the Cu pad with sand paper and clean it with isopropyl but yet the first bond doesn't stick with wedge bonding settings. I have tried ball bonding and there it only stick the first bond but not the second. I would appreciate any suggestions you have for me regarding any settings. Thanks for the video as well!
Nice job. I got my start in the microelectronics business as a field service technician for Kulick & Soffa about 30 years ago and threading the wire never gets easy.
Not the OP of course, but the one I use at work is at 63kHz, which is considered a "low frequency" machine. You have power (amplitude), downforce, and heat (of both the tool and the substrate) to play around with to get a good bond. The power isn't much, but when you're working with wire thinner than a hair you don't need much power to make a bond on clean substrate. You can find datasheets online manual and semi-automatic bonders, there's a lot to learn there.
2 роки тому+1
Given that those wires always go between fixed positions makes me wonder why there is not a more efficient process possible, like 'stamping' them on all together in one step.
In short: Tooling time and alternative bonding techniques. Long: every factory produces at least dozens of different chips, possibly hundreds. If you have to change the stamp at every change of product or buy a bonding machine for every product this will become very impractical very quickly. So you buy a machine from some supplier that can make it very fast. For chips with more than about 64 pins you use flip chip bonding - the package and its lead frame is prepared so that the connections of the pins already reach the right point under the chip. The chip has solder balls on top instead of simple flat pads. Then you flip the chip, put it on the lead frame contacts and apply heat. Done. The most aggressive version of this is BGA where you skip the package and bond the chip directly to the PCB.
the one we have in the lab doesn't get the hole for threading in the camera or the microscope, we have to do it basically completely blind which sometimes takes hours!
so basically this is like a manual 3d printer except the nozzle is supersonic. ive seen people use some ultrasonic "welder" or "cutter" these last months, cutting anything like butter. i have also seen some 5+ year old toshiba video showcasing cnc smd pulse laser soldering, the thing is so precise, makes perfect melts everytime, i almost cried. i can imagine some makers build a $150 wire bonding machine, seems like the gold wire is the most expensive thing here.
you know the thing to do is create an gue that kids can put on a piece of paper. each dot of gue can be an n type or p type to create a diode or a npn dots to create a transistor.
I know how difficult it is to thread the needle! In my institute, a lot of potential users failed training simply because of inability to thread the needle.
I just cannot believe humanity is arrived to the checkpoint where fabricating an IC at home (in a not 99.9999...% clean room) is documented on UA-cam. I hope you share every details about this in the near future, and we can learn everything. I am so proud of you, most precious channel, seriously.
With no laminar, hepa filtered flow of air, any photolithografic process would be severely compromised. It's alot like Iron Man building sophisticated technologies while held prison in a cave.
Charles Lavender There are other techniques for printing or “drawing” the circuitry onto the die. Of course photolithography is in common practice because it provides the best economy for mass production, but things get a little switched around when prototyping is the intended application.
you will need a wire bonding machine to fabricate the IC to play the youtube video which explains how to wire bond ICs.
@andy low - I do not think, that "progress" is the goal here. These kind of ICs are from like 50 years ago.
Until now it was: How can I put more transistors on a chip than Intel.
Here it is now: How can I put less transistors on a chip than Intel.
I think it's crazier that I'm probably going to end up learning bonding at home rather than at work (where we coincidentally have that same bonder lol)
Damn, I've been waiting for this video ever since Hackaday dropped that article about your homemade IC. Great to see you got the wire bonder and managed to get it running!
this channel is yet another indication of how bad UA-cam suggestion algorithms are. were it not for the thought emporium, i'd never find this channel.
@nicolas fritzges liddell me to
Yeah I got it suggested to me.. I almost didn't click on it, didnt really have a clue what it would be about
Alexander Sannikov UA-cam algorithm works for money, not for you
UA-cam algorithm doesn't care about the usefulness of the video
Considering UA-cam's AI algorithms have been demonetizing robot fighting videos for animal cruelty, I wouldn't put too much stock into them. Searching for STEM related stuff on Google, trying to get the results I want is like trying to reason with a brain damaged child half the time. Lately I'm finding a great many suggested videos on my UA-cam home page are ones I've already watched. I usually have to flag half a dozen or so. And it's not like it doesn't know I've already watched them. It's got the red progress bar at the bottom of the thumbnail showing I've already watched it. Does UA-cam really think an average user is going to watch a video more than once when people gripe about a video being more than a few minutes long?
The flat ribbon is also useful for RF applications, as it reduces the inductance of the connection from the die to the leadframe.
You need orthogonal bonds to get the L down.
You can also have a slightly larger bondpad (not too large to limit the stray capacitance), and use multiple gold wires on the bondpad to reduce the effective series inductance.
O so yoooou’re the smart one!
RF is wizardry compared to the quantum mechanics of silicon.
To what extent do the wire lengths and geometries here matter like they do in traces for chip antennas?
i think the ribbon is meant to reduce the skin effect rather than the inductance
I just found your channel today due to the thought imperium and I wish I'd know about you earlier, but such is how youtube works today...
You've made a very beautiful IC there!!! Modern IC's are boring to look at whereas the older ones from the 70's are works of art, but due to cost, power consumption, speed and reliability I'm willing to accept that they're just pieces of plastic with legs sticking out of them, but I have a deeply rooted fondness of old school components!
Now I'm going to binge watch all of your videos - which is the only positive thing about discovering a channel that's been operating for years without getting the recognition it deserves...
I'm very impressed with the setup you've got and I hope I'll learn as I'm watching your videos how & why you've acquired all of it, but like I said in the beginning I only wished I'd known about you a long time ago - and I'm even actively searching for content like this?!
Good information is sadly hidden amongst a vast majority of pure waste of storage and bandwidth today, but it is there and I appreciate those who link to interesting videos and thus providing access to information youtube doesn't care to show us via their "algorithms"... It's only promoting videos that has a potential to get the most views, thus bringing in that sweet ad revenue money.
I'm not going to watch home improvements, makeup tips and whatever stupidities people spend their time to watch instead of actually learning a specific scientific subject!
The world is slowly turning into what the movie Idiocracy depicts... Funny and sad at the same time...
Thanks for making your videos! Just the stuff my brain needs - especially being a retired engineer who's bored and isolated.
You're a God, Sam! I have to rebond the gold wires from my thermal camera microbolometer array to the chip under it. I have fabricated micro soldering heads and tips before but never tackled anything a 10th this small. I have so many questions and found this video. And of course, it's Sam Zeloof. So now I will have to brush up on some of your other vids to get my head around this task. Thank god only three of the wires are broken. I am not even sure if I can fit the assembly under my microscope. I am so much happier since you reminded me gold bonds to gold with compression and oxidation is not an issue.
Brings back such memories. I used a similar ball bonder for one off thick film circuits in the 80's. It definitely was a challenge to get things right when you only had a few ( if any ) devices to fail on. We had to import bare die from the US and it was expensive so it wasn't a case of give it a go and try again until you get it right. Impressive to see this stuff being used in home shops
i, too, dabble in ball bondage
Thanks for the upload. This is the first time I've ever seen this done.
That's amazing. You actually make a chip. Respect to you, man.
Right? What a freaking legend!
One of the most incredible youtube channels in the world! Congratulations!
This is so cool to see. I actually have been working on Wedge bonders for the last 25 years. The company I work for is now a subdivision of K&S. We make Large wedge bonders for aluminum and copper ribbon for high power applications. It is a very similar process, except that the wires are between 125 an 600 microns in diameter. This is used in high power mosfets and IGBT's, both leadframe and also hybird inverter modules. Thank you for sharing your journey into the semiconductor world with the rest of us :)
Since you mentioned copper pads: if you really plan on using it on the chip itself - be very careful - copper tends to migrate into the silicon and destroy the transistors. So it always needs an isolation layer that it cannot migrate through. Also: use some deposition machine for copper exclusively and make sure it is physically separated from all other machines, if possible by a wall and/or airflow. Even the slightest copper contamination can destroy your logic layers! That includes the almost undetectable copper vapor floating in the air around a Cu PVD site or the Cu traces deposited on a carrier or workbench.
I keep coughing up green stuff, must be the copper
I would think it is at the nano scale process nodes but here I don't know. Is that really an issue at this scales?
@@teresashinkansen9402 Yes, copper has been a problem since the beginning. Even very shallow surface defects can ruin your semiconductors. AFAIK, AMD was only able to develop the first copper based process in the mid 90'ies when they reached about 130nm. Before that it was just "shut up and use more aluminium".
🙄
I saw you started a fab company, programmers like myself are so excited for the arch and such you come up with!
Such soothing listening to him talk about the process
Outstanding work and adopted hobby. I come from the days when most all the chips were ceramic and gold. When they started to use plastic as the encapsulation material, there was a lot of problems with broken bonding-wires which was attributed to different expansion coefficients of the materials. So, if you ran a few extreme heat-cycles, it was not uncommon to have a disconnection between one of the outside pins and the die inside. So, for many years the military tenaciously held on to total ceramic-packages. Just ran across your channel from Huygens and SUBSCRIBED! I'm convinced that you could make some really high performance (and useful) IC's even if they aren't technically complicated or dense--matched transistor arrays come to mind. Seriously--Hat's off to you!
amazing work!! didn't expect youtube DIY has grown to complete such complex projects!!
I love your channel. I have watched most of your videos over the years and wish you would upload more often. Without regular uploads it is easy to forget great channels such as yours exist.
This is so cool! Watching your videos always makes me excited about building things. Thank you!
I’m about to start BSc in Industrial Electronics and automation engineering. I love your videos from Spain. Wish me luck guys :)
Love your channel, please keep bringing the secrets of industry to the masses.
thanks for taking the time in making these videos ,have a great day
I came from your recent lithography video, curious about the wire bonding process because I work for a company that makes those machines.... lo and behold you are using a wire bonder my company makes (or made)
what a world
Well thank you, Sir, very much for making this video. It has been made with professionality. It helped me a lot in understanding the basics of this technology!
4:57 If you dont have the chip holders then you can glue the chip to a board and wire bond it like a COB (Chip on board)
In the broadband RF/microwave world, manual bonding is still very much the way things are done.
Bare die are used rather than packaged parts to reduce the effects of parasitic inductance and capacitance. It also helps with miniaturisation, although this is generally a bit less important.
The geometry of some RF & microwave components, such as mmWave GUNN diodes are tiny by even a regular bonder’s standards and so I’ve seen some much larger machines floating on an air bearing for this purpose!
eBay is one of man’s greatest inventions.
Fantastic work man.
What an interesting series of videos. I have been curious about this stuff for years, and to now see it demystified is very satisfying. Thank you.
I am beyond impressed.. Its crazy that he did all that by himself! So skilled!
If i carry on doing the crazy stuff that i do on in my work shop for another 100 years , I might become as crazy as you.
I didnt even know what wire bonding is , nor did I know what a hp vacuum is either .
Thanks alot bro ,👍
I'm happy just getting my nets connected neatly and reliably on strip-board.
Wow.
It would be neat if you made a chip of Ben Eater's homebrew CPU!
Given the size of this relatively simple die, I don't think that would actually be possible, unfortunately. It would have to be made much more compact, and would also be very complicated!
@@CorvusPrudens Jup, the architerure "plan" of a CPU can have a size of multiple soccer fields...
an implementation of the gigatron cpu would be much more feasable. It has only about 300 transistors.
Great to see you posting again man, your work is increadible. Thank you so much for sharing!
I might just have missed it, but what IC is that? The last time I saw you messing around with fabricating MOSFETs and now suddenly there's a whole IC soldered to a carrier!
I also have a few questions about this process: Since most ICs we use specify maximum storage temperatures of around 85°C (and exhibit quite severe, almost 1ppm, parameter variation after soldering), how does keeping the substrate at 150°C for so long affect the chip's long term reliability and it's engineered tolerances?
And finally, I understand why Copper needs a diffusion barrier to adhere to, but why does oxidation of the copper pads interfere with the wirebonding process, but Aluminium (which obviously instantly forms a tenacious oxide layer) does not?
thanks for the kind words. I guess ive been neglecting youtube recently and mostly just posting on twitter and my blog. Yes, I have fully working ICs now! There will be a video soon. sam.zeloof.xyz/first-ic/. As far as the bonding, good questions. When the bonder is dialed in, it should take no more than a couple of minutes to bond the chip. So the die will only experience elevated temp. for a short amount of time. Idk how this effects the long term reliability of the chip, but I know the required substrate temp can be lowered substantially into the 100C range with more optimized bond parameters. This typically means using a higher ultrasonic power and thicker bond pads, which they have no problem doing because they have electroless plating solutions which I do not. The copper oxidation thing confused me as well to be honest, but I had serious trouble bonding to Cu pads until I saw this paper which said it needs a barrier to stop oxidation and it quickly stopped the problems link.springer.com/article/10.1007/s11664-004-0135-5 but I never had such problems with Al pads. Not sure why, they both oxidize at roughly the same rate to the same thicknesses in air (~5nm)
Total amateur hobbyist that knows nothing here... I think I recall reading a lot of the temperature issues post production are related to the chip packaging. Please correct me if I'm wrong. Again I'm no expert but I imagine the die is more robust on it's own and within a controlled atmosphere. How will it degrade physically/chemically under these conditions?
@@SamZeloof Thanks for the detailed response man, I always love reading the original sources for myself as well.
Also, thank you for sharing where you post regular updates now, I never knew you had a website, but I'll be sure to check it out now!
@@UpcycleElectronics So, basically, at elevated temperatures (usually > 125C when no power is applied). While the die itself is indeed pretty robust, the post-implantation annealing (if using ion implantation) is one low-temperature process that comes to mind as having a significant effect on the chip's performance.
Take a standard planar BJT for example; Even when only using diffusion doping to form the junctions, you will usually end up with a very thin thin base region that contains *BOTH* n-type and p-type dopants. For an NPN transistor, the base region just contains more p-type dopant atoms than n-type dopant atoms, however, the base region is also flanked by 2 n-type regions. Therefore, keeping the die at elevated temperatures could seriously impact the precise base width and the net dopant majority within it.
As for the wire bonding itself, some metals (especially Copper and to an extent also Gold) diffuse into Silicon extremely rapidly, thereby ruining your intended Ohmic contact and turning it into a crude diode with random characteristics.
One final thing of note: The longevity of silicon devices (which contain small/thin metallization layers) gets *EXPONENTIALLY* shorter with higher operating temperatures due to a combination of factors, with electromigration (which causes the wires themselves to fail, although not usually the relatively thick package-level bond wires) and wirebond delamination (where the intricate connection between the bonded wire, its matching adhesion layers which form the pad it's welded to, and the doped substrate) usually singled out as the chief causes.
Many Many Thanks for your great video . welcome to Egypt
10:48 Plasma cleaning and all your problems would be gone, extremely common practice in the vacuum industry.
You are just amazing.
Its fantastic how you do this allone.
Respect!
I miss micro so much lol. I did manual wire binding (ribbon/wire, wedge/ball) manual component attach, manual MMIC tuning, rework and manual substrate attach. Was in the defense manufacturing field for almost ten years:/
Amazing. Thank you for sharing. I was curious about how this was done. You helped me immensely.
I used to have a Baush and Lomb binocular microscope and it was one of my most treasured possessions... I think it's funny that you get one just as part of some other single purpose machine. ;)
How in the world have I not seen this yet? You did a great job of documenting the process of the internals… But how did you go about making the chip it’s self?
Or are they old reused chips that you just cleaned up exceptionally well and re-goldplated while replacing the insides? Amazing work!
Apologies, I think with my third time watching the video I got my questions answered. You referred to “chip carriers“ at one point in the video and I was mistaken in thinking those were the little glass dishes you had all the chips in. But I’m assuming you’re referring to the actual package/chip themselves…? I know the chip is the little wafer that you’re bonding wires to… I guess it’s just a habit for me to call the thing with legs the “chip.”
@@hullinstruments you can get the carriers on eBay
One part of me is going "Why am I watching this when I know I'm never going to do this?" while the other part of me is like "Amazing" and I just can't stop watching :D
did you tell Jeri Ellsworth?
she also made transistor on silica in her kitchen.
THIS! she made a few-transistor ics in her little kitchen-inspired lab & is maybe looking for a way to bond, package & test those things & do a little further experiments.
@@FrankHarwald try to find your self a girl than can actually make a transistor.
: )
This is amazing. Many, many, many props for taking this on, and for making videos about it.
Hi Sam, you mentioned at the end of this video that in the industry they wouldnt use such thick bonds etc you mentioned electroless plating with nickelless and soft gold, could you please mention in an upcoming video or posting here some good reeds related to semiconductors and how industry actually does stuff like I can find papers and books about how most of this stuff is supposed to be done in theory but very little (close to nothing) about industry standards and actual practices (well besides the occasional very general "how CPUs are made" youtube documentary)
Thanks.
The most epic DIY video I ever seen
Eu aqui, sonhando com um computador simples pra jogar paciência e o cara tem uma máquina de passar fios em microprocessador, na garagem de casa!!! Meeee.
Excelent channel friend… How does the wire adhere to the pad? just pressure?
Your content is amazing. Thank you for sharing your knowledge.
At the company I worked at, a plasma cleaner was running for a good chunk of each day removing contaminants from surfaces before die attach and bonding. Is this something you have and use?
i wonder if you can stack these chips with wire bonding then bending the wires up so that its above and use a resin to create a multi layer chip for more density.
Wow that was awesome. Thanks for sharing! Always wondered how they do that.
Oh i saw your post on twitter , forgot i was subscribed !
Cool stuff
Are there any plans to create a light emitting diode?
...got some expensive rare earth elements and low expectations lying around? ;-)
LEDs need weird materials, but the process is comparatively simple. Once you made multiple transistors on a single chip I'd imagine LED does not sound like a worthwhile challenge unless you are a materials guy with a fetish for tuning band gaps.
I love seeing the same chip carriers as we use at work for prototyping ASICs and new chips :) I just started as an intern, so I don't know much about the manufacturing steps of ICs yet, so thanks for showing how boning (used to) work ;)
Amazing stuff bud! I wish I had half or even 1/4 of your knowledge
Здорово! Спасибо за пример.
I am here for my own 101 in wire bonding. First, excellent video! Second, I am truly impressed by Sam's bottom line. I just wonder what his hobby projects are :)
Else, I wanted to ask him (or anyone here & in the know) if wire bonding works on thick Cu substrates, 500um strength wise or more. Is it enough to 24kAu-plate the substrate? Known issues or advantages? Comments?
increadbly good video , I highly interested in microelectronics , I love your videos , thank you for sharing
This is pretty dam impressive. home brew chip fab. Who needs clean rooms when you can do it your self.
Have you thought about combining 2 transistors into 1 qubit?
so thats what we cut when kamize hacking the xbox drive?
Honest question for a moment... does the lack of a clean room negatively impact the assembly process?
Hi Sam, May I know that whether we can bond aluminum wire to a platinum pad?
love the cheez-it in the background @ 11:51
Are those carriers beryllium oxide?
Hey, I have been trying to make a gold bond from Cu pad to Au pad using the machine 53xx Deep Ball Access by Delvotec. I scratch the Cu pad with sand paper and clean it with isopropyl but yet the first bond doesn't stick with wedge bonding settings. I have tried ball bonding and there it only stick the first bond but not the second. I would appreciate any suggestions you have for me regarding any settings. Thanks for the video as well!
4:00 finally found a use for that old broken lame bong pipe XD
I noticed that as well, the o-ring still being on there was a crackup.
Nice job. I got my start in the microelectronics business as a field service technician for Kulick & Soffa about 30 years ago and threading the wire never gets easy.
How does the aluminum bond to the copper legs? Because aluminum and copper cannot really be welded/soldered together easily
Seriously impressive! How much money did you have to invest to get this all working?
i want to learn wire bonding ultrasonic transducer part. how much khz of ultrasonics can bond wires?
Not the OP of course, but the one I use at work is at 63kHz, which is considered a "low frequency" machine. You have power (amplitude), downforce, and heat (of both the tool and the substrate) to play around with to get a good bond. The power isn't much, but when you're working with wire thinner than a hair you don't need much power to make a bond on clean substrate.
You can find datasheets online manual and semi-automatic bonders, there's a lot to learn there.
Given that those wires always go between fixed positions makes me wonder why there is not a more efficient process possible, like 'stamping' them on all together in one step.
In short: Tooling time and alternative bonding techniques.
Long: every factory produces at least dozens of different chips, possibly hundreds. If you have to change the stamp at every change of product or buy a bonding machine for every product this will become very impractical very quickly. So you buy a machine from some supplier that can make it very fast.
For chips with more than about 64 pins you use flip chip bonding - the package and its lead frame is prepared so that the connections of the pins already reach the right point under the chip. The chip has solder balls on top instead of simple flat pads. Then you flip the chip, put it on the lead frame contacts and apply heat. Done. The most aggressive version of this is BGA where you skip the package and bond the chip directly to the PCB.
man , where did you found a wire bonding machine ?
Ebay. I see them quite often.
Sam if you have any unanswered questions about the bonder, let me know, I work for the company that made it. BTW, great job how the process works!!!
This is really cool unit. Thanks for the explanation.
this is awesome, but do you offer services so i can make a chip?
Amazing, I just had to sub immediately!
What is the operating temperature of your chips?
Hey I have been looking to try this, here is a question. If one did not have a wire bonder, could you use flux and micro solder to the pads?
Andrew York it’s probably too small for you to do that
Yes, but it's difficult to get the power where it's needed.
the one we have in the lab doesn't get the hole for threading in the camera or the microscope, we have to do it basically completely blind which sometimes takes hours!
Amazing! You are an inspiration
!
Great video, amazing to watch .
hi can you make an arc reactor for 2022, after you finished solving chip shortage. thanks.
6:50 ...and I thought threading a needle for sewing was hard
I can thread the wire in 6 seconds that right there got me hired at a semiconductor company. 🙌🏻
I'm going to show this to my electronics students.
Yo dude, great stuff!
so basically this is like a manual 3d printer except the nozzle is supersonic. ive seen people use some ultrasonic "welder" or "cutter" these last months, cutting anything like butter. i have also seen some 5+ year old toshiba video showcasing cnc smd pulse laser soldering, the thing is so precise, makes perfect melts everytime, i almost cried. i can imagine some makers build a $150 wire bonding machine, seems like the gold wire is the most expensive thing here.
Nice! how much for wire bonding machine?
you mean homemade one? :D
I only feel maximum respect here ...
you know the thing to do is create an gue that kids can put on a piece of paper. each dot of gue can be an n type or p type to create a diode or a npn dots to create a transistor.
Idk I watch this but its satisfying
I used to grind wire bonding tools good times lol
Well you have come this far... I guess the next challenge would be to design and build your own 8 bit microprocessor, and then 16 bit......
Great Job! I love it!
I seem to have misplaced my own personal SEM and wire bonder... I HATE it when that happens!
Mad scientist!
I like it
Quer trocar essa maquina em um Celta? quer vender? Great job!
Se o cara entender o que você disse, vai trocar na hora, rsrsrs.
I know how difficult it is to thread the needle! In my institute, a lot of potential users failed training simply because of inability to thread the needle.
This is amazing.
Thank You for sharing 👍 this is amazing stuff