Edit | on duty Utada
| Tiger Sniff Technology Group
Cover from Wired
A Huawei Kirin 990 chip consisting of 10.3 billion transistors. If you zoom in to see the hierarchical structure of the chip, the arrangement layout rises and falls, like a city.
Needless to say, cramming tens of billions of transistors smaller than cells into chips is an extremely complex and costly engineering challenge that has even been described as "the most difficult and sophisticated manufacturing process in the world." The tens of billions of dollars invested in the construction of new fabs each year in TSMC and Intel's earnings report confirm everything.
It was at the time of the 2022 trillion fab plan to advance, a 22-year-old electronic engineering student at Carnegie Mellon University "preliminarily" completed his own semiconductor manufacturing grand plan in the underground garage of his home.
Successfully designed and manufactured a silicon chip equivalent to the level of the 1960s and 1970s.
A 22-year-old engineering electronics senior DIY has a chip with 1,200 transistors, pictured by Wired
This senior named Sam Zeloof, at home in August 2021, successfully DIY out his second homemade computer chip, Z2. That's why the famous magazine Wired reported on it.
However, this chip only attaches a chip of 1200 transistors, which is at least 50 years behind the current high-end process. But Ziluff joked: "Compared with the Z1 made in 2018, the number of transistors in this Z2 has definitely increased faster than Moore's Law." ”
It is very certain that a young boy completely reproduces the entire production and manufacturing process of semiconductors, although it sounds very remarkable, but there will be gimmicks in it - after all, it cannot be mass-produced, and the performance is backward;
But at the same time, many details show that this is not without commercial reference value, not to mention the success of Qiluf, in education and scientific research, there is also a very clear reference significance.
The cost and innovation of "eye-popping"
It's very interesting that when Ziluff started blogging about his chip manufacturing targets in 2016 inspired by some semiconductor DIY videos, many semiconductor people emailed him - "This is really impossible".
Wired quoted a senior engineer whose father had invited him to his home to check on security: "You should not do it, no one will do it." But then, seeing his progress little by little, he changed his tune to: "He did something that I never thought anyone would do."
And Chiluf's response to the objections was also interesting: "I just think DIY chips are interesting. When we hear 'impossible'... We should do it with more caution. ”
Chiluf explains his Z2 manufacturing process in the lab. Image from his Youtube video
In fact, in the early days of the semiconductor and integrated circuit industry in the 1960s and 1970s, many chips in laboratories and startups were "handmade". Therefore, for Qiluf, he needs to have not only the knowledge of electrical engineering taught to him in school, but more importantly, the historical experience of 60 years ago.
So he read all the patents and textbooks of the 1970s and realized that fairchild semiconductor, the first commercial integrated circuit, was to make chips on the ordinary bench— replacing a clean, clean room above class 1,000 with a clean garage; replacing tens of millions of dollars in instruments with blades, tape, and beakers.
Bottles and cans of various chemical solvents. He immersed the chip in a hydroxide solution and dissolved the photoresist on the silicon wafer.
Even if you are not familiar with the semiconductor manufacturing process, but in recent years, with the help of Huawei and Nvidia, you can more or less be popularized by the semiconductor knowledge bombarded by the network. In the seven production areas of the chip manufacturing process - diffusion, lithography, etching, ion implantation, film growth, polishing, and metallization, "lithography" and "etching" are undoubtedly the two most core steps.
However, precision instruments such as lithography machines, which cost hundreds of millions of dollars, are absolutely impossible to appear in a relatively rudimentary underground garage; in addition, these precision instruments also have a considerable degree of operating threshold.
But what you would never expect is that according to the netizens' public video of Qiluf (interested in watching Youtube, he filmed the whole process) and the value of the instrument in the data, the total cost should not exceed $20,000.
The basic principle of "lithography" is relatively simple: first apply a photosensitive material to the silicon wafer; then use a projector-like device to project the design template on it, and then dissolve the excess through a series of steps. Therefore, his solution to lithography can be called "knowledge transfer ghost"——
He bought a conference room projector on Amazon's website, modified it and mounted it on a microscope to form a simple "lithography machine."
He has optically shrunk his design to the right size and projected it onto each polycrystalline wafer of about half an inch that he cut by hand, smeared with UV-sensitive material.
A homemade lithography machine then projects a beam onto the "design": a grid of 12 circuits, each with 100 transistors, for a total of 1200.
The "etching" step is much like a chemistry lab class in college: each chip is corroded with acid (various solvents), then it is thrown into a stove at about 1000 degrees Celsius to bake to adjust the conductivity, and then sent to a vacuum chamber for further processing.
The above steps go back and forth three times to end the key production steps of the chip.
Spin-coat photoresist, bake, then expose and develop, using a variety of chemical reagents and instruments in between. Back and forth several times and then put into the vacuum chamber for evaporation, sputtering
But wired argues that even now, giant fabs produce chips in roughly similar basic ways — using a series of steps to add and remove material in different parts of the design. It's just that the complexity and cost are thousands of times higher than The manual workshops in Ziluff.
Interestingly, with the exception of "lithography machines" and "etching machines," almost all of Qiluf's equipment is a piece of technology scrapped from eBay and other auction sites — mostly made by Silicon Valley tech companies that collapsed in the last century and are now cheap and deeply repaired.
"One of his best discoveries was that the broken electron microscope, which sold for $250,000 in the '90s, only cost $1,000 to buy it and fix it." The main role of this machine is to check for chip defects, do not underestimate this machine of the century, it can even observe the nanostructures of butterfly wings.
Electron microscope, picture from Tsziluff's vlog video
It was under such a very low-cost experimental environment and software and hardware processing that Zieluff's second-generation chip Z2 finally ran a beautiful current and voltage curve on an HP analytical instrument produced 20 years ago - which represents a simple chip that has a short life.
Tsiruff believes that the results shown by the analyzer are ideal
In fact, everyone will see the superficial meaning of this thing -
Too good hands-on ability, execution, and a kind of disobedience and self-confidence that belongs only to young people. Chiluf's results were enough to shame engineering students who always complained about the inability to provide advanced instruments in schools.
As one developer praised Chiluff's execution: "What impressed me was how he drove the hobby. When I was 18, I had a lot of hobbies, but I rarely accomplished what I wanted to do, and learning was a bigger challenge. ”
In addition, Qiluf also won the encouragement and support of a group of semiconductor elderly people in the 70s on the Internet with real and confirmable video data.
But in fact, Ziluff's idea is relatively simple: he finds it interesting, so he does it (it is his habit to do it by hand — in 2020, he will also transform a Polaroid film machine into a digital camera); in addition, he wants to prove to the entire semiconductor industry a fact that is sometimes deliberately overlooked:
It's not about having a budget of millions of dollars to be an inventor or an innovator. This will drive away too many young people who are interested in semiconductors but who are too high a threshold for exclamation.
Of course, the chips he makes are absolutely impossible to fit into your phone and server; but this kind of behavior is precisely the goal of open source and human science - everyone can improve the overall outcome and promote the common good by copying.
Tsziluff also remodeled the camera: on the left is a photo of the film machine, and on the right is a photo taken by Tsziluff after transforming into a digital camera. Image from Carnegie Mellon University
In addition, from the perspective of the commercial market, Chiluf's behavior and results can prove that high-quality practical manufacturing will have high value for the small-scale production and operation of certain types of chips (such as ASICs).
Generally speaking, in the semiconductor market, only by achieving large-scale manufacturing can we reduce costs and obtain sufficient profits; in modern processes, achieving scale is more important than performance.
However, Qiluf told us that in the current development of the semiconductor industry for nearly 100 years, the existing open software and hardware tools have been able to meet the needs of some companies to achieve small-scale manufacturing under the premise of ensuring costs; and many of the old technologies of the last century are not completely devoid of merit.
Non-negligible issue
We found that some of the words in Wired's report on Ziluf were very much in line with the perspective of European and American culture. For example, they emphasize that the "garage" where Ziluf works is owned by his parents. Chiluf, on the other hand, is a young man in his senior year.
Yes, this is also the biggest point of contention on the Internet after Ziluf built the chip in the garage - the media only talked about his ingenuity, the process of making the chip, and his ultimate success, but ignored the obscure additional conditions behind his achievements.
For example, his family and his social environment are some hidden resources.
One netizen relentlessly criticized the public for the excess of Ziluf's behavior — it was nothing more than telling the story of the innovation of a smart rich man. The chips he made were of no use compared to what he has now.
"Looking at his garage lab, what kind of high school or college student can buy equipment with thousands of dollars?" At his age, I certainly don't have these resources to do anything. In addition, there is a 99.99% chance that his loved one has a physics or engineering related degree. ”
Image from Wired
This statement is precisely to the point of some facts.
Chiluf's father, an industrial designer who runs a racing parts workshop, and his brother is a robot engineer, it is difficult not to be exposed to a lot of engineering knowledge and hands-on opportunities in this family atmosphere.
In addition, we can't ignore that the senior semiconductor engineer who worked for the garage laboratory for more than 40 years was invited by Qiluf's father.
As a result, his success is well documented – although a large part of his personal ingenuity is not really starting from scratch.
This also confirms what Malcolm Gladwell, a veteran new Yorker contributor, wanted to prove in his blockbuster 2008 book The Outlier:
"Those who become outstanding seem to struggle on their own, but they are not. In fact, they have always benefited from some hidden innate advantage, or extraordinary chance, or the special advantage of a certain culture.
As biologists often refer to when discussing organisms: people usually think only of the best seeds of the highest quality, but rarely think of the fact that the finished timber must also have plenty of light, deep and fertile soil, and enough luck to escape rabbits and lumberjacks. ”
In fact, Qiluf also casually mentioned that he did this DIY experiment to further verify what position DIY manufacturing is in the modern technology ecosystem.
Robotics and 3D printers are too easy to buy and not too expensive at the 2022 node; the open-source electronic hardware and software platforms he uses, Arduino, and the small single-board computer Raspberry Pi are only around 2005. So, the technology of the 70s is now being validated by a college student, and it seems that it has become less difficult to imagine.
In other words, extraordinary achievement depends not so much on talent as on opportunity.
Therefore, this biggest controversy has also triggered the appeal of many foreign education netizens: if more young people with families and social conditions are allowed to obtain the same resources, time and guidance, then it is more conducive to universal technological innovation, and this must require the support of national legislation and policy guidance (perhaps also because semiconductors have matured into neglected industries abroad, until recent years).
And this, from the current trend, in fact, the domestic work is much better than abroad.
However, relatively speaking, the domestic semiconductor industry has won the full support of the state and public opinion, but it seems that there is still room for progress in the concept and system of family education and secondary and higher education.
Ziluf's selfie, pictured from Carnegie Mellon University
Write at the end:
Qiluf's success must be blessed with "having advantages that others do not have." But his dedication and creativity to science and technology has undoubtedly encouraged many young people who love science and have a passion for innovation.
"I think the lesson he taught me was that the answer to this conundrum is something we can touch, and that's it. This will be extremely useful for the next generation to accumulate scientific background and knowledge, and ultimately create something that will be useful to more people, just like the Raspberry Pi, passed down from generation to generation. ”
An engineering student who is currently studying at university doesn't care about the so-called hidden advantages:
"Hobbies are really useful! That in itself is a great learning! Even if you can't knock on this door, you may knock on other unexpected doors.
Whether these hobbies cost $1 or $1,000, Ziluff proved that the question is completely unimportant. ”
Finally, I would like to leave the reader with a question:
If Tsiruff's story takes place in China, what kind of commentary and story direction will there be?