Do you remember that in August 2021, the 22-year-old New Jersey boy Sam Zeloof announced that he had independently completed a 1200 transistor chip, and he said that he was developing a third-generation chip that could implement the adder function.
Using a series of recycled semiconductor devices, Zeloof thinned silicon wafers, microprogrammed them with ultraviolet light, then soaked them in acid, and documented the process on YouTube and his blog. Interestingly, his home is only 30 miles away from Bell Labs, the world's first transistor to be manufactured.
In 2018, the third year of high school made his own first-generation chip Z1 contains only 6 transistors, which is a pure process and equipment test chip. The Z2 integrates 1200 transistors on a 10 micron process. Although the process lags far behind Intel, Zeroof jokes that the second chip has 200 times the number of transistors than the first, which is far more than "Moore's Law".
Zeloof now hopes to catch up with Intel's 1971 4004 chip, the first commercial microprocessor with 2300 transistors for calculators and other commercial machines. In December, he began working on circuit designs that could perform simple additions.
Zeloof said making DIY chips easier will foster innovation in the tech sector. Photo by SAM KANG
Outside Zeloof's garage, the COVID-19 pandemic has triggered a global shortage of semiconductors, with supply of products from cars to game consoles struggling. After decades of offshoring, the U.S. is currently reinventing its own chip manufacturing industry.
Garage-made chips won't power your PlayStation, but Zeloof says his hobby has convinced him that even without a multimillion-dollar budget, enthusiasts have access to the chip-making process. "The high barrier to entry for chips will make you extremely risk-averse, which is not good for innovation." Zeloof said.
In 2018, in his third year of high school, Zeloof began making his own chips. He was impressed by the YouTube video by inventor and entrepreneur Jeri Ellsworth, in which she made a thumb-sized transistor that included cutting templates with vinyl. Zeloof set out to replicate Ellsworth's project and take what he thought was the logical next step: from isolated transistors to integrated circuits, which historically took about a decade. Ellsworth, who is now the CEO of an augmented reality startup called Tilt Five, praised Zeloof: "He took it [the DIY chip] a big step forward." Reminding the world that these seemingly out-of-reach industries all start out in an unremarkable place where you can do it yourself, and that's of tremendous value. ”
Computer chip manufacturing is sometimes described as the most difficult and precise manufacturing process in the world. When Zeloof started blogging about his project goals, some industry experts told him via email that it was impossible. "Honestly, the reason I did that was because I thought it was funny, but I did get more cautious when I heard it wasn't possible."
Zeloof's family expressed support, but was also cautious. His father asked a semiconductor engineer he knew to give some safety advice. Mark Rothman, an engineer who has worked at a semiconductor company for 40 years and is now an OLED manufacturing engineer, said: "My first reaction was that it was impossible, it was just a garage. After seeing Zeloof's progress, Rothman was shocked and said, "He did something I never thought anyone would do." ”
Zeloof's projects involve both history and engineering. Modern chip manufacturing in its expensive clean rooms, equipment up to billions of dollars. Zeloof couldn't match these technologies, so he read patents and textbooks from the 1960s and 70s, when engineers at pioneering companies like Fairchild Semiconductor were only making chips on ordinary workbenches. "They used X-Acto blades, tape, and beakers instead of these big devices." Zeloof said.
Zeloof must also equip his lab with vintage equipment. On eBay and other auction sites, he found inexpensive equipment from the 1970s and '80s that once belonged to a closed California tech company. Although many pieces of equipment need to be repaired, older machines are easier to repair than modern lab machines. One of Zeloof's best discoveries was a broken electron microscope, worth $250,000 in the early '90s, which he only spent $1,000 to buy and repair to check for defects in the chip.
Zeloof's garage lab
Sometimes Zeloof had to improvise. Just like in a real chip factory, he wanted to transfer his microscopic detail designs to his equipment using a process called lithography. It involves coating a photosensitive material on a wafer and burning in the MASK layer using a device like an ultra-precision projector. Lithography machines are expensive (up to $150 million), so Zeloof built its own lithography by buying an improved conference room projector on Amazon and fixing it to the microscope. It projects his designs onto silicon wafers on a tiny scale, and Zeloof applies them with UV-sensitive materials.
In 2018, Zeloof developed a simple amplifier with six transistors when a substitute teacher instructed students with their homework. Zeloof loves the Grateful Dead band so much that he has the dancing bear on his chip to show his love for the Grateful Dead.
Z2 and Bear, each Z2 contains 100 transistors, 12 make up a "package"
Z2 panorama
The transistors used by the Z1 are called "products of the 1970s" by Zeloof, and their characteristic line width is 175 microns, about the width of a hair. He puts the chips to work on a circuit board that flashes an LED and a guitar distortion pedal.
In late 2018, Zeloof began studying electrical engineering at Carnegie Mellon University while developing in his dormitory, which the university does not allow to use X-ray machines in his dormitory, despite his stated adherence to safety regulations. But he upgraded his setup at home to prepare for his second chip, the Z2. It uses a faster switching transistor design based on a raw material called polysilicon, which dominated in the 1970s.
Zeloof hand-cuts half-inch square polysilicon at 4,000 revolutions per minute on a small, homemade turntable, each becoming a separate chip, carefully coated with photosensitive materials, and then developed using its own lithography machine. A grid of 12 circuits, each with 100 transistors (and a dancing bear), for a total of 1,200 transistors.
Zeloof's first chip, the Z1, was made in 2018, when he was still in high school with six transistors.
His second chip, the Z2, was completed in August 2021 and has 1,200 transistors.
Zeloof is developing the Z3, a chip capable of implementing adders as a step toward a complete microprocessor. Photo by SAM KANG
Each chip is then etched with acid and cooked in a stove of about 1,000 degrees Celsius to roast the phosphorus atoms to adjust their conductivity. Then three rounds of lithography are carried out separately, including ultraviolet etching. Today's commercial fabs produce chips in roughly similar ways, using a series of steps to gradually add and remove material in different parts of the design. These chips are much more complex, with billions of transistors, and these steps are done by machines rather than by hand. The transistors on Zeloof's second-generation chips are approximately 10 times faster than the first generation and have dimensions as small as 10 microns.
In August, Zeloof connected the Z2 to a 40-year-old HP semiconductor analyzer for testing, and the current-to-voltage curve that appeared meant the chip could work. "The curve is amazing," Zeloof says, "and after you've been dipping these little pieces into a beaker filled with chemicals all day, this reaction is now like a sign of life." ”
How is the chip made celebrated? "Tweet!" Zeloof said. His project won a lot of Twitter attention and millions of YouTube views, and even semiconductor industry veterans in the 1970s paid attention to it and gave him some practical tips.
Zeloof said he's not sure what he wants to do after graduation this spring, but he's been thinking about the role of DIY chip manufacturing in the modern tech ecosystem. In many ways, DIY experiments have never been more powerful: robotic devices and 3D printers are readily available, and maker-friendly hardware such as Arduino microcontrollers and Raspberry Pi makes hardware development easier. "But these chips are still produced in some big factories," Zeloof said. "DiY progress in chip manufacturing is very limited."
Zeloof was inspired by Ellsworth's homemade transistors, and high-quality hands-on chip manufacturing can be valuable. "The tools we have today can be within reach of small-scale operations, and for certain issues, I think that makes a lot of sense." Ellsworth says chip processing technology, seen as obsolete by leading fabs, is still useful to engineers.
Zeloof recently upgraded his lithography machine to print sizes as small as about 0.3 microns or 300 nanometers, roughly on par with the process of commercial chips in the mid-90s. Now, he is considering the chip functionality that could be implemented for the Intel 4004. "I want to push the garage chip culture even further and get people interested in making chips at home." He said.
Source: Compiled from wired