One guy got angry at the RAM prices and decided to make it himself… himself. Dr. Semiconductor created its own circuit using amateur facilities. However, let me tell you right away: this is not ready-made memory to be plugged into a PC, and the process is extremely difficult.
Unfortunately. RAM is not something you can “talk about with your brother-in-law in the garage over a golden drink”. For this you need to have really solid preparation, a steady hand and amazing skills. I’m not saying you don’t have it, but I would like to point out that the whole process is not a guarantee of success, and there are many things that can go wrong. However, there is a man who managed to do it.
But let’s get back to the story. A YouTuber dealing with semiconductors tried to make RAM himself, using his own workshop and makeshift production facilities. Those who might see this as a recipe for brutal chip prices will be disappointed. In fact, its material is the truth given to us in the form of a “splat”: it tells us how many processes, chemistry and precision are contained in a component that looks inconspicuous, but we cannot do without it.
Home RAM starts with silicon and layers
RAM memory, simply put, is based on systems of cells in which transistors and capacitors work. The capacitor stores a charge corresponding to the information, and the transistor allows this information to be written or read. In modern components, everything happens on such tiny scales that dust particles, the effects of inaccurate exposure and minimal, accidental contamination can ruin the entire effort. Dr. Semiconductor (original spelling) tried to replicate some of these processes outside the factory, starting with small pieces cut from a large silicon wafer.
The first step involved preparing and cleaning the silicon, and then creating a 330-nanometer-thick oxide layer in a hot furnace. This layer was then covered with an adhesive coating and a photoresist film, a material that reacts to light. After exposure to ultraviolet light, the pattern could be transferred to the surface of the system and chemically removed selected areas. This is a home version of the photolithography process, although of course without the accuracy available in professional semiconductor factories.
How to “draw” a layout in… a house?
Shape mapping is a pain, because the memory chip must have appropriate electrical properties. The next stages concerned the formation of the source and drain of transistors, i.e. the places through which current flows. The film describes and shows the etching of subsequent layers, “finishing” the silicon and heating, which allows the dopant to be moved deeper into its structure. Such operations give a piece of silicon the behavior it needs to work as an electronic component. Before that, it’s basically useless.
After subsequent steps of deposition and removal of layers, the creator used a small stencil to apply the aluminum. Metallization is necessary because the finished structure must have electrical connections, although in this case we are still talking about a very small, experimental form. The excess material was removed and the layout was ready for inspection. And so a working memory element was created in the experimental sense, but not a full-fledged RAM chip ready to be plugged into the motherboard.
The 12 pF measurement looks promising
The newly created structures could not be conveniently checked using the classical method. The scale was too small, so Dr. Semiconductor used a CV Plotter device that examines the dependence of capacitance on voltage. The result showed a capacitance of 12 pF, or 12 picofarads. The created structure behaves like an element capable of storing charge and has the potential to function.
This does not mean, however, that the experiment solves the problem of expensive or hard-to-find modules. Memory requires huge, regular arrays of cells, controlled reliability, addressing, interface, packaging, testing, and compliance with multiple standards. A single successful capacity measurement is more of a “I can do it” stage, but I definitely won’t “run Crysis on it” yet. The creator himself has a plan to develop the project by combining subsequent structures into a larger matrix. One step at a time, iteratively.
A fascinating experience, not a recipe for memory prices
There is no replacement yet and there will not be one: RAM memory is not made of two or three cables and one chip so that you can do it yourself. I think that the author wanted to show a process that we are mostly unaware of, and that is why we have memory in computers. Even in this cottage industry, it is clear that the production of semiconductors is not “joyful printing”, but an incredibly difficult combination of physics, chemistry and enormous manual skills.
Read also: They bought RAM by the ton. Prices are falling and they are panicking
However, I am not surprised that this material has gained so much interest. The growing demand for memory, driven by, among others, AI, data centers and the ever-needed consumer equipment, is causing further price increases. The experiment will continue and even if Dr. Semiconductor will go through a stage with arrays, this will not mean ready-made memory to be inserted into the computer. I assume that even if he manages to finalize all stages (which would be something amazing), the produced system will not work like mass-produced components in factories. But still, respect!
