When I started my full time career five years earlier, I did not expect to work on hardware. It was something of the old times. My impressions were that hardware was packed away in a datacenter far far away never to be touched. When you think about it, this was always the case. Servers had always been far away, and there were a lot of reasons why it was.

Server specifications that I would have glossed over several years earlier were a smartphone that was almost in every pocket. Every one had a smartphone that became more powerful as we crossed years. They had become ubiquitous. Yet, in the lab, I saw that it was not the case. Servers were still there, more than ever, more powerful than ever, and more denser than ever. It was just hidden away in cold white rooms far far away.

At that time, I had already spent quite some time debugging hardware, following traces in PCB design tools, soldering wires, and looking at traces on oscilloscopes. I wanted more. I wanted to design these devices. I wanted to do hardware magic. I think it was around that time that I saw for the first time the presentation named indistinguishable from magic. The combination of working in hardware and seeing this video raised my interest to new levels. I wanted to work on these things.

Gregoire was giving courses on electronic design at school 42 near the office. He suggested I join the courses. The university was in session again and here I was, in class again. There were quite a lot of people interested in the topic. The course was structured with a few sessions of theory and then practice. It was so interesting. I think it's actually one of the first times in my life that I eagerly took notes. The way the subject was approached was interesting.

Electronics can almost be split into two categories. Analog and digital. When working with analog, there's a lot of math, a lot of things have to be taken into account. When working in digital, you almost just have to connect components together. It is much simpler and instead of having to juggle formulas, you juggle datasheets and wires.

As I was working full time, my practice time was at work. I was soldering, debugging, reading tracks, and reading datasheets. When I had joined the lab two years earlier, this was all magic, pure magic. As time went, I understood more and more. I had been shocked the first time Gregoire had looked at a hexadecimal dump and told me: "oh it's normal that it does not work, the 3rd and 7th bit are off". It floored me. They were reading the matrix. Yet as months went by, I was starting to be able to do the same thing.

Famous 3rd and 7th bit are off : 0xf3a79b

​PCB design for me was craft. As time went by, one would get better at it. Like a good potter would get better at pottery design over time. The design process was defining a spec; such as, choosing main components, working to connect them, adding components and wires until you had a complete setup. A wire is essentially a track of copper, and you draw it on your PCB. You would start with high frequency or high current tracks, then work out control and low speed signals. The less layers in your PCB, the less expensive it would be to build.

PCB layers, contemporary art, 2024

The amount of documentation that had to be digested, and the amount of context required to hold in your head to design this kind of stuff was staggering. When the design started, some of the components you were going to use were themselves still in design. You would go back and forth to review with chip vendors and adapt as they upgraded requirements. Some of the checks could be automated but a lot of stuff still had to be checked by hand.

Here is a task list if you ever want to design hardware one day. Design your ideal product, then check how much physical space it can take as well as how much electricity it's going to consume. Then you choose the chips and start talking to the companies who make them. Then you start writing code on paper. You have to spatialize your components and the connections between them on 5 to 15 stacked track layers. Then, adapt your design as the chip maker is itself designing its chip. After ending up with several thousands pages of code, you send that to someone else who will build your product. You have delivered some of the chips to their factory, and the rest they order. They physically craft the PCB, solder the components on it and have it delivered to your office. Once you get it, you check for short circuits, then work the way up, signal by signal, feature by feature. If something is wrong, if you're lucky, it's easy. You might have to get out a cutter to cut tracks or solder additional wires. Pray that you don't have to solder again a ball grid array (BGA) component. Sometimes a bug will be a chip errata, and the only fix is to get a new version of the chip.

The more complex the product, the harder it gets. Each additional feature will require more components, more layers, and more tracks. The more of anything, the more interactions you introduce, and the more interactions the harder it is to not make mistakes. As months went, I started to appreciate good PCB design work.

​

To pair with :

• Backtail Was Heavy - Lone
• Snow Crash by Neal Stephenson