My next assignment was making the 100 gigabit links of the router work. It broke my confidence and it eventually made me think I did not have the competence to work on hardware. Looking back, the issue was that even after all these successes, I still thought I was an imposter. This had caught back with me.

To understand why this happened, I have to take many steps back. I had joined the hardware lab not knowing how to do hardware. I had also skipped most of the math courses at university. Lastly, I didn’t clearly understand the relationship between ability, capability and work. When you combine these three with the assignment it becomes clear why I reacted this way.

As you continue to raise speed and thus signal frequencies in a link the smaller the margins become. At single digit gigabit speed you have to be careful about the track length to the micrometer, any difference between the pairs has an effect on the signal. As you continue to raise the speed more and more things have to be taken into account. The angles your tracks turn on the circuit board become important. Then, the electronic noise of nearby components has an effect. Then the purity level of the copper itself in the tracks has an effect. It keeps getting more and more difficult.

Our 100 gigabit links were actually 4 pairs of 25 gigabit links. Jumping from 4 pairs of 10 gigabit links for 40 gigabit. Even if you do everything right at the design level and the manufacturer does not make a single mistake, at those speeds you have to tune the settings on the links. A way to do this is to do what is called a SerDes eye.

I had never heard about this and my first encounter was almost like seeing the eye of Sauron. I stared at the ASCII drawing of the eye. I didn’t understand how it was made and what it represented. Because I had this imposter syndrome and had slipped math courses, I came to the conclusion that it was beyond what I could do. I could not make it work. I had reached my maximum ability. It was the first time I didn’t complete the assignment. I think Greg was surprised. With almost a decade since that moment, I now understand that my reasoning was flawed at the time. There was no logic in it.

Just by writing the first draft of this episode, I understood how the eye worked. You cut the waves of the frequencies and offset them to get them superposed. You do that with the positive wave and the negative wave. This creates the eye. It’s merely drawing all the waves upon each other. Then you use multiple “knobs” to make the eye as wide and tall as you can. The eye being the surface inside the curves. It’s almost as simple as this.

In essence, at the time, I had unconsciously decided that I could not perform this task. I had blocked myself from understanding it. I thought I was at the limit of my capabilities. But, I really wasn’t. I was at the limit of my then current abilities but not the possible future abilities. I didn’t know firmware, kernel or deep networking when I had joined, yet four years later I had acquired the knowledge and skills.

Why could I not do the same “trick” with this? It boils down to myself. I did not appreciate the path I had walked to that point. I did not understand that I was in effect in the right place. I felt like a fraud. I could not understand work and time would make my abilities stronger. I did not have the right mental debugging tools. This single assignment that I had failed because I was not confident initiated a process in myself that would end with me leaving the lab a few months later. But I didn’t know it yet.

​

​

(1) Many of the pictures for next episodes come from this wonderful thread about DC5 : https://lafibre.info/opcore/dc5/​

(2) https://www.rambus.com/blogs/overcoming-high-speed-serdes-ip-integration-challenges-part-2/

If you have missed it, you can read the previous episode here​

​

To pair with :

• Bora vocal - Rone
• The Light Ages by Ian R. MacLeod