It is done. We graduated and are ready to conquer the world of high technology. We are emboldened by the fundamentals of electromagnetics, linear circuits, and RF/Microwave. We know Green’s functions and advanced math. But six months into the new job, signal integrity issues are coming faster than we can fix them. What has gone wrong?
Designs that include functions like optical modules, supercomputing elements, and 400-800Gb Ethernet drive design complexity and signal integrity challenges exponentially. Knowledge of the fundamentals won’t solve today’s signal integrity issues. They require more—a foundational understanding of signal integrity methodology that only evolves through experience. Foundational elements for signal integrity include the following:
• Benchmarking 3D field solvers—simulation to measurement.
• Making quality measurements (TDR and VNA).
• Advanced understanding of non-simple crosstalk propagation.
• Understanding laminate systems.
• Ability to optimize electromagnetic structures using EDA tools.
• Ability to read both S-parameters and TDR as working languages.
• Creating sound stack-ups.
• Dynamic power delivery.
• And yes, all the fundamentals!
In the past, basic crosstalk over lower speeds and reflective loss were considered “the” challenges. Not so at today’s performance levels. High-speed digital components, including optical modules, switch and routing ASICs, SerDes IP, and AI accelerators transmit high-speed data over long distances. These long reaches and tight timing margins require minimized reflection and resonance. Impedance mismatch can come from anywhere. The unit interval—the time region that everything needs to happen for high-speed digital—is working down to less than 18ps!
Specialized companies focus on signal integrity analysis tools. It’s important to note that even with the best signal integrity analysis and test software available, nothing beats experience. In the rush to find and hire the smartest and most talented engineering grads from the best schools, recruiters, executives, and investors overlook one fundamental principle: staff with foundational experience matters. There is a rush for super-computer and AI development with huge capital investment, and those companies are staffed with droves of PhD folks with a finely tuned understanding of fundamental concepts, but no working knowledge. They cannot simply step into 112G PAM4 optical module design and expect to be successful. It’s not about being smart enough.
Those of us with signal integrity analysis expertise have lived through wild successes and epic failures that helped us to learn and develop expertise. It took years of practical experience to hone. We are already seeing the impact of low foundational skills. The result is that many super-computer and AI start-ups will fail. Where is the problem?
I’m afraid my take is that a lot of very clever people are ignorant with respect to the need of foundational mastery, and this is coupled with a lack of humility. The moral of this story is to determine what you don’t know and create a plan to address the gap between the fundamental understanding you have and the foundational knowledge you need.