[sflicht]

Genuinely asking (not snark, and I am not a physicist): what does it mean to conduct research in "electromagnetism" more than a century after Maxwell? Did you study superconductivity? Electrical power beaming? Magnetic properties of weird alloys?

I ask only because I thought people who study these things are more likely to say they research "electrical engineering" (if they want to be general and applied), or "condensed matter physics" or maybe "materials science".

I'm really just curious about the relevant terminology / sociology of academia.

[notalaser]

Uh :-). I said electromagnetism because it sounds waaay less pretentious than what it was generally termed -- computational electromagnetism -- and because claiming it was microelectronics, while technically correct, was somewhat frowned upon at our university because we were in the wrong department for that. Politics and whatnot.

We were doing research on simulation and model extraction techniques for passive structures in very high-frequency (think 30 GHz) integrated circuits. At these frequencies, the lumped-element approximations break down, so you can't describe the functioning of passive devices in the familiar manner, based on Kirchhof's equations et co.. That means SPICE-family tools can't give you proper results anymore, and you kind of need them to design the circuit.

The "proper" way to do it is full-wave simulation; unfortunately -- especially for very wide-band devices -- that's computationally very costly (simulating the behaviour of the device at a specific frequency isn't very cheap on its own, but we're talking about simulating it for a whole range of frequencies, which could span tens of GHz). You could plug those results into SPICE, of course, but running these simulations took hours, sometimes even days, for a single run, which is completely impractical. It's also not the level you want to work at when designing circuits -- where you're solving "circuit problems", not "field problems".

So what we were trying to do was:

a) Figure out how to make the simulations faster and more accurate, and

b) Figure out how to extract a model from them, i.e. an equivalent circuit, which could be readily used by the design engineers.

We did that through a combination of things like running as few simulations as possible and extrapolating, running simulations in parallel and so on. It wasn't some major scientific breakthrough, but then again, I worked there during my 3rd and 4th year of university; it was the most difficult kind of work I'd done until that point.

[dr_zoidberg]

Sounds very interesting, thanks for sharing :)