NOTES
Technical Notes
Ideas that come up again and again in RF and microwave design, written so they connect to the decisions you make in practice.
A Browser-Based Electromagnetic Simulator — What the Method of Moments Shows You
An electromagnetic simulator that solves planar circuits by the method of moments, running in the browser. How to use it, and what the three presets — line, gap and grounded stub — actually tell you, checked against numbers.
Inside the Electromagnetic Simulator — the Analysis Mesh, and the Checks That Make the Numbers Trustworthy
A simulator that passed every qualitative test was hiding an error that made capacitances up to 40 % too small. How the mesh is cut, what convergence testing showed, and a pitfall in the displayed current — what it took to be able to trust the numbers.
Solving Dense Matrices Fast in the Browser — Adaptive Frequency Sampling and Mixed Precision on WebGPU
What it took to solve the method of moments' dense matrices quickly inside the browser — a reduced-order model that cuts the number of frequency points, iterative refinement that gets double-precision answers from a GPU without double precision, and a shader compiler optimisation that was silently erasing the error terms.
Solving Pixelated Circuits — the Corner-Contact Problem Found While Reproducing a Paper
Trying to reproduce a pixelated filter designed by generative AI with my own electromagnetic simulator, at first no signal got through at all. The culprit was cells that touch only at a corner. How to model them, and how far the reproduction got.
Reading the Smith Chart — What Those Circles Actually Mean
A Smith chart is a map of the reflection coefficient with an impedance scale drawn on top. What the circles really are, and the ways of reading it that pay off in practice.
Designing an L-Section Match — Taking 200 Ω Down to 50 Ω
Why two elements are enough, and the textbook values worked through by hand. Along the way, the numbers show how Q ends up setting the bandwidth.
Stub Matching in Practice — Choosing the Line Length and the Stub Length
Single-stub matching of 100 Ω to 50 Ω, followed through to the line and stub lengths. When two solutions come out, the bandwidth numbers show which one to take.
What Conjugate Matching Really Is — The Chart Reference and the Target Are Different Things
Designing a match to 75 Ω or to a complex impedance on a chart normalized to 50 Ω. Textbooks quietly assume Z0 = Zt, which hides a distinction worth making explicit.
Matching Is Not Just About S11 — What Return Loss Hides
A circuit with 22 dB of return loss looks like a pass, yet it turns 14 % of the power into heat. The numbers show why S11 and S21 belong side by side.
Choosing a Load Model — How "Fixed R + jX" Flatters a Design
Two loads with the same impedance at the design frequency, but model it wrongly and the bandwidth looks nearly twice as wide. The calculation shows the difference.
When Bandwidth Falls Short, Add a Stage — Sharing the Q
With a large transformation ratio, a single L-section does not give enough bandwidth. Splitting it into two stages that share the Q widens the band, but a second passband appears higher up in exchange.
Using LC Resonators in a Match — Suppressing Harmonics While Keeping the Match
Replace a shunt C with a trap, and the match at the design frequency stays exactly as it was while a deep notch appears at the second harmonic. A way to add a function without adding a stage.
Drawing Spec Lines — "Out of Range" Is Not a Pass
Judging a design by spec lines rather than by numbers. Put a circuit with a deep notch through three specs and it fails two of them.
Using Smith Match — Designing Matching Networks by Dragging
How to use a matching-network designer that runs in the browser. Add an element, drag it on the chart, and its value is solved for you. Everything in the earlier notes can be tried here.