Wonderful example Cunningham's law, thank you. To summarise your point about decoupling capacitors. Use physically smallest capacitor you can, but with the lar

Wonderful example Cunningham's law, thank you. To summarise your point about decoupling capacitors. Use physically smallest capacitor you can, but with the largest capacitance (limited by the knee in capacitor price). High frequency response is dominated by parasitics, but you will get lower impedance for the vast majority of the frequency range until you hit the SRF. A lower capacitance will be better high frequency around the SRF, but worse at every other frequency. Don’t place ferrites down blindly. Ferrites vary wildly, impedance at 100Mhz more or less a useless specification. You need impedance graphs, and you really need to know when a ferrite lossy and not simply inductive, and when it loses effectiveness and becomes a capacitor. If you are going to put them down blindly use the exact same part number and manufacturer as on the development board. If you think you might need a ferrite, put a zero ohm down and measure later. Sprinkling ferrites blindly without a spectrum analyser is at best a placebo, doing it incorrectly is almost always worse than not doing it. Examples of what not to do, splitting a plane to add a ferrite, adding impedance to power pins, adding inductance to IO lines or filter network and creating resonance. As an addendum: Power delivery network, power plane stack up, and component placement matter far the than the precise value of decoupling capacitors. Thinking in terms of current loop area is vital. The PCB is a decoupling component, and the power planes may be your only decoupling at frequencies higher than your capacitors SRF.