Too many hardware teams treat EMI/EMC compliance as a testing phase — build the board, ship it to a certified lab, find out if it passes. This turns compliance into a gamble with expensive, late-stage consequences: a failed test at this stage usually means a layout respin, a new fab run, and a schedule slip measured in weeks.
The better approach is designing against EMI risk from the first placement decision, so lab testing becomes confirmation rather than discovery. This starts with switching regulator placement — every switching power supply is a radiated and conducted noise source, and its loop area (the physical area enclosed by its high-current switching path) directly predicts its emissions. We place switching components to minimize this loop area before anything else is routed nearby.
Ground plane integrity is the second lever. A continuous, unbroken ground plane under high-speed and switching signals gives return current the shortest possible path, which minimizes both emissions and susceptibility. Splitting ground planes for 'analog' and 'digital' sections is frequently done wrong — done well it isolates noise; done poorly it creates return-path discontinuities that make emissions worse, not better.
Cable and connector interfaces deserve particular attention since they act as antennas for both emissions and susceptibility. We add common-mode chokes and proper filtering at every external interface, and we route sensitive signals away from connector edges where possible.
Finally, we run pre-compliance testing internally — near-field probing on the bench — before a board ever goes to a certified test house. It's not a substitute for formal testing, but it catches the majority of layout-driven EMI issues early enough that the formal test is a confirmation, not a surprise.