Not that I know of; my approach is synthesized from physics principles, not from any particular source.
There's always Henry W. Ott,
Electromagnetic Compatibility Engineering, I forget if he goes into much detail as far as design-for-EMC-analysis, but many of the principles apply in general as part of signal quality and EMC control in the first place.
Principles like using reference planes, so that -- in my case -- a trace can always be analyzed as a transmission line over that reference plane, and its in-circuit equivalent is fairly obvious, and its consequences as far as EMC should be fairly straightforward.
Comparing between a well-grounded board and something a beginner would route (traces going every which way, and no ground pour, or a poorly connected pour), the intent is to reduce trace-to-trace coupling, so that instead of a random network of all fairly significant coupling factors (between pairs of traces), traces are reasonably well shielded from each other so they can be analyzed on a one-by-one basis. Consider all the couplings between all traces on a PCB, as the coupling matrix: in a naive layout, a lot of terms will have significant magnitudes; in a well-grounded layout, the matrix will be nearly sparse (i.e., most of the terms have magnitudes small enough to ignore).
The less heavy-weight things to work with, as far as conducted emissions goes, is to consider the supernodes in the system. Don't look at, e.g., the mains input as line and neutral; look at line and neutral together, with respect to ground. Effectively, the common-mode analysis shorts H+N together -- justified by the large value X1 capacitor bridging them at RF. Do the same for internal nodes, and any groups leaving the board (e.g., DC+/-, output +/-, cables..). Note any unbalanced sources (e.g., switching transistor tab screwed to heatsink: a capacitor from switch node (lots of RF voltage) to ground; similarly between windings in the transformer, for isolated SMPS), and reduce the circuit to its common mode equivalent. Now you have a topology with known (approximately) noise voltages and branch impedances, and you have something to build a filter around.
So, yeah, I'm dancing around E&M fields, network theory, linear algebra, circuit analysis theory and more here. It's... not easy to explain in a post.
Tim
Home
Help
Search
Login
Register
