I don't think you're likely too far off of "just work" in terms of EMI, but there's certainly stuff you can do to minimize the risk. Part of this is your application, the noise floor is going to be the first thing effected by EMI, but where you probably want to operate is considerably higher power levels, so you don't have to necessarily go all-out for noise reduction since the range it would effect probably isn't the one you'll be interested in.
That said:
1. I think it would probably be sufficient, an LC filter on the rail going to the analog section could keep a lot of the noise generated by your digital stuff, and the LDO with some extra caps on either side should do well for eliminating switching noise from the supply (though this depends on LDO choice, most should list PSRR rating). A second LDO for the analog section is often a good choice (in parallel with the digital one for a separate rail), but probably isn't necessary.
2. I don't think the ground loop would be big trouble, but you may want to try a star ground of sorts - ground the coupler's grounds and analog ground to the same tie point, then the digital to that (maybe at the power supply area), and join the chassis ground to that power entry point ground (or external chassis ground tie point, since the plug pack probably isn't doing it). This works with ground planes as well, you can have a plane for your analog section, a plane for your digital section, and a small break separating them except for where you want them to come together. Worth mentioning that if you're just powered from a plug pack, your device may just be floating unless you have an extra chassis ground point (and if it does float, then its ground will be through the coax connected to it).
3. Ferrite beads are single turn inductors through a permeable material, nothing particularly fancy. That means their inductance/filtering capability is a lot less than most inductors, but it also means their current handling/size/cost are better for a given equivalent packaged inductor. Most are rated by impedance at 100MHz, so while they are useful for keeping out upper range signals to prevent mixing products, they won't be filtering much of anything in the band of interest. Since they have a core, the bias current does reduce their filtering capability (like on a power rail), Analog Devices recommends about 20% of their rated DC current as a max for good filtering performance here: for a bit of a deep dive...
https://www.analog.com/en/analog-dialogue/articles/ferrite-beads-demystified.htmlRibbon cable will suffice, though parallel digital conductors do like to interact with each other. That said, the edge rates aren't going to be hugely fast, and adding ground conductors lowers the impedance and can physically separate noise producing lines - probably not necessary since none of the cables are mixed signal. If you're really concerned with edges from the digital lines, you can add a small LC (ferrite beads as the L can be good here), or small RC filters to signal lines to slow the edges... but don't go nuts with it or you'll slow the rise time to the point the digital stuff can't deal with it. Otherwise, you can try physically separating the analog section from the digital and signal lines to minimize coupling further. You could also look into shielding the analog frontend, but again I think this isn't required. You may also find that the speed your MCU isn't required, and going with a slower crystal or multiplier setting could be sufficient for your processing requirement but would lessen emissions again (though maybe not as much as one would hope since the edge rate would be about the same as the full clock speed).
And it's worth mentioning that your consideration is really only about the analog parts, so while you can do something to try and reduce emissions from digital sections, the digital stuff will probably be pretty happy on its own without much special treatment beyond decoupling caps. Just keeping the analog stuff on its own section of the board without digital lines going through/near it and a little LC filter on the supply line for the analog section will go a long way towards keeping things quiet. If your digital stuff was a lot faster, a lot more sensitive to power supply noise, or driving much higher currents, it would be much more of an issue.
Home
Help
Search
Login
Register
