My apologies for reacting to the 'Filter in the Ground Path? No!' alternative thread in this way. I do realise it is not relevant to the original thread (which no one has yet satisfactorily explained).

After far too many years solving EMC problems in noisy industrial environments, the more I learn, the less I know. I do find both Erik's and Kai's (both of whom I hugely respect) responses somewhat dogmatic. Nothing is that clear cut in the mad world of EMC.

But I liked Erik's rules and his glossary. Unfortunately it is not always practical to keep returns from gnd and opto-isolation is often an impossibility. As you say, 'often it will be a necessity'.

And I have solved supposedly intractible ground loop problems with a decoupling inductor between 'grounds' - we are talking situations where a 'solid' path is required at DC, but a high impedance at interfering RF. (With due regard for the possibility of current overload under fault conditions.)

I would also add that a PC is not a shining example of how grounding philosophy should be enacted. PCs are built with only one objective - to make a profit, and EMC components are expensive. I wouldn't dream of using a standard desktop in an industrial environment!

My current strategy (it's my age, I know), all too often, is to ask if some measure has been taken - if it hasn't; do it - if it has; take it off again! Sorry guy's; it works (sad to say).

Dave ;)