'How would you design a PC?'
I'm not even going to go there! And I'm absolutely certain that most PCs would never pass the required ENs. It's easy to print a label with 'CE' on it. I do agree though, that the manufacturers make the best of an almost impossible RFI situation.

I must relate an anecdote though: Last night I was testing a weighing indicator I designed that had been returned from site in Sri Lanka (one that HAS passed all of its tests and got a bono fide CE mark, not to mention International Weights and Measures Certification). The Crystal ADC had failed catastrophically and taken out the 10R Vcc decoupling resistors (PNPN action, clearly). With no other clues as to why, I had to suspect ESD - injected when load cells are swapped - the application they were sold for.

I replaced the ADC and its 10Rs and proceded to fire 7kV sparks from my Schaffner ESD probe into the load cell inputs!! The source is the 'human body model' of 150pF via 10K. I fully expected to damage something. The Winbond 78E516 derailed a couple of times, but continued to function after 'dogging'. The ADC never even blinked! Can you imagine; this device measures accurately to tens of nanovolts!? (Although now I must run it for some months to see if I have latently damaged any internal dielectrics.)

I finished the tests and turned on my two PCs. (From long experience with this kit, I had powered down and unplugged from the mains.) Both had 'lost' their Ethernet connections (which I had NOT disconnected). Worse, they were permanently damaged.

The only coupling was that the Cat 5E cables run alongside my office/lab mains supply for 50ft or so. The Schaffner mains input filtering/decoupling has to be seen to be believed - but still there was enough transient energy radiated from the incoming mains to damage the differential pair on the Ethernet receiver. Yes; the discharge 'loop' was carefully audited to prevent spurious external routes.

But the 8052 kit survived wonderfully. Heaven knows what they did in Sri Lanka!

Dave