Just a note on RFI and the Spectrum.. it failed FCC regs at first, because of the RAM bus architecture. One of the clever design features was the contended RAM below 32k, which was shared between the CPU and the video circuitry - basically, the CPU had its clock stopped when it was trying to access RAM that the video needed. The video buffer was in the CPU address space, saving the cost of two lots of chips, but at the price of performance - obviously, the video had to take priority, because when a scan line's starting on your television it needs its data there and then.
But the video buffer was fixed and small and lived below the 32k boundary. A 48k machine had 32k of memory above that , logically and physically separate, that the video circuitry would never need to see. So, by splitting the bus, it was possible for the CPU to run at full speed in the upper 32k while the video stuff could do what it liked in the lower. And - bonus feature! - you could do it with a bunch of resistors of just the right value to let both sides of the data bus be driven differently when appropriate, but track each other when one side relinquished driver duties. Cheap, fast, lovely.
Which is great, but it means when both sides are being driven independently you have two sets of PCB traces flailing around being two buses instead of one. When you have one length of conductor going one way and the other going the other, well - you have a lovely dipole And as you can see from the teardown, RFI shielding was as alien to Sinclair as juggling sand-fleas.
The FCC was unimpressed, which is why the Timex-Sinclair 2000 (and the other Spectrum derivatives) were late and had a lot more tin foil and that gritty grey conductive spray-applied coating on the inside of their cases.