Good point about the purple curve (possible artifacts of sweep rate of a spectrum analyzer).  I hadn't paid much attention to that plot, since in brochures I'd not be confident it's even from real data rather than an artist's impression.  I hope the block diagram of the device is fairly true to life, e.g. about measuring current in one conductor only, and not measuring voltage. 

The actual AFDDs are presumably based on simple filters and rectification in the analog part before the microcontroller, so I don't see what their trouble would be until their output near-saturates all the time due to non-arc disturbance, so that the correlation of HF detection to ac cycle wouldn't be seen. I see nothing about a level of external disturbance that would be needed for this, and don't recall any immunity requirement in the standard. 

This immunity is an interesting line to test, but I suspect the AFDDs would work as intended in "most" circumstances ... the bigger problem I see in justifying them is the familiar lack of evidence of the level of benefit they'd bring even if able to catch all/most arcing types of fault.

Hi Andy, also RCDs have also been known for sometime to offer protection against upstream arcing and loose connections!

I'm not sure RCDs offer reliable 'protection' from upstream arcing - I think it's more likely that the loose connection causes a temporary voltage rise on the N which in turn increases the leakage through a grumbling high resistance N-PE fault that in normal circumstances donesn't pass quite enough current to trip the RCD. I suspect that in the vast majority of cases an RCD wouldn't trip on an upstream arc.


    - Andy.

Impressive memory (the link to 2008)! I would be surprised at modern RCDs responding in this situation, but old ones with less transient immunity could well do so if the installation has filter capacitors L-E and N-E and particularly if the loose upstream connection is in a neutral.