Galvanic isolation on DC branch circuits

[jlgula]

In a recent email Chris asserted that 12V circuits need galvanic isolation when converted down from 48V:

Eventually, there may be 48V induction cookers, and there are already in India 48V 500W rice cookers. For audio-visual and computer equipment, I think it unlikely we will change from 12VDC/230VAC, with “wall-wart” power supplies. Almost all 12V equipment is Class-III, ie chassis connected to the negative supply. This actually demands that the 12V supply to each device is galvanically isolated – otherwise current flows down interconnecting signal cables. I think therefore that 48V-12V galvanically-isolated DC/DC converter “wall-warts” are unavoidable. Fortunately, we’re usually talking less than 30W (even for modern laptops). Computer printers are a tricky issue, as they have low standby current but high peak current. 12V lights are cheaper than 48V, but there’s the additional cost of the DC/DC conversion, so on balance 48V lights might be right.

I'd like to see some examples that demonstrate the need for isolation. I readily agree that ground loops can be established via signaling cables. For example, one could have a computer and a printer connected via a USB cable, both devices powered by ODG sources and having different negative pole potentials due to IR drops on their power cable. Yes, current will flow across the GND lead of the USB cable. The USB signaling is protected because it is differential. It is possible that the current flow could exceed the capacity of the USB GND circuits but this seems unlikely. Clearly audio noise is possible on old-fashion audio connectors but these seem like a disappearing technology.

Note that whatever issues occur at 12V seem just as likely or even worse when using 48V appliances because of the wider voltage ranges. That's why I made this issue generic for all branch circuits.

While researching this question I discovered that many laptop adapters have a zero ohm connection between the negative pole and the AC protective earth. This means that when a mix of ODG DC devices and AC devices with class I supplies are connected, a different type of ground loop can be established. This leads to a question about a galvanic isolation requirement on the point of attachment to an external grid.