Sorry
@Julie.
I have not looked in depth at how bi-directional RCD's overcome this issue, as the detection circuit would surely have to come from one or other side of the device.
Do you have any knowledge of how this is achieved?
Thanks
Sy
Well it's a pretty common issue with the cheaper microprocessors out there.
If there is no supply to the main controller, but there is to one of the i/o ports the current can flow into the controller and destroy it. If the supply is there the relative voltage across the protection diodes is not a problem.
If however you can limit the current to a much lower value, although the current backfeeds it isn't enough to cause damage.
So a redesign of the detection circuit may be sufficient, or changing to a microcontroller which doesn't suffer from this problem may be required.
In effect it's merely an electronic design flaw/feature the solution of which comes down to the actual implementation itself.
As a hobby I do design stuff using microcontrollers (often better ones like atmega4808 or atmega328, or stm32... but i have use things like the attiny range which are "cheap and cheerful".)
When I use an i/o port I tend to use a potential divider into a zener diode, then into the i/o port. The zener limits the voltage (usually to 4V9) and given the voltage a 10k resistance limits the current into the controller to 0.49mA. Of course this adds a zener and a resistor in addition to any potential divider on every i/o port - which for me is not a problem, and probably overkill for the better devices, but adds up for every production device.