As part of a discussion relating to the use of in-wire or power line networking type communications, Chris asked: how do you signal something with a hard short to isolate itself (and do you need to)?
I think the issue here is what should the grid to isolate and recover from a short, ground fault or other fault that would interrupt or corrupt a modulated signal carried over the power line wires.
Presumably this would be a fault upstream of the over-current protection in a device. If it was downstream of the over-current protection, that would trip and should isolate the communications in addition to the power. I think this is relatively easy to achieve by capacitor coupling the communications transceiver upstream of the switching transistors.
Also, this refers to a bus rather than isolated links that would presumably isolate communications and power and recover using redundant paths for both power and communications. Without redundant paths, the fault can be narrowed by following functionings links downstream to the first non-functional link which must then contain the fault. This assumes a functional communications link to the head end of the power path using a different technology like WIFI or GPRS.
I suggest that a bus needs a keep-alive message transmitted periodically from some designated bus master. Any device that does not see the keep-alive within a designated timeout period must disconnect power from the bus in a way that leaves communications intact. It is the lack of a signal from whatever cause that triggers the isolation. This technique will protect the sources and the bus itself from excessive current. It also works for most load device faults that don't trigger over-current protection but do interrupt communications. It doesn't by itself localize the fault if it is upstream of the isolation switches. That will need to be done manually by physically disconnecting devices or bus segments, if possible, until the fault is located. Once the problem part is isolated, both communications and power can be restored. A fault downstream of the isolation switches can be identified by selectively enabling devices and watching for communications failures.
As part of a discussion relating to the use of in-wire or power line networking type communications, Chris asked: how do you signal something with a hard short to isolate itself (and do you need to)?
I think the issue here is what should the grid to isolate and recover from a short, ground fault or other fault that would interrupt or corrupt a modulated signal carried over the power line wires.
Presumably this would be a fault upstream of the over-current protection in a device. If it was downstream of the over-current protection, that would trip and should isolate the communications in addition to the power. I think this is relatively easy to achieve by capacitor coupling the communications transceiver upstream of the switching transistors.
Also, this refers to a bus rather than isolated links that would presumably isolate communications and power and recover using redundant paths for both power and communications. Without redundant paths, the fault can be narrowed by following functionings links downstream to the first non-functional link which must then contain the fault. This assumes a functional communications link to the head end of the power path using a different technology like WIFI or GPRS.
I suggest that a bus needs a keep-alive message transmitted periodically from some designated bus master. Any device that does not see the keep-alive within a designated timeout period must disconnect power from the bus in a way that leaves communications intact. It is the lack of a signal from whatever cause that triggers the isolation. This technique will protect the sources and the bus itself from excessive current. It also works for most load device faults that don't trigger over-current protection but do interrupt communications. It doesn't by itself localize the fault if it is upstream of the isolation switches. That will need to be done manually by physically disconnecting devices or bus segments, if possible, until the fault is located. Once the problem part is isolated, both communications and power can be restored. A fault downstream of the isolation switches can be identified by selectively enabling devices and watching for communications failures.