Multiple isolated devices and GND

[jgauthier]

Struggled with the subject.  I'll try to explain.

I have three devices.   I will refer to them as A,B, and C.

Device A will power device C.  This will happen through a GPIO level trigger, through some relays etc.
Device B needs to communicate to/from Device C via TTL level UART comms.  Minimally, that is 3 lines.  TX,RX, GND.

However, device C GND will come through the power switching circuit through Device A.

Should I ultimately connect A and B GND together?  B is not my device, and "untrusted".

Thanks for the input.

[Vovk_Z]

B and C will be possibly connected so you don't have to make one more connection between A and B. The less ground connections - the better.

[coromonadalix]

Untrusted in what ?  network,  operating system ?

you can power them thru isolated power supplies instead of isolating connections  if they became at same ground levels  when connected / communicating ... you have fully isolated  dc-dc converters ??

Normally the ground is the common point in many things in communication, but not necessarily connected to the psu lines


are they automotive or industrial ??

define  device a will power b thru gpio ??   is it an raspberry ?

i think you do know  you can have  usb / rs232 / ttl or cmos isolators thru opto couplers or specialized isolator  ic's

[Terry Bites]

If you power something with iso power then you have to communcate with it over isolated lines. If not, the power isolation serves no purpose.
If you dont trust it then really isolate it.

Digital isolation is simple- optos are slow but things like iCouplers are fast and simple to use. You'll find made for purpose bidirectional versions for coms.
https://www.analog.com/en/parametricsearch/11037#/
https://www.ti.com/isolation/digital-isolators/products.html

You want to add protection on lines within the iso side circuit. ie signal to iso gnd and iso power.

[exmadscientist]

If you power something with iso power then you have to communcate with it over isolated lines. If not, the power isolation serves no purpose.
If you dont trust it then really isolate it.

This.

A lot of even senior engineers don't get this, so I'm going to say it again.

If you cross a "galvanically isolated" barrier with absolutely anything other than a galvanically isolated signal then you have no isolation at all. None. Nothing. What you have might or might not be useful, but it ain't isolation anymore. That's why best practice for isolated barriers on PCBs is to pull back solder mask so you can see clear through the PCB and verify immediately with your own eyes that only isolators ever make the crossing.

[jgauthier]

Coromonadlix> Untrusted device.  As in: It's not mine.  It's someone else's.  Untrusted is just a term - I mean I trust them. But it's not a known device, not my design, and not provided by me. So I call it "untrusted".

Since A and C are powered through the same PSU, and B is not - if I use GND for comms between B and C it's then connected to A.
Which I felt like was dangerous waters.


I have like 12V power source.  A is powered through a 3.3V regulator or DC/DC converter.
Device C is a 9V system. 

When I say A will power C I mean A will trigger a GPIO (not a pi) that will flip a relay, that will power device C from the 12V source.

I have recently read about some possible isolation ICs and I am digging into those.


Terry Bites>
Thank you. That was my gut feeling. I will look at the poroducts you mentioned.

exmadscientist>  Thank you! This exactly why I asked the question.  I was putting this down and when I saw the GNDs my brain went "Hm, nope. This is going to bind A and B and I do not want that"