Joining w/ external GND killing my PSU?

[bahstrike]

Hi all,

I have a device that outputs a regulated +12, -12, +5v  supply for it's own use.
The current design ties GND to the center-tap of the transformer.

This device detects a low voltage DC signal from an external source.
To provide a common reference, I tie my GND to external source's GND  with essentially no impedence.

Usually this works fine.  However in a couple instances, it causes what appears to be a dead short in my device-  the transformer/regulators get hot and the ICs are not powered up.  When I disconnect the external source,  the device works fine.


The external "active" signal is detected through 50k or higher ohms of resistance, so I cannot imagine that is the problem.  So the fact that the GND are connected directly together indicates to me that either my device, or the external device, has a rapidly changing GND potential and is sinking/sourcing substantial current-  appearing as a short.




Attached is the schematic of the power supply, with a representative model of the external signal detector.

Does anything here look obviously misguided?

Should I reduce the 100k input resistor down to 50k, and add the other 50k to the external GND?
I'm not sure if this would destabilize up my A/D converter.

View attachment 41597

 

[bahstrike]

I'm continuing some investigations and found that the external device ties its DC GND  to  the actual case Ground.  Should I be doing the same?

 

[Hero999]

You're generally better off not connecting the 0V rail to chassis because you don't know what the 0V rail of the device you're powering is connected to.

 

[bahstrike]

This particular one is solved.

Turns out the cable for another connected device was tying my -12v to chassis.
On it's own, this was "fine"  and didn't really cause a problem.

The problem arose when connecting to the external device, which ties its (and my)  0v to chassis,  thereby completing a short to -12v supply through mains GND prong.

Crazy to have a silly DC short going through wiring in the wall.



Unfortunately, I don't know whether this is the same scenario in a couple other reports of similar-sounding issue.