Earth ground and signal ground . . . connect or NOT?!?!?

[scoobyrollz]

I am designing a digital communications/controller PCB (w/ RS-485, RS-232, USB, Fiber Optics, Ethernet). I am designing this board to be inserted into an existing card cage design. The card cage has a switching power supply that generates 5V. The 5V signal ground is directly connected to earth ground. I disagree with this concept of directly connecting the signal and earth grounds for the sake of noise coupling into the logic and transients having a negative effect on the logic circuits.

I believe the signal ground and earth ground should be connected through a capacitor, ferrite bead, or resistor.

I was wondering if anyone has experience with the idea of connecting earth ground to signal ground through passive components, and what is the best method? Thank you in advance for your time and knowledge.

 

[VEBill]

"I believe the signal ground and earth ground should be connected >>through a capacitor<<..."

That would couple the noise without connecting the grounds. Less than useful.

The normal approach is to specify a single-point ground. This can apply even on the circuit card (not just the rack). Also, provide proper return paths where appropriate.

In general, when connecting grounds, finally, at the single point, one simply joins then up on the ground stud.

 

[europus2002]

Hi Scoobyrollz,
You are absolutely right about not connecting that Earth ground to your digital ground .There are many varibles to this question..
1) Where do you intend on using you circuits..ie.which environment?
2) SMPS would have supply isolation ?? So why do u want to bypass isolation level for merely no advantage?
3) Coupling small signal ground and earth ground
tend to create problems with your designed equipment should there be ground fault at the installation facility...
4) For equipments with high power rating, generally the ground connection creates inductive connection to ground which during leaky phase will give up some significant voltage which will create strong power surge in your equipment....Does your circuit have tolarance for this.....
Many manufacturers couple the ground to so called "ESD Protection" . At all if you "must" couple the 2 grounds , use an R-L-C low pass circuit in series and protect this circuit with TVS diode...Capacitor should be shunted ...I hope this helps..

 

[BobbyNewmark]

You are look at the trees and not the forest. You are driving your design with abstract ideas about purity and correctness. The design should be driven by performance requirements.

If your card is for use in a benign office or light industrial environment with everything plugged into the same power strip, then connect the grounds and be done with it. This is likely to solve more transient problems than it creates, will protect against several types of safety problem, and will keep small leakage currents from doing anything funny.

If your card will talk to multiple industrial machines that are subject to mega-amp fault currents, then every connector will need its own isolation transformer and optocoupler. In fact I would consider scrapping the whole thing and going all-optical.

If your card is for use near a 500 kW radar, or with some random person holding a 20 watt walkie-talkie against the cable, the ground connection is nearly irrelevant. Your design will live or die by filters and shielding.

Also consider fault modes. If you float the ground and the digital power rail shorts to earth ground, the digital ground will become a negative voltage that could have considerable current capability. (What is the peak short circuit current of your digital power supply?) Think about what would happen if you fried the other equipment with 20 amps on its RS-232 ground pin. Is it a $500k lab instrument? Is it critical to keep a $15k/day production line going?

Print this out and frame it: Requirements, Requirements, Requirements.