Connect ground one or two ends

[pcblearner]

Hello

Consider two circuit boards A and B, where:
1. DC power is fed in to board A from a power supply
2. board B is powered by a power cable from board A

There is also a shielded data cable (100kHz) going from A to B. Should the cable's shield be connected to ground on one side (PCB A) or both sides (PCBs A and B)? Is it possible to have a ground loop if both sides are connected?

[Richard Crowley]

Shielding and grounding is a "gray art".  Not completely magic, but not straightforward, either.  We don't know what distance you are talking about?  3mm?  30m?  We don't know what the board layout or power bypassing or interference sensitivity or a dozen other factors might be.  Consider that the current return path, either through the power ground wire, or through the signal wire shield may be the critical factor.

[edpalmer42]

I agree with Richard.  When it comes to shielding & grounding, even if you've mastered the art, the proper answer is "Do what works".

You might find that connecting the shield at both ends causes your circuit to stop working.

On the other hand, it might be the only way your circuit will work.

It might cause every other circuit within 100 meters to stop working.

You might find that it doesn't matter what you do, everything just works.

One fairly robust configuration is to use a twisted pair for power and seperate shielded twisted pairs for transmit and receive (if you have that situation) with each shield grounded at the receive side only.  Unless grounding at the transmit side works better.   

Other tricks are to use coaxial cable instead of twisted pair, differential drivers and receivers (maybe with shielded twisted pairs for the connections), or opto isolators or even fiber cables to connect the boards.

Ed

[Tandy]

Generally speaking to minimise RF interference cable shields for differential signals are grounded at one end.

[Dago]

Generally speaking to minimise RF interference cable shields for differential signals are grounded at one end.

Some people say this and some people say it absolutely needs to be grounded on both ends to be effective at all

Would be interesting to do some measurements some time.

[bruce273]

if you are grounding to the -ve on your dc supply then your ground will be conected to the other board anyway. So whatever is conviniant.
The usual time you only ground at one end of a signal cable is when another device is powered or grounded seperately, such as a pressure sensor connected to a pipe in a process plant which will be connected to a control system in a control room potentially a few hundred meters away. The pressure sensor chassis will be grounded by the pipe which will be at a different potential to the ground in the control room which can cause a ground loop.

[Sigmoid]

The usual time you only ground at one end of a signal cable is when another device is powered or grounded seperately, such as a pressure sensor connected to a pipe in a process plant which will be connected to a control system in a control room potentially a few hundred meters away. The pressure sensor chassis will be grounded by the pipe which will be at a different potential to the ground in the control room which can cause a ground loop.

I was under the impression that ground loop noise is pretty easy to get in plain home audio setting as well. Like by having two devices powered on separate circuits in the house.

[Richard Crowley]

I was under the impression that ground loop noise is pretty easy to get in plain home audio setting as well. Like by having two devices powered on separate circuits in the house.

If only it were that simple.  There are abundant examples of where widely-separated devices connected with grounded, unbalanced cable has no problem with ground loops.  And conversely, plenty of examples where you will find ground loops but cannot explain why.   

[T3sl4co1l]

ALWAYS ground at both ends.

I believe the common misinformation "ground at one end" arose from archaic vacuum tube designs, where impedances were high and the primary susceptibility hazard was electric field.  Thus, the one-end-grounded shield served to absorb electric field, which was dominant, while doing nothing about magnetic field or conducted (common mode) signals (which were also unimportant as data rates or bandwidths were too low to worry about).

Tim

[wraper]

Generally speaking to minimise RF interference cable shields for differential signals are grounded at one end.

Did you mean maximize interference? Do this for USB3.0 and make a nice radio transmitter, that wireless mouses, keyboards and, maybe, even wifi will stop to work. Though, quiet likely it will fail to work itself too.

[f5r5e5d]

how about ~2k pages of introductory reading:

http://www.amazon.com/Electromagnetic-Compatibility-Engineering-Henry-Ott/dp/0470189304

http://www.amazon.com/Grounds-Grounding-Circuit-System-Handbook/dp/0471660086

http://www.amazon.com/Electrical-Safety-Low-Voltage-Systems-Massimo/dp/007150818X

[bruce273]

Simply put ground loops are formed when two different devices are grounded in different locations then connected together. With two different house circuits, it is possible that each circuit has its own ground reference creating such a problem. You have to remember that ground does have a potential, and when you ground a piece of equipment you are bringing that equipment to that potential. This video explains it reasonably well.



In essence if your secondary piece of equipment is powered by the first then the signal and power earth should be grounded at only one end (which end varies by country, europe normally ground the source but it is not important), the chassis for both pieces of equipment should be grounded locally to prevent electrocution and the signal equipment inside shouldn't be grounded to the chassis