Dear All,

I have been pondering back and forth about this for some time so am just going to open up a discussion and see where it goes.

Basically, best practices tell me that for off board connectors it is generally best to have a clean 0V hard tied to your chassis where you 'attempt' to dump noise the cable has picked up and tie this clean 0V at one point to your internal ground planes. This premise is based on trying to give the noise the shortest return path to the chassis such that it can return through whatever path it can find which is hopefully better than your cables.

Now let's shift this thought to an impedance matched 3GHz link. The microstrip/stripline impedance matching will always be heavily dependent on the GND/POWER planes. So in my mind as described in the previous paragraph, this would result in an interruption to the microstrip/stripline and perhaps introduce 'more' noise in the form of reflections. I guess I'm just curious how you all out there deal with very high frequency off board connectors in terms of 'best' EMC practices.

Many thanks in advance,

Skeeb

 

Hello,

Perhaps the following article may give you an idea:
http://www.analog.com/library/analogDialogue/archives/39-09/layout.html

Bertus

 

If you have a 3GHz signal you'll be sending it down a coax. And you're correct, you would not want to route the coax shield through the chassis of your box. This would be a serious impedance mismatch and cause reflections and degredation of your signal.

In general, I would connect shields to the box chassis, but not signal returns.

 

Hey Bertus and JMac, thanks for the replies.

Bertus, yes I have this article in my design reference respository, it is a great one but unfortunately doesn't answer my specific question.

JMac, yes and no. TMDS (i.e. HDMI, DVI) are examples where this is not necessarily the case, although they do employ internal shields on the twisted pairs which use the shields similar to coax to improve their impedance matching capacity. I guess it is a similar consideration nonetheless as there are 'return currents' on the shields and not just noise.

I have attached two new items. "Off board connector filtering.png" and "Text book example.pdf".

"Off board connector - filtering.png" demonstrates the concept being discussed. I believe the answer to my question is likely to be, ans.: shielding. If you are doing high frequency you wouldn't filter and only shield. So the question likely shifts to the next item, however IS still relevant to the shield termination, where I agree with JMAC, you would want the shield of coax to be terminated to your signal ground as it IS part of the signal. However it begs the question of best practice when analyzing the next item (text book example.pdf)

"Text book example.pdf" demonstrates the concept of general best EMC practice for off board connectors, in particular unshielded. We see this concept employed to ensure ground bounce doesn't create off board antennas. So for Coax it would or would not be a good idea to terminate to Quiet I/O ground as shown? (I would argue it is bad, albeit for the noise picked up by the Coax shield [i.e. the outer non-field constrained portion of the coax shield conductor, keeping in mind coax is designed to try and constrain the field to the dielectric and subsequently the return currents to the inner part of the braid, on the outer part of the braid] it would be better).

All said and done I believe the following is the best practice:

1. Maintain a clean ground where 'non-signal' shields are connected (i.e. purely noise shields)
2. Filter unshielded off board connectors to the clean ground
3. Do not use unshielded connectors for anything high frequency
4. Terminate coaxial cable shields to noisey/digital ground
5. Terminate HDMI signal shields to noisey/digital ground
6. Terminate HDMI cable shield to the clean ground

Let me know what you all think,

thanks in advance,

James

 

Pretty much agree with your conclusions.

In our systems we had a "chassis ground" that connected to digital ground at only a single point. This chassis ground was equivalent to your "clean ground". Any filtering done on signal lines was returned to the chassis ground. Also any ESD protection was returned to the chassis ground. The theory was that noise and/or ESD would be returned to the source without traveling through the ground plane on our board and disturbing circuits. Anything carrying signal return currents would be connected to digital ground.