3

With HDMI or USB3.0/3.1, a cable assembly, in addition to the main braided shield, has multiple separate shields for each differential pair, here an example of a USB3.0 cable:

Spliced USB 3.0 cables Image source

The inner shields are connected to the GND_Drain pin on the plug:

enter image description here

With HDMI, it's even more complicated. There, every single one of the four differential signal pairs has its own ground connection on the plug:

enter image description here

How am I supposed to connect these additional ground connections together on a device? Coming from this question, the outer shield could be connected to power ground directly - a few more options are outlined here - and should be connected to the case of my device if that's made from metal. That's easy enough, but what about these additional connections? Would connecting them directly to the outer shield reduce their effectiveness?

Community
  • 1
iFreilicht
  • 671
  • 1
  • 6
  • 21
  • What do you mean by "deal"? usually you don't care how the cables are made up, you just plug them into the socket you have on your device, and then you "deal" with that socket... – PlasmaHH Nov 16 '16 at 12:21
  • I meant how should I connect them together on the PCB. Sorry, I'll update the question to reflect that. – iFreilicht Nov 16 '16 at 12:23
  • Since you have a photo showing how the cables are built, you probably noticed all these shields are touching themselves all the way through the cable. Therefore, can you please explain why you think an additional connection, on the PCB, of the inner shields with the main shield would reduce their effectiveness ? – dim Nov 16 '16 at 13:17
  • 1
    @dim That's not what I'm seeing. There is a thin, blue, presumably insulating film around these inner shields, which I suspect is done for a reason. – iFreilicht Nov 16 '16 at 13:24
  • @iFreilicht That is right, there is a blue side. But I don't think this it is really efficient as an insulator. It is easy to make the test, however: take a brand new cable, a continuity tester, and check whether the various shield are shorted together or not. – dim Nov 16 '16 at 13:32
  • @dim Ok, just tried that on new USB3.0 cable, there's no continuity between the connector housing and the `GND_Drain` pin. – iFreilicht Nov 16 '16 at 13:39
  • @iFreilicht Ok, my bad, then. Well, I'm still rather sure they're intended to be all connected together. – dim Nov 16 '16 at 13:53

1 Answers1

2

My tactic is to join the shields together and tie them to a ground connected to the power supply driving the interface chip. This is often the only ground on the board, but sometimes you have voltage isolation, and you don't want the shield to break that because the shield may be connected to signal ground on the other side.

You want to avoid two things: avoid current loops going through the separate grounds, and avoid voltage spikes that could capacitively couple into your signals. Having a resistance between the separate shields increases the odds of both of these issues.

The shields may be kept separate in the cable to reduce the change that the signals will capacitively couple into each other. Imagine that a mismatch in the length of the differential signals causes a voltage spike one pair. That spike can capacitively couple into the shield. The shield is grounded, but a long cable can have a noticeable resistance to ground. If the shields are joined together inside the cable, a signal can couple into the shield and then into another signal. Separating the shields puts a resistance (and inductance) between them that limits the high speed coupling, so the path for noise traveling from one signal to another is signal capactively going into the shield, which travels through inductance to go to the next shield, which capacitively couples into the next signal. Both events have small effects, but the second scenario offers more protection.

pscheidler
  • 1,079
  • 7
  • 13