2

I want to know the best approach to connect USB connector's (receptor) Shield to System Ground? Below are the two options.

PC(Earthed)<----- USB Cable ----> USB Connector female(PCB with NO metal casing or Earthing)

schematic

simulate this circuit – Schematic created using CircuitLab

Electric Monkey
  • 53
  • 6
  • 1
    WHAT did your EMI tests inhouse indicate were best for interference reduction? – Tony Stewart EE75 Apr 22 '21 at 12:46
  • @TonyStewartEE75 Product is at design stage and prototype is not yet available to test. – Electric Monkey Apr 22 '21 at 12:49
  • Then you must test the cables as I suggested with CM chokes for improvements to interference. Otherwise group delay distortion is unaffected. E.g. earth grounded at one end and floating SMPS noise device causing errors at other end from interference – Tony Stewart EE75 Apr 22 '21 at 13:03
  • Test on similar USB products – Tony Stewart EE75 Apr 22 '21 at 13:06
  • Why not option 3, just connect the shield to ground? (But you need to do it on one side of the cable only). – Math Keeps Me Busy Apr 22 '21 at 14:07
  • @MathKeepsMeBusy Since a standard USB cable is used to connect to a standard PC, cable shield will already be connected on one side to ground even if nothing is plugged in. – Justme Apr 22 '21 at 14:38
  • @Justme PC side is already connected to Ground and then earthed via case, so it should be left unconnected on Device (PCB side)? – Electric Monkey Apr 22 '21 at 14:41
  • What will happen if the cable is not connected at all, in terms of ESD protection? The connector is exposed and human touch is inevitable before plug-in cable to USB receptor on PCB. – Electric Monkey Apr 22 '21 at 14:43
  • I don't get the point of option 2. It's a DC short, and will prevent high-frequency distortions from being absorbed into GND. Rather just connect it directly to GND. I've only seen option 1, where the resistor ensures that the USB device GND potential won't float away from the computer' PE and the cap provides a high-frequency bypass to GND. – Ulminpoika Apr 23 '21 at 10:25
  • Does this answer your question? [Structure of a usb 2.0 connector](https://electronics.stackexchange.com/questions/269244/structure-of-a-usb-2-0-connector) – Ale..chenski Apr 23 '21 at 18:20

1 Answers1

0

You need to ask yourself "what is the purpose of the shield?"

In my mind, it serves 2 purposes. 1) Protection from external electromagnetic disturbances (ESD) and 2) Prevent internal EMI radiating out.

Note that an unconnected cable (a high impedance cable) will create electromagnetic fields and will act as a dipole antenna. It can be a very efficient radiator > 1MHz.

A loaded cable will create a loop currents (i.e. loop antenna) and generate magnetic fields. These can be highly effective radiators at low frequencies.

In both cases, the situation of concern is where the cable length is > 1/20th the wavelength of concern. So a cable is of concern whether it is loaded or unloaded, it's just different.

Personally I always keep the shield separated from DGND as I do not wish to provide a low impedance path for current to travel down my shields. (Currents will tend to the lowest impedance path). It is vital that your shields do not carry current. They should be tied to a "quiet" voltage potential. i.e. A they form part of your Faraday cage.

Also, due to skin effect, a shield will act like two conductors, on the inner and outer surface. You want that outer surface to be quiet, so it needs high impedance to your noise sources. (In my opinion). As a result, either of those options is acceptable, although the ferrite bead does not have to reference a ground in order to provide a high impedance path. The capacitor needs to be tied to a low impedance (quiet) GND point to be effective.

Personally, for ESD, I want to provide a path to EARTH for the fault current. I do not want to route this current to the DGND plane. As a result, you will often find that the shields are all connected to EARTH at some point in the system.

If the discharge path is via your DGND plane, then the capacitor is lower impedance and therefore more effective, with a 8kV strike being clamped to about 900V (4n7) for 10us with a peak of 30A current flow. The inductor will "clamp" to about 2.4kV for 100ns and the same peak current. So you get a shorter transient but higher transient with the bead. If you elect to parallel a 330R resistor with the capacitor, it will be subjected to up to 4A of current during the discharge event. All of which is to suggest it would be smart to use larger components and making sure your capacitor is "ESD rated".

Kram
  • 1