Colinb, an user from allaboutcircuits.com forum post this here:
I have been trying to understand how to ideally handle the cable shield
on a USB device. (Full Speed USB, in this particular instance.)
As seems to be the case with many signal integrity issues,
contradictory recommendations abound, each with its own unsupported
claims. Even authoritative-sounding sources such as Texas Instruments,
Intel, FTDI, and Cypress Semiconductor seem to disagree on the correct
way to handle the cable shield on USB devices.
Contrary to my initial supposition, the purpose of the USB cable shield
is not to protect the USB data lines from outside interference, but
rather to prevent the USB device from radiating EMI.
Here are some of the options that have been recommended.
Note that (2)series capacitor to pass high frequencies onlyseems to
directly contradict (3)series ferrite bead to block high frequencies
(1) Connect shield directly to signal ground.
(2) Connect shield to signal ground through a capacitor.
(Possibly with high-value parallel resistor approximately 1 Mohm.)
– Tying the shield directly to ground would create a direct path
from the ground plane to the shield, turning the USB cable into
an antenna. To limit the USB cable antenna effect, it is
recommended to connect the shield and ground through an RC
filter. Typically, R = 1MΩ and C = 4.7nF in Figure 3-5.
Atmel AVR1017: XMEGA – USB Hardware Design Recommendations.
Section 3.3.3 (p. 8).
(3) Connect shield to signal ground through a ferrite bead.
(4) Do not connect cable shield to ground on the device at all.
Whether or not the device has a metal chassis, and the handling
of chassis ground and signal grounds, (as well as how the USB cable
ground is connected to either one) is certainly important as well, but
this isn’t clearly discussed in most of the writings on USB cable shield
The device I’m developing is a bus-powered device which will likely be
in an unshielded plastic enclosure.
Thanks in advance for any bits of wisdom on this topic full of
contradictory information. I recently posted this question on si-list,
and even there I got little in the way of answers.