English is currently available.
USB DISCUSSION
Reference designs disagree on what to do with the metal shell of the USB connector on a device. I have seen it tied straight to signal ground, connected through a 1 MΩ resistor in parallel with a 4.7 nF capacitor, connected through a ferrite bead, and left floating. My product is a small USB-powered instrument in a plastic case with no other external connections.
What is each option trying to achieve, and which is appropriate for a design like mine?
The shell and the cable shield have two jobs: to keep the cable from radiating the noise of the signals inside it, and to give electrostatic discharges a path that does not pass through the electronics. Both jobs need a low-impedance path at high frequency between the shield and whatever your circuit uses as its reference. The host side bonds the shield to its chassis and ground, so the cable shield is already grounded at one end.
The disagreement between reference designs is about low frequencies and DC: whether it is wise to also make a direct DC connection at the device end, which creates a second ground path in parallel with the ground wire inside the cable.
The resistor-capacitor network is the compromise. The capacitor is a short circuit for RF and an open circuit for DC and mains frequency. For 4.7 nF the reactance is 1 / (2π × f × C): about 0.34 Ω at 100 MHz, but about 680 kΩ at 50 Hz. So high-frequency noise and the fast edge of an ESD event see a low-impedance path to ground, while low-frequency ground-loop current is blocked. The 1 MΩ resistor bleeds off static charge so the isolated shell cannot float up to a high voltage.
The capacitor should have a voltage rating well above the supply, since it takes the brunt of a discharge, and it must sit right at the connector with a short, wide connection to ground. A long thin trace adds inductance and defeats it.
Leaving the shell floating is the one option with nothing in its favour. An unconnected shield is a piece of metal capacitively coupled to the signal lines; it does not shield, it can radiate, and a discharge to the shell will arc to the nearest conductor, which may be a data pin. A ferrite bead between shell and ground is sometimes used to damp shield currents, but a bead is high impedance at exactly the frequencies where the shield needs a good connection, so it is a questionable choice unless testing of that specific product showed a benefit.
In products with a metal enclosure the picture changes: the shell should bond directly to the enclosure, ideally all around the connector opening, and the enclosure is then connected to circuit ground at or near the connector.
For a bus-powered device in a plastic case with no other cables, there is no second ground path to form a loop, so a direct connection from the shell to the ground plane at the connector is simple and works well; a great many small USB devices do exactly that. The RC network is equally acceptable and is the safer default if the device might later gain another connection to earthed equipment, such as an audio output or a probe ground, where a loop through the USB shield could carry noise current.
A practical approach is to lay out footprints for both: pads for a 0 Ω link and for the resistor and capacitor between the shell pins and the plane. Then you can choose during EMC and ESD testing without a board change. Whichever you fit, put TVS protection on VBUS and the data lines close to the connector as well, because the shell path does not protect the pins themselves.
Back to USB community
