USB shield at device.

Ground the shield at both ends, to the PCB ground plane (which should also be the case ground.)

That's because it makes no sense. If you want a good DC connection and a good AC connection, just connect them.

I don't have the luxury of a PCB ground plane, and I suspect that's true of a good many USB devices. In any case, the USB data signals reach the microcontroller within 5mm from the USB receptacle, and none of the rest of the circuit operates at high frequencies. The case is (OK, will be) plastic, which is also pretty much the norm.

This is not a commercial project, and I can get away with any amount of EMI that doesn't cause me problems directly, or result in a regulator arriving at my door. Still, I like to do things right if possible.

Sylvia.

This is good against electrostatically connected interference.

This is a bad idea, if the cable ends are connected to different grounds, creating different ground potentials and hence large currents to equalize the potentials. The current is often at mains frequency and harmonics.

Apparently the resistor is there to equalize any static electricity build up if one or both ends are floating.

The small capacitor at least breaks any mains frequency interference loops.

If the device has a metallic enclosure, the cable shield should be connected to the enclosure directly or through an impedance, not to the signal ground.

Well, just connect the cable shield to the uP ground and the cable ground wire.

Connect it to device shield. (in other words option a)

I'll sometimes put in two pads for a jumper. Then you can connect, play around after the fact. (Short/ open or an R//C.)

George H.

I always use (a).

We *want* to equalize those potentials! The shield helps do that.

When you connect a signal generator to an oscilloscope with a BNC-BNC coax cable, do you strip out the shield on the scope end of the cable?

This applies both to 120/240 V antiphase feeds in the US as well as to three phase 230/400 V systems in the rest of the world.

The principal idea in both systems is that with _balanced_ loads, the neutral wire current cancels and the neutral wire potential is the same as the ground electrode potential all over the place.

Unfortunately this is not so in the real situation.

Assume an apartment building with two stairs with separate feeders along the stairs. There are going to be unequal voltage drops in the phase as well as neutral conductor at both stairs.

At some higher floor, measure the mains neutral/ground connector potential against the potential of the same floor of the adjacent stairs. With absolutely equal and symmetrical load, the neutral/ground potential would be equal.

In practice, this is not the case. Connect an unisolated RS-232 or USB cable between the stairs and the shields might quite well be 1-5 Vrms at different potential.

Guess what is going to happen if you connect the shields together ?

With a bad unbalance load between the stairs feeders, a huge amount (several amperes) may flow from staircase 1 to staircase 2 neutral through your signal cable shield, causing _a_lot_ havoc to your communication.

For this reason, one should avoid connecting cable shields together.

As long as you are using a single mains outlet to feed a grounded extension cable for these two devices, the device ground potentials should be within a few millivolts.

I experienced a 60V !!! differential between two GenRad buildings about 100' apart, so I used a transformer at one end of the coax (Manchester encoded) to avoid ground currents. ...Jim Thompson

Not a surprise at all. In some countries the electric wiring conventions can be quite "relaxed".

These days, I prefer to use 10/100/1000BaseT Ethernet simply because of the simple (standard) galvanic isolation.

However, due to lighting protection and hence the different ground potential rise I would recommend using fibre links between buildings.

I have been working with industrial applications all over the world (including the tropics with high lightning rates), first my recommendations about isolation was deemed too expensive, but after a few weeks or months, my recommendations were accepted without objections :-)

USB data signals d+ and d- has CM range 0.8V to 2.5V and diff level of 200m V, so it has some protection from ground loops, but not much. Doesn't the U SB standard define maximum cable length so a connection to ground planes on each side is allowed since they assume small CM voltages?

Cheers

Klaus

Same here, now. My 60V case was ~35 years ago.

...Jim Thompson

Den fredag den 24. oktober 2014 23.20.00 UTC+2 skrev Klaus Kragelund:

maximum length is 5 meters. There is a ground wire in the cable so the grou nds will always be connected, it is whether the chassis are connected

-Lasse

In practice, where did you measure those figures ?

If you had a 230/400 Vac feed and run it through a six pulse rectifier and have two capacitors across the DC link, the mid-point potential varies by 30-40 Vac from nominal ground potential.

For a US 480 V feed you are close to 60 Vrms midpoint to ground potential.

Well, answer the question: go you break the shield in BNC cables?

We don't connect power neutrals to circuit common or to chassis; we connect circuits and boxes to GROUND.

If the commons of two USB units have, for some reason, some voltage difference, that difference will be applied across the super-skinny ground wire (pin 4) inside the cable. If that wimpy wire can't equalize the difference, we're going to get nasty data errors and maybe damage. The cable shield is a much lower resistance than that skinny ground wire; wouldn't we rather use the shield in parallel with that tiny wire to short the chassis potentials together?

Why would you WANT the shield, at the device end, to be a different potential from the ground wire? That's terrible for DC, AC, and RF.

The power outlet grounds on opposite sides of my office are about 50 mV RMS apart at 60 Hz, with a couple hundred more mV of fast trash. USB is not a long-distance bus.

Right. You don't run USB between buildings.

I have an outdoor webcam at our cabin that is 75 feet from the CPU, with two active USB cables and two passive cables strung together. Works fine, but the cam is not grounded.

Our philosophy: ground everything.

Local "ground" to local "ground" and it _was_ 60Vrms... fried the first cable I plugged in, with an enormous flash :-( ...Jim Thompson

Neutral current is assumed to be either the same or somewhat more than the maximum current in one phase. Max voltage drops along cables can be computed from this.

No-one connects signal grounds to the damn NEUTRAL !!! Doing that would be CRIMINAL !!

No-one connects signal grounds to the damn NEUTRAL !!! Doing that would be CRIMINAL !!

You need kicking off this NG pal.

Neutral current is not involved unless there is a dangerous fault in the AC supply wiring.

.... Phil