USB shield at device.

Ground loops is ground loops.

So an open ground on one end will cause violation of FCC? Might cause one?

Have a ground loop on USB cause bad performance.

???

Everybody uses that same 4-transistor circuit Mackie started in the '90s, outside of upmarket stuff. Just the transformer part cost alone for a Hammond 140NEX is $50ish bucks...

I haven't seen anything new that was xformer-coupled in a long time.

--
Les Cargill

A 5 m long quarter wave unipolar antenna will have the lowest resonance at 15 MHz as well as the 3/4 wave resonance at 45 MHz and

5/4 resonance at 75 MHz. At these frequencies the antenna is resistive and capable of dumping some real power into the losses at the "load" grounded end.

At other frequencies, the antenna is more or less reactive and re-radiates more or less of the captured power immediately to the surrounding space, without flowing and dissipated into the load.

Even with a resonant antenna, the impedance level along the antenna is not constant. The current is largest and voltage smallest at the load end (i.e. low impedance), while at the open end, the current is very low and voltage high i.e. high impedance of several kilo-ohms.

The USB twisted pair transmission line has an impedance of about 100 ohms, so how much interface do you expect the high impedance (kilo-ohms) is capable into the balanced data line ?

To add some real experience to all the theories posted in this thread:

I did DO-160 tests for EMC with a self developed system consisting of a CPU system and a USB device. With the shield unconnected (one end as well as both) we had a large, narrow peak at 120MHz, way above allowed levels. Connecting both ends of the shield to the device case which was also connected to signal ground on both sides there was no detectable peak anymore. The 120MHz came clearly from USB itself, some base frequency related to the 480MBit/sec.

The argument about voltage differences if the device is powered be some other supply do not make much sense to me. USB is not intended to be routed over long distances, max cable length is 5m. If there is such a big difference between the device cases/shields which very often are also connected to signal ground (more less directly) USB with its small signals will not work. You might even kill the USB ports. A customer managed to do this. He used an ungrounded lab supply. When unconnected the ground was at something like 120V-AC though capacitive coupling in the transformer (240V-AC mains). Plugging the USB device into the USB port of the test computer killed the port.

--
Reinhardt

No, they aren't. USB is not low-level analog, though there is the issue with sound cards (hardly the reason USB exists). The needs are very different.

A good possibility, sure. JL proposes that the grounds always be connected at both ends *because* of FCC requirements. If he said "usually", I'd agree. There are other needs, however. Engineering is about tradeoffs.

You're lucky. I see problems all the time, particularly with laptops and USB sound cards. It's one reason I got an isolated hub at work.

It's *very* common.

You're an idiot.

You haven't dealt with *any* audio, then. You really should stick to what you know.

I'm not sure I understand. If the AC was reaching the equipment through capacitive coupling in the transformer there should have been a pretty high impedance to any appreciable current flow. Even the fine wire inside the USB cable should have been able to conduct enough current to equalize the grounds to within a small voltage.

I expect what you saw was some sort of true fault which allowed a much higher current to flow and so a higher voltage to appear on the port.

--

Rick

There re multiple resonances at additional frequencies, well into the cell phone sort of range. Not to mention the wideband effect, namely treating the shield as an inductor/delay line, happy to twang whan hit with fast spikes, like ESD discharges.

If there is RF voltage on the ungrounded end of the shield, it is capacitively and inductively coupled into the wires inside. The effective coupling impedance will be something less than 100 ohms. So the RF voltage and ESD spikes get jammed into the electronics inside.

If you ground the shield, that doesn't happen.

--

John Larkin         Highland Technology, Inc 

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

EMI line filters do the same thing: couple half of the AC line voltage into the device ground. The cable shield can short that out, but the signal wires maybe can't.

--

John Larkin         Highland Technology, Inc 

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

There are low impedance resonances at odd multiples of the quarter wave resonance frequency and very high impedances at even multiples of that resonances.

What is wrong treating it as a transmission line ?

Interesting.

I suspect it's less common than it used to be.

There's this whole range of prosumer stuff being used now. Some of it is pretty good. Very strange...

We knew that. And your mama dresses you funny.

People just have options without transformers now.

I'd a never figured this say thirty years ago, but ...

I have dealt with what I have dealt with. For the record, I have never

*designed* any.

I'd say it's closer to observer bias. The last nice console I used ( a D&R ) was transformerless; many are now. There's just a heck of a lot more offerings now.

As proper solid state faux-balanced inputs became available, people moved away from transformers.

I am pretty sure Neve and SSL use transformers still but those are sort of legacy marques. Indeed; transformers sorta get sold now *as* a "vintage" accoutrement.

--
Les Cargill

I understand.

So we had a broken wall-wart that put hash into the case ground of a device. This caused performance problems on USB for this device.

If you put a scope on the ground of the device against the ground of the PC it was to be connected to, you saw a lot of hash.

I presume case ground and signal ground were isolated.

--
Les Cargill

Isolated warts are usually plastered with every UL/CSA/FCC/VDE sticker imaginable. And most shoot huge amounts of switching spikes into their outputs.

Consider a wart-powered USB hub. Would anyone suggest that the cable from the PC to the hub have an unconnected shield at the hub end? And unconnected shields at the ultimate device ends?

--

John Larkin         Highland Technology, Inc 

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

Less common, sure. Amplifier outputs aren't transformer coupled. They're still *very* prevalent in professional audio.

Don't know anything about "prosumer".

Ratty old clothes, too. ...since she's been dead for six years.

Sometimes.

They'll never go away.

For many uses, sure. Transformers still have many attributes that silicon doesn't.

Wrong again but you knew that.

IME they always are in PCs for some reason (a good reason, presumably, what with all those 100's of amps at multiple GHz).

--

John Devereux

They are only the same at DC which is the whole point of the concern with ground loops.

--

Rick

Faux-balanced inputs? What exactly is that? If you mean they are balanced, but not floating, ok.

--

Rick

It... depends. It's gotten all very strange over the last few years.

It's basically higher-performing MI store stuff.

I have one sitting here to my left; noise is -102dB and THD+N is like unto 0.001%, but it's low enough that I don't trust the cables I use...

It's not capital equipment. $500.

My apologies; I had no way of knowing.

Agreed.

Of course they do. There's nothing better for iso and they impart a sound.

That's... kinda how we got started here to begin with.

I mean something like this:

It's all about packaging options these days.

--
Les Cargill

Basically.

--
Les Cargill

Yep.

Well put. Probably not. This is where isolation as krw mentioned would be the thing to do.

--
Les Cargill

If that was the case with the wart unloaded, it should be removed off the store shelves. Only hardwired supplies should do that unloaded out of circuit and not in circuit. But as you say, it's a wall wart, so it should not be doing that.

Well, most hubs I've seen don't do much in the way of a grounding the shield inside, but more to your point, the USB cables are generally shield connected on both ends anyway, that is, if you have that type of cable. I have a couple over here that has no shields at all, just twisted pairs.

If the manufacture of the product being connected to the end of the cable sees a problem with DC ground loops, which can happen of course, but still needs an RF shield, they usually capacitor couple the shield to the device ground and also may include a series R to drop Q and introduce some resistance in a high RF area.

A DC ground loop via the shield can damage the front end of sensitive devices. When unexpected currents appear in between chassis, currents can pass through the shields, enough so that the magnetic coupling can pulse this current into the inputs and outputs of the devices on each end of the cable. In the industrial world, we always connect shield on one end of a data line and bring the shield close to the end of the leads at the other end, taped off.

I have always past comments to the guys I work with about checking the path of the SG (signal Ground) connection to device to true ground and if there does not seem to be a DC path, then simply connect the shield to the device and the other end. The problem is that a lot of devices out there do not supply a separate SG connection point, even if they do simply tie directly to chassis. When it's chassis ground only being supplied, we make sure one end isn't DC connected.

Have you ever tried putting a scope on the inside conductor of a coax and then pulsed some current through the shield?

Btw. I use shielded grounds in my shop for my computer, radios, amps etc, otherwise, I'd have so much RF noise I wouldn't be able to use any of my RF gear, due mostly to the computers.

I am sure you know all of this and I most likely wasted my time. ;)

Jamie