Rugged USB card?

That is something I had looked at but found none that contain a CM choke. Strange, for RJ45 you can find them all over the place but not for USB. Without CM chokes this isn't going to work because you can't have super-fat TVS'es on D+ and D- due to their capacitance.

Then we might as well build our own PCI card :-)

You can. But even 100pF is small consolation with level 4 jolts.

Good idea. Although I have seen USB with NEC chips hiccup on ESD. And I fixed that with CM chokes. Those chokes are missing on nearly all USB gear. I guess that 10 cents per port is asking too much.

--
Regards, Joerg

http://www.analogconsultants.com/

One trick I use for protecting lasers that I need to modulate is to put two Schottky rectifiers back to back parallel, in series with a big cap. Capacitance isn't too bad, and they're bulletproof for my purposes. How about a Schottky bridge with a TVS across it?

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC
Optics, Electro-optics, Photonics, Analog Electronics

160 North State Road #203
Briarcliff Manor NY 10510
845-480-2058

hobbs at electrooptical dot net
http://electrooptical.net

That is what we already have, in the form of two low-capacitance TVS'es per port. Problem is, low-C TVS'es have a fairly soft knee when several amps or more lean into them. Per datasheet they can go to around 9V which turns on the substrate diode pretty hard. It's not that anything is dying but the data gets messed up and then this gear trips up. In our case tripping up is considered a failure even if recovering.

So there has to be something that stretches out the pulse, into a longer waveform but at much less in amps. A CM choke is just great for that but next to impossible to rework into a COTS board.

--
Regards, Joerg

http://www.analogconsultants.com/

Are you sure that won't mess up the data either and still cause problems? I'd test that first. If the ESD pulse doesn't destroy the port then it certainly is a software issue. Its really difficult to get USB drivers rock solid under Windows!

--
Failure does not prove something is impossible, failure simply
indicates you are not using the right tools...
nico@nctdevpuntnl (punt=.)
--------------------------------------------------------------

I'm talking about a big fat TVS running inside a Schottky bridge, which gets you the capacitance of one Schottky, with the current capacity of the TVS. It charges up to the logic level, so you don't load down your signals, but it'll clamp as hard as you want, and because it's Schottky, it turns on as fast as the TVS. If you can stand 20 pF, you ought to be good to at least a few amps--the 7 pF BAT54 gives you an amp for 10 ms nonrepetitive.

You can even add a 100k resistor to the supply to make sure the TVS is charged up all the time.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC
Optics, Electro-optics, Photonics, Analog Electronics

160 North State Road #203
Briarcliff Manor NY 10510
845-480-2058

hobbs at electrooptical dot net
http://electrooptical.net

That's already in the plan. It will not mess up the data but the end-of-packet signal in USB is non-diff. Which is very unfortunate because that limits the inductance a common mode choke (or cumulative ones) can have.

There are common mode chokes that are specifically made of USB purposes, with a clean 90ohms TX line behavior inside. At 480Mbps that stuff begins to matter.

That is not certain at all. You can have diode paths turn on that don't recover fast enough, stages that hit saturation too hard, and so on. All this can snuff out bits here and there and then you've got a data error.

With USB I had surprisingly little in terms of problems. It's just that ESD and susceptibility tests of the more brutal kind can throw you a curve because the HW layer hasn't been optimized too well for that.

--
Regards, Joerg

http://www.analogconsultants.com/

Yes, I've done tricks like that. Nowadays you can buy devices with all that stuff in there. For example, you can pull up pin 8 on these:

Things is, a genuine ESD pulse just snickers when it sees 20pF and blows right past that. Now you can also hang a cap in parallel but you'd still need a pair of resistors to make sure the substrate diodes on the D+ and D- inputs won't take a serious hit. A Schottkly is by no means a guarantee that it won't. It'll be (slightly) better than Si diodes but many chips go into funky modes when the substrate gets tugged at with

100mA or so. And that's only 1% of the spike when the mother of all ESD hits comes along where the poor BAT54 goes to 2-3V and screams from the pain.

By the time all this is on the board you might be better off (meaning less $$$) with a CM choke.

The other issue is that the repetion rate on ESD tests can be in the tens of msec. It always reminds me of the machine guns we used in the army.

--
Regards, Joerg

http://www.analogconsultants.com/

It isn't clear to me where you are putting the common mode chokes. Data lines? Power? If in the power, this could lead to latchup.

Lastly, if you have a product this touchy, I hope you have a redesign coming. If something doesn't work with bog standard USB, it seems to me like a problem.

They are only on the data lines.

We are almost there. Our hardware is done, it's the PC that hicks up and that isn't under our design control.

Try zapping USB cable into a PC with level 4 ESD at machine gun repetition rates during data transmissions and watch. You may be very surprised :-)

Hint: The EOP message is transmitted by pulling both data lines low, a violation of the otherwise clever differential transmission scheme. So when the ESD pulse goes negative ...

--
Regards, Joerg

http://www.analogconsultants.com/

OK, inductance on the data lines isn't a latch up issue. Still if I had my druthers designing the usb line driver, I'd rather deal with a large amount of capacitance than see inductance.

Have you ever considered calling one of these usb chip companies and explaining your issues. They would probably rather design a beefy driver chip than see customers add inductors. You are probably not large enough to get PME visits, plus your locations isn't exactly "industrial", but calling apps with suggestions won't go into the infinite bit bucket at most companies. If they can design a beefier part with the same pinout as the schlock part, then the Chinese board stuffers can make your rugged box.

Most app engineers rather go into product marketing, so they are likely to try to peddle your suggestions.

How does something get zapped with ESD when it is already plugged in. Or is this the scenario:

Your box is sitting there doing whatever. I have to assume it has it's own power supply versus using the usb power. Then you zap the box at the usb port with no cable attached.

I know know to design chips for ESD, but not systems, so I don't have the test setup visualized.

If I had my druthers I'd design a whole new PCI card for USB and be done with the problem. But we ain't got the time to do that.

Not enough time. My location is pretty well known because Intel is 10mi from here but my client is in the heart of electronics county. The app EEs make their regular rounds here. Still, we can't busy ourselves correcting mistakes the PC card manufacturers made in the design of their products. Some day we might roll our own card but not redesign a USB chip.

The rule is this: Everything that is or can be connected during normal use must be connected. Then the test lab zaps it almost everywhere a user can potentially get a hand onto. For example into a USB cable shield. A seasoned tech there knows the typical hot spots and goes for those.

I do not know much about ESD on a chip but I do know board level stuff. Have designed some things where a major bolt of lightning can pulverize the fence post 100ft down the road and the unit keeps on humming.

One has to make sure that by the time ESD reached the pin of a chip the current is well below danger level for a substrate jolt. Ideally so low that the chip won't even act up and will complete the rest of the data transmission as if nothing ever happened.

Regards, Joerg

There are USB 3.0 PCIe cards in the channel. They are 100% backward compatible with USB 2.0.

You are pathetic attributing problems being introduced BY YOU as being "poor designs" or other bullshit claims. You were even so stupid that you thought the power headers were added as a fix of some sort.

You claim ESD and EMI issues, but I think you are pretty much clueless as to the source of your problems, AND as to ANY solution, if you are the one trying to get there.

If you are trying to create an ESD failure mode, you will succeed every time.

ESD causes failure in ALL equipment idiot. You are never going to find a proper device unless you cage everything from end to end.

I have serious doubts.

ESD and lightning strike abatement are two different animals, dork.

It doesn't matter that lightning is electrostatic.

Now I get it. You are zapping the ground. Doh! I've only heard of this testing, never saw it in the flesh.

When HP was HP (as in everything had to be top notch because people judge you on the crappiest thing that you put your name on), I was in a meeting with them and they wanted 25KV ESD. I think I quoted 5kV. They didn't blink and just said everything leaving the factory will do 25KV even if we have to make it do that ourselves.

Here is a surprise. Nobody cares what you think.

But that isn't a surprise.

Jamie

The usual ESD guns only go to 10kV contact or less and 15kV air. I tend to design for max zap, turned all the way up. It's actually not that difficult and not earth-shatteringly expensive to make everything pass. But it can be very difficult if you come in where it's too late for layout changes or when dealing with OEM boards.

--
Regards, Joerg

http://www.analogconsultants.com/