USB EFT

[daddario]

Hi!
I've been battling with an elusive issue of a small number of boards populated with Atmel's SAME70s have been coming back from the field with dead MCUs after a few years of service.
Everything seems to point towards having something to do with the USB - on many cases the USB TVS ( Würth's P/N 824015 ) is dead short in all directions and
so are the USB lines on the MCU. On the board that I currently have on hand the TVS has survived, but USB lines measure 15 ohms to gnd, VDDUTMII and VDDIO about 20
to ground and 10 between each other.
It's unlikely that it's an ESD issue as the setup is such that the USB is connected permanently to a computer. The computer seems to have NUP4301MR6 ESD diode on USB
and I don't have any info that USB ports have died in those instances.
It's an industrial setup and unfortunately the USB cable can in some cases share about a metre of trunking with cables carrying power - be it mains feeding an SMPS or,
in an even worse case, low voltage but several amperes of PWM modulated power (around 1kHZ, I believe). Unfortunately that's out of my reach to have it done otherwise.
The ground plane of the board is heavily bonded to the sub chassis, but proper bonding of that to mains earth can vary in quality between installations.
The shield of the USB connector is connected to ground plane via a 470nF X7R 0805 capacitor and about 5mm length of 0.4mm trace.
The TVS diode is located against the USB connector and so are capacitors and the Vbus inductor.

What could be potential design improvements in this case?
Different USB shield connection strategies - several capacitors in parallel, direct to gnd plane?
Lower Vth TVS diodes on data lines?
Something completely different?

I've also attached the schematic of the USB part, no other components between that and the MCU (the SAME70 has internal series resistors).

Datasheets:
https://www.we-online.com/catalog/datasheet/824015.pdf
https://www.onsemi.com/pdf/datasheet/nup4301mr6t1-d.pdf

[wraper]

Throw out 470nF cap and make a good solid connection between connector shell to GND plane. You can populate a 0 Ohm resistor but the best would be to respin the board. Here is what an EMC compliance professional says: https://www.unit3compliance.co.uk/hdmi-more-like-hdm-why-thoughts-on-cable-shield-grounding/

I’ve seen plenty of interesting EMC “solutions” over the last several years to deal with radiation from cables. A common one is to separate the shield ground from the signal ground with some combination of components (beads, capacitors, resistors). This approach appears to be particularly common on industrial touch screen display modules for some reason. This is (in 99% of cases) a bad idea. I’m not sure what you are hoping to achieve by this and, probably, neither are you 😉

I’ve yet to come across an EMC problem where floating or not connecting a shield ground has improved the situation.

[jkostb]

Since you are in an industrial environment, you need to ensure that the shield on your USB cables are properly terminated. Most USB cables are produced in China and the shield is terminated by use of a  pigstyle connection (which is easy to produce). Shield must be terminated 360 degrees around connector. You have to cut the cables in order to see whether shield is terminated properly. If you have a metal housing then you must ensure that the shield is directly connected to your housing. Noise, surge must be diverted to the housing and not through the ground plane of your PCB where it can disrupt your circuit. Problems start when you have a plastic housing. In this case the currents will go through your ground plane. Please note that if you connect the shield directly to  your groundplane then there is risk of ground loops. This can cause other problems. For this reason I would never select USB in an industrial environment where cables run through nasty EMI environments.

You wrote USB EFT in topic. Do you think that ESD protection chip is destroyed by EFT? EFT has not a high energy level and it is unlikely that it destroys your ESD protection chip. You wrote that power cables are running close to the USB cables. Is it not more likely that a surge event occured on the power cables and that this is capacitively coupled into your USB cable? In that case you can better fix the cable routing

[wraper]

Please note that if you connect the shield directly to  your groundplane then there is risk of ground loops. This can cause other problems. For this reason I would never select USB in an industrial environment where cables run through nasty EMI environments.

Ground loop will (or won't) be there regardless of the shield because there is also a GND wire. However by compromising the shield connection you only cause problems (both EMI and EMC) and cause ESD susceptibility.

Most USB cables are produced in China and the shield is terminated by use of a  pigstyle connection

If GND shield connection within the circuit is trash, a good cable won't save it.

[Benta]

This is only a partial schematic, but I have major problems understanding the rationale behind the 470 nF capacitor.
It's purpose eludes me completely.

[daddario]

Throw out 470nF cap and make a good solid connection between connector shell to GND plane. You can populate a 0 Ohm resistor but the best would be to respin the board.

If I can, I'll try tomorrow to do some A/B measurements bewteen the 470nF and direct short versions in a near real-life setup. Generally I like the idea.
Respinning new production runs isn't an issue but we also have to come up a solution for RMA cases - as they aren't exactly €20 boards, everything that can be repaired have to be - ideally improved.

Since you are in an industrial environment, you need to ensure that the shield on your USB cables are properly terminated. Most USB cables are produced in China and the shield is terminated by use of a  pigstyle connection (which is easy to produce). Shield must be terminated 360 degrees around connector.
You wrote USB EFT in topic. Do you think that ESD protection chip is destroyed by EFT?

Unfortunately I can only inspect what cables the units are shipped with today - I've not cut the connectors open but the cable is foil + braid. CCA conductors - as it it is nowadays. There's always a case that at some point in the past different cables have been used or that the customer has used their own cable.
As for semantics - I used EFT as a term for any ingress with a fast edge coupling in.

This is only a partial schematic, but I have major problems understanding the rationale behind the 470 nF capacitor.
It's purpose eludes me completely.

It is indeed a partial schematic but covering anything related to USB - if there's anything else useful I can provide I will happily do so.
Its purpose is to AC couple the shield to ground - it's a very diving topic indeed how it should be done. App. notes from the big players actually recommend coupling through capacitors in the few nF range or even... inductors(!?). True - these notes are usually written from the perspective of domestic appliances. The value itself was found years ago empirically - in a realistic setup there was a low-quality SMPS supply nearby, which was the source of nearly all the noise. We started probably from 1/2/4.7nF, increased the value until we got an error free bus and then increased the nominal maybe two times.

As for the ground loop issue - I've seen an application in instrumentation equipment where the USB shield was directly connected to ground, but the Vdd(actually a symmetrical setup for Vcc also) was coupled to ground through a ferrite bead. There was also a capacitor parallel to the bead, which is self-explanatory, but the idea of a ferrite is interesting. Obviously it's of no help if there's a DC imbalance between the nodes, but could help for CM noise and transients.

[T3sl4co1l]

Calculate the voltage drop along that 5mm trace, say for a typical EFT pulse, or ESD, of appropriate kV.

The shield must be hard grounded.

You don't have to do it galvanically, but it makes it a whole hell of a lot easier to do.  A good RF ground through bypass caps will need 4+ times more caps than you have here, and much shorter traces (and preferably multiple vias to save another nH or two).

The cable and connectors similarly need to be well (contiguously) grounded.  Cheap cables may have wire links between shield and shell, NFG (no something good).  Poor fitting connectors may have few if any shield contact points; check that the fingers are making contact.  Ferrite beads on the cables at both ends may help, though given the magnitude of this class of transients, it's not going to actually filter anything, the point is just to increase the impedance a little bit to reduce the peak currents through the shielding/grounding paths, thus reducing the CM error at the circuit.

As for cause, transients and poor grounding seem plausible enough; the environment must be truly horrendous indeed to apparently destroy a TVS.  Improved shielding will help, but this can only be ensured so far.  At some point you may have to go to a custom cable solution (say, circular connectors with solid braid coverage), or isolate the interface wholesale (USB isolator devices are available, though they often have high capacitance -- to reduce emissions -- which negates the whole purpose with respect to CM transients, so beware; a low-C isolation can be made on-board in any case).

Tim