Odd failure on GND-GND resistor

[Phaedrus]

So we have this system with about 10 boards, and suddenly all of them failed at once. All of them were burned in the same spot.

The boards have two inputs, USB for data, and +12VDC power. The two inputs have separate grounds, which are connected by a single SMD resistor. I'm not sure on the value of this resistor, as they were blown up on every board, and there is no marking. Best as I can figure, this was an attempt at noise isolation, to stop noise in the PSU ground from affecting the USB and vice versa.

The problem is, I don't see any way that this resistor could have failed. When I mean burned, I really mean it: it reflowed the solder, and shot flame and smoke. How the hell do you get that much potential between PSU and USB ground without anything else failing (as far as my testing can show)?

EDIT: PSU ground is connected to chassis ground. USB ground is isolated by this resistor.

[AndyC_772]

Maybe they're "0 ohm" links, intended to ensure a single point connection between USB and PSU ground.

They're never 0 ohms, of course, and can be surprisingly resistive. They'll certainly dissipate a fair amount of power, albeit briefly, if (say) you were to short the PSU +ve rail to USB ground. The full PSU current would go through the set of links, heating them all up until the first one goes, then the rest carry a bit more current and get even hotter, and the thermal runaway effect pops them all.

[Phaedrus]

That seems possible. However, the 1000W PSU has short circuit and over current protection. OCP (secondary side) should trip around 130mOhms, SCP (primary side) around 50mOhms. The impedance to ground has to be greater than that or the protection circuits would shut the PSU down within a couple milliseconds.

[Niklas]

Does this happen when you attach something to the USB port? If so, is that unit powered by something connected to the AC outlet?

Everything worked just fine with the battery powered laptop but when the laptop charger was connected, the problems began...

Grounding in cars can also be a problem. I had a thermocouple to voltage converter that failed but only when the motor was cranked. The unit was powered from the OBD connector and the sensor was connected to the transmission to measure the oil temp. During cranking, the starter motor current caused a difference in potential between the OBD GND and the motor/transmission GND. The result was a cooked thermocouple IC and the solution was to use an isolated sensor.

[tom66]

That seems possible. However, the 1000W PSU has short circuit and over current protection. OCP (secondary side) should trip around 130mOhms, SCP (primary side) around 50mOhms. The impedance to ground has to be greater than that or the protection circuits would shut the PSU down within a couple milliseconds.

How big are the 0 ohm resistors? I've seen 100mohm easily for a 1206 zero-ohm link. Add in some trace, via and plane impedance and you have a problem...

[ManateeMafia]

Did you measure the potential across the burned component?

Is the PS and laptop plugged into the same outlet?

Are you using a usb hub between all of the boards? If so, something like this https://www.olimex.com/Products/USB-Modules/USB-ISO/  may be a temporary solution.

[Phaedrus]

That seems possible. However, the 1000W PSU has short circuit and over current protection. OCP (secondary side) should trip around 130mOhms, SCP (primary side) around 50mOhms. The impedance to ground has to be greater than that or the protection circuits would shut the PSU down within a couple milliseconds.

How big are the 0 ohm resistors? I've seen 100mohm easily for a 1206 zero-ohm link. Add in some trace, via and plane impedance and you have a problem...

I think we're on the right track. It's only an 0805 package. If +12V shorts to the USB GND, or if USB +5V shorts to the PSU ground, this could give significant potential across that resistor. It's plane-to-plane by the way, no trace or via.

There was a USB hub between boards, it's being investigated as a potential cause as well.

[lgbeno]

Yes I'm sure that your isolation was compromised and some current (or all) return current from the power supply passed through that smd resistor.  I^2*R=P even with a few 100 mOhm and a few 100 mA, the 0804 resistor will go poof.

[free_electron]

not zero ohm. more likely a ferrite bead....

[lgbeno]

not zero ohm. more likely a ferrite bead....

Ferrite beads shoot flames when they exceed their rated current, it is pretty spectacular.

Reading the thread again, I'm certain it is a isolation problem.  First thing to note, this is a transient effect so you might not be able to catch it easily.  You've probably already realized it.

IMO, it's not about measuring potential, you want to measure impedances. 

First thing I would do (if not done already) is draw a diagram of your system with each connection.

Then identify which "ground" each block is referenced to, and where it's power is coming from (and returning!)

Then make sure that there is high impedance between "isolated" grounds.  You can just use a dmm at first to diagnose gross problems which it sounds like you have.  Otherwise this is also tested with a high pot or leakage test.

My thought is that you said you have a power supply and USB connection.  Normally current for the system comes from the positive terminal of the ps connection and returns to the negative.  That's great but USB has a ground too and if that ground is somehow connected back to the power supply negative terminal through a different system connection (or a black box like a USB hub) then there is a SECOND return path (I intentionally did not use gnd here) will share some of that return current. 

Maybe that is ok normally because the ps connection is very low impedance and hence takes most of the current but if that first power supply connection has an issue (vibration, pinched cable, some other failure) then all of the return current will flow through the USB connection.  If any element of that connection cannot handle that, something will smoke like said resistor or bead.

Power supplies are wired, especially the concept of "earth" or "protective" ground vs "neutral". Some supplies have a short from earth to the output negative (ones that are not labeled as "medical").  So even if you have two "isolated" supplies, they can share a ground through that "earth" connection.

I'm sorry it's terribly confusing and no simple way to explain, I've designed cardiac floating surgical devices before and we spent many months testing and certifying isolation so that's why I'm so familiar, it took me a long time to be proficient.

PM me with questions if you don't want to share diagrams with the rest of the forum, I would be glad to help.

[VK5RC]

To me it sounds like a ground issue,  a fuse gone somewhere else can lift ground potential,  best of luck.  Make no assumptions!  A similar issue in a 10GHz home brew transverter took me about 3hrs to find. AAAAAAARRRRRRGGGGGHHHHH.

[Phaedrus]

Gah! I was stupid. The part was numbered "FB27" -- ferrite bead! It's obvious now. It appears to be identical to some other ferrite beads on the +12V input; I don't have an LCR meter to measure inductance, but my Fluke shows about 1.5 ohms resistance.

Here's what I've been able to sketch of the inputs of this board. I'm still unsure about that diode, but otherwise it's about right.

[Phaedrus]

I think lgbeno is on the right track. That ferrite bead was the lowest impedance path between PSU GND and USB GND, and our PS cable was burned. We thought that was a symptom; but maybe the cable burned first, interrupted PSU GND, and caused current to return through USB.

Question is, was this a board design fault, ie. should the impedance between the two planes have been higher, or should the inductor have been rated for more current? Should they even have been connected at all?

[tom66]

What is the purpose of the 270 ohm resistor in parallel with the ferrite  bead -- since the bead has near zero resistance?

[Phaedrus]

What is the purpose of the 270 ohm resistor in parallel with the ferrite  bead -- since the bead has near zero resistance?

I have no idea. It isn't another ferrite bead, it's a smaller package (0603 vs. 0805 I think) and measures ~270R. A couple nF of capacitance according to my DMM, but it's definitely a resistor.

It's on the opposite side of the connector. The ferrite bead was directly adjacent to the +12VDC connector, and attached to its ground plane there. The resistor is on the opposite side from the +12VDC, and is connected to the same ground plane (or at least <0.5 ohms from it).

[lgbeno]

I think lgbeno is on the right track. That ferrite bead was the lowest impedance path between PSU GND and USB GND, and our PS cable was burned. We thought that was a symptom; but maybe the cable burned first, interrupted PSU GND, and caused current to return through USB.

Question is, was this a board design fault, ie. should the impedance between the two planes have been higher, or should the inductor have been rated for more current? Should they even have been connected at all?

The answer is "it depends" however I would say that this is not a good practice.  It can lead to emc problems and also as you experienced, safety issues (risk of fire).  At a minimum, I would expect a fuse in series to blow before something flames up.  If it wasn't that bead, then it will be the USB cable or the USB hub.  A fuse is a pretty ugly patch too.  First who has ever heard of a fuse from gnd to gnd and second, it may blow more often then you like and cause some real wired intermittent issues.

It really depends on your system, if USB is really chassis gnd and power has a different gnd and the two side are completely isolated with optos or something, you should have a very high impedance between the two planes at all and they should pass some high pot test.  IMO this is the best approach.

I've seen frequently that high voltage or safety rated ycaps are used to connect gnd planes but that would do the opposite for emissions that your bead is doing.  Another option is to tie the planes hard together.  Then use a fuse on USB gnd connection at the connector.  You will need to retest emissions and if there are issues, clamp a ferrite on the USB cable.

GFCI is another option for protection I don't see that a lot on low voltage dc circuits like this.

What market is the product in or is it just a hobby thing?  That will also determine how crazy you want to get at finding a solution.

[ignator]

My experience of a similar (I'm guessing) failure on a ground used for analog reference to a flight test pallet. On an aircraft, I had an avionics box that has a flight test connector on it. The flight test data acquisition system has an external +28VDC input power supply that distributed +5/+12/-12 to the flight test pallet.  The flight line technician connected the ground return to the aircraft (not the pallet). My avionics box uses chassis ground as DC return. It was connected via the flight test connector to the test pallet. The analog ground reference connection (to the test pallet) became the return path for the DC power. This analog ground reference smoked as it was not designed for the ground current of the test pallet.

I would check if your PSU ground return failed, and this ferrite bead "took it in the shorts". You indicated 10 of these boards in your system, so would would check where the common ground splits into the many paths for the boards. There is some path back to your PSU, which was probably intended to be an AC noise ground connection.

[Phaedrus]

I think lgbeno is on the right track. That ferrite bead was the lowest impedance path between PSU GND and USB GND, and our PS cable was burned. We thought that was a symptom; but maybe the cable burned first, interrupted PSU GND, and caused current to return through USB.

Question is, was this a board design fault, ie. should the impedance between the two planes have been higher, or should the inductor have been rated for more current? Should they even have been connected at all?

The answer is "it depends" however I would say that this is not a good practice.  It can lead to emc problems and also as you experienced, safety issues (risk of fire).  At a minimum, I would expect a fuse in series to blow before something flames up.  If it wasn't that bead, then it will be the USB cable or the USB hub.  A fuse is a pretty ugly patch too.  First who has ever heard of a fuse from gnd to gnd and second, it may blow more often then you like and cause some real wired intermittent issues.

It really depends on your system, if USB is really chassis gnd and power has a different gnd and the two side are completely isolated with optos or something, you should have a very high impedance between the two planes at all and they should pass some high pot test.  IMO this is the best approach.

I've seen frequently that high voltage or safety rated ycaps are used to connect gnd planes but that would do the opposite for emissions that your bead is doing.  Another option is to tie the planes hard together.  Then use a fuse on USB gnd connection at the connector.  You will need to retest emissions and if there are issues, clamp a ferrite on the USB cable.

GFCI is another option for protection I don't see that a lot on low voltage dc circuits like this.

What market is the product in or is it just a hobby thing?  That will also determine how crazy you want to get at finding a solution.

It's a commercial product, I'm not under NDA on the board itself, but I am on its application.

I imagine best practice would be that inductor plus, say, a 270R resistor in series. Reasonable impedance, but still low enough to prevent undue potential in most cases.

[lgbeno]

I imagine best practice would be that inductor plus, say, a 270R resistor in series. Reasonable impedance, but still low enough to prevent undue potential in most cases.

Maybe... This wouldn't be my first choice but  it could work.  To be safe, at minimum that 270r resistor should be able to dissipate 500mW.  You should also test your system with that configuration, that's the only way to be sure.  First configure it and test under normal conditions, then remove the PS gnd wire and see what happens.  My guess is that nothing will flame but there would be a noticeable shift in potential between the two planes, maybe that is ok.

[Phaedrus]

What would your preferred way be? There were some errors in my schematic by the way, I'll post the updated in a bit.

[Phaedrus]

We sorted it, by the way. Four boards pulled power from one branching cable on the PSU, instead of just two like it was supposed to be wired. This overheated the PSU connector until the plastic melted. Two GND pins lost connection and a third was darkened, it was up around 20 ohms end to end of the cable, and that was cold. It was likely higher under load. With the high impedance ground the current shunted through the ferrite beads and nuked them. We're going to pass the data along to the board manufacturer and suggest some improvements, alongside the RMA.

[lgbeno]

Awesome, nice debugging work!  I would say then that this falls into the category of "doesn't matter" under normal conditions.  It is a design enhancement for this particular failure mode that you stumbled upon.  Now that it is understood what happened, you can decide the best solution.

[VK5RC]

Good detective work, I recall thinking how odd it was that one radio I owned had two fuses, one in the active another in the earth UNITIL i had my first earth short, then I started to question why it isn't the norm.