PSA: ONTi L3 Switch - Circuit board charring/burning

[DavidWJohnston]

I noticed the distinct smell of burning electronics in the lab. It took a while to figure out where it was coming from, but I eventually found it.

The circuit board on my ONTi L3 switch is charring in a very specific spot:



It's very strange because the problem area is nowhere near the hot transceivers or switch ASIC, it's at the corner of the board, on the underside, below the console port.

At first glance it looks like there was an over-current situation on a trace leading to a via at the edge of the board.

That via is not grounded-out to the console port metal case, though I did measure it after it cooled-down. I haven't yet done a full analysis.

The switch was in service for months with no issues or physical changes. It just suddenly occurred with no apparent cause. The switch still works.

This is where I bought it:

https://www.aliexpress.com/item/1005005777395543.html

I have already contacted the seller, we will see what they say - I just wanted to warn anybody who has these to watch out for this. I doubt it could actually cause a fire, but it sure does stink. I still feel nauseous from smelling around trying to find where it was coming from.

If I learn anything new I'll post an update.

 

[klui]

Is that a cut on the PCB? Is there any burn marks on the bottom of the case? Does the console port still work?

 

[DavidWJohnston]

@klui - It's not a cut in the PCB, it's a diagonal burn mark apparently following a trace. The console port and everything else still works. There is a discoloration on the inside of the case. It is a gold colored area which doesn't rub off with a finger:



This is the response from the seller:

"My friend, this is because the device is not grounded, so it causes static breakdown, which will not affect normal use. You can give him a grounding treatment, and then in the re-use."

Somehow this response doesn't make sense to me. I think they are talking about this grounding screw on the back of the case:



Which I didn't have connected - But how could that make any difference? The whole metal case and metal SFP cages are at the same potential anyway, how could adding a ground to the case make a difference?

I turned on the bare board with everything connected, hanging by the network cables, and nothing is burning. No hot spot. And it works 100%.

So I don't know what's going on, but I guess I'll clean up the burned area, put some conformal on it, ground that chassis screw, and see what happens.

 

[mtg]

Possibly potential from two unearthed switches and one is dumping POE/mains voltage into that port. Or was nothing plugged in? That's even more scary.

IDK, personally once something starts burning or has the potential to burn if it's not bolted to ground...I'd stay far away!

 

[DavidWJohnston]

I have several things plugged-in, including metal SFP DACs connected to multiple other devices; some with grounded PSUs, others with ungrounded DC jacks. There is no PoE in this part of the network.

Even without the chassis ground connected on the burned switch, it is FULLY grounded to all my other equipment through those metal SFPs, including the ones with only DC barrel jacks. I measured for stray voltages between everything that has exposed metal - Everything is bonded, nothing is floating.

I connected the chassis ground as the seller suggested, but there is no current flowing through it, and it's bonded anyway through those DACs so after connecting it nothing actually changes grounding-wise.

I scraped away the charred mask and FR4, and I can see the remnants of a vaporized trace following that diagonal burn line. I don't like having burned circuit boards in service either, but I would really like to know what's going on. So far I can't reproduce it, and the damn thing still works perfectly even with that trace gone.

 

[klui]

The grounding post do make more sense for ungounded equipment since it should provide a path for stray current out of the way of electronics'. The problem here is I feel the voltage is probably coming from the switch's PSU. You won't know unless the PSU is opened, including dissecting the transformer.

My guess is the PSU. diodegonewild has done quite a few PSU tear downs and I just don't trust them from eBay, Aliexpress, Taobao, .... Maybe consider getting quality switching power supply? I've heard good things about Mean Well.

Maybe a problem with the PCB where the trace is marginal.

 

[DavidWJohnston]

@klui - I think you might be right. I'm aware of strays but I've never seen it burn a circuit board before.

The PSU that came with the switch appears to have a stray voltage. It triggers my Klein NCV tester above 70V like crazy. It actually lights up my Klein ET45 voltage tester to the 48VAC/64VDC range very briefly then extinguishes. So the stray does have some current behind it; not enough to feel though.

I poked around with the NCV and found another significant stray on the Ethernet cables from my FTTH modem (Bell HomeHub 4000 !!). I knocked that down temporarily by plugging a USB cable between it and one of my UPSs. (No bare metal anywhere on the thing)

I use a high quality DIN rail 12V supply for my CCTV. I think I might do the same with all DC devices and eliminate the individual crap OEM PSUs.

I suspect the manufacturer did not conduct proper Hi-Pot testing and design, and like you said that part of the board is marginal (has traces too close), which allowed the stray to discharge, recharge, then discharge again, over and over forming a spark gap. My mains is 120V - In areas with 240V this must be worse.

What a pain.

 

[klui]

These issues are a direct consequence of most consumers want pay the least for lots of things. Although my wife reminds me just because something is expensive doesn't mean it's "good" either. And I can see that with *phile, equipment and parts.

Did you test your modem in isolation or with cables connected? You may want to ensure the voltages are coming from the modem and not elsewhere. If the the voltage is coming from your ethernet cable are you using PoE++/PoH? I just performed a quick check using my non-contact tester and none of my network cables nor my sole network gear's wart trigger it.

The one device that intermittently triggers the tester is this Black & Decker portable vacuum charger wart.

My video cam's warts have long failed, replaced by more hefty and over-spec'ed ones. Rated at something like 1.2A, they failed after 2 years of continuous use. Their replacements are not the small warts; rated to at least 3A and have been working for over 4 years.

 

[DavidWJohnston]

Yikes.

I just tested stray voltage from my HH4000 FTTH modem with all cables disconnected except the DC barrel jack with stock ungrounded 12V brick. The NCV triggers hard (same as live Romex) if you approach the modem or DC line, or an Ethernet cable with the other end unconnected.

When I disconnected all of the Ethernet cables from it, the stray voltage at the modem got worse.

Incredibly, there is a 100V P-P truncated sine on all Ethernet ports - Not only the metal shields, but all the differential pairs too! I cut and stripped one end of a network cable, plugged the other end into the modem, and measured each copper core with an oscilloscope:



It has enough energy to light up the Klein ET45 briefly to the 48VAC/64VDC range. Here is the stray's voltage drop across my body, hand-to-hand, AC-coupled: (Maybe ~ 10s of microamps continuous?)



Well, I grounded it through a USB connector again, poked around everywhere with the NCV, and finally no more detectable strays.