12-250-DC Fried All Of My Radios

[krakrenterprises]

We have installed 2 dozen Netonix switches lately, all of them being 12-250-DC or 26-500-DC units. Today we get alerts coming from a 3 month old switch install. After arriving on site, we notice no Ethernet links on ports 7-12. When logged in to the unit, all of the ports show 24v ON, but no wattage being drawn.

After trying to power the radios with a separate 24V POE brick, we noticed that ALL of the radios that were plugged in to ports 7-12 have dead shorts. We then tested the output power on the ports and were astonished to see +50v on all of them. This overvoltage has fried all of the equipment plugged in to these ports, leading to an expensive replacement and tower climbs to replace the gear.

Before you tell us that this is caused by grounding..... This site has a #2 to earth ground, which is bonded to our cabinet and the switch via #6. Each cable run from the tower down has a separate surge protector, which is also bonded to the same point. And lastly it is the middle of winter and we have not had a windy day here in 10 days.

My questions are:

1. What would cause such a failure?
2. What are the chances of this type of failure happening again?
3. Why are there no safeguards for this?
4. Does the 26-500-DC suffer from this type of failure as well?
5. What is the process to RA this unit for replacement?

I am wanting to know these questions to decide if we should continue changing our remaining 50+ towers to 26-500-DC units. We cannot afford to install a $700 switch that cascades into $3000+ in damages when it fails. I would prefer to stick with a seperate POE per device, as the failure rate on those has been <10 on over 1200 devices in 5 years. This single failure has caused half of that damage in one problem.

Please advise.

[lligetfa]

I do not know why a MOSFET would fail in such a way as to deliver 50V regardless of PoE setting but that is what most likely happened. More over, I don't know why an entire bank of ports would all fail at once. Maybe Dave can enlighten us.

[lligetfa]

Each cable run from the tower down has a separate surge protector...

A surge protector works by shorting to ground. In the case of PoE switches it means also shorting the PoE to ground which causes damaging current through the switch.

[Dave]

surge protectors are dangerous for POE....we have several posts on this I believe...our POE have a common ground...if one of the external POE surge protectors decide to short the POE power to ground, it is then shorted on all of the POE lines, which could easily blow all of the mosfets attached to it.

[krakrenterprises]

So, no one should use surge protectors?

[Dave]

It greatly depends on the situation, but for POE switches, such as ours, we recommend not using surge protectors on the network lines.

Surge protectors are a lot safer to use in non POE uses, where the surge "feature" clamps the positive (or negative, depending on the actual surge protector used) to ground. In this case the ground is not a power ground.

When a surge protector CLAMPs in a POE switch case, it can actually put some nasty voltages on the actual POE ground, which is seen by the switch, and can cause massive ground bounce (or even positive bounce), which can cause global damage on the switch.

Chris, Netonix partner, has many WISP towers, for many years, and doesn't use surge protectors.

[krakrenterprises]

Update:

After returning to the site today to replace the radios and switch, we determined that none of the surge protectors have shorted and were not required to be replaced. Tower is back and fully functional after replacing the radios and changing the switch to a 26-500-DC.

The 12-250 we removed is still stuck at 50v on ports 7-12. Clearly something in the switch has failed.

Is there a way you can mitigate this problem? Having the switch default to 50v when something fails is a pretty lousy design. There should be a fail safe (if X voltage is detected, or a spike/short is detected, the port(s) is disabled). Do you offer a schematic of the switches so we could see which ports share power supplies/grounds/etc? This would help us layout the cables in a fashion that does not take down a whole tower if this happens again.

I am not certain removing the surge protectors is a smart idea for us. Is the grounding of the Netonix switch sufficient to drain off the static from the shielding on the cables? If we get a lightning strike, I would much rather have a surge protector pop, than the entire switch and all radios/routers be affected.

Also there is no ground points on UBNT Airmax radios. In most cases we ground the antenna, but I am unsure if this will drain the static buildup on the cable as well.

If you can advise on this, and also point me in the direction to have this switch replaced, that would be great.

[Dave]

When you have 50 volts on the output, with the POE power turned off, it means that the mosfets have been damaged..sometimes they will get damaged to the off, or on state. In your case they are physically damaged so they are stuck on....nothing we can do will stop the output from being 50 volts in this case...this is rather rare to happen.

And yes, our switch itself does have ESD protection diodes on all of the network data lines. ESD damage to our switches is also very rare.

Our new switch series coming out next month has a much better chance of surviving whatever caused this as it has micro second overcurrent detection that has a good chance
of turning mosfets off in time so they don't get damaged.

surge protector comment
viewtopic.php?f=17&t=5785&p=30810&hilit=surge+protector#p30810

RMA info …
viewtopic.php?f=6&t=1259

[krakrenterprises]

When the 24v is turned off, there is no power output on the ports. However, regardless of the voltage selected, +50v is what is output.

I will request an RMA for this unit.

Any dates for the new series? We have a large project upcoming and would hold off if they will be available soon.

[Dave]

because you have 50 volts always being output, it means that the channels 50 volt mosfet has been damaged, so it is always outputting voltage, even when its not suppose to..

new series first batch suppose to be ready for sale sometime in Feb.