How to [not] blow a port on CSR328 (and a hex poe) in easy 2 steps

Hardware MikroTik hardware general
1

Hi,
I managed to do a a very stupid thing, I blew up a port on my CRS328-24P-4S+RM and took down my RB960PGS as well.
To be fair, it took normal operations to do so…

Here are the steps, which I do not advise to follow:


  • Connect power to RB960PGS (48V adapter to drive 802.3af cameras)
  • Accidentally use crossover cable to connect CRS328 to the RB960PGS
  • ???
  • Magic smoke!

The carnage:

  • Dead PoE port in CRS328 (poe-out status: overload)
  • Dead RB960PGS, well, no PoE input, and if powered by the jack, and first port is connected the device would power off (short).

I managed to resurrect the RB960PGS, the PoE input diode was blown, replaced that and it worked. That is after I figured out that the cable I was using was crossed (ie: where the +24V supposed to be it was seeing -24V).


I am yet to disassemble the CRS328 (I hope there is a protection diode that is shorted… I hope), otherwise I am looking for a replacement PSE-24 board apparently Mikrotik doesn’t supply separately. Alternatively I would just live with a dead port.

2

A bit of an update regarding cable:
The cable is wired wrong (not even cross-over, ie TIA568A/B).

One end has correct A pinout:

  • White/Green stripe
  • Green
  • White/Orange stripe
  • Blue
  • White/Blue stripe
  • Orange
  • White/Brown stripe
  • Brown

While the other end is wrong:

  • White/Orange stripe
  • Orange
  • White/Green stripe
  • White/Brown stripe
  • Brown
  • Green
  • Blue
  • White/Blue stripe

The above doesn’t match A or B spec, the cable is factory made too.
This cable wasn’t thrown out because for unknown reason it managed to work (link at 1Gbit/s).

Because the Blue pair and Brown pair is swapped it would cause the polarity flip. I am fairly certain how the both devices failed:

the Hex was powered by 48V supply, while the CRS328 was sending 24V down the wrong cable, if the cable was correct, nothing would have happened due to input Schottky diode inside of the Hex.
Unfortunately the polarity was so the Schottky diode (SK56) blew up due to being rated only at 60V (as -24V and 48V add up to 72V). The input Schottky diode shorted (the common failure mode of diodes) which blew the port on the CRS328 due to the same 72V difference being applied across it.

I am tempted to replace both input diodes with 100V/5A ones.

I haven’t yet looked at PSE-24 board inside of the CRS328. I hope it is either a passive or a simple semiconductor (a diode or a MOSFET), otherwise I will be looking for a replacement board, broken CRS328 or ignore the dead port and live with it.

I am a bit disappointed that Mikrotik gear is so fragile.

3

Not that it helps you in any way, but I learned - through similar misadventures - to never trust a cable, no matter if new, from factory, self-crimped or whatever, whenever there is PoE involved[1], always test them, el-cheapo lan cable tester can be found for less than 10 Euro or so, it doesn’t make sense to risk frying an expensive piece of equipment (or even a cheap one) because of a faulty cable (besides the value of the equipment or of the repair, the time down does have value).
The real PITA (at least for someone that doesn’t use these tools often) is that they normally use a 9V battery, which usually is exhausted when you need to test a cable.
I have seen newer and slightly more expensive one that have a USB-C connector and a rechargeable battery (and can check also the ground for >CAT 6 cables, but had neve used one of those.


[1] Actually there are also small PoE testers, but they are IMHO a bit too expensive (40-100 Euro) for a non-professional use, I have seen also a simpler “Poe detector” that is in the 15-30 Euro range that would have been enough in a case such as yours (and even if it fried, it would have been better than the switch/router)

4

That gives me an idea: I could actually build one. Or go lazy route and get one off aliexpress…

5

A bit of an update:

It seems I didn’t blow up the PoE PSE-24 board but actually cooked the transformer module (part no: PGH5610RS LF).
Here is my trouble shooting methodology:

  • Checked every diode and transistor on the board for shorts (diode mode in multi-meter). None found.
  • Removed the PoE board and tried the ports without it, the dead port remained dead, rest of the ports worked.
  • Plugged in the suspect board into different set of transformer modules, all the ports on the suspect board worked.
  • Upon further inspection of the suspect transformer module (aka quad port magnetics module) noticed discoloration (burn) on the pin header.

I have ordered the part from Aliexpress, will see if swapping it fixes the dead port.

Here is the module in question: https://www.bothhandww.com/wp-content/uploads/2021/07/pgh5610rs-lf-reva4-200525.pdf

6

I’ve heard of these Ethernet PHrY things. Yours appears to be based on SET technology…smoke-emitting transformer. :rofl:

7

:laughing:

the 48V power source supplied about ~80W into that coil (which is rated to 802.3at, ~25W, or ~1/3), so yeah…

8

The technology is better known as MSEC (magic smoke emitting component), the magic smoke has even an entry in Wikipedia:
https://en.wikipedia.org/wiki/Magic_smoke

9

Update:

I received the magnetics module and successfully replaced it.

Desoldering was fun due to “hundreds” of through hole little pins, but it wasn’t too much trouble at the end.

All ports work now!

I have also replaced both of the input diodes with SS510 (100V 5A rating) on the hEX PoE, just in case.