Ethernet switch PoE repair

[joellinn]

Hi folks,
I got an Avaya 4850GTS-PWR+ ethernet switch which I bought on eBay dirt cheap some time ago. It worked flawlessly (except I couldn't download updates because no customer id ). However there was a storm and I presume some lightning nearby sneaked into the network.
The switch first stopped to deliver power to two ports, the others worked OK. So I restarted it. Bad Idea. Restart took for ever and no PoE was working on any port whatsoever (switching still working fine).
Of course I opened it and extracted the PoE board. I'm pretty sure the fault is on there, because it is separately powered with a >1kW 48V supply and uses optocouplers to connect to the switch board. The power to the devices is connected after the rj45 connector transformer I guess (typical setup). I tested the supply of course. 

Since it was partially working before I restarted it I think it can be a communication issue between the PoE driver chips and the main board. Maybe someone can point out if it makes sense to test the optocouplers (i.e. probe both sides for uC comm.) or has some other idea.

In the link please find an overview of the board and a nearer shot of the coupler region. The back is not that interesting apart from 48 diodesAreToExpensiveLetsUseZeroOhm Resistors
https://wolke.leuchtkanone.de/index.php/apps/gallery/s/YGn3UyAMNunrwWu

[raz221]

The board has two (obviously) bad electrolyte capacitors as far as I can see:

The top of a capacitor should always be flat, if it's even slightly bulged (raised) than that means that either
the capacitor is bad (or it will be soon), so such caps should always be replaced!

[joellinn]

Thanks very much, I thought I checked that... Obvious now... I would presume they are smoothing the 48V rail(?). If that's true how can they lead to such behaviour?
I will replace them and see what happens

[wraper]

The board has two (obviously) bad electrolyte capacitors as far as I can see:

The top of a capacitor should always be flat, if it's even slightly bulged (raised) than that means that either
the capacitor is bad (or it will be soon), so such caps should always be replaced!

To me they doe not seem bulged at all, just different capacitors with different vent style. When vent bulged, the center of it rises, nothing like seen on the picture. Also, as these capacitors are not a part of DC/DC converter it is not very likely that they will bulge.

[mariush]

80-90% sure they're not bulged, just a lousy automatic machine pressing the grooves on the top of that capacitor. Instead of just making clear channels down in the metal, it deforms the top slightly in the process.
A swollen capacitor will also have those grooves raised a bit.

[Kjelt]

When elcos wear out due to temperature the plastick jacket shrinks, does not look like this is the case here.
Lighting can do a lot of unseen damage, usually i scrap lighting damaged electronics, hope you have better luck.

[joellinn]

Hey, did some measurements. El. caps don't seem to be a problem, they are all just for smoothing the 48 V supply (measures at about 54.1 V , but I guess that's fine (?))
Broadcom being Broadcom there doesn't seem to be a full datasheet for the chips available, but i attached the diagram from the product brochure.
The chips include dc/dc conversion controllers and I think they use four of the chips for providing power to the others as well. There are four inductors, one being a little bit bigger, I guess its powering other logic components as well.
From the placement of L3 and D34 I conduct the left side of the inductors being the 3.3V rail. One 3.3 V Rail is down.
What do you suggest would be the next thing to test, desolder diode and test it / use lab supply to power 3.3 V rail / test inductor?

Maybe the controller board stops all PoE functionality because it can only communicate with some of the chips.

[station240]

From the circuit diagram you found, I see a protection zener on the 3.3V input. If those diodes are fitted to the PCB, it could simply be that one of more has shorted and shut down the rail.

[joellinn]

I checked the diode between the chip and L2, it indeed is blown. I also checked the diode D21 which I thought was blown and desoldered it. Unfortunately it is OK  I also desoldered D25 on PoE port 21 because it seemed shorted. It is OK as well so I desoldered the fuse F21 to and it seems the chip has shorted output between the legs of D25.

I think D21 is another protection diode for the 3.3V Rail. Can a cap blow to a short? Otherwise I think one chip is blown. 
I don't think there's a way to replace those because A) I don't think I can order broadcom chips and B) their diagram talks about RAM and ROM. I don't now for what they would need that but surely I don't have the image.