Ethernet Troubleshooting

[Ryu.Hayabusa]

I'm currently troubleshooting two Juniper SSG5 routers, both of which have 3 dead ports (#1 labeled 0/0, #2 = 0/1, #3 = 0/2) and one which has a intermittent 0/5 port (when a cable is connected, sometimes works, sometimes tx/rx light is solid).  So far I have found that the 75ohm resistors for the physical ethernet ports (after the magnetics) were way off.  I focused on port 0/0 which is for the WAN, but replacing them had no effect on the port at all, it still gives no lights when a cable is connected.  The resistance is now where it should be for that 0/0 port at 150 ohms between the TX/RX pins, which I read for a Mikrotik troubleshooting guide is what it should be, and confirmed it on a fully working Netgear switch.

On one router, port 0/1 and 0/2 tx/rx light flicker about every 5 seconds simultaneously, and on the other router port 0/2 flickers at about the same rate.

My current theory since the resistors changed nothing is the magnetics being the problem, though I am not sure how to test a transformer to ensure it is good (an equivalent part is the Pulse H1164NL, datasheet: http://www.mouser.com/catalog/specsheets/Pulse%20h1164.pdf).  Are there any tests I can do using just a multimeter?  I do not have an oscilloscope or waveform generator available to borrow at this point in time.  As a side note, the ports are split between two quad transformers, 0/0 0/1 0/2 are on one, and 0/3 0/4 0/5 0/6 are on another.

I have already checked lots of stuff before the magnetics such as resistors, voltages, chips for ethernet and CPU getting warm, port 0/3, 0/4, 0/5, 0/6 functionality (was able to login to the router), transistors, and a few other components, all of which seemed fine.

If there are any suggestions or ideas on other stuff to try or test as well I'd greatly appreciate them!  Thank you for your time.  

[NiHaoMike]

One common problem I have seen is that most of the Ethernet interfaces I have seen provide no path to dissipate stored charge, causing ESD issues. That sometimes causes the interface chip to fail, at which point it would likely be difficult to repair.

To prevent that, add 1M resistors across the caps that connect the midpoint of the transformer to ground. (In a pinch, drawing across the cap with a pencil also works.)

[Ryu.Hayabusa]

Thanks for the info, I'm going to give the 1Mohm resistors a shot tonight.

I removed the 2 magnetics chips, tested each with continuity on either side and both were the same, no internal shorts.  I re-soldered the known good ports magnetics chip to the dead ports side and no go, ports still dead.  Interesting part was the 4 good ports that had no magnetics chip installed yet, 1 port was doing that 5 sec light pulse, but not simultaneously with the dead port.  Re-soldered the other chip to the good ports and they all still work again, light pulse is gone, so that isn't the problem thankfully, those things are expensive.

If the 1Mohm resistors don't help I am going to order 75ohm resistors and 1500pf capacitors to replace all of the ports components at the same time.  If that doesn't work then I'm not sure what to do next. :/

[David_AVD]

I'm assuming that you've checked that the actual RJ connector pins are not damaged?   

[Rasz]

Its dead Jim.
Sounds like typical fried switch. Asic got ESD shock through the port.

[Ryu.Hayabusa]

@David_AVD: Thanks for the suggestion as sometimes the problem is the simplest thing that is easy to overlook, but unfortunately I have checked them already and they are all physically fine.

@Rasz: Well that's disappointing to hear.  I assume by ASIC you mean the Marvell ethernet controller.  Wouldn't all 7 ports be dead if an ESD shock went through any one of them?  I am able to use both of the routers with no issues on the 4 good ports.

If the resistors or caps located before or after the magnetics are bad/out of spec cause a port (or a sector of ports) to not work?

[NiHaoMike]

Blame it on whoever thought it was a good idea to have no DC path to ground. Some older Ethernet cards actually have that 1M resistor, but I guess someone decided that it doesn't do anything...

[Ryu.Hayabusa]

@NiHaoMike: Thanks again for that tip, I'll be adding it to any new gear I get, and probably to my existing gear if possible.  Just to be 100% sure, it is the capacitor after the magnetics that connects itself and the 75ohm resistors to ground correct?  Is it also possible to add this to ports with internal magnetics?

I did find a source for the Marvell ethernet chip, but it's in Shenzhen China, and their cheapo shipping quoted 2-3 weeks to arrive.

[MagicSmoker]

@NiHaoMike: Thanks again for that tip, I'll be adding it to any new gear I get, and probably to my existing gear if possible.  Just to be 100% sure, it is the capacitor after the magnetics that connects itself and the 75ohm resistors to ground correct?  Is it also possible to add this to ports with internal magnetics?

Yes, it is the (usually 1nF/1000pF) capacitor inside the magjack (ie - Ethernet jack with internal magnetics) that ensures the 75? termination resistors are at AC ground which needs the 1M bleed resistor. You can't get at this capacitor in most (all) magjacks I have seen.

[NiHaoMike]

You can often hack away at the plastic to get at it. Failing that, for a 10/100 connection, connecting the unused pins to ground through a resistor does the same thing.