Discuss what maybe happened to wispr - Re: cable short/fried

[lligetfa]

I don't mean to question wispr's methods insofar as what happened http://community.ubnt.com/t5/EdgeSwitch ... 1625#M1223 and I don't think we could have an open discussion there so I started this thread.

The complete sequence of events are a bit vague in that I don't know how 3 radios got fried but as the ports were set to auto detect for af/at, there is conjecture that af/at voltage got applied. I am no expert on how 802.3af/at detection works but reading http://www.eetimes.com/document.asp?doc_id=1271989 there are low voltage pulses sent out that look for a 25kohm resistance. I wonder what the chances are that a bad cable or Rocket could fool the logic.

I think a best practice would be to not trust the 802.3af/at detection logic and to set those ports that are known not to require af/at PoE, to either the known voltage requirement or off as need be.

[josh]

I agree. In operating in a Service Provider environment, we "know" what devices need power on what ports, and what power type they require. Anything "auto" in a SP network is dangerous.

[lligetfa]

Also, what is best practice WRT the built-in cable test? Should the far end be open? I would think that would give best results. It is not on par with a Fluke network analyzer. In wispr's case the end of the cable was already terminated to the Rocket so testing without the end device in place would not be convenient.

[rebelwireless]

for cable certification, the cable must be terminated and the test devices must operate on both termination points. So testing an open cable isn't 'certifiable' so to speak. I don't know how the rockets handle the testing. I'd assume, based on ubnt's products, that its a very basic and not nearly standards compliant test.

There are cheap IC's to handle multi-conductor testing but I'm sure were taking $5+ per unit to add which ubnt has proven time and time again is too much money to include for edge cases.


As far as the 25k ohm resister, I doubt that you could reproduce a crosstalk or termination or interference etc that would line up close enough. For instance, to get 25k ohms on 2 pairs of 26 gauge wire would be about over 100 miles of length. Any short is going to be very low resistant and any break, even a partial will go to infinite resistant really quickly.

I'm more inclined to think that either the termination was bad and power and signal conductors were mixed, or some logic breakdown on the ES.

[lligetfa]

For the cable test, I am not talking about a dedicated certifier but rather the cable diagnostic built into the switch. I don't know how the built-in test on the ES compares to the Netonix.

Further info has come to wispr's thread about there being inline surge suppressors. I know those can mess up even dedicated cable testers.

[lligetfa]

Here is an example of a cable test to a 24V PoE powered airCam.