On 3/25/22 07:36, Paul wrote:
> On 3/25/2022 9:49 AM, Rene Lamontagne wrote:
>> On 2022-03-25 12:51 a.m., T wrote:
>>> On 3/24/22 18:14, Rene Lamontagne wrote:
>>>> On 2022-03-24 7:45 p.m., T wrote:
>>>>> On 3/24/22 16:52, Jonathan N. Little wrote:
>>>>>> T wrote:
>>>>>>> Hi All,
>>>>>>>
>>>>>>> Had a weird on e yesterday.
>>>>>>>
>>>>>>> I installed a computer at a new location for a customer.
>>>>>>> The network cable from the wall worked fine
>>>>>>> directly into the computer.  But then I plugged it
>>>>>>> into a little 5 port hub.  Then no worky.  A
>>>>>>> different brand hub later and still no worky.
>>>>>>> Hmmmm.
>>>>>>>
>>>>>>> Well the master hub the wall cable was plugged into was
>>>>>>> across the room, so I strung a 25' cable from the
>>>>>>> little hub to the master hub.  Worked!
>>>>>>>
>>>>>>> Moral of the story.  Please do not use electricians
>>>>>>> to string data cable(s).  They always screw something up.
>>>>>>>
>>>>>>
>>>>>>
>>>>>> Who wires keystone jacks without testing them afterwards?
>>>>>
>>>>> Electricians
>>>>>
>>>>>> You don't own a tester either?
>>>>>
>>>>> Don't need one.  The low voltage guys I recommend
>>>>> all have thousand dollar units and their stuff in
>>>>> impeccable when they finish.    I have had them
>>>>> go in after electricians to fix things and they
>>>>> just shake their heads in amazement.
>>>>>
>>>>
>>>> There are Electricians, and there are wannabe electricians, at my
>>>> last place of employment (15 years)   As a licensed  Power engineer,
>>>> I also held An electricians license to 600 volts.
>>>>
>>>> As part of my duties I was in charge (hands on) of pulling in Cat5
>>>> cables through the raceways to some 90 PCs throughout the complex,
>>>> some pulls as long as 280 feet, Then terminating both ends with rj45
>>>> connectors, one at each computer and the other at the jack-field' in
>>>> the 2 wiring closets.
>>>>
>>>> My professional crimping tool cost about $100.00  cdn, non of your
>>>> DIY tools for me and I used only high quality connectors. I did all
>>>> my own crimping and testing, as this was a critical network, in all
>>>> that time we used about 4500 feet of  Cat5 cable and I never had one
>>>> failure during my time there.
>>>>
>>>> Rene
>>>>
>>>>
>>>
>>> Hi Rene,
>>>
>>> I have a bachelors degree in Electrical, Electronic, and
>>> Computer engineering.  I do not try to be an electrician.
>>> And I have professionals that are absolute artists do
>>> my cable stringing.  Again, I know better.  A man's
>>> got to know his limitations.
>>>
>>> I'd love to use you for cable stringing.
>>> This problem I encountered was probably do to the
>>> computer network card being more tolerant of
>>> crappy cabling/connections then both hubs.  Gave
>>> me a run for my money, but I figured it out.
>>>
>>> About 25 years ago, I got called into an electrical
>>> firm to find out why their network was trash.  They
>>> stuck their server in a small closed closet.  The
>>> heat nearly singed my eyebrows when I opened the door.
>>>
>>> I told them they had to take the door off and duct
>>> an air conditioning vent into the room.  I could
>>> tell they were getting agitated with me.
>>>
>>> Next step was to remove a wall plate to check the
>>> wiring.  No fooling, they splayed the cat 5 TEN
>>> inches back.  I told them they had to bring in a
>>> professional to fix all their wiring.  They
>>> then had had it with me, told me they did the
>>> wiring, and showed me the door.  Idiots.
>>>
>>> The current guys I use certify each drop with
>>> a live data tester.  They are awesome.
>>>
>>> -T
>>
>>   Thanks, I took great pride in my work, Good show, when you want good
>> work hire a good Professional team. a few extra dollars buys you no
>> callbacks and problems.
>>
>> Rene
>
> Even if the TP were splayed (the outer insulation layer removed
> allowing the wire pairs to spread out a little), that should not
> affect operation. If someone tried to untwist a length of each
> pair, that would be bad.
>
> The details around a connector matter, in the sense that
> the "connector impedance" can be a problem. That's where the
> notion of "RF connectors" came in, in discussions at work.
> "Is it an RF quality connector?" was meant to imply that
> looking into it with a network analyzer out to some number
> of GHz, it maintained a 50 ohm impedance, suited to matching
> things driving and receiving, around it.
>
> The TP would be 100 ohms or 110 ohms or so. The Transformer
> on each end of the link, would have a characteristic impedance
> to match that. The design of the connector, does not suggest
> any attempt at controlled impedance. There's no ground shield
> around it (Ethernet doesn't need that), and a ground plane was
> always part of making "RF connectors". Look at an SMA for an
> example of the quality that goes into 2.5GHz or 5GHz connectors.
>
> The wires are twisted, as part of handling crosstalk, and having
> the transformer at the end, nullify some of that. You can have
> large common-mode signals induced into those pairs, and it
> "all still works". I've taken oscilloscope pictures of 10V P-P
> signals riding (common mode, not diff) on pairs, and the transformer
> rejection takes care of it, like magic. The shocker for me,
> was not the fact the transformer was performing miracles,
> it was the voltage the oscilloscope was claiming was present.
> And that's just AC hum on the wires, but a hum of significant "fake"
> amplitude.
>
> I would sooner believe the network had a mixture of DHCP and
> static assignments, and perhaps a duplicate address was affecting
> things.
>
> If it's GbE, a certain amount of defectivity can be removed
> by the PHY during negotiation. It could drop the interface
> to 100BT, if a "lucky choice" is made for the wiring defect.
> I don't think MDI/MDIX can fix everything. It's sorta like
> PCI express, that can correct a number of faults, simply
> by automatically turning off some of the lanes and using
> the remaining lanes. A person might never know, that
> lane 9 on the x16 video card slot, had been busted for the last
> five years. There's no warning in software.
>
> One of the reasons for debugging stuff like this networking
> problem, is to build your catalog of experience.
>
> The Marvell NIC on one of the motherboards here, has a TDR on it,
> and it can tell you "open, short, matched" on each of the four
> pairs. You turn off the power on your Switch, so the transformers
> are quiescent, and the Marvell NIC on the PC sends pulses down the line,
> to evaluate the wiring. But that motherboard died, and my
> attempts over the years to buy a standalone PCIe card, failed
> to work out. One box from the computer store, said Marvell
> on the outside, but a RealTek was inside. so now I have
> no TDR to use.
>
> There is still a copy of the Virtual Cable Tester whitepaper
> at the bottom of this page. It has a couple pictures of
> waveforms. And the VCT can give an estimate of how many feet
> away the defect is. I don't know if it was stated anywhere,
> as to the measurement uncertainty (+/- nanoseconds or feet).
>
> https://www.teklib.com/library/vct-marvell-virtual-cable-tester-technology-for-gigabit-networks-white-paper/
>
>
>    Paul

Sounds slick.
When I come across Ethernet cable problems, I call in
the big guns.