Ethernet linked access point appears sensitive to cable length well below specified limits...

Have a question or want to start a discussion? Post it! No Registration Necessary.  Now with pictures!

Threaded View
Ethernet linked access point appears sensitive to cable length well
below specified limits.

I’m seeking to locate a wireless access point remote from my main
router, a D-Link DI-524 whose wireless function has been turned off. The
wireless access point is being implemented with a nearly identical
D-Link DI-524 wireless router whose DHCP function has been turned off.
The IP address of the second D-Link DI-524 wireless router has been
altered so as not to conflict with the IP address of the first router.
One of the LAN outputs of the first router is connected to one of the
LAN outputs of the second router so that it acts like a switch.

I have tested the above configuration at three locations linked by three
different CAT5 cables.

Location #1 – WORKS fine when linked by commercially constructed 10’
long CAT5 cable.

Location #2 – WORKS fine when linked by home built 40’ long CAT5 cable.

Location #3 – DOES NOT WORK when linked by home built 100’ long CAT5
cable. – Not only is there no communication over the link, but the
lights on both routers do not even indicate a connection. YET, if at
this same location #3, the 100’ long cable is plugged into any one of
several computers, the connection indicator lights come on immediately,
and full normal network access is quickly obtained.

As the cable length is well under the 100 meter (300 feet) maximum
length for Ethernet, at first I thought that maybe the third cable is
somehow wired differently, perhaps as a “Cross over cable” rather than
as a “straight through” cable.  I understand that some interfaces can
cross and un-cross a cable automatically as needed.  I am uncertain
whether or not the LAN ports of the DI-524 have that feature, but an
examination of all of my cables, both home made and commercially
assembled, and all appear to be wired “straight through.”

I’ve also sought to confirm that the problem is not associated with the
RJ45 connector at the end of the cable. The symptoms are unchanged when
a short (10 foot long) Ethernet extender (female to male) cable is used,
so the connector at the end of the 100’ cable is not disturbed as I
switch between the D-Link DI-524 and my lap-top computer..

Any Suggestions?

Re: Ethernet connection sensitive to cable length
added cabling & ethernet newsgroups -
Wonder if a small ethernet cable tester has been used to verify proper pairs
?
--

windsurferLA wrote:
Quoted text here. Click to load it



Re: Ethernet connection sensitive to cable length
Quoted text here. Click to load it

These homebuilt cables, which wiring pattern did you use?
T-568A or -B ?  If you don't know what I'm talking about,
it is highly likely you split a pair.

Please do not complain "but the other comp works".
Difference NICs and drivers are more error-tolerant
and have more robust fall-backs.

Electrons may be color blind, but they _do_ know
who their dance [twist] partners are.

-- Robert


Re: Ethernet connection sensitive to cable length
On Sat, 03 May 2008 20:22:07 GMT, Robert Redelmeier

Quoted text here. Click to load it

Huh?  Unless he did something really disgusting, like EIA-568A on one
end and EIA-568B on the other, either wiring standard will work.  The
color codes are different, but the pairing is identical.

Drivel:  I once had some hired help in wiring a medical office.  I did
568B while my hired help did 568A wiring.  Nothing worked when we were
done.  Never ignore the obvious.

Quoted text here. Click to load it

True.  I recently demonstrated that I can run 10baseT-HDX (half
duplex) through 2,000 ft of CAT5e (two rolls in series), without any
data degradation.  If I had a 3rd roll, I would have added it.  The
catch is that it would only work between my Cisco 1900 ethernet
switch, and a desktop with an Intel Pro100 card.  It would function to
other devices (several laptops, assorted junk around the office), but
these showed various errors in the switch SNMP logs.

Quoted text here. Click to load it

Perhaps a matchmaker would be appropriate?

Quoted text here. Click to load it

See item #6:
<http://www.LearnByDestroying.com/nooze/support.txt>

--
Jeff Liebermann     jeffl@cruzio.com
150 Felker St #D    http://www.LearnByDestroying.com
Santa Cruz CA 95060 http://802.11junk.com
Skype: JeffLiebermann     AE6KS    831-336-2558

Re: Ethernet connection sensitive to cable length
Quoted text here. Click to load it

Oh yes, fully agreed.  But I think it relatively unlikely
the OP would have followed either without some awareness.
I'm testing for that awareness.  

The intuitive wiring patterns (SBS and USOC) will split a pair.
All the correct ones are somewhat counter-intuitive.

Quoted text here. Click to load it

Easier to notice and fix with jacks.
 
Quoted text here. Click to load it

Nice data point.

-- Robert


Re: Ethernet connection sensitive to cable length
On Sun, 04 May 2008 03:04:43 GMT, Robert Redelmeier

Quoted text here. Click to load it

I enjoy asking the original questions backwards.  In this case, it's
"what would I have to do, to CREATE the problem"?  Creative wiring and
connector terminations are the probable culprits as the terminating
equipment is obviously working.

Quoted text here. Click to load it

I'll do the newsgroup(s) a favor and not rant on how Ma Bell, the old
TIA and the EIA created this mess.

Quoted text here. Click to load it

At the time (about 1998), if you purchased any manner of pre-wired
ethernet jumpers, you got EIA-568B wiring.  Never mind that EIA-568A
is the real standard.  Well, my accomplice was working on his BICSI
certification, and they were preaching EIA-568A.  Never mind that I
told him that I wanted the color coding to be consistent throughout
the entire building, all of which was EIA-568B.  He decided that BICSI
must be correct and was fully prepared to have me (not him)
re-terminate the entire building (about 400 wall jacks) to insure
compliance.  I paid him his fee and hired a day worker from the local
lumber yard.  He didn't speak much English, but he undid the damage in
amazingly little time, and finished the job in about half the time I
had expected.  The cable certifier found two wiring errors out of
perhaps 60 wall jacks.  I was going to pay him a bonus, but that was
before I noticed some of my tools had evaporated.  Sigh.

Quoted text here. Click to load it

Careful here.  That's not an endorsement for installing 2,000ft CAT5e
runs.  I once calculated the maximum cable length at about 1,200ft for
10baseT-HDX before timing becomes an issue.  Why 2,000ft worked is
still a mystery to me.  It shouldn't have unless the timing on the
ethernet devices is more relaxed than required.  Also, note that I was
using 10baseT-HDX (half-duplex).  Full duplex and/or 100baseT will not
work due to collision domain issues and cable near end crosstalk.  You
also have to use an ethernet switch.  Hubs (repeaters) will not work.

I do have several 900ft runs in service (one of which goes under some
railroad tracks).  No problems.  I do have a 500ft run that is giving
me problems.  I haven't had time to troubleshoot (due access issues),
but am guessing that I have some induced interference from rotating
machinery, transformers, ballasts, or something similar.

As for stretching the technology, I done my part:
1.  DSL over barbed wired.  1Mbit/sec SDSL.  Distance is about 3,500ft
of barbed wire, with 100ft of CAT5e at each end.  However, the
multiple splices tend to be noisy so it was replaced with a wireless
link about 2 years ago.
2.  10base2 (cheapernet) over CATV 75 ohm RG-6/u coax.  Distance was
about 1500ft at one location.  The other location was a radio station
that was stuffed full of RG-6/u coax runs.  Terminated with 50 ohms at
both ends (because 10base2 uses DC levels for busy detection).  Only
two transceivers (no taps or T connectors).  With such high losses,
the far end reflections just disappear and never become a problem.
3.  FTTS (fiber through the sewer).  Actually it's drain under the
road that dumps into the river, but it looks much like a sewer.  About
800ft.  I keep waiting for it to fail as some water propelled rock
cuts the outer jacket, but it's been up for 8 years and shows no
change in fiber attenuation.  I've lost two transceivers, but the
fiber is holding it's own.
4.  In the miscellaneous category, I've done ethernet through the
sewers (in order to cross the LATA boundaries), ethernet over 25 pair
telco bundles, ethernet over two 117VAC extension cords (I was
desperate), DSL over zip cord, and adapters made primarily from clip
leads.  900MHz data over G-Line.  I won't mention any of my other
wireless atrocities.

One of these days, I'll follow the standards and rules, but not this
week.

--
Jeff Liebermann     jeffl@cruzio.com
150 Felker St #D    http://www.LearnByDestroying.com
Santa Cruz CA 95060 http://802.11junk.com
Skype: JeffLiebermann     AE6KS    831-336-2558

Re: Ethernet connection sensitive to cable length
Jeff Liebermann wrote:

<snip>
Quoted text here. Click to load it

Starlan 1 Mbps?

Michael

Re: Ethernet connection sensitive to cable length
msg wrote:
Quoted text here. Click to load it

wow - StarLan -
That really takes me back...
We creating the video training and marketing materials...



Re: Ethernet connection sensitive to cable length

Quoted text here. Click to load it


Nope.  10Mbits/sec half-duplex ethernet over 25 pair telco bundles.
Works fine up to about 100ft.  It might work farther, but I haven't
tried it.

Gaaak.  Starlan brings back not very fond memories of doing battle
with 3B2-400 clunkers using Ma Bell 258A wiring on CAT3 shared with
Pre-Merlin phone systems.  I had no idea what I was doing, so it's
little wonder that I could never make it all work quite right.

Sorry, no LatticeNet.  Some Token Ring (4Mbit/sec), Moses Networks
Promise LAN, DECNet, Novell Netware, 3com something pure ISO stack,
Microsoft LAN Manager, Lantastic, and probably a few more that I'm
successfully forgotten about.  Add a bunch of TCP/IP implimentations
for Windoze 3.x that never quite worked right, and a mess of
non-802.11 schemes and protocols, that went nowhere.

--
Jeff Liebermann     jeffl@cruzio.com
150 Felker St #D    http://www.LearnByDestroying.com
Santa Cruz CA 95060 http://802.11junk.com
Skype: JeffLiebermann     AE6KS    831-336-2558

Re: Ethernet connection sensitive to cable length
Jeff Liebermann wrote:
Quoted text here. Click to load it

3B2 - ahhhhhh -
We also did the training video & product intro for the "Unix PC" -
along with the IBM 3270 replacement..... can't recall the number, maybe 6500
?
Lastly - we did the AT&T PC product materials & launch -

ahhhhhh -



Re: Ethernet connection sensitive to cable length

Quoted text here. Click to load it

I knew some of the developers there, and just *loved* the product name.
In the end, it was more of a promise than a LAN, however.

My all-time favorite communications product name was from a company
(whose name I forget) up in Marin county who built a very simple device.
"Back in the day", terminals connected to computers using modems, which
had a "DCE flavor" (modem-to-host computer) and a "DTE flavor"
(modem-to-terminal) for the wiring. If you wanted to connect a terminal
directly to a computer, you could eliminate the modem, but needed to
cross-connect the transmit/receive pairs in the cable. Today we all know
this as a "null modem" cable, and it is fairly ubiquitous, but it hasn't
always been so.

This small company marketed a "null modem" device; a small box that had
the transmit/receive cross-wiring, with connectors that would allow the
use of ordinary, uncrossed-wiring modem cables to the DTE and DCE.
Simple, but useful if you don't have a null modem cable.

No technological genius here, but great marketing savvy. Instead of
calling it a "cable crossover" or some such, they called the device a
"QuasiModem"--the box had a picture of a bent-over Quasimodo carrying
the device under his arm, with the slogan "We've got a *hunch* you'll
like it!"

Believe it or not, between the last two sentences I just went outside my
office and shot a bobcat who was chasing my house cat, whose name
happens to be "Cat 3"(he is my third cat). No, I am not planning to
upgrade him to Cat 5.

--
Rich Seifert              Networks and Communications Consulting
                          21885 Bear Creek Way
(408) 395-5700            Los Gatos, CA 95033
(408) 228-0803 FAX

Send replies to: usenet at richseifert dot com

Re: Ethernet connection sensitive to cable length
Rich Seifert wrote:
(snip)

Quoted text here. Click to load it

Shouldn't that be gato tres?

What do they call the wiring in Spain and Mexico?

(snip)

Quoted text here. Click to load it

-- glen


Re: Ethernet connection sensitive to cable length
glen herrmannsfeldt wrote:
Quoted text here. Click to load it

TIA/EIA 583-Ã



Re: Ethernet connection sensitive to cable length
Jeff Liebermann wrote:

<snip>
  4.  In the miscellaneous category...ethernet over 25 pair
Quoted text here. Click to load it

...or perhaps LatticeNet?

Re: Ethernet connection sensitive to cable length

Quoted text here. Click to load it

The propagation velocity of UTP is about 5.1 ns/meter, or roughly
1.5 ns/foot. Your 2,000 foot run has a round-trip delay of around 6 us,
which is far less than the 48 us allowance in half-duplex 10 Mb/s
Ethernet (i.e., 51.2 us less the 3.2 us "jam" time). As long as the
*signal characteristics* were still acceptable to the line receiver, it
should work, as there are no protocol timing violations in this
arrangement.

Much of the timing allowance in ordinary networks is allocated to the
repeaters; I suspect that your setup was a two-station back-to-back
connection, which has LOTS of timing margin.

Quoted text here. Click to load it

Backwards. The collision domain (timing) issues are related only to
*half* duplex Ethernet, not full duplex. You can (and we often do) run
full-duplex Ethernet over tens or hundreds of kilometers (using fiber)
with no timing concerns.

--
Rich Seifert              Networks and Communications Consulting
                          21885 Bear Creek Way
(408) 395-5700            Los Gatos, CA 95033
(408) 228-0803 FAX

Send replies to: usenet at richseifert dot com

Re: Ethernet connection sensitive to cable length
Quoted text here. Click to load it

A good diagnostic method.

Quoted text here. Click to load it

Probably because the US government requires -A

Quoted text here. Click to load it

While US industry historically has been -B

Quoted text here. Click to load it

10basedT has such relaxed timing because of the cascading
permitted. It became 100baseTX by tightening the circle.
Timing is mostly relevant for collision situations.  If you
are isolated by a switch, how can it matter?

Quoted text here. Click to load it

Keeping crosstalk out is probably the main benefit.

Quoted text here. Click to load it

I hope these installations have adequate lightening
protection!

Quoted text here. Click to load it

That's fine when expertise is available for troubleshooting
and the reliability is acceptable.

-- Robert



Re: Ethernet connection sensitive to cable length
reply at bottom

Robert Redelmeier wrote:
Quoted text here. Click to load it

Robert ... thanks for the hint.. the likely cause has been identified.

When you asked about T-568A or T-568B, it triggered a memory.  I then
recalled that the connectors were not assembled at the same time; one
end was rebuilt because of intermittent connection problems. The
intermittent end had been the very first RJ45 connector that I had ever
assembled. I further recalled being uncertain at the time I was
rebuilding the connector whether to use the "A" or "B" configuration.

Now doubting my wiring skills, I got out my jeweler's eye loop to very
carefully look at both ends to make sure one had not be wired "A" and
the other "B."  The result was not what I expected. A careful
examination of the newer end appears to reveal that the blue-white (#5)
and the green white (#3) were inadvertently interchanged at that end and
only that end.

I expect that repairing the cable will solve the problem. WHAT I FIND
SURPRISING is that the cable has worked with so many (guest's) computers
for so long that it was not until the A - B question was raised that I
thought to check the colors of the inner wires - a not so easy a task
for old eyes like mine. Obviously, there is enough cross talk between
the pairs to effect a connection with most, but not all, hardware.

Although I feel rather stupid for not spotting the wiring error earlier,
I never suspected a wiring error would be the problem. When ever I
assemble a connector, I keep a multi-colored wiring chart right in front
of me. Obviously, it was not enough to preclude the error.

Thanks for everyone's help..   Tomorrow, we try a rebuild.

Re: Ethernet connection sensitive to cable length
Quoted text here. Click to load it

That's the idea.

Quoted text here. Click to load it

This is the most common newbie wiring mistake.  Swapping
the whites turns the layout into side-by-side and splits
the green and blue pairs.

Quoted text here. Click to load it

It is hard for everybody, and impossible with some cable
mrfs where the whites are unmarked (some plenum).

Quoted text here. Click to load it

Crosstalk is not really required.  rememeber the signal is
differential and also present full-strength on the green.
Still, a bit surprising it ran as well as it did.  Of course,
it never needed to carry more than 1 MBps, so retransmits
would go largely unnoticed.

Quoted text here. Click to load it

Many years of experience have led to to _ALWAYS_ check the
cabling first.  Especially if there is any indication it is
home-made.  It is incredibly difficult to crimp connectors
correctly.

-- Robert



Re: Ethernet connection sensitive to cable length
Robert Redelmeier wrote:
Quoted text here. Click to load it

Which is why I like the Leviton Quick port for the part time do it yourself
crowd. You can buy the jacks at Home Depot on Saturday afternoon and they have
the wire colors on the sides of the jacks.

David

Re: Ethernet connection sensitive to cable length
Quoted text here. Click to load it

Agreed.  The key is _JACKS_ .  They're much easier
to get right than   _PLUGS_ .

-- Robert


Site Timeline