| Author |
Message |
Justin Time
Guest
|
 Posted:
Wed Jan 12, 2005 7:01 pm Post subject:
Re: Cat 5e twists |
|
|
James Knott wrote:
Why would running two circuits over the same cable, cause a problem
with
skew? You're still running on only one pair in each direction, for each
circuit.
James,
If I remember my Ethernet right, and it has been too many years since I
delved into all the specifications, there is a finite amount of time
between when a unit transmits and when it expects to hear something.
If the signal is delayed in one direction, then the timers begin to
kick in and another node begins to think the pathway is clear and
transmits. The signal that had been delayed is then stepped on and the
entire system goes into a wait state because a collision has been
detected. The clearing of the system after a collision causes the
entire network to slow down as it is a function of both hardware and
software waiting for the first node to do its collision detection
timeout.
Does this help?
Rodgers Platt |
|
| Back to top |
|
 |
James Knott
Guest
|
| Posted:
Sat Jan 15, 2005 3:22 am Post subject:
Re: Cat 5e twists |
|
|
Justin Time wrote:
| Quote: | James Knott wrote:
Why would running two circuits over the same cable, cause a problem
with
skew? You're still running on only one pair in each direction, for each
circuit.
James,
If I remember my Ethernet right, and it has been too many years since I
delved into all the specifications, there is a finite amount of time
between when a unit transmits and when it expects to hear something.
If the signal is delayed in one direction, then the timers begin to
kick in and another node begins to think the pathway is clear and
transmits. The signal that had been delayed is then stepped on and the
entire system goes into a wait state because a collision has been
detected. The clearing of the system after a collision causes the
entire network to slow down as it is a function of both hardware and
software waiting for the first node to do its collision detection
timeout.
|
Skew refers to differences in parallel signals. For example, in gigabit
ethernet, you're using all 4 pairs in both directions. This means that if
one pair took significantly longer than another to propogate the signal,
there would be a significant relative delay between the two, which might
cause problems. When you've only got one signal, there's no relative delay
to worry about. Collisions are another matter, which does not apply to
switches. Back in the days of hubs or coax based ethernet, collisions were
a concern, which is why the spec limits the maximum return distance of
collision domain to 512 bit times. This provides a process by which
collisions may be recognized and handled by the NIC. Collisions have
nothing to do with skew.
Bottom line, skew may be an issue with gigabit ethernet, where all four
pairs are used.. It is not an issue with 10 or 100 Mb, where only one pair
is used in each direction. |
|
| Back to top |
|
|
Justin Time
Guest
|
| Posted:
Tue Jan 18, 2005 11:45 pm Post subject:
Re: Cat 5e twists |
|
|
James,
I think we are beating different ends of the same dead horse. The
issues regarding delay skew and 10/100 Base-T date from when hubs were
more common than switches. As a historical note, delay skew did affect
a hub as a hub did not retime the signals.
Rodgers Platt |
|
| Back to top |
|
|
James Knott
Guest
|
| Posted:
Wed Jan 19, 2005 3:01 am Post subject:
Re: Cat 5e twists |
|
|
Justin Time wrote:
| Quote: | I think we are beating different ends of the same dead horse. The
issues regarding delay skew and 10/100 Base-T date from when hubs were
more common than switches. As a historical note, delay skew did affect
a hub as a hub did not retime the signals.
|
Once again. How do you have skew, when you don't have parallel signal
paths? Whether using a hub or switch, 10 & 100 Mb ethernet only use one
pair in each direction. This eliminates the possibility of skew. On the
other hand, gigabit ethernet uses all 4 pairs in each direction and as a
result, the possibility of skew exists. |
|
| Back to top |
|
|
Dmitri(Cabling-Design.com
Guest
|
| Posted:
Wed Jan 19, 2005 4:18 am Post subject:
Re: Cat 5e twists |
|
|
Justin Time wrote:
| Quote: | James,
I think we are beating different ends of the same dead horse.
|
Get off the poor animal! ;-)
| Quote: | The
issues regarding delay skew and 10/100 Base-T date from when hubs were
more common than switches. As a historical note, delay skew did affect
a hub as a hub did not retime the signals.
Rodgers Platt
|
In the context of the current cabling standards "delay skew" is referred
to as a UNIDIRECTIONAL parameter, so I think that you are talking about
DELAY, not DELAY SKEW. Of course, different delays on transmitting and
receiving pairs, if not controlled, can accumulate, especially across the
possible four segments of the Ethernet network based on hubs. I can
imagine that, with a cable bad enough, it could possibly get you to the
point where devices dis-synchronize, but I don't have an Ethernet textbook
handy to check on the actual timing values.
--
Dmitri Abaimov, RCDD
http://www.cabling-design.com
Cabling Forum, color codes, pinouts and other useful resources for
premises cabling users and pros
http://www.cabling-design.com/homecabling
Residential Cabling Guide
-------------------------------------
##-----------------------------------------------##
Article posted with Cabling-Design.com Newsgroup Archive
http://www.cabling-design.com/forums
no-spam read and post WWW interface to your favorite newsgroup -
comp.dcom.cabling - 1108 messages and counting!
##-----------------------------------------------## |
|
| Back to top |
|
|
|
|
|
|
FAQ
Memberlist
Register
Profile
Log in to check your private messages
Log in