🚀 go-pugleaf

In article <6sk60s$j7d$1@flea.best.net>, dillon@best.net writes...
> :In article <MPG.10572b78acbc04c39896ca@ispnews.walltech.com>,
> :Philip J. Koenig <For_Email@ddress.see.below_plea.se> wrote:
> :>In article <01bdd698$fe212500$8028a8c0@jb.zedak.com>, jb@zedak.keepyourspam.com 
> :>writes...
> :>> Daniel Erskine <netsup75@edwardjones.com> wrote in article
> :>> <35EACDBE.237D36B8@edwardjones.com>...
> :..
> :>
> :>The reason that using correct cable is important are as follows:
> :>
> :>1 - Terminators and NIC's are 75 ohm, if the cable is not, you end up
> :>    with reflections that play hell with the waveforms and thus, data 
> :>    integrity.
> 
>     10base2 uses 50 ohm coax and 50 ohm termination.


GACK... sorry my mistake.  Too many years working with cable TV stuff
I guess..  :)



> :>2 - The reason distance is a factor (and why the IEEE specifies a 
> :>    maximum length for 10base2 segments) is due to the rate of signal
> :>    propagation.  Since 802.3 collision detection relies on a known rate
> :>    of propagation and maximum segment length (tailoring it's "collision
> :>    listen window" accordingly), if you run out-of-spec cabling you run
> :>    the risk that collision-detection no longer works correctly, which
> :>    causes a host of problems including runt packets, late-collisions, 
> :>    etc.
> 
>     The collision window is not the maximum segment length,  It's the 
>     preamble period plus I think the MAC header (don't quite remember but
>     I think it's the first 64 bit times or something like that).  If
>     a collision occurs after that point, you get a 'late collision' which,
>     which detectable, is a very serious error.
> 
> 						-Matt
> 


Yes, "collision window" is not the maximum segment length.

To be more specific, the time allotted for listening for collisions is 
specific (I don't remember, possibly 64 bit times as you say) and it is
a direct function of A) the velocity of propagation in the coax, which
is specified in the 802.3 spec to be within a certain known range and 
B) the length of the segment.

Given standard "thinnet" cable, if the length is kept within the 802.3-
specified maximums, the propagation (or "round-trip") time is such 
that it does not exceed the "collision listen window".  The risk is of
course if the cable is too long, the transceivers on the wire will stop
listening for collisions on the wire when they may still be occurring.

Some NIC's had/have a mode for "extended segment length", it is not
technically 802.3 compliant.  And by the same token (no pun intended), 
if you really wanted to run a longer segment length, you could also run 
cable that had a faster velocity of propagation than the 802.3 standard
originally called for.

However in the absence of a thorough knowledge of the physics and 
transmission-line theory involved (and requisite test equipment), the
obvious solution is simply to use 802.3 compliant cable and adhere to
802.3-recommended maximum cable lengths. (and other recommendations
like maximum 30 taps per segment, no short stubs, etc.. :-)




Phil

-- 
Philip J. Koenig         The Electric Kahuna Organization          [see below]
-----------------Computers & Communications for the New Millenium-------------
References to my email address in this message have been modified to foil
address-collection robots.  If you wish to send email, use the following 
address by removing numbers and spaces:  pjkunet64 @  ekahuna27 . c o m