[MLB-WIRELESS] Cat5 max length

[Ben Anderson]

Actually, thats not quite correct either...

If you do the maths, 72% speed of light = 300,000,000m/s *.72 =
216,000,000m/s
Minimum jam signal duration is 32bits - and thats guaranteed to be long
enough to propegate to the colliding station before the end of the jam
signal.

At the .1us/bit (your number is good in this case, but only for 10mbit
ethernet) thats 3.2us or 3.2x10^-6 seconds

so 216x10^6 * 3.2*10^-6 = 691.2m maximim collision radius.
Practically, its less than this as one must take into account the delay of
the state machines in the phy and mac layers in the ethernet card.  The spec
allows for up to 4 hubs between end-end units, or 500m collision domain,
which agrees with my calculation+fudge factor extremely well.

The reason for the ~100m length is the signal-noise ratio - ie the amount of
bandwidth, or signal attenuation over that much cable.  So, in short, you
can use hubs to expand the radius of a single collision domain.  Using
higher quality cables also allow longer cable runs, as there is less signal
attenuation.

Also remember full duplex connections dont have any CSMA/CD as there is no
possibility for a collision.

the 802.3 spec is available for download from www.ieee.org for anyone whos
still interested.

Cheers,
Ben.


> The reason that the Cat5 standard mentions 100 metres is because of
Carrier
> Sense Media Allocation/Collision Detection or CSMA/CD. Or more to the
point CD.
> CSMA is tha part which listens to the connection, and if no-one is talking
it
> transmits (being the youngest of seven, this would have been very handy at
the
> dinner table when I was a kid). If CSMA hears a transmit, it waits for a
random
> amount of time, listens, and if all is well, transmits.
>
> CD (collision detection) comes in when both ends of the line hear nothing
and
> then decide to transmit at the same time. At this point both ends of the
> connection tell the other  (to use the dinner table analogy "Hey Pete, I
was
> talking here"). At this point both "talkers" wait a random number of time
and
> retransmit again.
>
> Now this is where it gets (only slightly) technical and I search back into
the
> memory banks of when I was A/NZ sales manager for Microtest Cable testers.
>
> Believe it or not, current travels on a wire at slightly less than the
speed of
> light. The speed of any cable is measured as an NVP (nominal velocity of
> propagation). NVP is the speed of the cable expressed as a percentage of
the
> speed of light in a vacuum. Most CAT 5 cable has an NVP of 70% - 75%.
Let's say,
> 72% (or, the speed of Craig Mead's Datsun on a Saturday night).
>
> Minimum frame size is 512 bits, each bit takes 0.1 microseconds (roughly
the
> time it takes a MLB-Wireless user to get a Galaxy dish off a house!) This
means
> that the minimum frame size takes 51.2 microseconds to complete.
Considering
> propagation times of various connectors and devices, 51.2 microseconds
means
> about 153 metres. Multiply this by the NVP of the cable and you get around
110
> metres.
>
> 100 metres is used because it is safe, a round number and easy to work
with.
>
> Probably could re-write this better, however it is early and the coffee
has not
> kicked in yet!
>
> -Michael
>
>
>
>
>
> To unsubscribe: send mail to majordomo at wireless.org.au
> with "unsubscribe melbwireless" in the body of the message
>
>



To unsubscribe: send mail to majordomo at wireless.org.au
with "unsubscribe melbwireless" in the body of the message