My first 10BASE5 network segment - test-drb@ccmp.vtda.org

26 Aug 2016

...
 On 26/08/2016 06:26, Warner Losh wrote:
  10base5 also had rules for minimum bend radius

True, because bending the cable alters the geometry and introduces
impedance discontinuities, though (to be picky) the allowable bend
radius varies between cable manufacturers because the precise cable
construction dictates the tightest bend that wouldn't upset the
impedance.  What I recall from the standard is that cables must support
a bend radius of 254mm /or less/ in order to be flexible enough for
reasonably easy installation.  A sort of "maximum minimum bend radius".

  as well as tap locations to be at the maxima of
the reflection point. 
Actually it's to /avoid/ maxima and thereby to ensure things are out of
phase, minimising adverse interference effects.  The node positions are
at 2.5m intervals, a distance which is chosen so that taps and
terminators are very unlikely to be exact wavelengths apart and hence
will /not/ be at maxima, so conflicting signals will be out of phase.
IIRC correctly it's deliberately not quite 1/19th of the wavelength.

For the same reason, cable sections are supposed to be odd multiples of
the half-wavelength of the signal (23.4m, etc).

> For early gear, failure to
> put it at a vibration node would often result in unreliable behavior, though
...
   I can't
recall if that included collisions or not. 
--
Pete
Pete Turnbull 
Yes, the whole reflection thing could get into a bit of a complex
discussion involving transmission line theory, but I am an electrical
engineer, so here goes:

Yes, you need proper termination at either end, or else you get the wrong
voltages on the line and cause 'collisions' (well, detect collisions
anyway, though there aren't really any until the first reflection comes
back from the cable end).

And yes, you're supposed to put the nodes 2.5m apart. With only a two node
network, and with a really short coax (<1m in my case here), the problem
this solves is practically nonexistant.

In theory, a properly terminated cable should have virtually no reflections
at either end of the cable. In reality, imperfections in the cable,
tolerances in your terminators, etc. can cause a small bit of reflection,
but it should be mostly negligible. At each node, however, the transceiver
places a small load on the coax, and some (small) amount of reflection will
occur here. The reason you want the 2.5m spacing is so that you stop any
nodes from being a perfect wavelength apart (23.4m minimum) and having
their reflections build on each other enough to cause a problem with the
signal levels and corrupt data (and trigger a collision).

Also, the wavelength of ethernet signalling is 23.4m. This is a full
wavelength, not a half as pete said. Ethernet is 10Mbit/second (more
accurately, also 10Mbaud), which means normally that the maximum frequency
of the raw data signal would actually be 5MHz, but since ethernet uses
manchester encoding, there are sequences of bits which can produce a 10MHz
signal, as well as sequences of bits which can produce a 5MHz signal. The
minimum propagation speed of 802.3 compliant 10BASE5 coax is (IIRC) .78C (C
being the speed of light, 3E8 m/s), so taking .78C/10MHz = 23.4m

Regards,
Joe Zatarski