<< Previous Message Main Index Next Message >>
<< Previous Message in Thread This Month Next Message in Thread >>
Date   : Mon, 29 Mar 2004 00:22:25 +0100 (BST)
From   : Pete Turnbull <pete@...>
Subject: Re: MDFS incomplete startup

On Mar 28, 13:52, Johan Heuseveldt wrote:

> Actually there are terminators on board: All lines have a serial
> resistor of 100K and pull up of 10K.

No, those are bias for the comparators, not terminators.  The
characteristic impedance of the Econet bus is 110 ohms.  The serial
resistors are there to make the whole thing high impedance, and to
*prevent* them having any effect on the line, like termination.  Look
at any data sheet or example circuit which shows how to use opamps or
comparators and you'll see what they're for.

In fact, Issue 2 and Issue 3 BBC B boards *do* have places to fit
termination components (and a clock) but Acorn removed them from later
boards because they caused a lot of trouble (think what happens if you
have extra terminators, or two clocks; or if you add an extra machine
to the end of the net; or even have the right number in the right
places but one of the machines is switched off).

> Compare the connection to the net with Ethernet Coax, using T-pieces.
But
> the 'T-piece' is not at the machine side but at the network side,
leaving
> some length which cannot benefit fully from the main network
termination.
> Normal BBC connection leads are 1m to 1.20m. The connection lead for
the
> Risc PC is almost 2m!.
>
> So, every connection lead between net and machine is a small
> subnetconnection which is not part of the main net in this context of
> termination. The network side is already ok, but then the machine's
side
> needs proper termination too!!
> So a small termination at the machine side /is/ necessary!

No, you're wrong.  You're not allowed/supposed to have stubs more than
1m long on an Econet network.  That's comparable to thin Ethernet,
which allows stubs of 15cm (6") -- almost all of which is taken up by
the "equivalent length" of the interface.  You can get away with more
on Econet because the speed (pulse width, actually) is an order of
magnitude slower.

As for termination on the machine side, if you understood what
termination does, and why it is necessary, you'd know it must match the
characteristic impedance of the cable -- 50 ohms for Ethernet, 110 ohms
for Econet.  Those resistors you see have nothing at all to do with
termination.

The Econet or Ethernet cable is a transmission line.  As with any
transmission line, a signal passed along it will be reflected when it
reaches the end.  This reflection is A Bad Thing; to put it
simplistically, the point of the termination is to "soak up" the signal
so it doesn't get reflected.  To do this it must be the same impedance
as the impedance of the line.

> It is small enough to prevent too much load on the network cable with
many
> machines connected. So termination in the machine is as small as
possible to
> allow for some length of the connection lead.

The length has nothing to do with it!  The resistors you mention are,
however, there to minimise the load.  They're designed so the interface
places as little load as possible on the line; they are there to do
exactly the opposite of providing any "termination".

> > BTW, phone cable *is* twisted, just not nearly as tightly as proper
> > Econet cable, or Cat 3/4/5/5e.
>
> It is/was very simple cable, also usable for telephone purposes, and
not
> twisted at all. Very cheap.

All the telephone cable I've seen, except of course flat connection
cord, is twisted, and indeed the BT spec requires it to have a twist
rate not less than a certain amount.  Even common alarm cable (which is
often stranded, telephone cable is not) has some twist.

-- 
Peter Turnbull
<postmaster@...                 >
<< Previous Message Main Index Next Message >>
<< Previous Message in Thread This Month Next Message in Thread >>