Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!mailrus!tut.cis.ohio-state.edu!gem.mps.ohio-state.edu!apple!motcsd!hpda!hpcupt1!hprnd!pat
From: pat@hprnd.HP.COM (Pat Thaler)
Newsgroups: comp.dcom.lans
Subject: Re: Grounding Ethernet coax lines.
Message-ID: <2230027@hprnd.HP.COM>
Date: 11 Oct 89 17:02:31 GMT
References: <3824.25292578@zeus.unl.edu>
Organization: HP Roseville Networks Division
Lines: 71

> 
> >Wrong. A given Ethernet (thick or thin) *must* be grounded *EXACTLY ONCE*,
> >no more, no less.  To do other wise can *increase* your noise problems or
> >even -- are you ready for this? -- melt your cable!
> 
> Hmm. The thin ethernet cables I'm familiar with all have two ends. Neither
> end is distinguished from the other. Both have the coaxial braid firmly
> attached ( grounded ) to the connector ... I don't see how you could only
> ground it ONCE. It has to be connected to two devices.

The connector is attached to the shield of the cable, but the connector
is not ground.  Ethernet devices are required to provide DC isolation
between the cable shield and ground.  The IEEE 802.3 standard does 
specify a low AC impedance to ground, but there are to be no DC ground
loops. The cable is only to be grounded in at most one place. The connectors
are to be protected from inadvertant contact with ground in other places.

There is an ECMA LAN Safety Standard ECMA 97 which recommends the level
of isolation depending on the situation. In a nutshell, when in a single
building with one ground system, 500 V. When going between buildings,
2250 Vdc (or alternate AC or impluse tests). See the standard for elaboration.
> 
Stuff about ground potential differences within a building deleted.
> 
> Here's my understanding of [thin] ethernet :
> 
> 
> 	------------------------------------------------------------	[1]
> 	============================================================	[2]
> 	------------------------------------------------------------	[1]
> 
> 	[1]	coaxial cable, external braid, reference voltage
> 	[2]	coaxial cable, shielded inner conductor
> 
> ... where the data is conveyed via [1], and said data is kept from being
> radiated by [2], which is grounded.
> 
> [1] connects to the transceiver. [2] connects to the transceiver, perhaps the
> case, but the transceiver, nonetheless.

Actually, the shield of the cable is not used as a "shield" in the strict
sense of the word. It is signal return. The transmitter sends a current
into the inner conductor and it returns on the shield.  There is no
radiation (well, little radiation) because the two currents are balanced.
Since there might be some imbalance, an _AC_ ground is provided but not
a DC ground.

A true shield would not carry signal currents and would require using
triax or some form of cable with two conductors inside a shield. 

> 
> This is a complementary situation to that found in RS232, where
> 
> 	1	is frame ground, and
> 	7	is signal ground
> 
> ... and while you may or may not need [1], it being optional, [7] is mandatory,
> as There Can Be No Signal Without (some sort of) Ground.
> 
The difference is that RS232 uses a single-ended signal.  That is, the 
receivers look at the signal received relative to their ground.  LANs
normally use differential signals since they are designed to cover
greater distances and link multiple devices.  RS-422 is also differential.
A differential receiver looks at the potential difference between
two signal lines.  Most coax Ethernet receivers do this by having
the internals of the transceiver isolated from the chassis ground
and referenced to the shield of the coax.  Most twisted-pair Ethernet
transceivers use an isolating transmformer at the twisted-pair to 
remove any common mode offset.

Pat Thaler