[comp.sys.next] Connecting a NeXT to twisted-pair

rhp@INEL.GOV (Robert Powell) (12/13/89)

Our NeXTs arrived last Monday, so far our expectations have been exceeded.
However, the real test will come when we start developing software on them.
Our biggest problem right now is connecting our cubes to our ethernet.  It
is a problem because our building is only wired with twisted-pair thick-net
ethernet.  Our networking people tell me that I need a thick-net to thin-net
bridge or repeater, at a cost of around $2,600!  Has anyone else out there
done this type of connection?  If so, how did you do it and how much did you
pay?  Please note:  Our building is wired with TWISTED-PAIR thick-net, not
coaxial thick-net.  I would much rather use $2,600 towards memory, Frame, and
other useful things.  Any help would be greatly appreciated.

Bob

  /--------------------------------------------------------------------------\
/ Bob Powell      Internet: rhp@inel.gov      US Mail: P.O. Box 1625           \
| EG&G Idaho, Inc.        Phone: (208) 526-8107        M.S. 1206               |
\ Idaho National Engineering Laboratory                Idaho Falls, ID 83415   /

izumi@violet.berkeley.edu (Izumi Ohzawa) (12/13/89)

In article <341@egg-idINEL.GOV> rhp@INEL.GOV (Robert Powell) writes:
>
>It is a problem because our building is only wired with twisted-pair thick-net
>ethernet.  Our networking people tell me that I need a thick-net to thin-net
>bridge or repeater, at a cost of around $2,600!  Has anyone else out there

We, at UC Berkeley, have the same problem, and I am anxiously
waiting our computer center people to come up with a solution.

However, I noticed several articles back about a router/bridge
made out of a cheap PC XT clone + 2 Western Digital Ethernet cards
for under $800-900.  If this box can have 1 twisted-pair Ethernet
card and another thin-Ethernet card, it should do the job.

Whoever posted that article, would you clarify this point?
If you also know Model numbers of such Ethernet cards, please
let us know.
Thanks.

Izumi Ohzawa, izumi@violet.berkeley.edu

lee@uhccux.uhcc.hawaii.edu (Greg Lee) (12/13/89)

From article <341@egg-idINEL.GOV>, by rhp@INEL.GOV (Robert Powell):
" ...
" Our biggest problem right now is connecting our cubes to our ethernet.  It
" is a problem because our building is only wired with twisted-pair thick-net
" ethernet.  Our networking people tell me that I need a thick-net to thin-net
" bridge or repeater, at a cost of around $2,600! ...

I thought thick-net and twisted-pair ethernet were different animals.
There is a twisted-pair to thin-net converter in the DEC catalog
for, I think I recall, $400 for a pack of 8.  Something like that.
I don't know whether that would be suitable.

The subject of connecting thick-net to thin-net came up before
in this newsgroup, and it was said that you can just
stick them together directly with a connector.  Following is some
of that past discussion.
				Greg, lee@uhccux.uhcc.hawaii.edu

>From: jbn@glacier.STANFORD.EDU (John B. Nagle)
Newsgroups: comp.sys.next
Subject: Re: My ramblings on the NeXT machine
Message-ID: <17845@glacier.STANFORD.EDU>
Date: 15 Nov 88 18:09:46 GMT
References: <26812@ucbvax.BERKELEY.EDU>
Reply-To: jbn@glacier.UUCP (John B. Nagle)

In article <26812@ucbvax.BERKELEY.EDU> munson@renoir.Berkeley.EDU (Ethan V. Munson) writes:
>3) Thin Ethernet -- NeXT has a Thin Ethernet connector.  Here at Berkeley,
>	neither the EECS department or the academic computing service has
>	anything but thick ethernet.  As of last week, even NeXT didn't
>	know where to find converters.  Each converter will cost about
>	$200-300.  Since these machines are targeted at professors, rather
>	than students, it may be hard to build short Thin Ethernets which
>	share one converter.

       Thick and thin Ethernet cable can simply be interconnected with
coax adapters.  Losses are higher in the thin sections, but the adapters
themselves seem clean; I've looked at a mixed cable using a time-domain
reflectometer, and I can see tranceivers and sharply bent cable, but not
thick/thin adapters.  Usually, you convert from thick to thin inside a wall,
box, floor, or plenum, then bring two thin cables up to the machine, where
they meet at a T connector attached to the back of the machine.  Remember,
the T connector must be directly on the back of the machine; any cable
between the T and the computer causes reflections that trash data.  If
the entire cable is within the thin Ethernet length limit, everything
should be safe. 

      The appropriate adaptor is a constant-impedance UHF (female)
to BNC (female) adapter.  However, this is a hard-to-find adapter, and
it may be necessary to combine a UHF (female) to BNC (male) adapter
with a BNC-BNC barrel.  Amphenol part 12025 is the adapter.

					John Nagle


>From: jbn@glacier.STANFORD.EDU (John B. Nagle)
Newsgroups: comp.sys.next
Subject: Re: Followup to Who Has One?
Message-ID: <17916@glacier.STANFORD.EDU>
Date: 21 Dec 88 19:36:31 GMT
References: <18496@agate.BERKELEY.EDU>
Reply-To: jbn@glacier.UUCP (John B. Nagle)


     The simple, cheap solution to connecting thin to thick Ethernet is
just to get the appropriate connector hardware.  Thick Ethernet uses what
are called "UHF" connectors, and thin Ethernet uses "BNC connectors."
Adapters are available for under $10.  See a Newark, Zack, or Arrow
catalog for details.

     I've actually used such conversions, and couldn't see an impedance
mismatch at the connector when examining the cabling with a TDR.  Sharp
bends (4" radius) in the thick Ethernet cable cause more reflections
than a thick/thin transition.

     Remember how you hook it up.  BNC T-connector on the back of the 
machine, two thin Ethernet cables to the place where you reach the big
cable, two adapters to convert to thick Ethernet.  Never put a section
of thin Ethernet between the machine and the T-connector; the short
arm of the T can't be more than a few inches before everything stops
working due to reflections.

					John Nagle
>From: kchen@Apple.COM (Kok Chen)
Newsgroups: comp.sys.next
Subject: Re: Followup to Who Has One?
Message-ID: <22678@apple.Apple.COM>
Date: 22 Dec 88 02:31:10 GMT
References: <18496@agate.BERKELEY.EDU> <17916@glacier.STANFORD.EDU>
Organization: Apple Computer Inc, Cupertino, CA

In article <17916@glacier.STANFORD.EDU> jbn@glacier.UUCP (John B. Nagle) writes:
>     The simple, cheap solution to connecting thin to thick Ethernet is
>just to get the appropriate connector hardware.  Thick Ethernet uses what
>are called "UHF" connectors, and thin Ethernet uses "BNC connectors."
>Adapters are available for under $10.  See a Newark, Zack, or Arrow
>catalog for details.

No, no.  You want to use a "Type-N" connector, not a "UHF connector" for 
thick Ethernet.

There IS such a thing called the "UHF connector," but that is NOT what
you will find normally used on Ethernets.  Those ancient enough may remember
using the Tektronix 503 oscilloscope in undergrad labs (they were popular
in the '60's). THOSE animals used UHF connectors.  A "type-N" is VERY 
different, and definitely will not mate with a UHF type.  


Kok Chen			{decwrl,sun}!apple!kchen
Apple Computer, Inc.

P.S. Someone (a Harvard prof., no less [but Physics, not EE :-) ] ) once
     claimed to me that BNC (the connector used in thin-Ethernets) stood for 
     "Baby N-Connector."  Said person is also author of the "The Art of 
     Electronics" book you see in bookstores and once (40+ years ago :-)
     held the record for being the youngest amateur radio operator in W-land, 
     so he may be right. (I doubt it, Paul :-) :-))


>From: gore@eecs.nwu.edu (Jacob Gore)
Newsgroups: comp.sys.next
Subject: Re: Followup to Who Has One?
Message-ID: <12670010@eecs.nwu.edu>
Date: 22 Dec 88 04:02:26 GMT
Organization: Northwestern U, Evanston IL, USA

/ comp.sys.next / jbn@glacier.STANFORD.EDU (John B. Nagle) / Dec 21, 1988 /

>BNC T-connector on the back of the 
>machine, two thin Ethernet cables to the place where you reach the big
>cable, two adapters to convert to thick Ethernet.

Just to make sure I understand you correctly:  are you actually splicing
a piece of thin Ethernet into a thick Ethernet?

Jacob Gore				Gore@EECS.NWU.Edu
Northwestern Univ., EECS Dept.		{oddjob,gargoyle,att}!nucsrl!gore


>From: jbn@glacier.STANFORD.EDU (John B. Nagle)
Newsgroups: comp.sys.next
Subject: Re: thin/thick Ethernet
Message-ID: <17925@glacier.STANFORD.EDU>
Date: 22 Dec 88 18:08:22 GMT
References: <18496@agate.BERKELEY.EDU> <12670010@eecs.nwu.edu>
Reply-To: jbn@glacier.UUCP (John B. Nagle)
Organization: Stanford University
Lines: 26

In article <12670010@eecs.nwu.edu> gore@eecs.nwu.edu (Jacob Gore) writes:
>Just to make sure I understand you correctly:  are you actually splicing
>a piece of thin Ethernet into a thick Ethernet?

     Yes.  It works.  Loss per unit length is higher in the thin cable, but
unless you are pushing the length limits of the Ethernet cable, adding a
few feet of thin cable isn't that important.  You might think that the
transition would produce reflections, but my experience is that it doesn't.
Find a TDR and check for yourself.

     If you have a farm of NeXT machines or other small machines with thin
Ethernet connectors, it might be better to tie them all together on thin
cable and only have one thick/thin transition.  But this is not mandatory.

     Always check the machines at the ends of the cable for high retransmit
rates.  This is the first symptom of excessive losses, and indicates
that collisions are being undetected due to cumulative loss in the cable.
This means the cable is too long, or too lossy; time to get out a TDR or
put a repeater in the middle.

     If your thick/thin mixed cable falls within the length limitations for
thin Ethernets, everything should work fine.  Only beyond that length
should it be necessary to worry much about the electrical properties
of the cabling.

					John Nagle


>From: gore@eecs.nwu.edu (Jacob Gore)
Newsgroups: comp.sys.next
Subject: Re: Re: thin/thick Ethernet
Message-ID: <12670012@eecs.nwu.edu>
Date: 22 Dec 88 21:29:25 GMT
References: <17925@glacier.STANFORD.EDU>
Organization: Northwestern U, Evanston IL, USA
Lines: 23

/ comp.sys.next / jbn@glacier.STANFORD.EDU (John B. Nagle) / Dec 22, 1988 /
> In article <12670010@eecs.nwu.edu> gore@eecs.nwu.edu (Jacob Gore) writes:
> >Just to make sure I understand you correctly:  are you actually splicing
> >a piece of thin Ethernet into a thick Ethernet?
> 
>     Yes.  It works.

Oh, I'm not surprised that it works.  It just struck me as a rather
inconvenient thing to do.  When we have several thin-connector machines
near each other, we do connect them all with thin-net, as you suggested.

Except that our thick-cable networks are usually large-spread nets, with
multi-user hosts on them as well as scattered workstations.  We just don't
like splicing into that net, even for inline transceivers.

We use thick/thin repeaters for this kind of stuff.  A repeater with one
thick port and one thin port plus its transceiver and drop cable (for the
think side) cost us about $1,000.  If you are putting 5 or more stations on
the thin side, the cost is quite reasonable.  But for a single workstation,
it's rather steep...

Jacob Gore				Gore@EECS.NWU.Edu
Northwestern Univ., EECS Dept.		{oddjob,gargoyle,att}!nucsrl!gore


>From: jbn@glacier.STANFORD.EDU (John B. Nagle)
Newsgroups: comp.sys.next
Subject: Re: Re: thin/thick Ethernet
Message-ID: <17934@glacier.STANFORD.EDU>
Date: 23 Dec 88 17:01:57 GMT
References: <17925@glacier.STANFORD.EDU> <12670012@eecs.nwu.edu>
Reply-To: jbn@glacier.UUCP (John B. Nagle)
Organization: Stanford University
Lines: 12


      I certaintly agree that a repeater to isolate important multi-user
systems and servers from farms of little machines where the cables are
accessable to many people is a valuable addition to a network.  I was
merely pointing out that thich/thin Ethernet transitions are not difficult
to accomplish, should one wish to do so.

      Some people have found repeaters, as an active device and a single
point of failure, to cause more trouble than they prevent.  But that 
information is a few years old, and repeaters may be more reliable now.

					John Nagle

jacob@gore.com (Jacob Gore) (12/14/89)

/ comp.sys.next / izumi@violet.berkeley.edu (Izumi Ohzawa) / Dec 13, 1989 /
> I noticed several articles back about a router/bridge
> made out of a cheap PC XT clone + 2 Western Digital Ethernet cards
> for under $800-900.  If this box can have 1 twisted-pair Ethernet
> card and another thin-Ethernet card, it should do the job.
>
> Whoever posted that article, would you clarify this point?

I'm checking it out with the author of PCroute.

You should be able to get a single-port thick-thin repeater for well under
the $2,600 that Robert Powell quoted.  Try Cabletron -- if I remember
correctly, their educational price for one is about $800 (sorry, I have no
clue about the non-educational price).

Jacob
--
Jacob Gore		Jacob@Gore.Com			boulder!gore!jacob

izumi@violet.berkeley.edu (Izumi Ohzawa) (12/14/89)

In article <5719@uhccux.uhcc.hawaii.edu> lee@uhccux.uhcc.hawaii.edu (Greg Lee) writes:
>The subject of connecting thick-net to thin-net came up before
>in this newsgroup, and it was said that you can just
>stick them together directly with a connector.  Following is some
>of that past discussion.

Thank you for the collection of past articles.

Unfortunately, this splicing of Thick (Yellow) and thin (usually
black) Ethernet coax cables is not an option for many sites
with NeXT connection problems.
Our computer center would not allow this, and I would not
either if I were the caretaker of the network. 

The problem is NOT the connection to thick coax, but the
connection to the BLUE transiver cable (containing multiple
twisted-pairs) usually terminated by a 15-pin D connector.
It's really a connection problem to a "DELNI-net".

For this, some sort of active device is needed, and a simple connector
and a splicing kit won't do.

Izumi Ohzawa, izumi@violet.berkeley.edu

jacob@gore.com (Jacob Gore) (12/14/89)

In article <341@egg-idINEL.GOV> rhp@INEL.GOV (Robert Powell) writes:
>It is a problem because our building is only wired with twisted-pair thick-net
>ethernet.  Our networking people tell me that I need a thick-net to thin-net
>bridge or repeater, at a cost of around $2,600!  Has anyone else out there

/ comp.sys.next / izumi@violet.berkeley.edu (Izumi Ohzawa) / Dec 13, 1989 /
> ... I noticed several articles back about a router/bridge
> made out of a cheap PC XT clone + 2 Western Digital Ethernet cards
> for under $800-900.  If this box can have 1 twisted-pair Ethernet
> card and another thin-Ethernet card, it should do the job.
>
>Whoever posted that article, would you clarify this point?

OK, here it is.  It certainly can be done, but it's probably not the best
way to do it.

Your best bet is calling the supplier (or manufacturer) of your TP hub and
asking them if they have a TP to Thinnet converter.  I'm told that
Cabletron has one for their TP for about $500 (I don't know if that is
educational or "normal" price).

Also, I know HP sells (or at least it did last year) an adapter that hooks
in between a drop cable from a transceiver on one side and a Thinnet outlet
on the other.  Sorry, I don't have the part number or the price.  You would
still need a TP transceiver between the TP wire and the drop cable.

If you do want to use a PC for this, here are the considerations:

The way things are done now, you pretty much need to buy your TP
transceiver and the TP hub from the same manufacturer.  So, if you do buy a
TP Ethernet card from Western Digital, you have to make sure that it is
compatible compatible with your hub (i.e., ask them before you buy).

In the likely event that it's NOT compatible with your hub, you can simply
buy the standard card (WD8003E, a.k.a. "EtherCard Plus", I believe) and use
the TP transceiver from your hub's vendor and a drop cable to connect to
the Thicknet port on the card.  The card also has a Thinnet port, so you
can just buy two WD8003E.  (No, you can't get by with just one card, you
need one for each network.)

Unless you really want to have a separate IP subnet on your side of the PC,
instead of PCroute you should use its sibling, PCbridge.  It is available
from the same places (definitely accuvax.nwu.edu and probably
uunet.uu.net).  It is simpler to set up (no configuration required), and
you won't end up having to assign an IP subnet just for one NeXT -- it will
be on the same IP subnet as all the other hosts on that TP network.

For PCbridge you may want a faster PC than for PCroute, because unlike a
router, a bridge forwards every single packet it gets.  Packet forwarding
in PCroute/PCbridge is several times slower than packet filtering.

By the way, I should mention that you do NOT need a monitor card or a
monitor, as long as the boot PROMs of the PC clone you buy do not insist on
checking for one (most don't).

So, here are the details about how you can use a PC between networks (not
the only way to do it, there is other software too).  I personally would
opt for a TP to Thinnet adapter from the TP vendor if it was available, and
if not, I'd go for a single-port receiver from Cabletron (not the only
source, but my personal favorite).  A PC clone just seems like a rather
bulky transceiver...

Jacob
--
Jacob Gore		Jacob@Gore.Com			boulder!gore!jacob

jacob@gore.com (Jacob Gore) (12/14/89)

Those considering using PCroute or PCbridge should read Vance Morrison's
posting in comp.protocols.tcp-ip.ibmpc under subject "Re: Request for
Summary Info on 'PCRoute'" (<2117@accuvax.nwu.edu>).

Jacob
--
Jacob Gore		Jacob@Gore.Com			boulder!gore!jacob

hue@netcom.UUCP (Johathan Hue) (12/14/89)

In article <1989Dec14.035818.5544@agate.berkeley.edu> izumi@violet.berkeley.edu (Izumi Ohzawa) writes:
>The problem is NOT the connection to thick coax, but the
>connection to the BLUE transiver cable (containing multiple
>twisted-pairs) usually terminated by a 15-pin D connector.

You could always do what a user here tried.  Hook the transceiver up to the
DSP port.

Jonathan (my LANalyzer plays Bach) Hue