Sign in with
Sign up | Sign in
Your question

Sharing broadband across half mile

Last response: in Wireless Networking
Share
Anonymous
June 27, 2004 4:18:20 AM

Archived from groups: alt.internet.wireless (More info?)

I want to share my broadband connection with a friend who lives about
half a mile away.

I was hoping I could do this by installing antennas in both our lofts
and connecting it to a wireless access point connected to my Telewest
cable modem. There is a raised railway embankment between our two
locations, so direct line of sight is not possible. Is this a problem?

Can anyone tell me if this is a feasible option. Also, any
tips/websites where I can build the kit cheaply.
June 27, 2004 4:37:34 AM

Archived from groups: alt.internet.wireless (More info?)

Dave D wrote:
> I want to share my broadband connection with a friend who lives about
> half a mile away.
>
> I was hoping I could do this by installing antennas in both our lofts
> and connecting it to a wireless access point connected to my Telewest
> cable modem. There is a raised railway embankment between our two
> locations, so direct line of sight is not possible. Is this a problem?
>
> Can anyone tell me if this is a feasible option. Also, any
> tips/websites where I can build the kit cheaply.

You can buy ethernet cable for pretty cheap, but about every 1,000 feet
or so you need to use a hub (repeater).

Seriously, sometimes non-line-of-sight works when it isn't supposed to.
The only way you'll know is to try it.
Anonymous
June 27, 2004 8:37:44 AM

Archived from groups: alt.internet.wireless (More info?)

Dave D <frank.praverman@bigfoot.com> wrote in
news:ro0sd0dm1373tgjlv9k47v1utf03vj2204@4ax.com:

> There is a raised railway embankment between our two
> locations, so direct line of sight is not possible. Is this a problem?

If you are using a directional antenna with a focused beam pattern it will
likely cause a problem. Can you get your antennas any higher?

--
Lucas Tam (REMOVEnntp@rogers.com)
Please delete "REMOVE" from the e-mail address when replying.
http://members.ebay.com/aboutme/coolspot18/
Related resources
Anonymous
June 27, 2004 1:18:57 PM

Archived from groups: alt.internet.wireless (More info?)

On Sun, 27 Jun 2004 00:18:20 +0100, Dave D
<frank.praverman@bigfoot.com> wrote:

>Also, any
>tips/websites where I can build the kit cheaply.

http://www.freeantennas.com
http://www.saunalahti.fi/%7Eelepal/antennie.html
http://www.seattlewireless.net/index.cgi/DirectionalAnt...
http://www.seattlewireless.net/index.cgi/DirectionalPar...
http://www.qsl.net/k3tz/ (2.4GHz patch antenna)
http://flakey.info/antenna/biquad/
http://flakey.info/antenna/omni/quarter/ (don't)


--
Jeff Liebermann jeffl@comix.santa-cruz.ca.us
150 Felker St #D 831-336-2558
Santa Cruz CA 95060 AE6KS
Anonymous
June 27, 2004 1:22:27 PM

Archived from groups: alt.internet.wireless (More info?)

> cable modem. There is a raised railway embankment between our two
> locations, so direct line of sight is not possible. Is this a problem?

Chances are that it's a non starter as a direct link but there's another
way. This is what we had to do:-

www.nodomainname.co.uk/Equation/equation_broadband.htm

In other words, you'll either have to over or around.

> Can anyone tell me if this is a feasible option. Also, any
> tips/websites where I can build the kit cheaply.

www.nodomainname.co.uk/cantenna/cantenna.htm

David.
Anonymous
June 27, 2004 6:33:29 PM

Archived from groups: alt.internet.wireless (More info?)

In article <G_udncMcTa5l00PdRVn-sw@pghconnect.com>,
=?ISO-8859-1?Q?R=F4g=EAr?= <abuse@your.isp.com> wrote:
:You can buy ethernet cable for pretty cheap, but about every 1,000 feet
:o r so you need to use a hub (repeater).

Which ethernet standard were you thinking of, that would support
such long distances between repeaters? 10BaseTX, 100BaseT, 1000BaseT
are limited to 100 meters per segment. One of the older standards
(5Base2 or something like that) allows 220 meters per segment.
But I do not recall any standard that allows one to approach 300 meters ?
--
WW{Backus,Church,Dijkstra,Knuth,Hollerith,Turing,vonNeumann}D ?
Anonymous
June 27, 2004 6:33:30 PM

Archived from groups: alt.internet.wireless (More info?)

On 27 Jun 2004 14:33:29 GMT, roberson@ibd.nrc-cnrc.gc.ca (Walter
Roberson) wrote:

>In article <G_udncMcTa5l00PdRVn-sw@pghconnect.com>,
>=?ISO-8859-1?Q?R=F4g=EAr?= <abuse@your.isp.com> wrote:
>:You can buy ethernet cable for pretty cheap, but about every 1,000 feet
>:o r so you need to use a hub (repeater).

>Which ethernet standard were you thinking of, that would support
>such long distances between repeaters? 10BaseTX, 100BaseT, 1000BaseT
>are limited to 100 meters per segment. One of the older standards
>(5Base2 or something like that) allows 220 meters per segment.
>But I do not recall any standard that allows one to approach 300 meters ?

Standards are made to be conservative and following them is a good
assurance that things will work under all circumstances and
environments. Standards do not consist of a brick wall limit, where
everything falls apart if slightly exceeded. The trick is to know
just how conservative they are, and what controls the real limits. As
always, breaking standards are risky, educational, offers a few
suprises, is lots of fun, and must be tested thoroughly.

I have installed wired network links running well beyond the various
specifications. For example, I have a 10base2 coax cable network
running on RG-6/u (CATV coax) at 1200ft. No problems or errors. Note
that there are commerical products that also do this:
http://www.multilet.com/us/baseband/product_range/produ...

I routinely test 1000ft rolls of CAT5 cable by crimping connectors on
both ends and connecting them to a switch and laptop. No problems at
10baseT speeds, but not a great idea at 100baseTX. We had a
neighborhood wired ethernet system running 7 houses with about 500ft
of CAT5 daisy chained between houses. End to end was about 3000ft.
We tried to follow the 5-4-3 rule, but the topology was not very
helpful. I used switches instead of hubs (repeaters) and all was
well.

I've also used existing 25pair telco cable bundle at 600ft for
10baseT.

There are a bunch of other tricks to using overly long wiring
segments, and understanding their limitations, but since this is a
wireless newsgroup, methinks I'll avoid the inevitable flame war and
leave out the details.


--
Jeff Liebermann jeffl@comix.santa-cruz.ca.us
150 Felker St #D 831-336-2558
Santa Cruz CA 95060 AE6KS
Anonymous
June 27, 2004 7:39:00 PM

Archived from groups: alt.internet.wireless (More info?)

> But I do not recall any standard that allows one to approach 300 meters ?

No but we used to run our thin ethernet segments to 430m before they'd
fall over but then:-

a) The NIC's were manufactured by the company I worked for and they had
a special 300m mode (although most weren't set to run with that)

b) We were the tech support department

c) Had a time domain reflectometer to measure it when it all stopped

:) 

David.
Anonymous
June 27, 2004 8:14:49 PM

Archived from groups: alt.internet.wireless (More info?)

"David Taylor" <djtaylor@bigfoot.com> wrote in message
news:MPG.1b48ec5c6092976e989c8e@news.individual.de...
> > But I do not recall any standard that allows one to approach 300 meters
?
>

a thick ethernet segment (10Base5) is good for 500m - but this is thick
expensive co-ax, and you need separate transcievers.

> No but we used to run our thin ethernet segments to 430m before they'd
> fall over but then:-
>
> a) The NIC's were manufactured by the company I worked for and they had
> a special 300m mode (although most weren't set to run with that)

3com - their NIC cards for thin ethernet were certified to 300m
>
> b) We were the tech support department
>
> c) Had a time domain reflectometer to measure it when it all stopped
>
> :) 
>
> David.
--
Regards

Stephen Hope - return address needs fewer xxs
Anonymous
June 28, 2004 12:49:31 AM

Archived from groups: alt.internet.wireless (More info?)

"Walter Roberson" <roberson@ibd.nrc-cnrc.gc.ca> schreef in bericht
news:cbmlrp$80n$1@canopus.cc.umanitoba.ca...
> In article <G_udncMcTa5l00PdRVn-sw@pghconnect.com>,
> =?ISO-8859-1?Q?R=F4g=EAr?= <abuse@your.isp.com> wrote:
> :You can buy ethernet cable for pretty cheap, but about every 1,000 feet
> :o r so you need to use a hub (repeater).
>
> Which ethernet standard were you thinking of, that would support
> such long distances between repeaters? 10BaseTX, 100BaseT, 1000BaseT
> are limited to 100 meters per segment. One of the older standards
> (5Base2 or something like that) allows 220 meters per segment.
> But I do not recall any standard that allows one to approach 300 meters ?

10BASE5 runs up to 500 meters.
10BASE2 is certified to 195 meters.
Early fiber segments as far as 3500 meters.
Anonymous
June 28, 2004 12:52:36 AM

Archived from groups: alt.internet.wireless (More info?)

"David Taylor" <djtaylor@bigfoot.com> schreef in bericht
news:MPG.1b48ec5c6092976e989c8e@news.individual.de...
> > But I do not recall any standard that allows one to approach 300 meters
?
>
> No but we used to run our thin ethernet segments to 430m before they'd
> fall over but then:-
>
> a) The NIC's were manufactured by the company I worked for and they had
> a special 300m mode (although most weren't set to run with that)
>
> b) We were the tech support department
>
> c) Had a time domain reflectometer to measure it when it all stopped
>
> :) 
>
> David.

195 m for 10base2 was rather restrictive. If you knew what you were doing
(i.e.
use good quality cables, connectors and terminators) then it is not to
difficult
to calculate the maximum possible cable length. Having a TDR does help :-)
A TDR used to be pretty expensive around 1985, how are prices these days?

Hans
Anonymous
June 28, 2004 12:55:26 AM

Archived from groups: alt.internet.wireless (More info?)

"Jeff Liebermann" <jeffl@comix.santa-cruz.ca.us> schreef in bericht
news:5uotd0hfdeqsdmmgih4got8i9ook1lde3m@4ax.com...
> On 27 Jun 2004 14:33:29 GMT, roberson@ibd.nrc-cnrc.gc.ca (Walter
> Roberson) wrote:
>
> >In article <G_udncMcTa5l00PdRVn-sw@pghconnect.com>,
> >=?ISO-8859-1?Q?R=F4g=EAr?= <abuse@your.isp.com> wrote:
> >:You can buy ethernet cable for pretty cheap, but about every 1,000 feet
> >:o r so you need to use a hub (repeater).
>
> >Which ethernet standard were you thinking of, that would support
> >such long distances between repeaters? 10BaseTX, 100BaseT, 1000BaseT
> >are limited to 100 meters per segment. One of the older standards
> >(5Base2 or something like that) allows 220 meters per segment.
> >But I do not recall any standard that allows one to approach 300 meters ?
>
> Standards are made to be conservative and following them is a good
> assurance that things will work under all circumstances and
> environments. Standards do not consist of a brick wall limit, where
> everything falls apart if slightly exceeded. The trick is to know
> just how conservative they are, and what controls the real limits. As
> always, breaking standards are risky, educational, offers a few
> suprises, is lots of fun, and must be tested thoroughly.
>
> I have installed wired network links running well beyond the various
> specifications. For example, I have a 10base2 coax cable network
> running on RG-6/u (CATV coax) at 1200ft. No problems or errors. Note
> that there are commerical products that also do this:
> http://www.multilet.com/us/baseband/product_range/produ...
>
> I routinely test 1000ft rolls of CAT5 cable by crimping connectors on
> both ends and connecting them to a switch and laptop. No problems at
> 10baseT speeds, but not a great idea at 100baseTX. We had a
> neighborhood wired ethernet system running 7 houses with about 500ft
> of CAT5 daisy chained between houses. End to end was about 3000ft.
> We tried to follow the 5-4-3 rule, but the topology was not very
> helpful. I used switches instead of hubs (repeaters) and all was
> well.
>
> I've also used existing 25pair telco cable bundle at 600ft for
> 10baseT.
>
> There are a bunch of other tricks to using overly long wiring
> segments, and understanding their limitations, but since this is a
> wireless newsgroup, methinks I'll avoid the inevitable flame war and
> leave out the details.
>
Jeff, a little theory does not harm anybody. Wired or wireless, technology
obeys the laws of physics, right?
Hans
Anonymous
June 28, 2004 12:55:27 AM

Archived from groups: alt.internet.wireless (More info?)

On Sun, 27 Jun 2004 20:55:26 +0200, "Hans Vlems"
<hvlems.dotweg@zonnet.nl> wrote:

>> There are a bunch of other tricks to using overly long wiring
>> segments, and understanding their limitations, but since this is a
>> wireless newsgroup, methinks I'll avoid the inevitable flame war and
>> leave out the details.

>Jeff, a little theory does not harm anybody. Wired or wireless, technology
>obeys the laws of physics, right?
>Hans

Ok, I have an audience. Good excuse to not do the laundry.

First, skim through the ethernet FAQ at:
http://www.faqs.org/faqs/LANs/ethernet-faq/
There's quite a bit in there on standards and calcs.

The big headaches are NEXT (near end crosstalk) and line loss. While
this is mostly a function of the CAT5 cable, it's also seriously
affected by the sensitivity of specific switches and ethernet cards.
I gotta dig through my notes for the exact numbers. Maybe later.

In general, FDX (full-duplex) vs HDX (half-duplex) is a compromise.
FDX removes collision issues from the problem, but is affected by NEXT
because it simultaneously transmits and receives. Forget about using
hubs, which cannot do FDX. For very long end to end runs (1000
meters) through multiple switches, collision timing is the major issue
and therefore should use FDX. For shorter single runs of about 300
meters, cable loss and NEXT are the issues. Therefore HDX should be
used.

Late collisions are a problem. This is where the 32 bit jam signal
arrives after the first 512 bits of an ethernet frame. For CAT5, with
a velocity factory of 0.59, the speed of propogation is 177,000 km/sec
(300,000km/sec * 0.59). At 10mbits/sec, one bit = 17.7 meters.
512bits of the header results in a 9000 meters ( 512 * 17.7) limit for
late collisions. Delays through the switches are not an issue because
both the original packet and the jam signal are delayed identically.

For coax, there are two solutions. One is to use a twisted pair to
coax balun (xformer) and the 10baseT collision detection mechanism as
in:
http://www.multilet.com/us/baseband/product_range/produ...
The velocity factor is 0.65 which results in a slightly longer limit
of 10,000 meters.

Another coax solution is to use 10base2 (cheapernet), which uses a DC
level shift for collision detection. The official end-to-end maximum
is 1000 meters which methink is rather conservative. The correct
cable is 50 ohm RG-58/u. However, I've been using RG-6/u 75 ohm coax.
In theory, there would be substantial reflection problems with voltage
nulls and peaks along the coax. Therefore, it only works on
relatively long runs (>100 meters) where the reflected signal is
sufficiently attenuated to be insignificant. I only use end-to-end
runs of RG-6/u with no "tee" connectors for additional machines in
between. Therefore the topology is a star, not the more cenventional
bus. For interface, I prefer 10baseT to 10base2 converters, but will
use an old ISA 10baseT ethernet card if desperate.

There are a few more complications, but I gotta get back to the
laundry.

--
Jeff Liebermann jeffl@comix.santa-cruz.ca.us
150 Felker St #D 831-336-2558
Santa Cruz CA 95060 AE6KS
Anonymous
June 28, 2004 3:37:30 AM

Archived from groups: alt.internet.wireless (More info?)

"Jeff Liebermann" <jeffl@comix.santa-cruz.ca.us> schreef in bericht
news:v27ud0pveqfnlcgemnfgqs9n792naa1bd8@4ax.com...
> On Sun, 27 Jun 2004 20:55:26 +0200, "Hans Vlems"
> <hvlems.dotweg@zonnet.nl> wrote:
>
> >> There are a bunch of other tricks to using overly long wiring
> >> segments, and understanding their limitations, but since this is a
> >> wireless newsgroup, methinks I'll avoid the inevitable flame war and
> >> leave out the details.
>
> >Jeff, a little theory does not harm anybody. Wired or wireless,
technology
> >obeys the laws of physics, right?
> >Hans
>
> Ok, I have an audience. Good excuse to not do the laundry.
>
> First, skim through the ethernet FAQ at:
> http://www.faqs.org/faqs/LANs/ethernet-faq/
> There's quite a bit in there on standards and calcs.
>
> The big headaches are NEXT (near end crosstalk) and line loss. While
> this is mostly a function of the CAT5 cable, it's also seriously
> affected by the sensitivity of specific switches and ethernet cards.
> I gotta dig through my notes for the exact numbers. Maybe later.
>
> In general, FDX (full-duplex) vs HDX (half-duplex) is a compromise.
> FDX removes collision issues from the problem, but is affected by NEXT
> because it simultaneously transmits and receives. Forget about using
> hubs, which cannot do FDX. For very long end to end runs (1000
> meters) through multiple switches, collision timing is the major issue
> and therefore should use FDX. For shorter single runs of about 300
> meters, cable loss and NEXT are the issues. Therefore HDX should be
> used.
>
> Late collisions are a problem. This is where the 32 bit jam signal
> arrives after the first 512 bits of an ethernet frame. For CAT5, with
> a velocity factory of 0.59, the speed of propogation is 177,000 km/sec
> (300,000km/sec * 0.59). At 10mbits/sec, one bit = 17.7 meters.
> 512bits of the header results in a 9000 meters ( 512 * 17.7) limit for
> late collisions. Delays through the switches are not an issue because
> both the original packet and the jam signal are delayed identically.
>
> For coax, there are two solutions. One is to use a twisted pair to
> coax balun (xformer) and the 10baseT collision detection mechanism as
> in:
> http://www.multilet.com/us/baseband/product_range/produ...
> The velocity factor is 0.65 which results in a slightly longer limit
> of 10,000 meters.
>
> Another coax solution is to use 10base2 (cheapernet), which uses a DC
> level shift for collision detection. The official end-to-end maximum
> is 1000 meters which methink is rather conservative. The correct
> cable is 50 ohm RG-58/u. However, I've been using RG-6/u 75 ohm coax.
> In theory, there would be substantial reflection problems with voltage
> nulls and peaks along the coax. Therefore, it only works on
> relatively long runs (>100 meters) where the reflected signal is
> sufficiently attenuated to be insignificant. I only use end-to-end
> runs of RG-6/u with no "tee" connectors for additional machines in
> between. Therefore the topology is a star, not the more cenventional
> bus. For interface, I prefer 10baseT to 10base2 converters, but will
> use an old ISA 10baseT ethernet card if desperate.
>
> There are a few more complications, but I gotta get back to the
> laundry.
>
Well, you've got the laundry and here it's nearly midnight, so just a short
comment
on impedance. 10base2 uses 50 ohms coax, and on a site I used to work at an
electrical engineer wired up his own office with RG61 (91 ohm impedance).
We used repeaters then (DEC's multiport repeaters) and all other segments
connected
to that same repeater started to have problems.
The question is: can you get away with non-standard coax provided it is the
only type
you use?
Even then, connecting 50 ohm gear to a cable plant with a different
impedance is just
asking for problems. Network load is important too, lightly used LAN's can
get away
with cutting some corners.
Anonymous
June 28, 2004 3:37:31 AM

Archived from groups: alt.internet.wireless (More info?)

On Sun, 27 Jun 2004 23:37:30 +0200, "Hans Vlems"
<hvlems.dotweg@zonnet.nl> wrote:

>> Late collisions are a problem. This is where the 32 bit jam signal
>> arrives after the first 512 bits of an ethernet frame. For CAT5, with
>> a velocity factory of 0.59, the speed of propogation is 177,000 km/sec
>> (300,000km/sec * 0.59). At 10mbits/sec, one bit = 17.7 meters.
>> 512bits of the header results in a 9000 meters ( 512 * 17.7) limit for
>> late collisions. Delays through the switches are not an issue because
>> both the original packet and the jam signal are delayed identically.

My last line is more than a little bit incoherent. What I meant to
mumble was that a late collision will cause a switch to propogate a
corrupted packet through the switch. Jam packets and packets
corrupted by the jam signal normally do NOT go through a switch.
(They do go through a repeater/hub). However, if the jam packet is
late by 64 bytes (header), the switch will merrily pass the following
data frame even if was intentionally trashed by the jam signal. This
is why it is a good idea to limit collision domains with switches.

>Well, you've got the laundry and here it's nearly midnight, so just a short
>comment
>on impedance.

I change my mind. It was about 1PM here and I grab lunch and catch up
on my magazine reading pile.

>> Another coax solution is to use 10base2 (cheapernet), which uses a DC
>> level shift for collision detection. The official end-to-end maximum
>> is 1000 meters which methink is rather conservative. The correct
>> cable is 50 ohm RG-58/u. However, I've been using RG-6/u 75 ohm coax.

Argh. I forgot to mention that you must use 50 ohm terminators, not
75 ohm. The DC levels used by the 10base2 collision avoidance
mechanism uses the terminator as part of a voltage divider. (Well,
it's actually a current source driving a fixed 50 ohm load). If you
use 75 ohm terminators, the transceiver will think it's permanently
busy.

>10base2 uses 50 ohms coax, and on a site I used to work at an
>electrical engineer wired up his own office with RG61 (91 ohm impedance).
>We used repeaters then (DEC's multiport repeaters) and all other segments
>connected
>to that same repeater started to have problems.

That's considerably different from running long lengths of RG-6/u.

I think you mean RG-62 which is 93 ohm coax normally used for Arcnet.
To the best of my limited experience, DEC never made an Arcnet
repeater (hub) so my guess is that it was either DECnet or ethernet.
I don't think it will work. My guess(tm) what may have gone wrong.

1. 93 ohms is a 2:1 VSWR from the nominal 50 ohms. That's a bit
much. That will create standing waves (voltage peaks and nulls) along
the length. A bad choice of cable length and the end points could
have ended up in a null resulting in no signal delivered.
2. For short lengths and 2:1 VSWR there's going to be a substantial
reflected signal, which will interfere with data transmission. If the
length of the coax segments were considerablly longer, the line
attenuation would absorb most of the reflections.
3. 93 ohm coax is rather lossy.
4. Hopefully, he terminated each segment with a 50 ohm terminator,
not a 93 ohm arcnet terminator.

Without knowning the specific topology and lengths, it's difficult to
speculate beyond the obvious potential problems. 75 ohm to 50 ohm is
only a 1.5:1 VSWR, which is tolerable. 93 ohm to 50 ohm is about 2:1
VSWR, which methinks is too much.

>The question is: can you get away with non-standard coax provided it is the
>only type you use?

I think you're asking can you mix different types and impedances of
coax. Sure, as long as they are the same impedance on a given
segment. I would not mix coax impedances on a given segement. Of
course different ports (segments) on a switch can be different.

>Even then, connecting 50 ohm gear to a cable plant with a different
>impedance is just
>asking for problems. Network load is important too, lightly used LAN's can
>get away
>with cutting some corners.

Two of my RG-6/u installs were at a tv production studio and a
broadcast studio. The network was carrying digitized video traffic,
spots from the satellite feed, streaming content, giant bloated files,
and a VoIP intercom. This was not a lightly used network. There were
two seperate networks in each building. One for user traffic and one
for server to server replication and backup. I used RG-6/u coax
because there was plenty of it in the cable trays and a fiber upgrade
was allegedly planned. They were suppose to be temporary, but ran
without incident for about 4 years. Two other TV stations and studios
learned of my method and implimented similar temporary solutions.
Incidentally, I monitor both of them remotely with various SNMP tools
and generate MRTG graphs of key traffic statistics and errors. Both
had successfully saturated the LAN at times, but SNMP showed minimal
errors at the central ethernet switch.

Hmmm... laundry. Never mind.

--
Jeff Liebermann jeffl@comix.santa-cruz.ca.us
150 Felker St #D 831-336-2558
Santa Cruz CA 95060 AE6KS
Anonymous
June 30, 2004 2:10:47 AM

Archived from groups: alt.internet.wireless (More info?)

> 3com - their NIC cards for thin ethernet were certified to 300m

No it wasn't 3com but either way it's like anything, the standard is
there for a reason, go outside it and you're out of bounds and on your
own.

It used to amuse me the conversations that we had with customers who'd
ask "but will it work if I go another 10 metres?"

Stock answer "It will be outside spec." We weren't prepared to endorse
out of spec networks whether they worked or not.

David.
Anonymous
June 30, 2004 2:17:28 AM

Archived from groups: alt.internet.wireless (More info?)

Jeff, you haven't mentioned the electrical effects of increasing line
lenght by increasing resistance and capacitance either yet.

Increasing these causes the rising voltage to equate to Ve(-t/CR) and
decaying voltage of V(1-e(-t/CR)) and so instead of nicely ramped
waveforms you end up with exponential curves that drag out over time and
may fail to trigger the threshold voltage of the receiver.

David.
Anonymous
June 30, 2004 2:20:18 AM

Archived from groups: alt.internet.wireless (More info?)

> use good quality cables, connectors and terminators) then it is not to

:-) Good quality? You should have seen some of the shite attempts at
making cables by junior tech support staff!

> to calculate the maximum possible cable length. Having a TDR does help :-)
> A TDR used to be pretty expensive around 1985, how are prices these days?

Dunno, pretty cheap these days I'd have though give the processing
advances. We had a couple, one was a rather large instrument about
twice the size of a current laptop and interfaced with other instruments
and the other was basically a "cable tester" but still did the TDR
function in order to calculate distance to fault.

David.
Anonymous
June 30, 2004 3:42:44 AM

Archived from groups: alt.internet.wireless (More info?)

On Tue, 29 Jun 2004 22:17:28 +0100, David Taylor
<djtaylor@bigfoot.com> wrote:

>Jeff, you haven't mentioned the electrical effects of increasing line
>lenght by increasing resistance and capacitance either yet.

CAT5 and coax look like a distributed LC network also known as a
transmission line. Attenuation follows the typical 3dB per octave
(10dB/decade) rolloff curve. Actually attenuation depends mostly on
the loss tangent of the dielectric. 10Mbits/sec needs at least the
3rd harmonic to be properly decoded. That means 30-50MHz of -3dB
bandwidth is usually sufficient depending upon the sensitivity of the
ethernet receiver. While dielectric losses are directly proportional
to the frequency, resistive losses are proportional to the square root
of the frequency, making resistive effects less signifigant.

>Increasing these causes the rising voltage to equate to Ve(-t/CR) and
>decaying voltage of V(1-e(-t/CR)) and so instead of nicely ramped
>waveforms you end up with exponential curves that drag out over time and
>may fail to trigger the threshold voltage of the receiver.
>
>David.

Ethernet waveforms are suppose to square waves, not ramps, although an
ugly triangular waveform is what usually comes out of a bandwidth
limited coax cable. Should be a photo around somewhere...
http://www.bitzenbytes.com/101/n115-an-442.pdf
See figure 5. I've seen far worse come out of a very long piece of
coax and still get decoded.

At 300 meters RG-6/u has a loss of:
MHz Loss (-dB)
1 2.5
10 8.5
30 15.
50 20.
I gotta check my overpriced printed copy of the ANSI/TIA/EIA-568-B.2
(addendum 1) specs to see what propogation delays, delay skewing, and
insertion loss can be officially tolerated. I suspect that 300 meters
is well over the loss spec and only the conservative design of the
transcievers and conservative cable specifications is what allows it
to work. With -10dB loss at 10Mhz

Hmmm... I found a partial reference to allowable skewing for EIA-568A:
"Cable delay skew shall not exceed 45ns/100m between 1 MHz and the
highest frequency in any specified category." At 10Mbits/sec, that's
not going to be a problem even to 300 meters. At 100baseTX, it's
probably way over the spec.


--
# Jeff Liebermann 150 Felker St #D Santa Cruz CA 95060
# 831.336.2558 voice http://www.LearnByDestroying.com
# jeffl@comix.santa-cruz.ca.us
# 831.421.6491 digital_pager jeffl@cruzio.com AE6KS
Anonymous
June 30, 2004 8:50:20 PM

Archived from groups: alt.internet.wireless (More info?)

> Ethernet waveforms are suppose to square waves, not ramps, although an
> ugly triangular waveform is what usually comes out of a bandwidth

Yes except that a square wave produces ugly harmonics hence the aim
shouldn't be a square wave but a very rapid rising voltage but not quite
square and that was all I was referring to. However, that's splitting
hairs and doesn't add to the discussion. My interest/recollection of
transmission line theory appears to be nowhere near as good as yours. :) 

> highest frequency in any specified category." At 10Mbits/sec, that's
> not going to be a problem even to 300 meters. At 100baseTX, it's
> probably way over the spec.

Yes and in 1985 100Mbps wasn't exactly run of the mill 10base2.

David.
Anonymous
July 1, 2004 4:55:25 AM

Archived from groups: alt.internet.wireless (More info?)

On Wed, 30 Jun 2004 16:50:20 +0100, David Taylor
<djtaylor@bigfoot.com> wrote:

>> Ethernet waveforms are suppose to square waves, not ramps, although an
>> ugly triangular waveform is what usually comes out of a bandwidth

>Yes except that a square wave produces ugly harmonics hence the aim
>shouldn't be a square wave but a very rapid rising voltage but not quite
>square and that was all I was referring to.

Well, I know what you're trying to say, but you're doing it all wrong.
Fourier worked it out 200 years ago.

A square wave is composed of a fundamental sine wave and a mess of odd
order harmonics, all of which have a zero degree phase shift to the
fundamental. For an ideal 50% duty cycle square wave, 1/3 of the
power it tied up in these harmonics. Varying the duty cycle move some
of the fundamental power into the harmonics. The narrower the pulse
width, the more harmonics.

Drivel: If the waveform is asymmetrical, there are even harmonics
present.

What a coax or twisted pair cable does is screw things up in two ways,
amplitude and phase. In amplitude, the -10dB/decade frequency rolloff
of the typical coax cable causes more attenuation at higher
frequencies. Therefore, the harmonics get reduced in output more than
the fundamental. The resultant waveform ends up with longer
(non-vertical) rise and fall times, which are the high frequency
components of the waveform. (The flat parts are the low frequency
components). Eventually it starts to resemble a trapazoid or triangle
instead of a square wave. Fig 5 in the previous URL is a good example
of what comes out.

The phase shift of the cable is more pronounced at higher frequencies
than at lower frequencies. This causes the harmonic components of the
square wave to shift in phase relative to the fundamental sine wave.
This results in a rather ugly waveform. However, if the cable is
properly matched, and of good quality, the degree of phase shifting is
fairly nominal for 10baseT and little distortion results. As long as
the 3rd and perhaps the 5th harmonics are in phase and present, the
waveform will sufficiently resemble a square wave to be decoded
properly by the 10baseT or 10base2 transceiver. The higher harmonics
are of little consequence.

Incidentally, I find harmonics to be quite beautiful and not at all
ugly.

>However, that's splitting
>hairs and doesn't add to the discussion. My interest/recollection of
>transmission line theory appears to be nowhere near as good as yours. :) 

Y'er doing fine. However, I cheat. In a past life, I played RF
engineer for a variety of radio and telecom companies.

>> highest frequency in any specified category." At 10Mbits/sec, that's
>> not going to be a problem even to 300 meters. At 100baseTX, it's
>> probably way over the spec.

>Yes and in 1985 100Mbps wasn't exactly run of the mill 10base2.
>David.


--
# Jeff Liebermann 150 Felker St #D Santa Cruz CA 95060
# 831.336.2558 voice http://www.LearnByDestroying.com
# jeffl@comix.santa-cruz.ca.us
# 831.421.6491 digital_pager jeffl@cruzio.com AE6KS
Anonymous
July 1, 2004 1:12:02 PM

Archived from groups: alt.internet.wireless (More info?)

> Incidentally, I find harmonics to be quite beautiful and not at all
> ugly.

Can I scream RFI?! :) 

> Y'er doing fine. However, I cheat. In a past life, I played RF
> engineer for a variety of radio and telecom companies.

My interest when I did any of this was as a digital electronics hardware
designer and obviously, we're only interested in avoiding those
harmonics and getting the nice clean 1's and 0's. If in doubt sample
the bugger and process it to death. ;) 

Last time I did FFT's was oooh, about 20 something years ago and I
wasn't particularly interested in them at the time for some bizarre
reason, think it was girls. :) 

David.
!