Two years ago, I bought a house that already had coaxial plates installed on almost each rooms. Each plate uses a single wire (no T conectors are used), which makes the whole installation a physical star coaxial network.
Of course, unless I plan to buy 18 TVs and put them everywhere in my house, this setup is kind of useless. However, if I could use these coax wires as an ethernet network, it could help me transform my not- so-reliable wireless computers network to a fast reliable wired network.
I've found RJ-to-coax adapters that seem to be able to do the conversion. But, as a certified CISCO administrator, my knowledge of coaxial wires consist of knowing that they have been used an eternity ago in physical bus topologies barely reaching 10mb with a lot of instability.
So if I put these physical rj-coax converters, will I be able to reach fast ethernet speeds and standards? Will the coax wires ever be able to reliably support a lan-type computers network?
To my understanding, coaxial cables, such as RG-6 and RG-58 or 59, can be used for analog broadband networking as promoted by Chipcom and Cabletron in the 1970s/80s. You need a Headend to be able to create the different channels, and each run has to have a terminator. Speed was like what you said (10Mbps) unless you coupled the channels together. I am assuming you have RG-58 or RG-59 (older coax) and not the newer quad-shielded RG-6.
Sounds like the coax runs were for the TVs, not for computer- communication.
Personally, it would be better to have a secure wireless than try to tackle coax cabling adaptation. Run a couple of CAT 5es and install wireless access points so that you get better signal distribution. Or run CAT 5e or CAT 6 wiring and change the faceplates to have both coax and CAT 5/6 RJ-45 receptacles.
In my residence, all of my wiring are "home-run" to a main distribution frame (MDF). Each faceplate has 2 CAT5e and 2 quad- shielded RG-6. The MDF is where the main network switch, firewall/ router, and DSL modem reside (with UPS battery backup). I am planning to add a dual fiber run to my relay rack on the second floor, which has my main servers, to provide better electromagnetic frequency isolation between floors due to power surges.
I think it is possible with a transformer, but it isn't easy. Also, there will be some loss which will reduce the distance. How well it can be done depends on how close the impedance match is, but coax is usually pretty close. Are they designed for the impedance of your cable?
Otherwise, there used to be eight port 10base2 repeaters. One of those, with each cable terminated at 50 ohms might work fine. Even better if you can change the collision detect threshold to the appropriate level for 75 ohm termination.
What kind of connectors on the plate? Screw-on ~ 3/8" diameter normally used by TVs (type 'F' connector)? BNC?
If that is for television, it's almost certainly RG-6/U or RG-59/U coax, which is 75 Ohm. Ethernet uses cables similar to RG-8/U (for the original thick-net) or RG-58/U (thin-net) both of which are 50 Ohm. That means Ethernet doesn't work to well (if at all) using television coax.
Not very likely. The cables come together somewhere (the "center" of the "star" using splitters - which are basically resistor networks to "match" the impedances. These will not work on Ethernet.
There were media converters that allowed connecting coax to one (or more) twisted pairs. These worked for 10Base5 (Thicknet) or 10Base2 (thin-net) ONLY.
That would come as a bit of a surprise to Bob Metcalfe who co-invented Ethernet, and founded 3Com. It might also surprise a few people who were using coax networking for a dozen or more years.
They did for years - but not using the configuration you have.
Google for 'media converter'. For the simple minded, think of a
10Base2 NIC connected back-to-back with a 10BaseT NIC.
If the coax is in conduit - perhaps. More often, the coax is secured to internal wall support structure (studs) with staples or bent nails.
having lived thru StarLan and ArcNet - along with IBM mainframe 3270's it's always interesting to stumble across coax related discussions.
Depending upon what you have found to "convert" the coax into 10baseT, you might not be looking at the right stuff....
We used to use coax to twisted pair "baluns" for the IBM 3270 terminals, to run the coax terminals across the internal twisted pair phone wires....
You are looking at the other way around - you need to have TWO pairs of signals - xmt & rcv - going across the coax. That would seem to indicate that an "active" device would need to make the transformation from the RJ-45 based Ethernet to the single conductor based coax.
Worse case - you would need; xmt, rcv, common ground - 3 conductors - the coax has center + shield = 2 Therefore, a plain direct connect RJ45 solution won't seem to work, it would appear you need something that puts a "signal" on the coax - xmt channel + rcv channel ?
In that case, they will likely have the wrong impedance.
It is theoretically possible to do with a transformer, but the impedance has to be pretty closely matched. It has worked for telephones for many years now. Newer phones may do it with active circuitry, but older ones used a special transformer.
The phone design leaks some of the signal back to the receiver, as it makes people more comfortable.
Easiest for current technology, where wireless is out of range and coax is available, is to take the center conductor of the coax and wrap it around a wireless adapter. The coupling doesn't need to be large for it to work.
A very common device in the 10Base? era was a hub - specifically one that came with two or more interface types - 10BaseT, and 10Base2 coax and/or AUI. It was a simple matter to plug in your twisted pair on one side, and use the appropriate interface on the other. These would probably be considered a repeater, but as we commonly used them to connect those systems that only had a 10BaseT interface to our coax network, under the worst case, there would be two repeaters between the furthest hosts on the wire.
Results 1 - 10 of about 157,000 for Ethernet hub 10Base2. (0.28 seconds)
Results 1 - 10 of about 97,300 for Ethernet hub 10Base5. (0.27 seconds)
For a coaxial Ethernet, the collision detection depends on having a very high return loss, which translates to a very low VSWR. I don't believe this was specified directly, but a 1.08:1 (26 dB return loss) was starting to push things.
Hmmm.... transmit power +20 dBm max, receiver sensitivity for 10 MB about -80 dBm - so total available loss is 100 dB max. 802.11 is at
2.4 to 2.48 GHz, and single braid coax is really out of range at those frequencies, but the insertion loss of RG-58 is about 40 dB/100 feet. Sure hope the coupling is relatively tight, or you aren't trying to go very far. I wouldn't even bother trying with a 5 GHz wireless.
at one Megabit - sure that's possible (though in planning a link, I'd take those figures with a bag of salt, not a pinch), but it doesn't speak to bit error rates. More bits/second means wider channel, which means more noise heard, so the sensitivity is worse. Notice the numbers dropping as you widen it up.
No, that's a non-standard cable - notice the presence of the "Beldfoil" shield in addition to the 95% tinned copper braid. MIL-C-17 (actually M17/29-RG59) specifies a single braided copper shield, and a copper coated steel center conductor. I know Belden was on the QPL for many coax, but I'm not sure which of the "RG-59*" types they list is actual M17/29-RG59. Type 9240 looks close. This obsolete specification also only goes up to 1 GHz. Single braid coax becomes very erratic above that frequency due to the poor shielding. I stop using single braid coax (whether 50, 75, or what-ever Ohms) at 900 MHz, and use double braid coax up to about 10 GHz for short runs. Longer runs above 1 GHz should use a semi-rigid (solid metal shield) if practical, and waveguide is preferable above 3 GHz.
Same problem - M17/28-RG58 is pushing it above 1 GHz. The problem about having a star network of RG58 is how do you connect it at the center of the star. Power combiners/splitters? The insertion loss is prohibitive.
I can see you haven't tried piping 5 GHz over any distance. RG214/U (formerly RG-9B/U, a double braid coax 0.42" diameter) is running around 28 dB/100 feet, and you're really getting into the range where a semi-rigid coax is highly desirable (except for cost, flexibility, and weight).
True - but does your S.O allow you to run coax - heck, even ordinary cable - directly from A to B? When I moved into the current house, I spent several days putting plastic conduits into the walls up to the attic so that I could install networking cables at my convenience, and be able to upgrade later without tearing up the house.
First, I would be really careful about arguing with Rich on anything related to ethernet.
As for collision detect and reflections, there are stories that using 75ohm cable with 50 ohm terminators works reasonably will, where with 75 ohm terminators it doesn't work at all. (I believe that is with only two stations, one at each end.)
That points to voltage being more important than low reflections.
The reflection coefficient is (75-50)/(75+50) or 0.2 so that
20% of the signal is reflected back. There is a good chance that won't be enough to trigger a collision at the sender.
That goes back to the source, and again reflects at 20%, for 4% of the original signal. That is a fair amount of noise for the receiver, but maybe not an unreasonable amount.
There are many kinds of cable impedance variations that are fairly gradual, and will have much less effect on the signal.
First, thank you for all your answers. Really sorry to come back so late on that topic but I had to hold my projects for several weeks.
OK, here it goes for the questions you've asked :
1- My home network consists of 8 RG6 individual coaxial runs.
2- The adapaters I found are about 25$/piece ... so with a cost of
50$/ run, the whole project would cost around 400$
3- Using the existing coax cables as "pull string" for LAN wires or passing new LAN wires on the walls are impossible ; it would ask me to open several walls and, shortly after, negociate an expensive divorce with my wife
From your readings, I understand that the physical properties of RG6 wires are incompatible with fast-ethernet requirements and that those cables were used during base-2 and base-5 era, where physical networks were in bus configurations.
I decided to go for another strategy. I recently moved my phone number on cable-modem technology, along with my Internet connection. Since I am using wireless phone units connected to a single phone base (which connects to my modem), my house phone wiring is completely unused. I will therefore use these 2-pairs wires for fast ethernet.
The only thing I havent figure out yet is where is the phones patch- panel. I suspect the company put all terminations in a locked box, outside the house.
You seem to have this concept that just because you have access to a "wire" that it will serve just about any purpose.... WRONG
Why do you think they are all different to begin with ???? Can you run your TV from the coax thru the "telephone wires" and connect to a TV ?
Each "wire" has it's own set of attributes or properties that makes it work for a given application...... coax, Cat5, telephone twisted pair It's doesn't work like on TV where they tie knots in fiber optic to wiretap..
Read my post correctly. If you don't know that a 2-pairs wire (like those used for voice) also supports data, go read more on networking.
And for you information, wires are *not* made for *given* application. In all standard models (like OSI), layers are independant and a physical specification (like a 2-pairs wire) can support more than one higher specifications.
There is a good chance that phone wiring should work for 10baseT, most of the more recent wiring (less than 20 years or so) should be Cat 3 or close. Shorter wiring doesn't need to be as good as longer wiring. I wouldn't be surprised if it worked at 100baseTX for normal house distances.
The biggest problem is that phone wiring is not always done in star configuration running all back to a central wiring spot.