jyoull@bgsuvax.UUCP (Jim Youll) (10/23/90)
I hope this doesn't sound too dumb, but here goes... On a Thin Ethernet segment, is there significant signal loss at each "T" connector? What if the connector is not attached to an Ethernet card, but is just left unconnected? How about through the "barrel" BNC connectors used to join two pieces of thinnet coax in the absence of a "T" connector. We are looking at options for prewiring a building. 10base-T is preferred, but cost is a big problem and Thinnet is fast and cheap. I've been asked to find out about leaving ready-to-go connections in the walls with barrel connectors or T-connectors at each workstation, for extraction when needed. Suggstions invited. Thanks.
kwe@buit13.bu.edu (Kent England) (10/24/90)
In article <6464@bgsuvax.UUCP> jyoull@bgsuvax.UUCP (Jim Youll) writes: >I hope this doesn't sound too dumb, but here goes... > Dumb questions always welcome on this station. >On a Thin Ethernet segment, is there significant signal loss at >each "T" connector? > Anytime you insert a connector in thin-net you get some loss and reflection. The spec tells you how many you can have on any given 185m segment (but I always forget the numbers and have to look it up). Isn't it fifteen stations and 30 connections? >What if the connector is not attached to an Ethernet card, but is >just left unconnected? > Not really any different. The open side of the T is not supposed to present a problem. It is like powering down an attached node. >How about through the "barrel" BNC connectors used to join two pieces >of thinnet coax in the absence of a "T" connector. > Looks like a T. >We are looking at options for prewiring a building. 10base-T is >preferred, but cost is a big problem and Thinnet is fast and cheap. >I've been asked to find out about leaving ready-to-go connections in the >walls with barrel >connectors or T-connectors at each workstation, for extraction when >needed. > >Suggstions invited. Thanks. Thin-net is no faster than 10BaseT, either in network speed or speed of installation. Daisy-chaining thin-net thru your walls will end up costing you much more than 10BaseT. It is just that it is a different line item in your "budget" called "labor". You cannot attach workstations to thin-net T connectors with any length of patchcord. The T must go onto the back of the PC. If you must daisy-chain thin-net thru walls I recommend the AMP solution, but you won't like it because the cables and wall jacks are expensive. Here is a compromise. If you have labs or classrooms to network, link them up with thin-net daisy chains. Then use an adaptor like the Cabletron CT-100 to attach this small daisy-chain to a 10BaseT port. Install isolated nodes directly onto 10BaseT. This way the thin cable is outside the walls and entirely within one room. Easy to maintain and fix. If you take the low end 10BaseT concentrator gear, you should be able to reach a target of about $100 per port now or in the near future. The workstations will come with 10BaseT already in place. NeXT done it already. --Kent
imp@jhereg.osa.com (Charles T. Lukaszewski) (10/25/90)
In article <6464@bgsuvax.UUCP> jyoull@bgsuvax.UUCP (Jim Youll) writes: >I hope this doesn't sound too dumb, but here goes... >On a Thin Ethernet segment, is there significant signal loss at >each "T" connector? There is loss and attenuation, which is why the 10-Base2 specification limits thin ethernet segments to 30 "breaks" which include either barrel connectors or T connectors. However, it makes no difference whether the T is connected or not. You should give careful thought to how you plan to run your thinnet if it is to move between rooms. One implementation we frequently find at client sites is the use of a double-bulkhead on the wall, with one connector being used for 'in' and one used for 'out'. Remember, each barrel counts as a cable break, so a room with such a contraption and one station would have used 1/10 of your available breaks. >We are looking at options for prewiring a building. 10base-T is >preferred, but cost is a big problem and Thinnet is fast and cheap. >I've been asked to find out about leaving ready-to-go connections in the >walls with barrel connectors or T-connectors at each workstation, I urge you to really work out the costs of twisted pair versus thinnet. On every building-size install we have done, the costs ended up being comparable at worst, and twisted-pair far cheaper at best. As Kent Englund and others have pointed out here, there are a lot of hidden costs with thin net, and even if the up-front cost is somewhat less, I guarantee you that the medium and long-term investment in new cable, new connectors, move/add/change management and aspirin will be *substantially* more. You may want to seek out a qualified LAN consultant in your area that understands the types of systems you want to connect that can lay out the issues and costs more clearly for you. One final thought - be *extremely* careful about what kind of cable you install and make certain that the craftsmanship of the installation team is excellent. Virtually every thinnet system we've troubleshooted has had at the root of its problems poorly-installed, damaged or wrong cabling. Make sure you are using 50-ohm RG58/U, rather than 53-ohm RG58. We have had excellent luck with Belden 8240 (solid core) and Belden 8219 (stranded core) cabling. Carol also makes a good cable (model C1178 or C1155). There are others, but be careful. -- _______________________________________________________________________________ Charles T. Lukaszewski imp@osa.com 612 525-0000 Managing Partner & Chairman Open Systems Architects, Inc. "Who needs a disclaimer? I liked the opinions so much, I bought the company!"
bruce@ccavax.camb.com (10/25/90)
In article <6464@bgsuvax.UUCP>, jyoull@bgsuvax.UUCP (Jim Youll) writes: > On a Thin Ethernet segment, is there significant signal loss at > each "T" connector? It isn't the signal loss you care about but the slight mismatches that cause reflections. These add up. For every splice or tight kink or storm window pintching your TV antenna cable, you add to your flock of ghosts. You can sit at the end of your cable with a good TDR, and 'see' each T, and count them off down the cable to the break or short you are hunting for. Much easier to say go 5 taps down the hall rather than saying 123' of cable. Those same visible mismatches will give you grief if you have too many. Without looking it up, I think the rules are 60 connections. If you just had that segment, you could squeeze in 30 Ts, but, since you normally start at something like a DEC DEMPR (which counts as a connection), you get 29 Ts (2 connections each). On an 8 port DEMPR, that could get you 232 connections! > What if the connector is not attached to an Ethernet card, but is > just left unconnected? Harmless, but don't let anything get in there and short. Also, realise that the T must go to the card, no stub cable off it is legal. > How about through the "barrel" BNC connectors used to join two pieces > of thinnet coax in the absence of a "T" connector. Should count the same as a T. Is a little safer from accidental problems, but forces you to break the segment to replace it with a T. A normal T with a plastic cap of some sort (CapPlug(tm) type thingie) would be nice. > I've been asked to find out about leaving ready-to-go connections in the > walls with barrel > connectors or T-connectors at each workstation, for extraction when > needed. How fancy and how secure is your environment? How technically sharp are your users? You can use el-cheapo BNC'd wall plates with an IN and an OUT that get a jumper between them when not in use. You certainly CAN'T do 30 wall plates this way, but the 8 limit I am about to mention for another product should be ok. BNC equipped through connections always allow the segment to die when some turkey decides to rearange his office. AMP/DEC decided to make a relatively user-friendly / user-proof system. The wall plate has a connector that takes the in and the out coax and connects them on through when nothing is plugged in. The cord that plugs in really is a twin coax that loops out to the BNC at the far end (the BNC is electrically a T at the center of this loop) and then back. As you plug this in, it breaks the through connection and the signal detours out past the BNC and then back in again to go to the next wall plate. DEC says 8 such wall plates MAX. Amp says 8, but I have seen some of their curves from tests that MIGHT let you sneek a couple more wall plates if they were not all actually to be used. Stop at 8 and be very safe. On an 8 port DEMPR, you get up to 64 connections. YES Coax IS cheaper that twisted pair. Amp even makes a 'cute' box that has several of the wall jacks (all in series) and has the cord to plug into the wall. This lets you expand your single tap to several! Obviously DON'T do this at every room on a segment, but is a handy 'fix' for that VP or programmer, or demo room that suddenly needs several enet connections. DEC's version (made by Amp) has a trap door to keep crud ourt and mounts in their latest DecConnect universal face plate. The AMP version mounts sort of the way a single toggle wallswitch mounts to its plate. The guts of the jacks, and the cords they take are identical, though. The wall jack is ~$7.xx if bought 'right', but often is $15-25 or more in those glossy catalogs that clutter your mail box. The special cable from the wall to your computer IS much more expensive than the jack, but get them through an AMP distributor (Anixter, Graybar, Clifford, TW, etc) who thinks you are either a BIG industrial account or are in the wiring trade and you will get the right price. Buying the IDENTICAL cable through DEC is needlessly expensive, and AMP offers several additional lengths that DEC doesn't. N.B. that Anixter IS DEC's DEConnect distributor to the wiring trade as well as being an AMP distributor. Just check relative pricing and clearly specify what you want. For wire, you have a problem. RG58 is NOT really good enough for all 185 meters and all 30 Ts. Many LOVE it, because it strips easily and doesn't have the foil-needing-stripping problens of the 'right' stuff. DEC and Belden went through some 25 or 26 tries to get the right cable. They didn't do this for the fun of it, but because rg58 just wasn't good enough. The glue pattern to allow easy foil stripping, and returning to 93% braid coverage (from 75%) mostly for mechanical crimp strength issues were typical problems addressed. There are several other vendors that supposedly make the cable to DEC's spec, but for the piddling few dollars difference I would definitely stick with Belden. The Belden numbers are 9907 for vinyl, and 89907 for teflon. I don't know if Belden yet makes this cable in their new "Flamarrest"(tm) jacketing. If they do, it makes a nice limp easy-strip (like vinyl) cable with plenum rating without Teflon's price! The generic mil-spec hex-crimp tool for BNCs on rg58 size cable is fine IF you get connectors made to be so crimpped. The standard AMP ones (also what DEC sells) crimp with a standard (to AMP) ROUND die (+ square for center pin). Use the wrong tool, and they barely stay on! Modtap's contract wiring group did some work for DEC at the preceeding DECWORLD (when the QE2 was there), and only learned AFTERWORDS why the DEC connectors barely stayed on. They were using the generic HEX tooling... Note well, though, that AMP also makes BNCs for hex crimping (they sure would not want to miss a sale to someone tooled for using them), so brand alone does NOT determine what tools to use. Buy where the support folk know what they are talking about... And buy a few of anything you are considering and play with them until you are comfortable that you REALLY understand them.