Date: Mon, 13 Aug 2001 23:28:28 -0700 From: "Ted Mittelstaedt" <tedm@toybox.placo.com> To: "Doug Reynolds" <mav@wastegate.net>, <freebsd-questions@FreeBSD.ORG> Subject: RE: Network throughput Message-ID: <001401c1248a$54b02000$1401a8c0@tedm.placo.com> In-Reply-To: <200108140135.f7E1Zlb28792@mail.freebsd-corp-net-guide.com>
next in thread | previous in thread | raw e-mail | index | archive | help
>-----Original Message----- >From: Doug Reynolds [mailto:mav@wastegate.net] >Sent: Monday, August 13, 2001 6:33 PM >To: freebsd-questions@FreeBSD.ORG; Ted Mittelstaedt >Subject: RE: Network throughput > > >On Sun, 12 Aug 2001 23:10:47 -0700, Ted Mittelstaedt wrote: > >> I hate to disagree with you here but DSL service is delivered over >>POTS lines that are generally buried in a 50 or 100 (or often much >>higher) cable binder that is in no way shape or form anywhere NEAR >>CAT-5, or even CAT-3 for that matter. The 50-100 feet of wire he >has control >>over is going to be less than 1% of the total cable run of the average DSL >>installation (about 7K feet) It makes absolutely no difference >what CAT-level >>the cable is going to be as long as there is some light twisting in >the pairs. >>Remember this isn't Ethernet that we are talking about. > >oh I agreed. I've popped open more than pick-terminal that didn't meet >any standard at all, except for maybe the yarnball standard. the only >reason i suggest CAT5 is because a lot of houses i goto have anywhere >between 14 and 24 jacks that home run from the garage, and are at >least 200 feet long. 20 x 200 is 4000ft of wire that are stapled in >place. > Your absolutely right with NEW wiring installs - I was mainly pointing that out from a theory standpoint anyway. Actually I think that for the average homeowner that they can pick up a spool of 1000ft of CAT-5 from Home Depot for less than $40 these days as well as all the tools to terminate it pretty like in jacks and such. That's probably cheaper than a 1000 feet of bellcore wire. Existing wring installs are something else though. You always want to pull and replace - but I'm reminded of my cousin who lives in Phoenix and bought one of those spec-built houses last year with his wife. We toured the place while it was being built - they use some very different construction techniques down there. Basically the systems and cabling you put into the walls of the house better darn well be what you want for the next 100 years because the wall is the equivalent of a styrofoam block with fiberglass painted over it - drywall on studs construction is completely unknown. I guess it's cheaper and faster to build and cooler in the heat. They use steel studs too. He was showing me some CAT-5 that he was having the builder put in and I opened my yap and suggested instead of burying the wire that they run plastic interduct with string then pull the wire later. I got some blank stares and snorts of laughter at that. OK I guess if they want to screw themselves... > >>For ADSL there are 6 critically important things that MUST be followed: >> >>1) Distance of the entire run must be under 18000 feet. (actually, you can >>train at greater distances than this, but the training speed will be so low >>that you may as well use a modem) This of course is the length of the cable >>run, not the "as the crow flies" distance. > >that i do know, what kind of speed would you get at 18k ft? just >curious, that is how far i am from the co. > I've got a Qwest DSL line that measures out at 1750 or so feet from the CO. I had to bugger some strings to get that in - Qwest does NOT care to attempt DSL installs on anything longer than 1400 feet because of the risk of multiple truck rolls as well as the customer getting impatient and jettsoning the installation a week before the line comes up. In my case I paid an install fee and signed a 1 year contract which was enough to convince them I was serious. Even then they had to do 2 truck rolls, one for a load coil that they knew about initially, and the second for a bridge tap that the DSL tech and I discovered on the line after the modem was plugged in. It's very humorous but I can to this day key in my phone number that has the DSL circuit on it into the Qwest qualifier and it comes back no-go. In my case I get a full 640K down. The thing is that DSL can be pushed - but everything has to be optimal for it - no funky baloney in the wiring. The big issue is the recieved power. It decreases a set amount over the line length that's due to the guage and wire resistance. Once you drop below 20db the remote modem will fail to train on a standard noise level line and that is that. Some lines have less internal resistance than others. I've also heard of SDSL going to 21000 feet. According to the Cisco docs on their DSL CPE's, you can train at past 18000 feet and it will still work but the further away you get the lower the top speed. Also the noise level is very important - they assume I think 6db of line noise because that's average. If you can get that figure down by cleaning the copper than you can go past the 18000 feet limit. I would guess that you could probably get 128K down at 20000 feet on a Cisco 678 running DMT if the line were clean. I wouldn't trust the cheaper internal DSL modems, though. > >I've seen 60k dls w/ 618k dsl, on old 18 gauge single pair. I don't >know how much that is followed, or how it affects it. most cables are >24gauge in my area, and the drops are 22gauge. > The closer you are to the CO the higher the power levels and the more crap you can get away with. I've heard of DSL installs on customers within 5000 feet that were done on bridge taps and it still worked. That's why Qwest prefers to not qualify things further away - less need to modify their own network by cutting out those bridge taps, and fewer truck rolls. The engineering documents I've read on DSL though all are really hard on guage changes, it's like the 3rd worst thing behind bridge taps and load coils. But few techs and fewer end users seem to be aware of this problem. In my case I previously had an ISDN line and I made them replace the drop when they came out to install that and I'm pretty sure the new drop was also 24 guage. I did look closely at it and remember thinking it was surprising that the wire wasn't more bulky. > >>3) There must NOT be any unterminated sections of the cable (ie: >Bridge Taps) >>teed off of the main cable. Note that a homeowner can create bridge taps in >>their home that will be just as bad as the ones the Telco creates. > >hehe.. you'd be surprised, really. > We've all got our stock of "You wouldn't belive this" stories. I just added another one 2 weeks ago - customer complaint on DSL being slow, I went there and found a y in the pair and told them to remove it - but as I was driving away I happened to look up and see that the 200 pair coming into the building was _hanging_ from it's steel support wire, with half the wrapping unraveled. A subsequent call found that right around when the DSL started having problems that that week the shipping guys were partying (July 4th) and caught the cable on an overloaded semitruck they were driving around. I could just picture this big, fat 200 foot long 200 pair cable being drawn back like a bowstring, stretching further and further and further and suddenly snapping free and wanging back and forth - it's amazing the fools didn't pull down the power poles and blow up the transformers. The damn assholes didn't want to call Verizon and report it though - the figured that since their phones still worked that they had got away with no harm done and they were scared they would be charged for a new lead in. So like all the assholes in this predicament do they pulled the old "blame the ISP" crap, as if it was our fault. >>4) There must not be any load coils or other inductive loads in the circuit. >>(such as an extra 30 feet of cable wound up into a loop somewhere, or a >>ferrite bead on the cable somewhere, or passing through a "surge >protector" or >>some such) > >yeah, most long cable runs here have a lot of load coils down here. >they also have a lot of lightspanned boxs, and PG boxes over here, >which according to the production techs i've talked to say that they >still havent figured a way to switch dsl through them.. i find that >hard to believe with fiber.. i could see it being a problem with copper >though. > They are starting to put in a number of remote DSL terminals in around here, the only problem is that you have to special order the circuits off of them. There's a big rush to do it though. >>6) There must be no connections from either side of the pair to >ground, or to >>other pairs. > >that goes for anything- have a tip to gnd or ring to gnd short, you'll >have a little hum to no dialtone, depending how bad it is. > This can be a problem with phonesets too - some of the real cheap 2-line phones I've seen fail with this kind of problem. I don't know if circuitry fries inside or what. >>One thing he needs to do immediately is call the Telco and get ahold of a >>DSLAM tech. Modern DSLAMS have testing circuitry and you can do shorted and >>open pair tests on the pair that will discover most of these problems. >>(assuming the DSLAM tech knows what they are doing) > >most splicer techs where i work carry good fluke meters to test for >that. > It's been my experience that the Telco field techs are very helpful when you can get them - but they are always on the run and you have to be waiting for them when they show up. The problem are the people back at the office and the CO. I've had more than a few T1 circuts that worked the day the field tech installed them then 2 days later stopped working because some bozo at the CO buggered it up. They must put all the inexperienced telco techs to work in the CO and give the experienced ones the field work. ;-) Ted Mittelstaedt tedm@toybox.placo.com Author of: The FreeBSD Corporate Networker's Guide Book website: http://www.freebsd-corp-net-guide.com To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-questions" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?001401c1248a$54b02000$1401a8c0>