Slashdot Mirror
Handmade vs. Commercially Produced Ethernet Cables
An anonymous reader writes "We have a T1 line coming into our satellite office and we rely fairly heavily on it to transfer large amounts of data over a VPN to the head office across the country. Recently, we decided to upgrade to a 20 Mbit line. Being the lone IT guy here, it fell on me to run cable from the ISP's box to our server room so I went out and bought a spool of Cat6. I mentioned the purchase and the plan to run the cable myself to my boss in head office and in an emailed response he stated that it's next to impossible to create quality cable (ie: cable that will pass a Time Domain Reflectometer test) by hand without expensive dies, special Ethernet jacks and special cable. He even went so far as to say that handmade cable couldn't compare to even the cheapest Belkin cables. I've never once ran into a problem with handmade patch cables. Do you create your own cable or do you bite the bullet and buy it from some place?"
And I have never had any problem with them. Even on 50 servers running at full Gig. No errors.
I've learned the hard way when setting up a couple of clusters: You MUST use custom-made, cut to length cables to prevent a huge rats nets in the server room. Buying precut cables is a disaster. I had to rip out and completely rewire one cluster because I made that mistake.
However, you need to TEST the cables. And not just by plugging in and making sure it works, but a full ethernet validation tester.
I've been very happy with the Fluke Cable-IQ qualification tester, which doesn't just make sure that the wiring is correct, but actually tests the cable up to gigabit speed to make sure everything is kosher.
Test your net with Netalyzr
I buy cables because I would go through 5 - 10 cables a day and by the time I made them, tested them, labelled them, I could be doing 101 other things.
It's not to say that you can't do it, you can. It's just a matter that the amount of time you spend doing it just makes it a hell of a lot cheaper in the long run to buy them.
This is ESPECIALLY true when dealing with CAT7 or STP. On a 20Mb line (Probably a 100Mb link) the chances of having a problem though are pretty low provided you terminate it cleanly.
Curiosity was framed; ignorance killed the cat. -- Author unknown
We had a contractor come in and rewire our facility. They ran raw CAT 6 and hand terminated it, then TDR'd each run.
Your boss is unclear on the tools needed and the difficulty...just simple hand crimpers were all they needed. There's going to be
an impedance bump at the RJ anyway...the cable's not twisted there.
As to making them yourself or buying patch cables? It's way cheaper to buy them (I like L-Com) but if you need one *right now*,
(or a custom length) it's cheap to have a crimp tool, some RJs and a roll of cable handy in the corner of the office.
While it may be cost effective to crimp and cut your own cable when you are making less than 20 dollars an hour once you are making 20 dollar+ just buy it.
I promise you I can make more than $20 worth of test-worthy cables in one hour.
Dewey, what part of this looks like authorities should be involved?
I used to work in a large shop that mostly made their own cables. We honestly rarely had issues. Sure, sometimes you write off a problem to a bad cable but if you know what your are doing (without a TDR) you can still be fine most of the time. Recently, I've moved to a different employer who prefers to buy everything. Nothing handmade. They literally have bins of cables in every size in a store room and backfill what is taken out so you never run out and you don't have to go to the store. The result is a cleaner solution IMHO. I guess it depends on the scale and mindset of the company....
Aren't the commercial ones also hand made? I find it hard to imagine an automated way of doing it.
Commercial cables have going for them: rubber injection/overmold for more ruggedness, and they're pre-tested. Aside from that, I don't see exactly what should stop you from making decent ones yourself, assuming sufficient skill.
Depending on wages and such, the commercial cables could easily have 10x the labor and still be cheaper.
Nerd rage is the funniest rage.
Yes, you can use handmade cables that are as good as mass-produced factory cables. But that really isn't the issue.
It's just not worth the time spent to cut and crimp your own lines anymore. In my experience, it was a more common practice years ago in IT. That may have had something to do with the fact that there weren't nearly as many PC's or ethernet ports in buildings as there are today.
My advice: Find a good supplier (i.e. not one that charges $800 for a 6 ft. adamantium-coated cable) and do something else with the rest of your time.
While I would agree that hand making Ethernet cables is a fine way to go about things, if the boss says says to buy them pre-made, then just buy them. That is, unless you can convince him, for example, that hand making cables is necessary for keeping a tidy data room (precut sizes aren't going to work for running a building's worth of Ethernet jacks).
What you've posted him saying sounds like a lot of I'm-afraid-of-doing-that-myself jargon. By crimping the cables yourself you're opening yourself up for any data or transmission speed problems to be blamed on you and your shoddy Time Domain Reflectometer tests.
I've bought thousands of dollars of cable. Full disclosure, it has been BNC cable, and not ethernet, but I think my experience is likely germane. This cable has been used to construct installations of scientific equipment that gets reconfigured pretty frequently (and I've been the primary user on most of this equipment). I have never, ever had a single cable-related failure using ITT/Pomona cables. My peers, on the other hand, use hand-made cables and are constantly debugging their setups.
I spend my time doing my job (collecting data), while other people in my lab spend their time fixing problems. (Really full disclosure, I'm the only one with an EE degree.)
Good cables can be found inexpensively. These are the ones you want. Cheap cables can be found for less money, but these are the ones you do not want. Custom cables, unless you have high-quality crimping tools (the $39.99 variety don't cut it) and a proper means for doing testing, which means TDR and bandwidth testing in your case, just are not worth it for general-purpose use.
Look at it this way: how long does it take you to generate a qualified cable? Not how long does it take you to make one cable, but how long does it take you to make one cable that you will use, including all of the failed crimps, cables that were cut too short, too long, were miswired, or must be discarded, for some other reason. How many cables will you be making? Total that up and use 1/2 of the time to search for low prices on high-quality cable instead. You will be ahead in the end.
Put my fist through my alarm clock with its ding-dong death inside my ear. - The Blackjacks.
I never ever make my own cables for long-term PC connection patch cables, only for patch panelswitch connections and cross-overs.
I wired all of my own in-wall cabling at home, but I pay others to do it for work (just for time reasons).
I always use packaged stranded copper patch cables for connecting PCs to wall jacks though, as they're more flexible and resilient to breakage when twisted or bent repeatedly. Solid core cables will snap or degrade rapidly if bent repeatedly or at sharp angles.
- Michael T. Babcock (Yes, I blog)
If pre-made cables from a good vendor are available that meet your needs, then buy them.
Times they may not meet your needs:
*You need it NOW
*They don't come in your size and the nearest sizes are not acceptable
*The only vendor that makes the one you need charges a fortune
Examples of when custom is good:
*Temporary installations, and it's faster or a lot cheaper to make than to go shopping
*Runs over 50 feet that aren't close to pre-made sizes. A 63 M run in tight conduit and little room for slack cable is probably better done custom than using a 75 M premade run. Usually, runs under 50 feet can use a 50' or smaller pre-made cable.
For permanent installations, test your custom cables. Test your store-bought ones too for that matter.
For temporary installations or installations where it's easy to yank-and-replace if there's a problem, it's frequently good enough to just power it on and see if "it works."
Knowledge is how to play a game, intelligence is how to win, wisdom is knowing what game to play.
Solid core has slightly better propagation properties (the 100M limit implies solid core for example) however it also acts similar to a wire coat-hanger. Like any metal it weakens as it bends and after a period of time it'll grow weak, thin and even completely break.
Stranded is similar to a braided rope, it can withstand constant reconnections (user area, especially common with laptops), movements (telcom closets when you're moving the cable mess to access equipment ports) and the stress that will wear down the solid-core cables.
Do yourself a favor and make sure that if you create your own patch cables:
There's nothing wrong with making your own patch cables, and it could potentially save you big bucks (compared with buying a $35 patch cable at a local store). However if it's not done right you will kick yourself down the road -- or more likely blame the network electronics, server, network cards, or whatever you normally blame. :)
Most of the time it comes down to what wiring inspectors will say is good and they usually do other tests that most of us don't have the tester for. I would go for the professional cables also. In a professional setting. In a non professional its safe to go with your own cabling.
As an EE: What the FUCK are you talking about? Please read a book on transmission lines, rather then repeat vaguely technical words.
I can't agree with this - if the termination of a transmission line is correct at each end, then the length has no matter at all for any frequency (in theory, not accounting for increasing losses with frequency, but then there's a reason for length restrictions in the CatX/Ethernet standards).
If you're talking about a *tuned* line (eg a stub or a tuned antenna feeder), then length is important. But we're not. If you've got problems with harmonics or matching and reflections then your ethernet cards are probably bottom-shelf knock-offs.
The problem with premade-lenght cables is you're going to run into tangles if many changes are made, and are going to end up coiling. Make that coil too tight and you're going to cause crosstalk. A custom job with all cables neatly following defined routes with no coils, twists or kinks is going to make life easier in the long term.
-- Sig Sig Sputnik
Why is this modded interesting? I thought the parent was trying to be funny. Somehow I doubt EMF harmonics has anything to do with the nice round values like 6 and 10 ft, that are commonly found at compusa, best buy, etc
I've used both handmade and commercial cables and there is big difference in the connectors. When I crimp it myself (or get cheap pre-made ones) the only thing holding the connector on is the little metal spikes inside it. With decent commercial cables, there is moulded plastic connecting the ends on to the cable, which is a lot stronger.
Whether it's worth going for the expensive option depends on whether the cable will be plugged in and unplugged a lot or whether the ends are likely to have sharp bends in the cable near them.
I am TheRaven on Soylent News
Orange and white, Orange. Green and White, Blue. Blue and White, Green. Brown and White, Brown.
Use pieces (cable, plugs, jacks) certified for the speed you want to carry.
Once you get those two down, understand not to untwist more of the cable than absolutely necessary to get it into the connector, get it correctly into the cable, and get a good solid crimp on it - and TEST IT after you crimp both ends - odds are it's more than sufficient to carry as much GigE traffic as you care to move.
Once you have a stock of pieces on the shelf, it's WAY more cost effective from an employers perspective to make a single cable than to sit down, fill out a purchase order, have that purchase order pass through several hands during processing, follow up with the paper order, wait a week to have that single cable shipped to you. ESPECIALLY if that cable is a statistical anomaly and needs to be replaced.
If you're wiring a patch panel for the first time, however, order a hundred or so cables of various length and save yourself the hassle.
Glonoinha the MebiByte Slayer
Here's a link to a page explaining the reasons for this.
3 things about computers: they're alive, they're self-aware, and they hate your guts.
Not quite sure how talking about the characteristics of UTP regarding reception of external noise relates to standing waves on CAT5 cables?
I was just pointing out to the AC above that UTP cables do have different behaviour as you change the length. If you fancy a fun experiment, get a fast enough scope, and some 100M Ethernet kit and see what happens as you change the length of the cable by small increments relative to the wavelength (100MHz = 3m).
I'm well aware of how to crimp CAT5 and how UTP works though, thanks all the same.
You miss the point.
The domain is successful Ethernet via the TIA TSB specs using the IEEE recommendations for connectivity with the signaling method employed. Good signal, no weird phase shifts or nullings, and no discriminator problems (e.g. via NEXT).
Standing waves are inevitable in non-DC cable connectivity and are a red herring unless propagation effects signal discrimination, or unwittingly becomes an antenna for other problems. In my experience (50K+ end point terminations), it's not been a problem. With a few discrete components, I can make any Ethernet cable into a wicked antenna. But the question would be: why would I do that? Standing or sitting waves (pun intended) may change ground-level, but that's when STP or 'screened' cabling is an alternative. If you need shielding because of ground-based level shift, use fiber. In fact, fiber is just about as easy to terminate as UTP these days.
---- Teach Peace. It's Cheaper Than War.
That's funny.
I mean, a couple of weeks ago I finished up a job where I went into a mess, with a mix of premade cables and mixing A and B pinouts. I re-did most of the connections - by hand - and installed all new patch cables - made by hand, and tested every link with a TDR. A couple failed - turned out the oh-so-slight crosstalk between T568B patch cables and the old T568R runs was just enough to break the link so I switched those old connections to T568B and all was well.
I've seen articles which claim the crosstalk from mixing A and B only sometimes cause link problems, but I've seen it often enough to make it a blanket rule to always, always, always go 568B. 568B is supposedly deprecated but every cable I've ever bought off the shelf, aside from crossover cable, has been wired 568B so I always stick with B.
Most of the premade patch cables that were on site tested bad BTW. I've since installed a few premade cables but they were brand new and those tested fine.
If you're going room to room, don't go with premade patch cables. Get a spool of CAT-6 and use keystones (jacks) on the PC side and a patch panel (or keystones if the boss is too cheap - although once you do more than 20 jacks the patch panel becomes much cheaper so just tell him to STFU and do it right, and skip one appetizer and alcoholic beverage at a meal to recoup the cost) on the other side. Just hanging a patch cable out of the wall is really hack. It works, but it's fugly.
The Christian Right is Neither (Christian nor right). See: Matthew 23, Matthew 25, Ezekiel 16:48-50
That's a fib. Cables over 100m are verboten.
I do agree with 100% certainty your experiment would prove you correct. Assuming you leave out one critical item, the 100 Ohm termination resistor that the switch adds to each end of the CAT5 cable to minimize reflections. Add the resistor in, and your experiment would then require $100,000+ equipment to find the proverbial meaningless needle in a haystack.
I've gotten a lot of CAT6 cables from monoprice.com and they are cheaper and better than I could make myself. They are seriously the cheapest place I have ever seen by a long long margin.
Cool! Amazing Toys.
I have a cat6 cable certifier. You can make cords by hand that certifies to cat6 the majority of the time. Something that isn't cat6 compliant isn't going to hurt your 100base, you only need cat5e for gigabit, cat5 for 100base. You can't tell if you meet cat6 spec without the $10k certifier.
A lot of people put rj45 mod ends on solid wire (instead of stranded). Then when the wire moves it pulls on the pins and 'goes bad'. Premade patch cords are always stranded, ones you make usually are using solid wire. If you use solid wire from jack to jack and tie it down, then use premade patch cords made from stranded wire from jack to device, you're fine. Or buy stranded wire and make your own. Putting a rj45 mod end on stranded wire is a little bit challenging. Your best bet is solid wire from jack to jack, tied down, prebought short patch cords from jack to device.
The rule of thumb I use: After you cram each wire into the connector, if you can't see light glint off the cut end of each wire when looking through the end of the connector, they aren't in right. Looking at the sides or the top or bottom of the connector won't do.
There is a minimum length, and a maximum length, and you are WELL advised to not screw that up when doing Gigabit Ethernet (and you'd need to be insane to do it to 10G). Especially when doing borderline stuff like gig-e over cat5e.
The problems with the lenghts are related to the minimum and maximum delays expected by the NIC. It gets especially important when you have more than one signal path (gig-e uses four, with reflected signal cancelation to use them bidirectionally), it is not just SNR or cross-talk that matters. Remember that the length of the four twisted pairs are NOT the same, because the twisting is NOT perfect, so even if the NIC can deal with the maximum delay for the group, it will NOT like if the length difference becomes expressive enough to exceed the maximum jitter allowed inside the group.
Yes, gigabit ethernet can work over cat5e in some cases, but here where I work, we have seen it fail more often than not, especially on the extra expensive, extra-crappy stuff IBM uses on their p-Series servers. How well the NIC will deal with crappage is also important, and just because it is coping, doesn't mean it is not doing a lot of error correction and retransmits.
Gigabit ethernet certainly doesn't do as well in cat5e as in cat6. Unless your cat5e is actually exceeds (for the better) the cat5e specs a lot, which is _common_ if you use top-quality cabling. And exceeding the maximum length is just stupid, get a pair of media converters and go over fiber if you are too cheap to have proper optical ports...
Hint: The injection molded ones are made by robots.
I used to work for a company that builds large machines to do exactly that.
You are not the customer.
Gigabit Ethernet all but eliminates half-duplex (it's technically in the standard, but nobody implements it properly, and autonegotiation is required), so collisions are no longer a concern.
The length restriction is still in the standard, but most of the time it will still establish a link at 110 meters of Cat5, let alone Cat5e or Cat6.
Convert FLACs to a portable format with FlacSquisher
Ethernet networks are increasingly 100% full-duplex. Such a network can't have collisions.
Strictly speaking, 100m is a fuzzy limit.
It's a hard limit for half-duplex. The CD part of the CSMA/CD Ethernet standard can't wait indefinitly. So they picked numbers that were just longer than the standard. Back in the old days, they would actually let you run fiber at half-duplex. I was working on an old 3-Com network that was set up by the lowest bidder. They had copper (full duplex) running at 200+ meters, but the fiber they had laid (before the copper, they had to retrofit the copper because the fiber didn't work) would fail because it would sense collisions when there weren't any and retransmit until it collapsed. One quick setting to "full" and they went from 10 Mbps copper to 100 Mbps fiber and things magically got better. So yes, the distances as far as signal level and quality are fuzzy, but if you run half-duplex (not that anyone does anymore), there is a hard limit just past. And no, I don't remember the number. Back then, I looked up the spec and calculated it by hand, but I don't remember it and figure someone has it on a web page you could find if you were interested.
Learn to love Alaska
But ethernet is a broadband signal with a huge range of frequencies.
No it's not. Ethernet is baseband (manchester coding). It needs very high frequencies to ensure a fast slew rate, but it is not broadband.
The hard limit on length applies only to a given shared segment or collision domain. For switched full duplex, there is no collision domain, so the only limit is signal integrity. 100M is a safe rule of thumb for copper, but with fiber you can go several km.
100-Base TX Ethernet runs (physically) at 31.25Mhz, not 100Mhz.
FWIW when doing comm/nav maintenance on (rather expensive) Air Force aircraft the harnesses, coax, and other connections are almost all hand-crimped. Been there, done that since a TDR weighed about 60 pounds not counting accessories.
It's easy to make good crimped connections.
"This post is an artistic work of fiction and falsehood. Only a fool would take anything posted here as fact."
> Just Remember: OverWeight Olga Gives Willingly,
> But Betty White Gives Bitchin'-Wild Blowjobs.
Or in haiku with where C = Cyan instead of B since we have brown and blue:
Oil Water Ocean,
Glowing With Color Crazy Whorls.
Get Back When Burning.