Description: Cat getting out of a bag and other observations
File name: Cat getting out of a bag and other observations
Category 6 cables -- and to a lesser extent, Category 6a cables -- are widely available from a variety of vendors, including variety stores, office supply stores, consumer electronics stores, and online shops. While a great many people are content to run wireless networks in their homes, the fact is that if you need truly high data speeds and a secure network, nothing beats a wired connection. But is that cable in the store, or in the online shop, really Cat 6?
It's a vexing problem, if you've got a cable in a bag and you'd like to know how well it conforms to Category 6 specifications. None of the tools you're likely to own, such as a volt-ohmmeter, will shed any light on the subject. Indeed, if you haven't spent a lot of time digesting cable specifications (we do that here, but it's definitely an acquired taste), you may have no idea at all what makes a cable "Cat 5e" or "Cat 6" or "Cat 6a." You can pick bulk cable up at the hardware store, and physical examination is rather unrevealing; in fact, the similarities are much greater than the differences, and you may well wonder what makes one bundle of eight little wires better at running high-bitrate data than another.
So, What's the Difference Between Categories?
Network cabling, in whatever "Category," is at root a simple paired-conductor cable design. Signals are sent down each data pair, and the two wires in the pair are twisted together to improve noise rejection and to ensure that they have very similar length; these are important aspects of the cable because signals are sent in a "differential mode" where neither wire is grounded and the receiving circuit has to measure the difference in voltage between the two wires in the pair. Good control over twisting results in excellent "common mode" noise rejection, where noise which affects each conductor equally effectively disappears from the signal, and control over lengths is important both to common mode noise rejection and to keeping the signal from falling apart due to intrapair skew (this is not the "skew" you are likeliest to be familiar with--it's the difference in timing between the two wires in the pair). Because the frequencies involved here are high, the dimensions of the pairs, and their spacing and relationship to other pairs in the bundle, are all important. These attributes of the cable affect its characteristic impedance, which must be kept as stable as possible to prevent signal degradation, because impedance variations cause portions of the signal to reflect within the cable or connector.
These simple features--the physical layout of the cable, the consistency of its dimensions, the rate and consistency of twist--cause a variety of effects upon high frequency signals run through the cable. To perform reliably and up-to-spec, network cabling needs to be able to meet some rather exacting specifications, which relate directly to just that: what happens to high-frequency signals run through it. High-speed data signals are not at all easy to convey reliably from point to point. If you've heard people say that it's very easy to do because, after all, it's digital, and "just ones and zeros," -- well -- that's one of those ideas which is intuitively very appealing and which turns out to be terribly wrong. To put it simply, the difficulty of actually getting those ones and zeros, at very high bitrates, through a cable and out the other end is surprisingly great, and sometimes for reasons which are far from obvious. In fact, contrary to common belief, it's not even really just ones and zeros--the different encoding schemes involve as many as sixteen distinct signal levels. There are, in fact, very good engineering reasons why data centers are generally not built using the cheapest available imported data cabling.
Okay--So, How Do We Measure Compliance?
Category 5e, 6 and 6a patch cords are governed by specification; the usual spec cited is the TIA spec, TIA-568.C.2, part 6.9, though there is also an ISO spec which is somewhat more stringent. By contrast to the standards that apply to such things as "horizontal" installed cabling and the like, the specification for patch cords is fairly simple. First, it requires that patch cords be built out of cable and connectors which comply with the relevant cable and connector specifications. Second, it requires that the cable meet or exceed performance standards for two things: Near-End Crosstalk, or NEXT, and Return Loss. Both of these attributes measure how much signal degradation a cable causes, and both are highly dependent upon the dimensional characteristics of the cable, which in turn depend upon good manufacturing practices. Consistent drawing and extrusion of the cable, consistent twist rates, and consistent dimensions and materials make for good cable. Additionally, the connectorization of the cable is a major issue: if there is too much untwisting, or other rearranging of the conductors in order to get into the connector, both crosstalk and return loss will suffer; likewise, if the connectors that are used are not up to snuff, they can cause all manner of trouble.
If you're not familiar with these attributes, some short definitions are in order. Crosstalk is the tendency of the signal in one pair to induce a signal in one of the neighboring pairs, and the specification requires that crosstalk be measured from each pair to each other pair, as crosstalk frequently is much worse between some pairs than others. In particular, due to the unfortunate decision made long ago to make network cable pinouts compatible with telephone wiring pinouts, the pair which is split between pins 3 and 6 (green, in T568B) tends to have higher crosstalk with the pair it passes around, which is on pins 4 and 5 (blue, in T568B). Return loss is the loss caused by signal reflecting when it hits impedance discontinuities in the cable. Both crosstalk and return loss are heavily affected by cable manufacturing quality, and in particular by the internal layout of the cable, the consistency of the dimensions of the pairs, and the consistency of twist rates and spacing. Both crosstalk and return loss are also affected by termination; the worst impedance bump in any cable, well-made or not, is encountered at the connector, and the worst crosstalk performance occurs at the connectors as well because conductors need to be straightened out and, in the case of the 3-6 pair, split up.
So, the difference between Cat 5e, Cat 6, and Cat 6a cable is not so much in the basic design as it is in the tolerances. As the bitrate, and correspondingly the frequency, of the signal increases, smaller and smaller discontinuities and inconsistencies in the cable become relevant. Cat 5e is required to meet certain specs for signals up to 100 MHz (one "Hertz" is one complete wave, e.g., a sine wave, per second, and a "Megahertz" is one million of those per second). Cat 6 is required to meet tighter specifications, and to meet those specs to 250 MHz. Cat 6a must meet the same specifications as Cat 6, but must also meet similar specification limits all the way out to 500 MHz. So, while the basic cable architecture doesn't change between Categories, the demand for consistency and quality in manufacturing does; for example, a sloppy connectorization that's "just good enough" for a Cat 5e cable will almost certainly cause the cable to fail at Cat 6.
The Slippery Side of the Cable Business
The fact that it's hard to tell the difference between Cat 5e, Cat 6 and Cat 6a by looking at them opens the door, unfortunately, to some deceptive practices in the industry. The door is further opened by the fact that, unlike some specifications which have a licensing agency and an enforcement squad (e.g., HDMI Licensing, which licenses the use of the HDMI trademarks and enforces the specification), these Ethernet specs are operated purely on the honor system. If a manufacturer wants to sell "Cat 6" cable, all he has to do is change the jacket lettering on his Cat 5e cable to read "Cat 6." Beyond that, it's pure caveat emptor -- it's the buyer's job to figure out whether he's being scammed.
The idea that somebody would just change the jacket lettering on his Cat 5e cable and call it "Cat 6" might seem just a bit too brazen, even for a sharp operator. If you have just a bit of faith in humanity, you'd think that nobody would label a cable "Category 6" on the jacket and sell it in a major national store chain without ascertaining that the cable actually met Category 6 specifications--but if you thought that, you'd be wrong. A few years ago, Fluke corporation, who make various Ethernet test devices, announced that in its survey of the market approximately 80% of the patch cords sold as "Cat 6" did not meet the specification (see Fluke article). This can be a severe problem in data networks, even when the permanent link cabling is of high quality -- for more detail on this point see this excellent Fluke online presentation.
The Ugly Truth
When we began work to develop our own Cat 6 and Cat 6a cables, we knew we would need to test every assembly; while Cat 5e cables are easy to assemble without a lot of cause for worry over compliance, Cat 6 and Cat 6a are another matter. Defects in assembly that one barely notices when putting the connector on -- things like just a bit too much split of the members of a pair, or too-abrupt bends in conductors as they route into the connectors -- can cause failure. Near-End Crosstalk and Return Loss are both sensitive to termination quality, and a failure of either means failure to meet spec. Accordingly, we invested in a Fluke "certification tester," model DTX-1800. The DTX is basically a purpose-built network analyzer programmed to evaluate network cables; instead of needing to do four passes through a network analyzer to evaluate return loss on the pairs, and then six more passes to evaluate crosstalk between all pair combinations, the operator can just press a button and the DTX runs all of those tests, sweeping through the whole range of frequencies and measuring the cable's compliance with relation to the specification. You can see a sample report, and explanation of what it means, at this page. The Fluke is a lovely piece of gear. We'd tell you that everyone who ever uses network cable should have one, if the unit with its various patch cord adapters didn't cost over ,000.
Fluke's claim that 80% of so-called Cat 6 patch cords were noncompliant seemed surprising, and we wondered whether things had gotten any better -- surely they had? -- since then. With the DTX on the desktop, this became a fairly easy question to answer, and we decided to go shopping.
Now, let's have a look at some Ethernet cables in common circulation. We've gone around and purchased a variety of patch cords that were labelled "Cat 6" and "Cat 6a" from four brick-and-mortar and four online vendors, and we've tested these on the DTX against the actual spec. We don't want to get a sheaf of letters from lawyers threatening litigation over our publication of these results, and so have made the decision to keep the vendors, and the brand names, anonymous (and please don't ask us about it over the phone; the individual who performed the testing has intentionally not informed the others at our office as to who these vendors are, or which results correspond to which vendor). What we will say, however, is that this is very, very far from being a list of no-name fly-by-night offshore vendors. These are serious, large companies with well known names and with reputations to lose, and if you are a US resident familiar with places to buy cable on the street and online, you know and probably have purchased something (even if not a cable) from most if not all of them. We did not cherry-pick these results; every cable we tested is shown.
A couple of quick testing notes: we used the TIA specification (easier than ISO) for all of the Cat 6 and 5e testing below. For 6a, our tester supports only ISO, so we used ISO there. Also, we have in a couple of cases noted "marginal pass" results; what this means is that while the cable passed testing, it did so by a margin narrower than the tester's margin of error. All of the test reports, in .pdf form, are linked to the result column and, where applicable, the "Additional Tests, Notes" column.Vendor Description Length
(feet) Category Result in
Category Additional Tests in Lower Categories, Notes
Well, That Wasn't Very Good, Was It?
No, it wasn't. A few observations are in order.
To start with, of course, the near-unanimity of the results is pretty surprising. Out of twenty cables tested, four met spec, and of those, two did it by a hair. This 80% failure rate is, as it happens, exactly what Fluke reported a few years back, and our sample certainly shows no improvement over that time. Plainly enough, most of these cables aren't designed to meet Cat 6 or 6a specifications, and they're certainly not tested for compliance before leaving the plant.
Perhaps the most surprising aspect of the results is that, out of the sixteen cables which failed their stated spec, all but five also failed Cat 5e testing. Cat 5e is not a difficult standard to pass; using conventional, non-bonded, twisted-pair Cat 5e of good quality (e.g., Belden 1583A) and ordinary RJ-45 connectors, not only is it easy to build compliant Cat 5e assemblies, but it's common to see them pass the specification by several dB both on NEXT and RL. Contrast that with some of the shockingly bad test results--e.g., the 50-foot major-brand cable from the hardware store, which failed the 5e spec's return loss requirements by a whopping 8 dB. Plainly, many of these vendors are using very, very poor cable stock. Anybody with an extruder and a wire-twisting machine can make paired data cables; making them meet spec, however, is another matter entirely.
Note that of all of the brick-and-mortar store cable purchases we made, the grand total score is zero passes, eight fails with only two of the purportedly "Cat 6" cables even passing Cat 5e. That's shameful, and what it means is that there is a good chance that, in your community, there is not a single place you can go and buy an honest-to-gosh Cat 6 patch cable; the closest you can come is to buy a cable that says "Cat 6" on the package and which may, if you're lucky, be Cat 5e.
The online vendors didn't do all that much better, though there are some passes amid the fails. We found no vendors who consistently passed, but some who consistently failed. What this suggests is that even those who do a better job than the others fail to consistently test their assemblies against the specification, and whether any particular assembly passes the spec is purely a guess unless you've got the gear to check it.
We've had a few calls and contacts from people who really just do not believe the testing results above. Surely, they think, it can't be that bad, because data professionals who build server farms have got to rely upon quality patch cords. But it really is that bad, and when we talk to people who build data centers, one of the things we hear often is that the quality of commonly-available patch cords is very poor. There are other companies, like us, who build individually-certified patch cords for data networks, but we don't know of anyone else who does that for the consumer market. If you're buying a thousand patch cords through a professional electronic, broadcast or data supply house, yes, you can get quality patch cords; but unless that's the business you're in, chances are that you wouldn't recognize the names of the vendors or the manufacturers of those cords, and you won't find them at the typical consumer-market price-driven cable vendor.
Our sales pitch at this point seems a bit obvious, if not redundant, so we'll keep it short. With every Cat 5e, Cat 6, or Cat 6a cable we sell, you will receive the test report for your cable, run on the same test apparatus on which these cables were tested, and confirming that your cable meets the specification. If we build a cable that doesn't pass, it doesn't ship.
Back to Articles Index
Back to Blue Jeans Cable Home