That was some great research, Jim, thanks. And of course, you are correct in your conclusion. Copper "is," in this respect, faster.
I was a little surprised to see John Curtis in the NWFusion article trivializing minute differences, although he "was" comparing different media types, and not parallel links of the same media. In the latter sense, let me explain what happened several years ago in these regards.
Circa 1995, when a shortage of certain resins that were needed for plenum-rated insulation hit the industry, we had what became known as the "Great Category 5 Shortage."
Whether or not this shortage was real or fabricated is still suspect, in my eyes, albeit it was certainly real for those who needed to put in cabling systems at the time.
In order to conserve on this resin material, and to actually get some percentage more product out the door than would ordinarily be possible with the same amount of resin during a period of incredible growth, manufacturers, with the approval of TIA/EIA committees, decided to change the composition of the fourth pair in every the Cat5 cable coming off the run. Remember, each Catx cable has four pairs in total.
100 Mb/s Ethernet made installing Cat5 an imperative for many firms at the time, because up until then many enterprise networks were still using the old "level" system of cabling (pre-category x), or they were still using Cat3 and 4 at the time, which are now relegated to voice only, and virtually-non-deployed at this time, respectively, as far as new implementations of those grades are concerned.
Well, in order to conserve on these resins No Problemo, they thought. As long as it met national fire protection standards for plenum rated spaces, it would be fine. Or, would it?
By changing the composition of the insulation on pair number 4 they also changed its dielectric constant, making pair #4 slightly less capacitive (to ground, and between other pair elements within the same sheath, hence affecting other parameters, as well) than the other three pairs in the bundle. This would have no bearing whatsoever on protocols that used a single pair for "serial" transmission. But just about at that point in time (as is the case today with advanced line coding techniques that didn't exist then), there were no fewer than four candidates for 100 Mb/s Ethernet being weighed by the standards committees, or Ethernet-like protocols, and some of them used two pairs in each direction, simultaneously, while others used all four pairs in a parallel scheme of transmission, in half duplex mode.
Here, the fourth pair "did" cause problems for the parallel transmission schemes (those using more than one pair in each direction), because the net velocity of propagation (NVP**) of pair number 4 was slower than the other three pairs, throwing the envelope delay of the entire bundle of symbols being sent into unacceptable territory.
The net effect of this was that links of appreciable distance would wind up delivering garbage, because of the unsynchronized bit bundles being received, caused by excessive envelope delay distortion between pairs.
Eventually the shortages that caused this kneejerk were eradicated, but for a while these compromised cables were getting into the supply chain, allowing them to enter enterprise mainstream infrastructures, where they still exist today.
This is important to note for those companies who intend running super-100 Mb/s protocols (such as GbE) over their existing Cat5, because these newer protocols do, in fact, use all four pairs of the Cat5 or Cat5e, both ways.
**NVP, or net velocity of propagation, is a craft-friendly term for the unit of absolute delay measurement facilitated by handheld Category 5 field test units. NVP is similar to the index you mentioned in your post where you compared the speed of light in a vacuum to the speed of signal propagation in various other media.
I've seen NVP used in ANSI/TIA-approved test paradigms for the index, referencing the speed of light in a vacuum to those of various cable constructions. Usually, UTP NVPs are in the range of .72 and .78, where 1.0 represents the speed of light in a vacuum.
I've not actually measured light in this manner, although I'm familiar with other forms of time domain measurement that usually show propagation times, such as through the use of an optical time domain reflectometer, or OTDR.
But I suspect that fiber-optic NVPs would be closer to .60 or .70. Again, making them slower than copper.
Comments, corrections, welcome. Anyone?
FAC |
