Improving Bit by Bit

Chia sẻ: Dang Thach Quan Koliadang | Ngày: | Loại File: PDF | Số trang:4

Thêm vào BST

Báo xấu

94
lượt xem 2
download

Download Vui lòng tải xuống để xem tài liệu đầy đủ

To move signals of that frequency around efficiently within my transmitting station needed what can only be described as plumbing transmission lines made out of thin copper pipes and ceramic insulators—the alternative being expensive, specially constructed, but very high loss, coaxial cable. Yet here we are twenty years later happily shoving similar frequencies down twisted pairs without really thinking about the physics involved! Three factors about radio frequency signals are very important in understanding our problem:...

Chủ đề:

Bình luận(0) Đăng nhập để gửi bình luận!

Lưu

Nội dung Text: Improving Bit by Bit

Improving Bit by Bit . many installers and customers To move signals of that frequency around efficiently within AS DATA are finding that physical my transmitting station needed what can only be described R AT E S networks designed, installed and tested to the current Cat 5e as plumbing transmission lines made out of thin copper pipes and ceramic insulators—the alternative being RISE, or proposed Cat 6 standards simply won’t work properly at 100 Megabit Fast Ethernet, expensive, specially constructed, but very high loss, coaxial cable. Yet here we are twenty years later happily shoving similar frequencies down twisted pairs without really let alone at Gigabit speeds. Independent Management thinking about the physics involved! Consultant Philip Turtle explains some of the reasons why this is happening and how one manufacturer seems Three factors about radio frequency signals are very to have come up with the solution. important in understanding our problem: 1. They become attenuated very rapidly in a transmission There have been numerous reports during the latter half of line system such as our twisted pair, and the power loss 1999 about Cat 5e and “proposed” Cat 6 systems being increases with both frequency and line length. So the installed and passing category testing. Yet when presented signal you get at the receiving end is much, much less with real network traffic they have failed to pass data than you started with, a mere 0.005 watts out for every sensibly at all, even though the link lights on the network one-watt you put in at Cat 5 limits. interface cards (NICs) are “on,” both on the PC and the hub 2. They treat every wire as an antenna radiating energy or switch. into other conductors and receiving energy from them, too. Just think, you are trying to receive a signal which The practical solution in many cases has been to switch the is only five thousandths of the high power signal in an 100 Mb/s NICs from Full Duplex to Half Duplex, effectively adjacent wire, a fantastic recipe for the interference we halving the theoretical maximum bandwidth to 50 Mb/s know as “crosstalk.” and in fact, taking the practical bandwidth much lower. 3. They reflect or bounce back quite dramatically off any “discontinuity” that gets in their way, and any poor It’s all about radio waves. connections and impedance change or “mismatches” along the way prove to be very good reflectors. Every As a teenager, I was a keen ham radio practitioner and built bit of power that is reflected doesn’t get to the far end, a VHF radio station operating on the “two meter” band at reducing our power received to even less than five frequencies around 144 MHz. Why do I tell you this? thousandths of what we put in. And if that’s not Because Cat 5 systems use 100 MHz rated components and enough, the reflected signal interferes with and distorts cable, Cat 6 uses 200 MHz—both very close to my radio the signal which is going in the correct direction and station frequency. can often turn a “zero” into a “one” or vice versa. Figure 1 Attenuation, crosstalk, reflections NEXT and digital signal processing. Without the wizardry of digital signal processing, high speed NEXT data over twisted pairs just wouldn’t work. If you look at figure 1, you’ll see how the signal goes in and how it comes out, and that it is unrecognizable. The trick is that most of KRONE: 800-775-KRONE www.kroneamericas.com www.truenet-system.com. No part of this document may be reproduced without permission ©2000 KRONE, Inc.
the crosstalk or noise is from a known or predictable source. The Cat 5/5e specifications require every component to have In fact, most of the noise (or unwanted signal on the receive nominal impedance of 100 ohms, but they allow a tolerance pair of a Cat 5 cable) is generated by the adjacent transmit of plus or minus 15%. This means that at any of the points pair. So, since we’re generating the transmit signal in the we’ve mentioned, there can be an impedance mismatch of NIC, it is a relatively easy job to up to 30 ohms (or 30%) and the system is still within spec. subtract the interference signal Yet as we now know, reflections are a very serious problem BIOGRAPHY (which may well be greater than at higher data rates and a 30% mismatch would have Philip Turtle is an independent the signal you are trying to terrible results. Actually, if it wasn’t for the fact that the management consultant in the receive) and recover the weak reputable manufacturers use closer tolerances and build Internet, communications and signal that we wanted. significant “headroom” into their systems, many installations public transport sectors, wouldn’t work at all! as well as a freelance That’s great. At least it was when we were dealing with 10 Mb/s But even though the reputable manufacturers take these technology writer. applications like 10Base-T precautions, the result is only assured if you use their He is immediate Ethernet. Unfortunately as we components throughout the system. If you mix and match, past national get faster, 100Base-T and then you are absolutely guaranteed to have mismatches. As an chairman of the 1000Base-T, the effects of example, one manufacturer might produce cables at Institution of Electrical unpredictable noise become 110 ohms (±5) ohms because they’re easier to manufacture Engineers’ (IEE) Group and worse. Alien crosstalk signals, and they’re inside the specification. However, use the cables is actively involved in the induced from other Cat 5 cables, in a network with another manufacturer’s 95 ohm revitalization and become significant; and since we connectors, and you can just see those reflections building up. restructuring of the IEE’s didn’t generate them in the NIC, Knowledge Services Division. we cannot correct for them. We Recent lab experiments at KRONE’s Laboratories (with call this “alien crosstalk.” deliberately mismatched components that were all well Likewise, reflected power from impedance mismatches is Impedance vs. Distance proving to be an absolute killer and in the specifications, 130.6 Return Loss (the measurement for reflected power) is listed 121.1 Mismatch Patch Cable Horizontal Connection Point as “For further study,” meaning no one knows what figure 111.7 Horizontal Cable is acceptable and so no one tests for it! And a further Ohms 102.2 impedance mismatch problem occurs where two mismatch 92.8 or reflection points are particular distances apart: standing 83.3 waves are set up. (Remember holding one end of a rope still 73.9 and waggling the other until it formed a constant sine wave 0.1 9.9 19.7 29.4 39.2 130.6 58.8 68.5 ...... pattern?) These standing waves can absorb lots of energy Meters making sure hardly any arrives at the far end! Figure 2: Impedance versus distance of mismatched channel configuration. Note the impedance mismatch occurs at the near end, approximately 4.2 meters into the channel. Reflection points within the Cat 5e spec) showed just how real these We have talked quite a lot about reflection points, and impedance mismatch effects are. Looked at in the time recent lab tests have shown just what a disastrous effect domain (Figure 2), the system falls within the specification they can have, but what are they? limits. But you can see the mismatch effect and you can start to imagine the potential for reflections at the patch and In a Cat 5 or Cat 6 cabling system, the obvious impedance horizontal cable connection point. change points are: s RJ45 plug to RJ45 jack (particularly if provided from Looked at in the frequency domain (Figure 3) you can see different manufacturers). that the impedance varies wildly far outside the specified s RJ45 jack to horizontal cable. limits at particular frequencies. In fact, you can see some of s Horizontal cable where it is severely bent, kinked the effects of these reflections in the associated attenuation or squashed. versus frequency graph (Figure 4), a characteristic not tested s RJ45 jack to patch cord. on site, which takes the received signal well below the spec s And the worst culprit of all, patch cords. level at frequencies which correspond to those wild KRONE: 800-775-KRONE www.kroneamericas.com www.truenet-system.com. No part of this document may be reproduced without permission ©2000 KRONE, Inc.
impedance swings. So it is no wonder installers and users In fact, they quickly become the major part of network are having problems on site! traffic (Table 1). And just for good measure, of course, in a system with lots of reflections, quite a few of the retransmitted packets will also get garbled and have to be Impedance vs. Frequency retransmitted a second or third time. 130.6 121.1 In practice it is quite possible to produce horizontal cable 111.7 102.2 and connectors to the exact nominal impedance of Ohms 100 ohms (not some other nominal figure because it is 92.8 83.3 easier to manufacture) and to control the manufactured 73.9 products’ impedance to within a few percent of that 100 64.4 ohms. In separate tests, KRONE labs discovered that, in sending one million bits through a Cat 5 system with only 55.0 1.0 28.6 56.3 84.0 111.7 139.3 167.0 194.7 six ohm variations, some 365,000 (or 36.5%) were actually Frequency (MHz) being sent again. Of these 365,000 retransmissions, a Figure 3: Mismatched channel configuration frequency domain. further 36.5% had to be sent again and so on. At the end of the day, Ethernet got the data through perfectly, but it Attenuation vs. Frequency had to send and receive 1.6 million bits to get one million -0.0 correct bits. That meant a reduction in network throughput -4.4 of nearly 40%. KRONE engineers claim to have seen far Insertion Loss -8.9 Deviation worse on site. “Field experience has shown that some low -13.3 end systems are operating at as little as 4% of supposed dB -17.8 capacity, that’s only 4 Mb/s from a supposedly 100 Mb/s -22.2 system,” KRONE Technical -26.7 Services Manager -31.1 SKIN EFFECT 1.0 28.6 56.3 84.0 111.7 139.3 167.0 194.7 Karl Tryner claims. Frequency (MHz) At radio frequencies, electrical energy or power does not flow Figure 4: Mismatched channel configuration (attenuation versus frequency) shows insertion loss deviation greater than the standard. along the inside of the Is there a solution? conductor, but actually travels around the outside Bit errors Yes, radio to the rescue! as an electromagnetic field. But you knew I would say The ultimate effect of all of the problems we’ve looked at is This is why crosstalk is such that. The theory, at least, is that bit errors occur. “Ones” arrive as “zeros,” “zeros” arrive a problem, because every quite simple. If you closely as “ones,” and either way, the chunk of data that we were conductor in the system is a impedance match all of the trying to send ends up as garbage. Ethernet and other radio transmission antenna. elements of the system, protocols are thankfully not stupid and can detect when then the amount of power they receive a packet of garbage. Then they shout back to reflected is reduced, so is the distortion it causes, and so are the standing waves that produced all those. The big problem % of Retransmissions Data Rate is with the patch cords because the cable used is flexible, the 0% 100Mbps geography within the cable changes every time the cable 1% 20Mbps moves. And therefore so does its impedance. In fact, a 2% 4Mbps 100 ohm patch cable can quite readily change from 3% 800Mbps 85 ohms to 115 ohms in the space of a few seconds as you 4% 160Mbps re-patch it or run it around the cable management. The only 5% 32Mbps (modem speed) solution is a new design of flexible cable, which does not exhibit these wild impedance changes. KRONE has done this, Table 1 the other end, “Sorry, didn’t get that one. Can you send it indeed it is one of the main advances in KRONE’s “TrueNet™ again please?” This is okay for one or two garbled packets, Technology,” which has been demonstrated at trade shows but as the incidence of garbling increases, so do the requests around the world and with which they are claiming they for retransmissions—and the retransmissions themselves can guarantee zero bit error rates on both Cat 5e and become a significant but invisible part of the network traffic. Cat 6 solutions. KRONE: 800-775-KRONE www.kroneamericas.com www.truenet-system.com. No part of this document may be reproduced without permission ©2000 KRONE, Inc.
TrueNet technology Impedance vs. Frequency 140.0 KRONE has applied my beloved radio theory and produced a 130.6 patented TrueNet Technology patch cable wherein the 121.1 individual copper strands are somehow glued to each other 111.7 Ohms to maintain an overall cylindrical shape and the internal 102.2 geography of the cable is very closely constrained while 92.8 83.3 maintaining flexibility. The end result is a patch cable that 73.9 has an impedance of 100 (±3) ohms — a factor of ten times 64.4 better than the spec. 1.0 28.6 56.3 84.0 111.7 139.3 167.0 194.7 222.3 Frequency (MHz) But you cannot buy KRONE’s patch cable except as made-up Figure 6: Well balanced channel configuration (frequency domain). patch leads because they believe that the quality of termination, strain relief, and testing as a finished assembly The overall result is a physical layer solution, closely is the only way to be sure of total system performance. impedance matched throughout, both in the time domain (Figure 5) and the frequency domain (Figure 6), which Impedance vs. Distance 121.1 Attenuation vs. Frequency Mismatch -0.0 Patch Cable Horizontal Connection Point 111.7 Horizontal Cable -4.4 102.2 -8.9 Ohms No Deviation 92.8 -13.3 83.3 -17.8 dB 73.9 -22.2 0.1 9.9 19.7 29.4 39.2 130.6 58.8 -26.7 Meters -31.1 Figure 5: Well balanced channel configuration -35.6 (impedance versus distance). -40.0 1.0 28.6 56.3 84.0 111.7 139.3 167.0 194.7 222.3 250.0 Frequency (MHz) They make them only to specific lengths so as to avoid known standing-wave wavelengths and their associated Figure 7: Well balanced channel configuration (attenuation versus problems. KRONE connectors have always been frequency). No insertion deviation. manufactured to 100 (±3) ohms so they match the patch cords well. KRONE’s cable operation has also developed exhibits such low levels of reflection, attenuation deviation TrueNet Cat 5e and Cat 6 horizontal cables that are not only (Figure 7) and alien crosstalk, that KRONE is prepared to 100 (±3) ohms, they also include the neat device of a guarantee the network to be bit error free for five years. helically round central core which significantly reduces the They even test your as-installed network free of charge to effects of alien crosstalk. verify the performance warranty. KRONE, Inc. North America Headquarters Impedance* 6950 South Tucson Way Englewood, CO 80112-3922 Impedance is a measure of the total opposition that a circuit or a part of a circuit presents to electric current. Impedance includes both resistance and reactance. The resistance component arises from the Telephone: (303) 790.2619 collisions of the current-carrying charged particles with the internal structure of the conductor. The Toll-Free: (800) 775.KRONE reactance component is an additional opposition to the movement of electric charge that arises from Facsimile: (303) 790.2117 the changing magnetic and electric fields in circuits carrying alternating current. www.kroneamericas.com The magnitude of the impedance Z of a circuit is equal to the maximum value of the potential difference, www.truenet-system.com or voltage, V (volts) across the circuit, divided by the maximum value of the current I (amperes) through the circuit, or simply Z=VII. The unit of impedance, like that of resistance, is the ohm. Depending on the nature of the reactance component of the impedance (whether predominantly inductive or capacitive), the alternating current either lags or leads the voltage. *“electrical impedance” Encyclopaedia Britannica Online KRONE: 800-775-KRONE www.kroneamericas.com www.truenet-system.com. No part of this document may be reproduced without permission ©2000 KRONE, Inc.