FOGGY DOCSIS AN ENABLENCE ARTICLE WRITTEN BY JIM FARMER, CTO APRIL, 2010 www.enablence.com
The whole cable industry is in a fog. It used to be just me in the fog, but since I saw the light and went over to FTTH, now all of cable TV is in a fog. And not just any fog: an R- fog! RFoG stands for Radio Frequency over Glass. A couple of friends of mine have claimed they originated the term, and to preserve decades-old friendships, I ll not name them. There have been moves to change the name, but so far no one has proposed anything that has caught on. Besides, you just about can t find an acronym that you can have more fun with than that one. What RFoG is, is a different form of FTTH that looks more like cable TV. It uses RF transmission in both directions, and doesn t use any baseband digital transmission, as do both EPON and GPON. Data goes by cable modem (the infamous DOCSIS standards). Video is normal broadcast, which works really well, you know. Voice, as usual in cable TV, travels over the DOCSIS infrastructure. Upstream transmissions back to the headend (CO to you Bellheads) travel on an analog RF-modulated laser just like that on which the downstream RF is transmitted. Since it is RF (modulated onto light) upstream, you can send any signal upstream, including RF return from set tops. The catch is that you have to detect RF coming out of the home, and you have to turn on the laser when you detect RF, then turn it off when the RF stops. You have to do this in order to avoid interference that would be generated if you left all the upstream lasers on. You also have work to do in order to keep RF and optical levels within their operating ranges, going in both directions on the same fiber. The RFoG standard is being worked on within the SCTE, the Society of Cable Telecommunications Engineers. While there are companies out there claiming conformance with RFoG, in fact there is no standard yet with which to be compliant. And the folks claiming compliance are not interchangeable, so the claims of compliancy really don t ring too true. There are also folks building GPON or EPON systems with bolt-on RF upstream transmitters at a fourth wavelength, who claim they are RFoG compliant. But of course, they really are not compliant, since the spec doesn t exist yet. So why RFoG rather than GPON or EPON? Well, that depends on who you ask. The basic idea is to preserve cable TV s investment in DOCSIS infrastructure, and to do something that looks just like cable TV. There are those who feel it is easier to manage RFoG, since you do it just like you manage DOCSIS, but there are others of us who feel it is so easy to simultaneously manage DOCSIS and GPON or EPON, that why bother with the extra foggy hassle or the lower data speeds? The cable industry sees RFoG as an easier way to meet the demands of developers who demand fiber-to-the-home. What we don t really know yet though, is whether or not the advantages of RFoG will be enough to carry the day in these developments. True, with RFoG you get some of the benefits of FTTH you get lower maintenance, no radiation or ingress, and probably a lot cleaner (read, higher-speed) upstream transmission. But you are still DOCSIS-bound, and that means that you don t really get all the advantages of FTTH. This gets us to cycling back to the subject of the last issue, DOCSIS 3.0 vs. FTTH. Again we d like to thank Dave Russell and Matt Schmitt for their very professional comparison of the technologies. 2 P age www.enablence.com
With DOCSIS, either over normal cable TV HFC architecture or over RFoG, you share RF frequencies between data and video. The laws of physics say that you can only get so much information into so much RF bandwidth. Now admittedly, when I watch certain TV shows or when I visit certain web sites, I question that there is any information (i.e., intelligence) transfer at all taking place. But in technical terms, bits are being transferred even if I think they contain no intelligence. But I digress. It works out that with current standards and technology, you can get into one traditional 6 MHz-wide TV channel, either about 10 standard definition (or two high definition) video programs, or 38 Mb/s of data. You can t put both in the same 6 MHz TV channel, even if neither contains any real intelligence (but I m digressing again). An operator has only a certain number of such TV channels available (158 RF channels for up-to-date FTTH and RFoG, 116 for most HFC networks due to when they were built). Doesn t really matter at this point in the conversation whether we re talking about workhorse DOCSIS 2.0 or channel-bonded DOCSIS 3.0. We ll get to that issue shortly. So for every 38 Mb/s of data an HFC or RFoG operator wants to send to a group of subscribers, he loses the ability to program two high definition or 10 standard definition programs. He has to figure out the optimum revenue and/or competitive split between data (including voice) and TV. In the upstream direction, things are grimmer for the DOCSIS guy, though helped with RFoG. The frequency spectrum available for DOCSIS upstream transmission is limited by what frequencies are available as a practical matter, and by interference. The frequencies used for upstream transmission are partially inhabited by high powered shortwave transmitters, atmospheric noise, and man-made noise from fluorescent lights, light dimmers, electric motors, and who knows what else. As you go toward the higher end of the upstream spectrum things get better, but you see that the operator is getting squeezed. What it means is that he has to be careful with how he uses the upstream spectrum. DOCSIS specifies a number of modulation formats in the upstream direction, with the tradeoff being more bang for the buck (that is, more bits per second per Hertz of upstream RF bandwidth) vs. more chance for the data to be corrupted by interference. The HFC guys have gotten really good at figuring out just what they can and can t do in the upstream direction, and they have gotten really good at optimizing plant to increase bang for the buck. But the laws of physics still limit what can be done in the upstream. RFoG is an advantage in the upstream direction, in that for technical reasons we can t go into here (Dave would kill me for making the article too long), the interference in the upstream spectrum is much less than it is for HFC. So someone deploying RFoG may actually get within about an order of magnitude of the upstream bandwidth available with EPON or GPON. Here s a good place to get into DOCSIS 3.0, which some misinformed folks think may be an FTTH killer (pardon, my prejudice is showing). Let s go back to talking downstream for a minute. DOCSIS 2.0 and 3.0 all specify exactly the same modulation formats, that is, the same bang for the buck, er MHz. So if you took several TV channels worth of DOCSIS 2.0, you would have the same total bandwidth you have with DOCSIS 3.0. What you would not have with 2.0 but would have with 3.0, is the ability to deliver that bandwidth to one subscriber. 3 P age www.enablence.com
With 2.0 you would not be able to deliver more than about 38 Mb/s to one subscriber, because that s all the bandwidth you get in an RF channel. With 3.0, you can bond several channels (four seems to be the sweet spot folks are talking about), to get more bandwidth. Bonding simply means I can split one data stream over several RF channels and put it back together again at the cable modem. This is the same thing as pair bonding in DSL, but in the frequency domain rather than in the pair domain. Upstream transmission in DOCSIS can do the same thing by the spec, and will do the same in practice when the industry gets to producing bonded upstream gear. The claimed downstream speed from bonding four channels is 160 MHz, and the claimed upstream speed is 120 Mb/s. Both numbers are playing a bit loose with rounding, but 3.0 is faster than 2.0 in both directions, at the expense of using more spectrum. Four downstream channels bonded will cost a total of 40 standard or 8 high definition programs, compared with 10 standard or two high definition programs for a single DOCSIS channel. But you get almost four times the data bandwidth. And RFoG will give you more downstream RF bandwidth than most HFC networks have, and a better shot at using the available upstream bandwidth. In both directions, 3.0 would allow DOCSIS to get within an order of magnitude or so of GPON or EPON. Of course, with either GPON or EPON, you don t loose any TV broadcast channels for your data, since data is on different wavelengths. So how many subscribers share data in DOCSIS? It really does depend. With RFoG, you have the same exact physical architecture you do with other FTTH (or so we presume, since the spec is not finished yet, and thus is subject to change). So one idea would be to share the data among 32 subscribers, just as we usually do in GPON or EPON. But that requires a rather expensive port on a CMTS, the equivalent of an OLT in GPON and EPON. It is technically possible to combine several PONs to one CMTS port (Also called a DOCSIS channel). In fact, it is not unusual to put a few hundred customers on one DOCSIS port. This drives the average downstream speed down to tens of kilobits per second. But it works because not every subscriber uses the bandwidth all the time. Maybe at a peak hour one third of subscribers are on line. Then many of them are downloading email, which doesn t take a lot of bandwidth (unless my wife has sent pictures of the grandkids again), and besides, if a packet gets delayed a few milliseconds for something else, who cares? Other folks are downloading web pages, which takes a bit of bandwidth for a few seconds then takes nothing for a long time. OK, some clown is watching a video, and that takes some continuous bandwidth, but today the video is optimized for small windows on a computer screen, not a big-screen TV (and indeed, if you put the video on a big screen you would not like the result). Those little pictures don t take all that much bandwidth. In FTTH (differentiating from RFoG for a moment), we have much more bandwidth, but tend to not get quite as much efficiency out of our bandwidth as does a DOCSIS channel with several hundred subscribers. This is because 32 subscribers may not be a big enough group to make the statistics we described above work as well as they might. Let me illustrate what I m talking about with a disgustingly oversimplified model. Don t take figure 1 as being representative of real data usage. 4 P age www.enablence.com
Figure 1. Percent channel utilization vs. time for different numbers of subscribers Rather take it simply as an illustration of how statistics work in bandwidth sharing. What this figure shows is two data channels, one having 5 subscribers sharing the bandwidth, the other having 25 subscribers sharing the bandwidth. In each case, if all subscribers are using bandwidth at the same instant, 100% of the available bandwidth will be used, but no more. For the 5-subscriber study, the percent of bandwidth utilized in any interval goes from 0 t 100 percent. There was a time when all of the bandwidth was being used, and there was a time when none of it was being used. Of course, there were also all of the in-between situations. For 25 users, the utilization also had the ability to go between 0 and 100%, but it really stayed between about 35 and 65% usage. Why? Because statistics were working better: the probability that no one wanted data, and the probability that everyone wanted data at the same time was a lot lower than in the 5-user case. Sure, it could have happened, but it didn t. If you were to expand this to compare the peakiness of a 32-user PON vs. the peakiness of, say, a 200-user DOCSIS channel, you would see that the 200-user DOCSIS channel was less peaky because of the larger number of people using the channel. Of course, real models are much, much more complex than this oversimplified example, but it serves to illustrate the advantage of large numbers of subscribers. Here s another tidbit that is non-intuitive at first, but makes sense when you think about it: as you give people more speed, the utilization of your network does not go up proportionally. That is, if you suddenly give each subscriber on a PON twice the bandwidth, your data load on the network will not go up. That s because each subscriber transfers the same number of bits he would have had he had the lower speed. Over time, of course, applications will be written to take advantage of higher speed, but until they get deployed, more bandwidth per subscriber results in the same utilization of the network. 5 P age www.enablence.com
In FTTH we have developed the bad habit of not taking statistics into account in thinking about PON data capacity. On the other hand, there are unknowns in how data will be used in the future, as we put on more IPTV and as we deal with more over-the-top video plus new applications I m not smart enough to think of. We can get a pretty good idea of what IPTV will do to bandwidth it is and will be the data hog, though not quite to the extent some people say but what else is going to happen? The lesson is to monitor capacity utilization, and make sure you have enough, without spending money on unneeded capacity. So now hopefully you are not in too much of a fog about RFoG, and maybe you have a bit more perspective on the difference between DOCSIS and EPON/GPON. And maybe I kind of justified the time I spent developing the over-simplified model to illustrate the advantage of statistics in data loading. For more information visit www.enablence.com 2010 Enablence Technologies Inc. The information presented is subject to change without notice. Enablence Technologies Inc. assumes no responsibility for changes or inaccuracies contained herein. Copyright 2010 Enablence Technologies Inc. All rights reserved. 6 P age www.enablence.com