The great bulk of the copy already written about DSL Phantom mode, and the announcement this week that Nokia Siemens has eked out 825 Mbps over bonded twisted pair Telco lines, seems to say ‘don’t hold your breath,’ and the sentiment is why not get on with shifting to fiber. An immediate shift to fiber just does not add up.
The same was written, more or less, about VDSL, which is now the backbone of the AT&T network, alongside ADSL2+ and VDSL2, and the pathetic 500 Kbps that used to qualify as a broadband line in 2002/3 has now improved by an order of magnitude and then some. Most homes that are within reach of DSL can now order a 10 Mbps line, and with a small amount of Capex this can shift to 24 Mbps for many telcos.
Radio and communications engineers are a clever bunch. The idea of Phantom circuits first came about around 100 years ago and is the process of using spare capacity on two lines to create at third virtual line, which has the benefit of automatically cancels out crosstalk noise between the two. While neither Nokia Siemens or Alcatel, which released data on its own Phantom DSL in April, have fully explained how their version works, we have to remember that there are two lines in a DSL twisted pair and in a bonded set of lines this means at least four copper connections. Boring everyday bonding, which is now well understood, using ADSL2+ across these lines is likely to yield something in the order of just under 50 Mbps up to 3,500 feet from an exchange.
Let’s be clear about this, US homes on long loops are already a problem and this doesn’t solve that. Varieties of DSL manage to deliver a half decent service at around 6,000 feet from a DSLAM or IPSLAM, and this newer technology does little to solve this issue. Some US loops are 18,000 feet long, and can never have broadband put on them. So much of the disillusion out there is with the simple problem that copper cable inconsistencies and signal attenuation prevent DSL broadband spreading. But while many US (as well as French and Spanish) homes remain in limbo where broadband is concerned, those which are within that magic 6,000 foot limit, get an embarrassment of riches, even from telcos.
Cable in the US has deeper fiber than telcos, so that the fiber part of its lines are closer to US homes. It’s not magic? It’s history of how they built their networks. Telcos are building fiber closer and closer to homes and AT&T’s Project Lightspeed took the US average down from 12,000 feet to 5,000 feet for over 20 million homes and that upgrade is permanent and continuing. Cable, because it has a more uniform, purer and thicker copper cable inside a protected housing, can use more powerful, but also more sensitive, modulation schemes.
In original DSL it was 1.1 megahertz bandwidth of signals coded in Discreet Multi Tone, in ADSL2+ we used 2.2 megahertz. There is enormous complexity within the DSL standard and at set up each frequency is tested on that particular wire, signal to noise ratios observed, which result in a data allocation being made for each tiny slice of spectrum. The more spectrum used, the more data carried, but the shorter the line length. So in upgrading ADSL2+ to use 30 MHz of spectrum, the maximum benefit was reduced to homes within 3,500 feet of the exchange. This process HAS to continue, and while, with no fiber extension, this will be useful to less and less homes in a given country or territory, fiber is gradually getting closer and closer to homes, so the benefits do spread further afield over time.
In Cable, DOCSIS 3.0 allows for bonding of multiple channels in a Hybrid fiber coax, and no-one has said that this is fiction or complained that it took around 3 years from its final specification until operators began installing it in droves, but when Telcos talk about bonding and using newer techniques like crosstalk cancelation and Phantoming, they are pilloried by analysts who appear to be cable apologists or who think that overnight telcos should shift fiber to every home.
It is in the light of all this that you have to examine the Nokia Siemens Networks announcement that its ‘Phantom DSL’ line can offer 825 Mbps over 400 meters of copper lines. Quite clearly this is a headline figure. A service with 825 Mbps would instantaneously reverse the drift to Cable for broadband, but it would require far better fiber reach from all the world’s telcos. So we need to look beyond the headline figure and think about what can be done today, with the existing loop lengths. Another statistic NSN released was that Phantom DSL using bonding provides an increase in throughput of 50% to 75% over existing bonded copper lines.
Given that ADSL2+ performs at a maximum of 24 Mbps, and even supposing that when bonded it can reach 48 Mbps, then an increase of 50% takes this to 72 Mbps and for 75% this would be 84 Mbps. That genuinely puts it within the range of DOCSIS 3.0 real world performance.
We have to remember that DOCSIS 3.0 can bond up to 8 separate 6MHz channels, and that while the attenuation and interference issues are less in cable, there has to be that number of channels built into the system. The cable companies are harvesting spare capacity by using Switched Digital Video and other measures, and a lot of this is being used up on HD, 3D and VoD service provision, but for the most part there is enough left over for multiple slices of TV channel sized spectrum to be used for broadband for each home. It does depend on the configuration of QAM and the number of loop destinations on any given loop, and capex plays its part in keeping this down, so that spare capacity is available for each home.
Also backhaul and end to end internet performance (and broadband server capacity for instance) for cable and for telco is likely to remain the constraining issue, not the line, for some time to come. So the overall performance of an 84 Mbps line that reaches some 3,500 feet from the exchange, may be in the ball park of the 150 Mbps theoretical maximum of DOCSIS 3.0 when it comes to actual home experience. Certainly is could carry 3 or 4 HD channels, one of which might be 3D, and as much html and voice traffic as you like.
The next, and most important step here is to come up with a standard way of producing Phantom DSL so that telcos are not held to ransom by a proprietary system. Then, from the standard, first generation chips need to be made, after that these chips can slide into any DSLAMs which are shipped from now on, and once there are a noticeable percentage of such DSLAMs installed, and help desks are trained to cope with the different problems and performance issues, then a new service level can be marketed. Maybe that timescale is 2 years, maybe it is 3 years, maybe it is less.
So holding your breath is not an option, but at the same time, had you tried to hold your breath for DOCSIS 3.0 and bonding in general within DOCSIS, then you would have turned blue some time ago. And of course this is not a solution for reach DSL and we have to perhaps look to the economics of first or perhaps second generation Wireless LTE, for rural wireless broadband to change that equation. What these technologies do is take the loyal telco base and allow existing metropolitan DSL customers to remain on the leading edge for new service provision, which in turn makes them immune to the drift to cable broadband, once such a service is ready. There will be many more desertions over the coming couple of years, but price cuts and bonding will go a long way to slow this, and by then Phantom DSL should be ready.
Nokia Siemens says that Phantom DSL will become an integral part of the company’s DSLAM products hiX 562x/3x and Alcatel back in April revealed a similar R&D benchmark which talked about delivering 50Mbps to all users, peaking up to 100Mbps using VDSL2 bonding. Back in September Huawei also showed its SuperMIMO, which uses four twisted copper pairs to deliver a 700 Mbps downstream at a similar distance of 400 meters and said it can increase DSL bandwidth by 75%, the same as NSN is now claiming.
Where are all of these new twisted pair lines going to come from? That one’s easy, from the mobile fixed line substitution and the cable desertions that the US telcos have witnessed. That has left them with more and more unused copper pairs within their networks.
So perhaps DSL will cease to be an acronym for ‘dedicated slow line’ after all and it can continue to hang on to the coat tails of cable for some time, without a complete capitulation to fiber.