Fast Net News, DSL Prime, DOCSIS Report

Stephen Cooke offers on DSLR one possible explanation.: 

"In general each pair is a ground wire & a signal wire. What 'Phantom Mode' is is simply sharing a common ground wire in a multi-pair configuration. They re-purpose the ground wires from each additional pair, after the first one, and convert those into signal wires. Therefore for 2 standard pairs they get 1 ground wire & 3 signal wires. This takes out the capability to use common mode rejection between the individual wires of the pairs to reduce noise. Vectoring helps with that." 

Chipmakers working on the required high performance chips haven't gone beyond FPGA demos, so the 2012 hope for production gear isn't assured.  They are being coy and apparently all have different schemes. Promises of the best chips are supplemented by hints their competitors will fail miserably.  An additional problems will be integrating the new vectoring chips delivering the DSM Level 3 features with an effective DSM Level 1 control system. Cioffi tells me ASSIA is confident their system will do the integration well and that integration may prove particularly difficult for others. That's unproven until we have results from the field.

      Some schemes do not "repurpose the ground" and maintain common mode rejection capacity. An engineer tells me the "asymmetric phantom" as described above by Cooke,  has way too many drawbacks. 

   Cooke adds some of the news reports had things wrong.

Bonding allows you to make the 3 signal wires look like a single communications channel. Some of the reports seem to mistakenly imply that the 300 Mb/s happens over a single pair. Not true. It is a multi-pair implementation.

John Cioffi sent a note perhaps a little miffed because I hadn't credited him or his company, ASSIA, who are the leaders in this kind of work. Besides the ads, I've reported so often about Cioffi and vectoring/DSM that I assumed readers could make the connection without my including it in this story. But John has a point, especially in light of his article from 2004. Regardless of the history, Alcatel's physical demonstration of the technology is a welcome confirmation.

Phantom of the DSL
Figure 19 illustrates the basic concept of a phantom signal for two twisted pair (the use of the term “phantom” is somewhat misguided as the signal is very real and exists – indeed the phantom is what creates what we normally call crosstalk, which exists). For quad situations, the phantom component can be high. Typically the phantom is defined between hypotentical “center tap voltages” of the two pair, denoted as ΔV p in Figure 19. In reality, there are 4 conductors that create 3 closed loops, not two, if appropriately terminated and excited as a group. The third extra loop can have very real and significant data-carrying capacity. It can thus be exploited in bonded and vectored schemes. The gain can be more than 50% or less depending on loop geometries, imperfections, and coupling. Bonding
lines with significant coupling may indeed be physically and or practically difficult, except for the case of so-called “quads” where the 4 wires of a single quad twisted as an ensemble (so a twisted quad) may in many cases be maintained to a customer. The concept of a quad generalizes to any number of conductors but would then require yet further bonded and vectored use of multiple pairs. In reality, this is a random possibility for transmission that is very real (i.e., not a phantom) but would not be reliable unless pairs were selected carefully for customers. The data rates for vectored quads (based on French cable) are shown in Figure 20.

Such auto-selection and routing of pair in a cable in a distribution frame and at customer premises may be possible in multiple-dwelling units or elsewhere and likely represents the last known physical capacity in the copper loop plant that can be exploited to squeeze the very last highest rates from copper. It would be highly adaptive and capacity would need to be assessed by a DSM center on a binder-to-binder basis for the various customers. Given the high costs of fiber in the very last segments of the network, this last frontier of bandwidth in phantoms may need to be someday exploited. Fortunately, DSM offers significant bandwidth increase of very large factor even without exploitation of this last opportunity.

 

From International Symposium on Subscriber Loops - Edinburgh, Scotland - March 2004 
DYNAMIC SPECTRUM MANAGEMENT
J. M. Cioffi1 , M. Mohseni

Big conflict of interest note: As the ads make clear, I have a financial relationship with ASSIA and also am on their advisory board.