Use the play controls and links above (on the right) to play the audio now or download the audio to your computer.
ASSIA's John Cioffi discusses the physical limits of DSL, how telco DSL is improving and what's preventing us from getting more upstream bandwidth on our DSL connections
What I think Cioffi is recommending for AT&T is to build out U-Verse from FTTN to FTTC. I imagine you could work through issues like reusing the existing fiber, powering, node placement, and so on. I'm less clear on how a business case analysis for doing so would go. My bet would be that the optimum business case would be to hang on to what they have for as long as competitively possible, and then overbuild FTTH with GPON.
Cioffi points out that 1G downstream can be achieved with DSL by bonding 4 pairs. I'm not aware of any point in the network where there are 4 available pairs per subscriber; as far as I know, most distribution was engineered for something less than 2 pairs per home passed, and drop cable is typically 2 pairs. So if we assume linear scaling, the highest downstream rate we can practically expect is about 500 Mbit/s. I've argued that from a user experience perspective, we're starting to look at diminishing returns from increasing data rates in that region. But from a marketing perspective, when the cable guys start selling 1G downstream service, a hard limit of 500 Mbit/s becomes a liability.
Anecdotally... I spent last weekend in a U-Verse served community, and was struck by the number of AT&T trucks I saw on both Saturday and Sunday. These were clearly not new installations. And I never cease to be struck by the physical size of the VRAD, power supply and cross box, and the creative engineering that goes into placing them. Homes in this particular community routinely sell for a couple of Mill, and I think if I owned one of those homes I'd be more than a bit annoyed looking at all those huge beige metal boxes.
John is an amazing and very brite guy (Stanford what else?) but I worked with SBC and was in San Ramon, Ca. when Randall Stevenson decided to drop the TriBOC FTTH project in favor of FTT VRAD for Lightspeed (now uVerse) based in part on John's work.
In your interview Phil you ask about Gigabit speeds for DSL and John's response was to make the comparison to Gigabit Ethernet (IEEE802.3 GE) over copper. I worked on that committee and I can tell you that we spent a lot of time on inestigation of media capabilities for GE. Cat 3 was thrown out early as an option and as John said, we needed 4 pairs of cat 5 and that was for a nominal distance of 100 meters. We also were careful to define how to attach the cable to the connector and did not recommend taps or other commonly used telco "Bell wire" proceedures.
But Cat 5 or even Cat 3 is not what is deployed CONSISTENTLY in our legacy telco networks and this is where uVerse has suffered so much. Therefore making a comparison of DSL to Gig E is an interesting lab project for PhD candidates at Stanford, it is definitely apples and oranges to use for telco apps like HD video. And copper being an active element, changes properties especially in a hot deployment like sitting on a pole in the summer. So the speed potential question is very much dependent on the age and condition of the copper installed loops amongst other things. Yes there are new technologies that allow for overcoming the inconsistencies in the copper plant but those become IMHO a case of diminishing return on investment.
It was either Duh or Brook7 who posed the concept of a cost study for FTTH like FiOS vs FTTV and I for one would pay to see a valid study. He brought up some good critical study issues like IPTV churn etc but I fear this would be a futile effort. It is too easy to bury opex costs for a service like uVerse in generic POTS expenses.
Those who know me know that I have been working on ATSC HD video over Enet and to the prem via Ethernet GPON for over 10 years and this issue has a personal element for me. I'm in SBC/ATT country with a VRAD less than 100 meters from my home demark. We also have new cat 3 copper in my local loop. I signed up as an early trial customer for uVerse TV but was very disappointed with the HD video quality. ATT has set the max loop rate and the per channel video quality for the lowest common DSL loop denominator. I have been advised by some ATT lab people to ask for pair bonding and while that should improve my VDSL speeds, their video headend HD compression rates will still be set for their worst case loop condidition. My neighbors are execs for Yahoo, M-soft and Google with that Steve guy down the street. We need FTTH for uVerse here!
There was no lie nor slipperiness in the podcast as implied by some bloggers' comments. Let me try to respond here to each of the alleged misrepresentations in an attempt to clarify that indeed 100 Mbps at 1 km is possible on category 3 twisted pair:
Category 3 and category 5 twised pair differ from transmission point only in their twist rates (category 5 is tigher twisting so has greater immunity to crosstalk from other pairs). DSL vectoring as in ITU standard G.993.5 allows crosstalk to be cancelled and thus this twisting difference becomes irrelevent. The limit at 1km of .5mm (24 gauge) twisted pair is roughly 100 Mbps (even allowing for upstream speed simultaneously of about 25 Mbps). The single-pair limit at about 5000' is 35 Mbps. Of course, bonding of two pairs with phantom vectoring allows yet much higher speeds and indeed 100 Mbps is possible on two phantom-vectored pairs (so like 3 channels) at 5000'. These results depend on no crosstalk (because it is cancelled), -140 dBm/Hz background noise, and very good dynamic management of other customer-side noises. It is not easy, but these results have been known and accepted in the DSL community for many years. It really requires exceptional management and fully ITU-compliant (both modem and modem management) equipment to achieve these results beyond test labs (where it has already been demonstrated by several equipment vendors). A symmetric Gigabit is possible on 4 phantom-vectored pairs at 300m at least for fiber to the curb, and that is unshared physical-layer bandwidth simultaneously to each and every customer (unlike cable or PON where speeds are shared).
These results all consider the allowed power levels by the FCC and the various standards groups that limit power on lines, and there are no violations of emissions' limits implied by the results above.
I hope this helps understanding of the basic issues and clarifies that there was no misrepresenation nor slipperiness in the comments.
That is correct with one caveat. The investment is not linear so it is not like an exponential decay like linear returns. The investments come in chunks based on loop length on the drop. If you are at a CSA reach, you are at one number. At the DA, you are at another.
On top of that EACH time you make a change it is an investment. Think about U-verse. If they ever decide to do FTTH, AT&T tosses away all their access investment in U-verse. The amount of fiber laid is not enough and so you get to do it all over again.
Phil,
No - what I am trying to tell you is I can give you GigE today no new DSM if you are willing to redo your entire copper plant. Of course, you could redo it to fiber for less money.
See also Cioffi's guest column in the May issue of IEEE Communications.
No question that Cioffi is a self-promoter, maybe a bit of a huckster, as well as being one of the most significant communications engineers of the past three decades. He's done great work, but you always have to look for the caveats in everything he says and mentally divide everything by a hype factor. Perhaps that's the difference between a brilliant technologist and a great one: having the self-confidence to dismiss doubts, alternative approaches, and caveats.
In any event, I'd argue that the problem Seven raises is one of diminishing returns. Each increment of DSL speed comes at a higher cost, with deeper fiber construction, more remote electronics (and powering), more signal processing and probably more risk to get smaller increments of improvement. How far ahead do the planners have to look to see a point that it would make better sense to invest up front in FTTH than continuously tweaking copper? That's a hard question.
It would be interesting to do a retrospective comparative business case analysis of U-Verse and FiOS, insofar as actual normalized CAPEX, OPEX, ARPU, churn and similar factors.
Ah, I see. So you think he's inflating what's possible by talking about two different scenarios so close together that an average listener would conflate the two. Well, I don't think that was his intention but I'll ask him if he wants to reply to the charge here.
My point in interviewing Cioffi is that I think it's amazing how much more bandwidth CSPs could potentially wring out of copper. Just a few years ago copper was a dead end street. Now it may still be a dead end eventually, but it's a much longer street with much nicer homes. :)
No actually Cioffi talked about vectoring and ADSL providers. He is quite slippery in what he says and basically he says two things that are unrelated:
- Look the GigE guys can run over copper
- Hey we can run DSM over Cat-3 and get better rates
The problem is the question has to be asked "At the same time"? Yes it is possible to run Gigabit over Cat-3. Yes it is possible to get 5Kft service out of VDSL. The implication of what he said was that a generic Cat-3 network could get 100 Mb/s VDSL service at 1Km. He never said that in 1 sentence. He said that in three sentences closely placed. What listeners here is "At the same time" but he never says it. It is just implied.
So, the way to ask that is "On the exisiting U-verse network is it your view that every subscriber will be able to achieve 100Mb/s with a Vectoring implementation with no new construction - Yes or No." Leave no wriggle room. There was even a chip company (rhyming with Ikanos) that used to have a marketing slide that had 2 lines on their slides:
- Rates up to 100 Mb/s symmetric!
- Reaches up to 9 Kft!
The uneducated reader might assume that this was again "At the same time." Simply untrue. The chipset in different implementations (not even the same implementation) could do that.
One other power transmission factor that was left off of the discussion is that the FCC mandates the maximum power transmitted down a line. So even if you wanted to go above a certain limit, it violates the rule.
The gains from DSM are not from 1.5 Mb/s to 8 Mb/s or 25 Mb/s. Its to 3 Mb/s. Important for the carrier - absolutely. Solving world hunger....uh no.
No one said you wouldn't need fiber running to some point. I thought it was pretty clear we were talking about the last part of the last mile. As in, making something like U-verse even more competitive to something like cable broadband.
So, the whole notion that you are not going to have to run fiber into neighborhoods to support 100 Mb/s vectored VDSL is such a load of BS that I almost fell off my chair.
That is the whole frickin' problem. You can NOT do 12Kft at 100Mb/s. That is the requirement to not lay fiber (read that again - single pair 100 Mb/s at 12 Kft). Anything less than that you are in many cases having to do network construction to lay fiber.
Once you are laying fiber - the question is when does the last few hundred feet break the business case. There are other operational issues - the number of cabinets and battery trays that need to be supported. Those are your trade-offs. There is also the problem of housing density. Go run the short loop VDSL case in rural America where you will have 1 - 2 houses per remote.
Where the vectored solution works is for customers like AT&T where they have U-verse and they are already using short loop VDSL. If you are on ADSL at a FULL CSA then you have to run fiber. Now the question is how much fiber you run. NOT that a full CSA reach ADSL system can run 100 Mb/s without new construction.
I also like the idea that the use of the specialized twisted pair used in Gigabit copper networks validate the use of CAT-3 fiber in telco networks. Yes we can do that on ANY CAT-3 fiber that is installed and will NEVER require any new copper plant in any circumstances. That is what is implied but not stated. You can do this stuff on nice clean plant that doesn't have nasty things like T-1s on them. The cost to retrofit networks if they are problematic are so high that it basically means that you might be better off constructing a new fiber network instead of trying to use your copper network.
Remember how the loop lengths on U-verse grew shorter and shorter over time to compensate for this kind of stuff (They started if you look back calling out 5kft loops). 1Km = 3000 ft or about where U-verse ended up getting its 25 Mb/s service out there. That's what happens when clean room technology hits the real world copper network.
The blogs and comments are the opinions only of the writers and do not reflect the views of Light Reading. They are no substitute for your own research and should not be relied upon for trading or any other purpose.
To save this item to your list of favorite Light Reading content so you can find it later in your Profile page, click the "Save It" button next to the item.
